Reactor Analysis (Holmes and Meghreblian)

Reactor analysis [by] Robert V. Meghreblian [and] David K. Holmes.
Meghreblian, Robert V. (Robert Vartan)

New York, McGraw-Hill, 1960.
Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Public Domain, Google-digitized / http://www.hathitrust.org/access_use#pd-google
http://hdl.handle.net/2027/uc1.b4118385
Public Domain, Google-digitized

http://www.hathitrust.org/access_use#pd-google
We have determined this work to be in the public domain,
meaning that it is not subject to copyright. Users are
free to copy, use, and redistribute the work in part or
in whole. It is possible that current copyright holders,
heirs or the estate of the authors of individual portions
of the work, such as illustrations or photographs, assert
copyrights over these portions. Depending on the nature
of subsequent use that is made, additional rights may
need to be obtained independently of anything we can
address. The digital images and OCR of this work were
produced by Google, Inc. (indicated by a watermark
on each page in the PageTurner). Google requests that
the images and OCR not be re-hosted, redistributed
or used commercially. The images are provided for
educational, scholarly, non-commercial purposes.


UNIVERSITY OF CALIFORNIA
LIBRARY
DAVIS





REACTOR ANALYSIS
McGRAW-HILL SERIES IN NUCLEAR ENGINEERING

Walter H. Zinn,
Consulting Editor
Jerome D. Ltjntz, Associate Consulting Editor
Benedict and Pigford
Bonilla
Nuclear Chemical Engineering
Nuclear Engineering
Hoisington
Nucleonics Fundamentals
Meghreblian and Holmes
Price
Schultz
Reactor Analysis
Nuclear Radiation Detection
Control of Nuclear Reactors and Power Plants
Stephenson

Introduction to Nuclear Engineering
REACTOR ANALYSIS
ROBERT V. MEGHREBLIAN
Oak Ridge National Laboratory
DAVID K. HOLMES

Oak Ridge National Laboratory
McGRAW-HILL BOOK COMPANY, INC.

New York
Toronto
London
1960
LIBRARY
UNIVERSITY OF CALIFORNIA
DAVIS
REACTOR ANALYSIS
Printed
Copyright 1960 by the McGraw-Hill Book Company, Inc.
This book, or
All rights reserved.
in the United States of America.
parts thereof, may not be reproduced in any form without permission
of the publishers.
Library of Congress Catalog Card Number: 59-15469
THE MAPLE PRESS COMPANY,

41328
YORK,
PA.
PREFACE
is,
The purpose of the present work is to provide a complete and consistent

mathematical development of what is commonly known as reactor analy
the mathematical study of the nuclear behavior of reactors
sis, that
based on certain approximate physical models.
The subject of reactor
analysis differs somewhat from that of reactor physics in both viewpoint
and content.
Whereas reactor analysis deals primarily with the mathe
matical tools for treating the physical behavior of reactors, reactor physics
models and emphasis
placed on
aid of elementary mathematical
is
is
is
it
is
The formal level of the presentation

directed primarily toward the
first- and second-year graduate student in engineering science although
useful.
expected that students in physics will also find
Consider
able pains have been taken to provide textbook which could also be used
in a first course on reactor analysis.
The introductory sections of each
of the principal chapters have been organized and written with this
thought in mind.
In these sections the treatment
initiated with the
it
dis
it
is
cussion of the principal physical concepts to be developed.

The more
sophisticated mathematical considerations and the development of the
Thus
expected
broader theory are left in each case to later sections.
that this work will serve as an elementary text which can also be used in
an intermediate course by simply including the complete treatment.
The material presented in this book was developed from course organ
period of five years at the Oak
ized and presented by the authors over
Ridge School of Reactor Technology at the Oak Ridge National Labora
tory. The development of the subject matter in the ORSORT course
constitutes one-third of this book.
It presumed that the reader who desires complete understanding
of the contents of this book has had at least a course in advanced calculus
general course also in partial differential equations and
and preferably
boundary-value problems, or a first course in the methods of mathe
It also assumed that he acquainted with the funda
matical physics.
mental concepts involved in modern physics and has been introduced to
is
is
is

is
places much more emphasis on the physical aspects themselves of these

The tone of this book
systems.
therefore much closer to that of
in
advanced treatments
engineering analysis rather than to that of books
on physics.
vi
PREFACE
the use of analytical methods in the solution of engineering problems.

For those who desire only an introductory knowledge of the subject and
therefore limit their study to the elementary sections, the usual under
graduate course in differential equations will suffice.
In this treatment the subject matter of reactor analysis has been
The models
developed with the aid of various mathematical models.
selected for this purpose are those which have proved to be useful in
describing the various neutron phenomena peculiar to nuclear reactors.
Emphasis is placed upon detailed presentations of each method in order

that the reader becomes sufficiently well equipped to treat new and differ
In nearly every instance the mathematical treatment

ent situations.
has been extended to include the derivation of working formulas, and
these are usually followed by numerical examples which display the com
In a few
putational techniques which may be used in application.
instances only a formal presentation is supplied, and in these cases the
intent is merely to exhibit the principal physical ideas involved.
The authors have attempted to present a discussion of all the principal
topics of reactor analysis, with an entire chapter devoted to each.

In
several analytical methods are presented in order to

a treatment as possible.
as
wide
These include Chap. 2 on
provide
probability concepts; Chap. 3 on the neutron flux; Chap. 4 on slowing
many
instances
Chap. 5 on diffusion theory; Chap. 6 on the Fermi age model;

Chap. 7 on transport theory; Chap. 8 on reflected reactors; Chap. 9 on
It is important to men
reactor kinetics; and Chap. 10 on heterogeneity.
in
main
part extensions and
that
the
remaining
represent
tion
chapters
down;
applications of these general topics.

The material in Chaps. 1, 2, and 3 and in the first section or so of Chaps.
4, 5, 6, 8, 9, and 10 constitutes a comprehensive first course in reactor
A complete coverage of this text would constitute a second or
analysis.
intermediate course.
The authors wish to express their appreciation for the assistance and
encouragement given by their friends and colleagues.
To L. Nelson they
are especially indebted for his penetrating criticism and gentle tolerance.
To R. R. Coveyou, L. Dresner, R. K. Osborn, and H. Schweinler they are
for many suggestions and hours of stimulating discussion; to
R. A. Charpie, W. K. Ergen, E. Guth, G. Leibfried, L. W. Nordheim,
A. Simon, A. M. Weinberg, and T. A. Welton for reviews and comments;
and to H. Honek, D. H. Platus, and D. L. Platus for help with the numeri
grateful
cal examples.
Finally, the authors thank Mrs. Yvonne Lovely for her expert assist
ance in the preparation of the manuscript.
Robert
David
V. Meghreblian
K.
Holmes
CONTENTS
v
Preface
Chapter
1.1
1.2
1.3
1.
Introduction
Chain Reactions
Nuclear Reactors
The General Reactor Problem
Chapter 2.
Probability
1
17
22
Concepts and Nuclear Cross Sections

Neutron- Nucleus Interactions

Probability of Collision
2.3
Macroscopic Cross Section
2.4
Macroscopic Cross Sections and Physical Properties
Problems
29
2.1
29
2.2
31
Chapter 3.
3.1
3.2
3.3
3.4
Multiplication
Constant and Neutron Flux
Collision Density
Neutron Balance
Infinite-medium Criticality Problem .
to Safety Storage Calculation
Application
49
50
51
55
of a Chemical-processing
Plant
60
Neutron Flux
Problems
Slowing-down
40
50
3.5
Chapter 4.
36
63
68
Process in the Infinite Medium
Mechanism of Energy Loss by Scattering Collisions

Slowing-down Density in Pure Scatterer
4.3
Slowing Down in Hydrogen
4.4
Slowing-down Density with Absorption
4.5
Mixtures of Nuclear Species
4.6
Infinite Homogeneous Multiplying Media
4.7
Thermal-group Cross Sections
Four-factor Formula
4.8
Problems
69
4.1
69
4.2
85
Chapter 6.
5.1
5.2
5.3
5.4
5.5
5.6
Diffusion Theory: The Homogeneous One -velocity Reactor
One-velocity Diffusion Equation

Elementary Source in Infinite Media
Diffusion in Finite Media
Application of One-velocity Model to Multiplying Media
Diffusion Length
Sample Calculation with One-velocity Model
vii
98
102
112
118
125
148
157
160
160
180
189
198
223
230
viii
CONTENTS
Applications of the Kernel Method

Problems of Neutron Population in Localized Absorbers
Problems
5.7
237
244
5.8
Chapter
6.
Fermi Age Theory
The Homogeneous Multivelocity
262
Reactor
268
The Combined Slowing-down and Diffusion Equation

Fermi Age Theory and Elementary Solutions
The Unreflected Reactor in Age Theory
6.3
Applications of Fermi Age Theory
6.4
Reactor Temperature Coefficients
6.5
Application to a Heavy-water-moderated Reactor
6.6
Problems
6.1
268
6.2
273
Chapter
7.1
7.2
7.3
7.4
7.5
7.6
7.
302
308
321
327
Transport Theory
330
Introduction
The One-velocity Transport Equation
The Boltzmann Equation with Energy Dependence
Methods of Integral Equations: One-velocity Model
Applications of Spherical-harmonics Method
Computational Methods for the Neutron Age
Chapter 8.
286
330
331
352
366
387
399
Reflected Reactors
Introduction
The One-velocity Model
8.3
Reactors with Spherical Symmetry : Serber- Wilson Condition
The Two-group Model
8.4
Sample Computation Using Two-group Model
8.5
Completely Reflected Systems
8.6
Feynman-Welton Method
8.7
Multigroup Treatment of Reflected Reactors
8.8
Problems
418
418
8.1
422
8.2
Chapter 9.
9.1
9.2
9.3
9.4
9.5
9.6
456
470
476
486
520
543
Reactor Kinetics
546
Time-dependent Behavior of the Neutron Flux with Delayed Neutrons

Neglected
Reactor Parameters by Pulsed Neutron Beam
Effect of Delayed Neutrons
Reactor Kinetics with Temperature Dependence
Kinetics of Circulating-fuel Reactors
Decay of Fission Products and Burnout Poisons
Problems
Chapter 10.
10.1
10.2
10.3
10.4
10.5
10.6
440
547
556
566
577
590
610
624
Heterogeneous Reactors
Introduction and Survey

Disadvantage Factor and Thermal Utilization
Resonance-escape Probability
Fast Effect
Effects of Cavities and Fuel Lumps on Migration Area
Feinberg-Galanin Method of Heterogeneous-reactor Calculation
626
626
645
660
692
698
.
704
ix
CONTENTS
Chapter
11.1
11.2
11.3
11.4
Control-rod
Theory
721
Central Rod in One-velocity Model

Central Rod in Two-group Model
Eccentric Control Rod in Two-group Model
Ring of Control Rods in Two-group Model
Chapter
12.
Hydrogenous
The Special Treatment Required for Hydrogen

Elementary Development of the Goertzel-Selengut Equations
Solution of Boltzmann Equation for Hydrogenous Systems
Numerical Results
Chapter
13.1
13.2
13.3
13.4
13.5
13.6
13.
Perturbation
727
738
743
748
12.2
12.4
721
Systems
12.1
12.3

11.
....
.
Theory
The Scope and Methods of Perturbation

Perturbation in the Infinite Multiplying
Perturbation in the One-velocity Model
Perturbation in the One-velocity Model
Perturbation in the Two-group Model
Perturbation in the Age-diffusion Model
748
750
752
759
763
Theory
Medium
of the Bare Reactor
of a General System
....
....
763
771
776
781
785
787
Name Index
791
Subject Index
795

CHAPTER
INTRODUCTION
1.1
An understanding
Nuclear Chain Reactions
of the properties
and behavior
of nuclear
chain
reactors is achieved through a study of the neutron population which

Information about the neutron population is con
supports the chain.
veniently expressed in terms of the neutron-density-distribution function.
The detailed features of the chain reaction are determined by the
processes which can occur between the free neutrons
As in chemical chain reactions,
and the materials of the reactor system.

various
nuclear
the rates of the reactions involved in the chain are directly dependent
upon the density of the chain carrier, in the present case the neutrons.
Thus in order to determine the various properties of a reactor, such as
the power-production rate and the radiation-shielding requirements, it is
necessary to obtain the fission reaction rate throughout the system and,
In fact, all the basic nuclear
therefore, the neutron-density distribution.
and engineering features of a reactor may be traced back ultimately to a
knowledge of these distribution functions.
The subject of reactor analysis is the study of the analytical methods
and models used to obtain neutron-density-distribution functions.
Since
these functions are intimately related to various neutron-induced nuclear
reactions, a knowledge of at least the basic concepts of nuclear physics
is essential to a thorough understanding of reactor analysis.
The first section of this chapter
is a brief discussion of those aspects

reactions which are of principal interest to reactor physics.
This presentation assumes that the reader is equipped with an intro
ductory course in nuclear physics.
The second section is an outline
of the basic nuclear components of reactors and of the various types of
reactors, and the last section is a summary of the principal problems of
of nuclear
It is intended
reactor physics and the analytical methods of attack.
that the last section be used primarily for purposes of review and to aid
the reader in orienting the various topics with regard to the over-all
structure
and scope of the subject.
In introducing the general subject of chain reac

Fission Reaction.
will
tions it
be helpful to begin with a review of some elementary but
basic notions about nuclear reactions, in particular the fission reaction.
a.
REACTOR ANALYSIS
This information is included here primarily

explaining
reactors.
the
physical
features
[CHAP.
as a convenient
of neutron
phenomena
reference for
relevant
to
The bulk of the neutrons participating in the chain reaction within a

reactor possess kinetic energies which range from thermal energy1 (hun
dredths of an electron volt2) up to fission neutron energies (several
million electron volts). Even though this range extends over some eight
orders of magnitude, it is nevertheless possible (and convenient) to
describe nearly all the important neutron-induced
reactions within a
reactor by means of a single conceptual model, namely, the compoundnucleus idea of Bohr.3
This model is especially useful in studying the
fission process.
The formation of the compound nucleus constitutes

the reaction between the neutron and a nucleus.

symbolically by
XA + n1-*
It
the first step in
may be represented
(1.1)
{XA+1}*
where XA denotes some nucleus of mass number A which has captured

The symbol { ( * indicates that the resultant
(absorbed) a neutron n1.
The excitation
compound nuclear structure XA+1 is in an excited state.
of
this
binding
nucleus is the combined kinetic and
energy
energies
of the captured neutron (in the compound nucleus).
If the captured
neutron had exactly zero velocity relative to the nucleus, then the exci
tation energy would be precisely the binding energy Eb. This point is
easily demonstrated with the aid of the inverse to the complete reaction
implied in (1.1). The complete reaction is accomplished when the com
pound nucleus achieves one of several possible stable states by simply
energy Ey in the form of electromagnetic radiation
ejecting the excess

(gamma rays)
thus,
XA +
(X^1)* XA+l + Ey
(1.2)
Now consider the situation in which the nucleus XA+l in the unexcited
(ground) state acquires an energy Eh just large enough to separate a
neutron.
The appropriate reaction would be
XA+l +
E-*
{XA+1}*
XA + nl
(1.3)
This reaction is called the photoelectric liberation of a neutron.

Now, by the theorem of detailed balance,4,5 this reaction is just the
1 Energy comparable
to the thermal motion of the nuclei in the medium supporting
the neutron population.
*
One electron volt (ev) = 1.6023 X 10"1S erg = 1.6023 X lO-" watt sec.
3 N. Bohr,
Nature, 137, 344 (1936); J. M. Blatt and V. F. Weisskopf, "Theoretical
Nuclear Physics," pp. 340-342, John Wiley & Sons, Inc., New York, 1952.
4 Blatt and
Weisskopf, op. cit., pp. 601-602.
5 E.
Fermi, "Nuclear Physics," rev. ed., pp. 145-146, course notes compiled by Jay
Orear, A. H. Rosenfeld, and R. A. Schluter, University of Chicago Press, Chicago, 1950.
SEC.
INTRODUCTION
1.1]
inverse of (1.2); a comparison
of Eqs. (1.2) and (1.3) therefore reveals
that
Ey = Eb
(1.4)
concept is essential to an understanding of the

A short summary of the idea is therefore in
nuclear-fission process.
total binding energy El of a nucleus is given
the
order.
By definition,
by the difference between the mass of the nucleus and the sum of the
If M is the
masses of its constituent nucleons (protons and neutrons).
The binding-energy
nuclear mass, A the mass number (number of nucleons) , Nn the number

of neutrons, m and mp the masses1 of the neutron and proton, respec
tively, and c the velocity of light, then
Ef* = c\mnNn + mp(A
~c2(Am-A/)
- Nn) - M]
= c2[JVn(m
mp)
+ mpA
- M]
(1.5)
where the approximation follows from the fact that mn ~ mp. In order
to obtain the average binding energy per nucleon Eb, we divide through
by
^ ~ c*(mn - ^
(1.6)
This is the average energy required per nucleon to separate the nucleus
The value of Eb varies from about 1 to
into its constituent particles.
In the mass range of interest to
9 Mev over the entire mass scale.2
reactor physics (A > 70), Eb decreases monotonically from 8.6 Mev at
A = 70 to 7.5 Mev at A = 238. It is this variation which determines
It is shown later
two fundamental features of the fission reaction.
that this indicates (1) that there is a positive energy release if a nucleus
with A > 85 is caused to disintegrate and (2) that nuclei in this range are
theoretically unstable with regard to the fission process.
Consider first the question of the energy released when a nucleus
As an example, let us take the case of the neutrondisintegrates.
induced fission of the nucleus XA which results in the formation of two
On the basis
fragments YAl and ZA' of masses M i and M2, respectively.
of the compound-nucleus concept, this reaction may be written
nl +
XA-+ \XA+1}*-> YA' +
(1.7)
ZA>
is,
In general, the binding energies of the fragments will differ from the
the combined masses
binding energy of the original nucleus; that
of the fragments will not equal the mass of the fissioned nucleus XA+i.
The difference appears as an energy release Es which may be determined
mv
= 1.00758 amu, where
These are: m = 1.00893 amu (atomic mass units) and

amu = 1.66 X 10_M g.
See, for example, C. F. Bonilla, "Nuclear Engineering,"
Hill Book Company, Inc., New York, 1957.
J

Eb m
Fig.
3-5, p. 63,
McGraw-
REACTOR ANALYSIS
from the Einstein mass-energy relation.
E,
In the present
- (Mx + Mt)]
cWc
[CHAP.
case
(1.8)
where the subscript c refers to the compound nucleus.

This expression
in
of
various
binding energies if we use the
terms
the
may be written
relation (1.6); thus, in general, for nucleus i
"
- =U
(1-9)
The substitution of this equation into (1.8) yields
Ef ~
{A
- ] + AJfi* - El]
l)[Ep
(1.10)
Mass number A
Fig.
1.1
Fission energy
of the mass number.
and electrostatic repulsion energy of a nucleus as a function
at what value of A, Ef becomes positive depends, of course,

In the case of symmetric division
upon the mode of disintegration.
For mass numbers sub
the critical value of
about 85.
(A i = A
stantially greater than 85, there are many modes of disintegration which
result in
is
2),
Precisely
positive energy release.
The principal observation to
be drawn from the fission-energy
relation
is
that, in the case of the heavier nuclei, the fission fragments

(1.10)
This would
lower energy state than the original nucleus.
represent
imply that the heavier nuclei are inherently unstable toward fission and
a

where we have used Ac = A + 1 = A\ + A2. The variation of this

function with mass number has the general shape shown in Fig. 1.1.
could
conceivably
shows, however,
undergo
Experience
spontaneous
disintegration.
fission does not occur at anything
that spontaneous
SEC.
INTRODUCTION
1.1]
like the rate one might be led to expect on the basis of the fission ener
The explanation rests with the fact that a nucleus cannot
getics alone.
disintegrate until it has first acquired a certain activation or threshold
Thus there exists an energy barrier between the state of the
energy.
whole nucleus and the fragmented state.
If
one were to take the poten
tial energy of the fragments at infinite separation to be zero, then the

merged state of the fragments (forming the original nucleus) would
have energy Ef. The variation of the potential-energy curve between
If the
these limits would exhibit a maximum, as shown in Fig. 1.2.
maximum value of the potential energy is
then 2?*1 Ef = ETn
would be the threshold energy for fission.
A nucleus can acquire the necessary excitation energy to overcome
this barrier by absorbing either a nuclear particle or electromagnetic
r2
Potential energy of fission fragments as a function of separation.
1.2
radiation.
r,+
Fig.
Of the two possibilities,
it,
the former leads more easily to

fission since the absorption of a particle makes available not only its
kinetic energy but its binding energy as well. This point was previously
noted in connection with the neutron-capture
reaction.
If the energy
acquired in this way is sufficiently large, the nuclear structure experiences
increasingly violent oscillations which eventually rupture

forming the
various fragments.
An estimate of the fission threshold can be obtained from the energy
required to distort the nucleus into an extreme shape which results in
complete separation into fragments.
It has been shown12 that this calcu
lation can be based on the liquid-drop model of the nucleus. The two
principal contributions to the distortion energy of the nucleus are the
"surface-tension" effect from the nuclear forces between the constituent
J.
elementary
treatment
Physics," pp. 417-421, John Wiley
66, 426
(1939);
J.
Frenkel,
given by D. Halliday,
"Introductory
Sons, Inc., New York, 1950.
&
'An
A. Wheeler, Phys. Rev.,
J.
Phys.
(1939).
is
N. Bohr and
(U.S.S.R.),
125
1,

max
Nuclear
REACTOR
ANALYSIS
[CHAP.
nucleons and the electrostatic repulsion due to the charge on the protons.
When the nucleus is set into oscillation, any departure from its original
shape results in an increase in its potential energy because of the "surface
tension."
Such distortions tend, however, to separate the proton popu
lation and thereby create centers of electrostatic repulsion; these forces
the potential energy of the system and increase the distortions
If the distortional oscillations lead to a "dumbbell-like"
further.
decrease
nuclear configuration, the electrostatic repulsive forces may eventually

overcome the nuclear attractive forces and the nucleus will divide.
When the separation results in two major fragments, the threshold
energy for fission is given by the potential energy of the fragments at the
instant of separation.
If
A representative curve of the potential energy

forces predominate.
Ep of this system is shown in Fig. 1.2.
The variation of the potential energy E* with mass number has the
It is seen that for all A < 250 the
general shape shown in Fig. 1.1.
threshold energy i?"

Ef > 0; energy must be added in order to
The figure reveals, however, that the

produce fission in all nuclei.
threshold energy becomes progressively smaller with increasing mass
As the mass number approaches 250, there is a rapid increase
number.
in the probability for spontaneous fission through the mechanism of
When A > 250, the probability is essentially 1,
penetration.1
and the nucleons do not remain together long enough to be described as
It is not surprising, therefore, that nuclei with such
a nuclear structure.
large A do not exist in nature.
barrier
Nuclear Fuels.
Nuclei with mass numbers in the range 230 < A

Thus in these cases
< 240 have fission thresholds of some several Mev.
fission can be brought about by the absorption of radiation or neutrons
of only a few Mev kinetic energy.
There are a few nuclei that can even
be caused to fission by thermal (very slow) neutrons.
The specific energy
requirements for a particular nucleus depend strongly upon the excita
tion energy which the captured particle can impart to it. It was shown
previously that, in the case of the neutron-induced reaction, the binding
not all) of this excitation
energy of the neutron represented a large part
(if
b.
If
M{A,Z)
the mass of
even-odd term in the nuclear-mass formula.

Blatt and Weisskopf, op. cit., p. 567.
is
is
However, even though the mass numbers of the more easily

energy.
fissionable nuclei differ but little, the binding energies vary by as much as
50 per cent; hence, the variation in the fissionability of the various nuclei.
due primarily to the influence of the
This relatively large variation
1

one assumes that separation yields two spher

ical collections of nucleons with Z, protons and of radius r,-, then the
maximum value of the potential energy
due to the electrostatic field
is proportional to Z\Zi(t\ + r2)-1.
When the separation r exceeds
T\ + r2, the coulomb forces predominate, and when r < T\ + r2, nuclear
SEC.
INTRODUCTION
1.1]
nucleus containing
A nucleons of which Z are protons, then, in atomic
mass units,1
M(A,Z)
= 0.99395.4
0.00084Z + 0.0141
A' +
0.000627^
0.083
2)2
(A/2A~
&
The term of interest to the present discussion is the quantity
It
so-called even-odd term.

5
5,
the
is defined
when A is odd
= 0
0.036
A*
(1.11)
0.036
Ai
when iV is even, Z is even
(1.12)
when Nn is odd, Z is odd
the dependence of the mass of a nucleus, and

energy, upon the number and type of nucleons it
contains.
An accurate computation of the binding energy of a neutron
in
nucleus is obtained from the equation
a
compound
Et
These relations indicate

therefore its binding
Eb =
c\M
mn-
Mc)
(1.13)
which may be compared to the approximation (1.6).

Equations (1.11) through (1.13) reveal that neutron absorptions
which result in compound structures with even numbers of protons and
neutrons acquire the largest excitation energies since the 5 term is nega
tive for these nuclei. Compound nuclei with an odd number of nucleons
acquire the next largest excitation energies, and odd-odd nuclei, the least.
It is on this basis that the isotopes IP33, U236, and Pu2M can be made to
fission by neutron captures of any energy, whereas Th232 and U238 will
In the case of the first three nuclei,
fission only with very fast neutrons.
a neutron capture leads to an even-even compound structure, and the
excitation energy due to the neutron binding energy alone (~6.8 Mev)
is equal to the fission threshold.
Thus these nuclei can fission by thermal
(very slow) neutron capture, as well as by captures of fast neutrons.
It is this characteristic which makes these nuclei especially important as
As will be discussed later, these nuclei fission with such
nuclear fuels.
relative ease in the thermal-energy range that it is well worth while to
provide means to moderate (slow down) fission neutrons
energies so that this characteristic may be fully exploited.
to thermal
In fact, the
problem of neutron moderation is a principal consideration in reactor

analysis.
In the case of Th232 and U238, the compound nucleus has an even-odd
collection of nucleons and the binding energy for this state (~5.3 Mev)
1 Bonilla,
op. cit., pp. 69-72.
REACTOR
ANALYSIS
[CHAP.
than the threshold energy (~7.1 Mev). Evidently, neutroninduced fission is not possible with these nuclei when the neutron kinetic
This limited fissionability makes these
energy is less than 1.8 Mev.
isotopes much less attractive as the primary fuel component for a steady
chain reaction, although their presence in a chain supported by any of the
"thermal fuels" can lead to significant augmentation.
The energy released from the fission reaction of any of the isotopes
This
mentioned above may be computed from the equation for Ef (1.8).
expression, which applies for the two-fragment divisions, will suffice for
is less
most situations
of interest
since a three-fragment
(or more complex)
A rough
a very low probability.
estimate of the fission energies generally available from such nuclei
The resulting
may be obtained from a sample computation for U236.
value will be representative since the various mass numbers differ but
little, and the principal dependence on binding energy occurs in the
Consider, therefore, the reaction

Ebe) term [cf. Eq. (1.10)].
division of the compound nucleus has

ni + u23i-->
|U236
)*-* Kr"
Ba142
(1.14)
The average binding energy per nucleoli Eb for a nucleus of mass num
ber 236 is about 7.5 Mev, and for nuclei of mass numbers 94 and 142,
8.6 and 8.3 Mev, respectively.
The approximate relation (1.10) yields
for E,:
Ef
= (236) (8.6
Thus, roughly
200
7.5)
+ (142) (8.3
Mev is released in
- 8.6)
= 217
Mev
typical fission reaction involving
a heavy nucleus.
The energy from fission appears principally as the kinetic energy of

As these fragments speed outward from the point
the fission fragments.
of reaction, they encounter the various nuclei of the surrounding
environment.
Such encounters are relatively frequent since the frag
ments are usually highly ionized and therefore experience strong coulomb
with the electron clouds of these nuclei. These interactions
are primarily scattering collisions, and each collision results in the trans
interactions
fer of some of the kinetic energy of the fragment to the struck nucleus.
A series of such collisions eventually slow the fragments to thermal
equilibrium with the environment.
Approximately 85 per cent of the
fission energy is liberated in this way and must be removed by a suitable
The remaining 15 per cent appears either as radiation

cooling system.
or as the kinetic energy of neutrinos1 and neutrons evaporated from
A detailed break
the fragments or released at the instant of fission.
down of the energy distribution for U235 is given in Table 1.1.
The reaction of Eq. (1.14) gives one mode of division of the U235
nucleus as a consequence of fission.
Of course, this is not the only one
1
Halliday,
op. cit., p. 94.
SEC.
INTRODUCTION
1.1]
possible and, in fact, any division consistent with the conservation of

mass and energy can occur.
The fragments which appear cover the
mass
certain
divisions are favored, depending on
scale;
however,
entire
The mass
the target nucleus and the energy of the captured neutron.
distribution of the various fragments resulting from the fission of U235,
For a given nucleus, the shape
U238, and Pu239 is shown in Fig. 1.3.
of the yield curve is highly sensitive to the energy of the incident neutron.
The figure shows clearly the characteristic grouping of the fission frag
ments into two distinct regions of the mass scale for the case of a thermal
It is seen that such reactions most frequently
neutron-induced fission.
produce two fragments, one of mass number around 95 and the other
Although other subdivisions can occur, their
of mass number 140.
very improbable.
Of these alternative schemes, the division
of the compound nucleus into two fragments of equal mass occurs in
likelihood
is
about 0.01 per cent of the time.

Table
1.1
Distribution of Fission Energy for Um

Form
Energy, Mev
Kinetic energy of fission fragments
165 + 15
Prompt gamma rays

Kinetic energy of fission neutrons
Fission product decay:
Gamma
Beta
Neutrinos
Total energy per
fission
197 15
The fragments formed in fission are generally very rich in neutrons.

This is a consequence of the fact that among the stable nuclei1 the ratio
Nn/Z increases with Z. Thus the fissionable nuclei with Z ~ 90 will
possess far more neutrons than are required in the nuclear structures
Nearly all these excess neutrons3 are released
of the stable fragments.2
at the instant of fission.
These are the prompt neutrons.
The remain
ing small fraction of neutrons to be released are evaporated off from the
fragments at various time intervals after fission; these constitute the
In the event that the emission of a neutron
delayed neutron groups.
or two by a primary fragment does not leave the nucleus in a stable
configuration, further nuclear readjustment may then occur through the
emission of /3 radiation (electrons).
For the purpose of studying the neutron economy in a reactor, it
will be convenient to rewrite Eq. (1.7) so as to include the neutrons which
For the present, we can omit
are emitted by the primary fragments.
1
Bonilla, op. tit., Fig. 3-6, p. 68.

Halliday, op. cit., pp. 408-412.
99.245 per cent in the case of U"'.
REACTOR ANALYSIS
10
[CHAP.

Of
>
Mass number
Fig.
Yields from fission of U2", U"8, and Pu2" versus mass number A.
[Re
printed from C. DuBois Coryell and N. Sugarman (eds.), "Radiochemical Studies: The
Fission Products," Part 3, National Nuclear Energy Series, Div. IV, vol. IX, McOrawHill Book Company, Inc., New York, 1951; Uin from Paper 219 by Steinberg and
Freedman, and Uut and Pu*n from Appendix B.]
1.3
SEC.
INTRODUCTION
1.1]
11
the distinction between the neutrons that appear instantly and those
which appear at later stages in the decay processes of the fragments.
If we denote by v the total number of neutrons which appear from fission,
then it is clear that the nuclei Y and Z will contain fewer neutrons by
exactly
this amount;
we now
therefore,
write
reaction
(1.7) in the
form
representative
n1
\XA+l\*-^
YA*
For the case of U235 considered in Eq. (1.14),

the following division:
ni +
tj236
(U236)
m1
(1.15)
we might have,
for example,
ZA~A>-'+1
- Kr93 +
Ba140
3n
(1.16)
We have assumed here that three neutrons are emitted by the fragments.
The number of neutrons emitted from fission is dependent upon the
In first approximation, it is permissible
energy of the incident neutron.
to neglect this dependence and to assume that the average number of
neutrons produced per fission is a function only of the target nucleus.
It should be recognized that this is not a true picture and that, in fact,

Table
1.2
Neutron
Yields Due to Thermal
Fissionable
nucleus
neutrons per fission
Us
Un
2.54
2.46
2.88
Pu"
Fission in Nuclear Fuels
neutrons per absorption
0.04
0.03
0.04
2.31
2.08
2.03
0.03
0.02
0.03
the average number of neutrons increases with the energy of the incident
For many reactor calculations, however, this dependence is
particle.
neglected and some average number is selected for the entire energy
range.
The actual value of
for
material can be obtained

from experiment.
In such measurements, even if the incident neutrons
were all of some fixed energy, the number of neutrons produced by fission
reactions would be a statistical quantity and would vary from one fission
to the next.
As a matter of fact, the actual number of neutrons released
v
a given fissionable
by any one reaction is of little interest in reactor physics, and one would
certainly prefer an average value which could be assigned to a large
number of such reactions.
Measurements of this type have been made,
and the presently acceptable values of v (the average number of neutrons
per fission) for the three principal nuclear fuels are listed in Table 1.2.
c. Nuclear Reactions and Neutron Economy.
So far in our discussion
of the chain-reacting system, we have focused attention on the fission
Fission is but one of several neutron-induced nuclear reactions
reaction.
REACTOR
12
ANALYSIS
[CHAP.
a fissionable material; however, essentially all the

by which we
reactions are nonproductive,
neutron-capture
and
additional
do
not
lead
to
releases
significant
energy
mean that they
It is the competition between these nonproductive processes
neutrons.
and fission which in part determines the relative merits of a specific
The other essential factor is the
nuclear species as a "fuel material."
that can occur in

remaining
number of neutrons produced per fission (v).

The two competing nonproductive neutron-capture
radiative capture and
(2)
reaction can be symbolized
XA +
n1
inelastic
processes
are(l)
The radiative-capture
scattering.
by the equation
\XA+l)*-* XA+l + Ey
(1.2)
The symbol Ey here denotes energy liberation in the form of gamma rays
(electromagnetic radiation of wavelength in the order of 10-10 cm) hav
ing about 1 Mev energy. In this process, a neutron is captured by the
fissionable material and forms a compound state.
However, instead of
releasing its excess energy by fissioning, it does so by the emission of some
This emission may remove all
radiation immediately after the capture.
the excitation energy of the compound nucleus and thereby leave it in a
If this initial emission leaves the nucleus with some
state.
residual energy, further emissions must eventually follow until a stable

stable
configuration is achieved.
reaction could be written
For the
case of U235,
ni + u5 |U2!6!*->
the radiative-capture
+ Ey
U236
(1.17)
The symbol Ey represents all the
excess energy of the compound state

this
the
excited
U236
case,
nucleus) which must be released in order
(in
to achieve the stable state U236.
The second nonproductive process, inelastic scattering, does not phys
ically remove neutrons from the system, it merely degrades (as a rule)
the kinetic energy of the neutron population.
For the nuclei of primary
to reactor technology (heavy nuclei), the inelastic-scattering
process is conveniently described with the aid of the compound-nucleus
In these reactions the nucleus "captures" a neutron at one
model.
interest
energy and releases

n1
it at another; thus,
+ XA-> {XA+i\*-> \XA\* +
A compound nucleus
is not
always
formed,
scattering reactions are properly symbolized
n1
however;
(1.18)
some inelastic-
by
nl + XA-> \XA}* +
n1
(1.19)
But, in either case, the essential feature of the inelastic-scattering reac

tion is that the total kinetic energy of the original particles differs from
If the struck nucleus XA was before
that of the product particles.
SEC.
1.1]
collision
INTRODUCTION
13
in a stable state (the ground state), then after collision it will

The energy of excitation, which appears as
an excited state.
occupy
an increase in the internal energy of the nucleus, is precisely the energy
Thus, the sum of the kinetic
loss suffered by the incident neutron.
energies of the emitted neutron and the struck nucleus (after collision)
will differ from the total kinetic energy of the system before collision by
exactly the amount of the excitation energy of the residual nucleus.
A third nonproductive neutron-nucleus reaction which is of some conse
Although reactions of this type are of

quence is elastic scattering.
great importance where nonfuel components of a reactor system are
concerned, they do not represent truly competitive processes in the case
When an elastic scattering occurs, the incident neutron leaves
of fuels.
the region of the target nucleus with a different kinetic energy1 from

that which it had before the collision.

However, the total kinetic energy
of the particles involved before and after collision is conserved.
From a
interactions
this
viewpoint,
type
neutron-nucleus
of
phenomenological
have the characteristics of both a "capture" type of reaction and a
"deflection" type. The so-called potential scattering component is a
deflection-like process and does not involve the formation of a compound
structure.
Resonance scattering, on the other hand, may be
interpreted by means of the compound-nucleus concept;2 thus in terms
of Eq. (1.19) it can be described as an inelastic-scattering reaction which
nuclear
Whereas this process

leaves the struck nucleus with no residual energy.
is limited to relatively narrow ranges of incident neutron energies, the
Of
potential-type scattering extends over the entire energy spectrum.
primary importance to the present discussion is the fact that the nuclear
fuels are generally of large nuclear mass, and elastic collisions with
neutrons produce little change in the kinetic energy of the neutron.
Thus, elastic-scattering collisions with fuel nuclei effect little or no
change in the neutron population, either by altering the neutron energies
or by serving as neutron removers.
Inelastic scattering and radiative capture are the two processes which
compete with the fission reaction.
The extent of this competition is
determined by the relative magnitudes of the rate at which nonproduc
tive reactions occur as compared to the rate at which fissions occur. If
the nonproductive reactions for a particular fissionable material constitute
only a small fraction of all neutron-nucleus reactions, then the possi
bilities of this substance as a nuclear fuel are greatly enhanced. The
deciding factor, however, is the number of neutrons released by fission.
a given substance, the likelihood of fission taking place may be very
For
In the laboratory system of coordinates. The kinetic energies in the center-ofmass system are conserved.
1 See, for
example, Halliday, op. cit., pp. 396-402; also, Blatt and Weisskopf, op. cit.,
chap. VIII, especially pp. 325-329.
1
REACTOR ANALYSIS
14
[CHAP.
high, but for the purposes of a chain-reacting system, this is of little value
if on the average the number of neutrons released per fission is negligibly
Thus it is the combined characteristics, reaction rates of non
small.
productive processes and number of neutrons per fission, which determine
the merits of a fissionable material as a nuclear fuel.
The focus of attention in reactor analysis is upon the fate of the
neutrons produced by the fission reactions, rather than upon the fission
process proper with its resultant fragments and electromagnetic radia
For most situations in reactor analysis, the nuclear-fission process

may be adequately represented by picturing a given neutron as disappear
ing at a point and being immediately replaced by v neutrons whose kinetic
energies and directions of motion are very different from those of the
The principal objective is to trace the life histories
incident neutron.
of these neutrons, of the neutrons which they in turn may produce, and
tion.
complex is determined by the purpose and application of the device;

thus, the configuration and composition will vary from system to system,
Nevertheless, there
and an all-inclusive description is not meaningful.
of
reactor
which
are common to all.
components
every
are certain primary
These are:
The region (usually central) which contains the nuclear

fuel, along with various structural materials, moderator, and coolant.
2. The reflector: A shell region surrounding the core which usually
consists of moderator-type material, supporting structure, and possibly
1.
The
core:
(if
coolant materials.
3. The shield: An outer shell which serves as a stopping medium for
the biologically dangerous radiations from the core and the reflector.
any) constitute the principal
The core and the reflector region
From the viewpoint of the reactor

nuclear components of the reactor.
analyst, the main interest rests with the distribution of the neutrons
throughout these two regions. Neutrons which pass through the outer
boundary of the reflector are effectively lost to the system. Attention
therefore focused on the neutron population within the confines of the
is
is
is
core and the reflector.

the source of the neutrons which
The core region of the reactor
large fraction of the neutrons produced
support the chain reaction.
lost to the chain process, either by passing through the
by fission
the reactor and thereby escaping from the system, or by
of
boundaries
is
is
is
Unless the reactor configuration

removal through radiative capture.
lost by
unusually small, the fraction of the neutron population which
to
due
absorption
relatively small and the bulk of the losses
escape
is

of all the subsequent generations.

A detailed study of the life history of the neutrons in a chain-reacting
system depends upon a knowledge of the materials and geometry of the
A complete description of a reactor
assembly which supports the chain.
SEC.
INTRODUCTION
1.1]
15
We have already noted

(capture) reactions which do not produce fission.
a competitive process with fission and,
therefore, serves as a neutron-removal process (sink) in the chain-reacting
that radiative capture in the fuel is
This phenomenon is by no means confined to the fuel nuclei.

Radiative capture of neutrons by nonfuel components of the reactor is a
As in the case of the fuels, the nonpro
significant loss factor as well.
ductive capture process in these materials can be described by means
system.
of an equation of the type (1.17).

It was mentioned previously that the core region may contain some
moderator- type material.
(This will not be the case, however, in all
thus in the so-called fast reactors, the absence of moderator
The primary function of a moderator is to
materials is greatly desired.)
degrade the neutron energies through the mechanism of scattering
collisions.
We observed previously that scattering collisions, both elastic
and inelastic, cause the incident neutron to transfer part of its kinetic
reactors;

Whether or not collisions of this type

energy to the struck nucleus.
tend to improve the conditions for maintaining a chain reaction depends
upon the type and function of the reactor.
The primary purpose of the reflector is to return some of the neutrons
which escape from the core. Generally speaking, suitable reflector mate
rials are also good moderators; thus, intentionally or not, the reflector
region serves both to degrade the neutron energies and to reduce the
escapes from the system.
In undertaking a first crude attempt to understand the various phys
ical features of a chain-reacting system, it will be helpful to picture the
life history of a typical neutron as consisting of the following stages:
(1) The neutron is born from fission somewhere in the core.

(2) It
will then wander about through the core and possibly pass back and
forth into the reflector several times during its lifetime.
(3) Eventu
ally, it will disappear by one of two processes, absorption by a nucleus
in either the core or the reflector, or by escaping from the reflector.

Only if the neutron is absorbed by a fuel nucleus is there a chance
that additional neutrons will be produced.1 If there is to be a stable
neutrons produced by a given fission must
excite, on the average, one further fission, thus replacing these v neutrons
and thereby perpetuating the chain.
neutron chain reaction, the
This simple model may
be used to describe
in rather broad terms the
If n denotes the total number

requirements on a stable chain reaction.
of free neutrons available at any time instant in the reactor, then it is
clear that if the neutron population is to be stable the number n must be
maintained for all subsequent time.
1 This is
A certain fraction of these, however,
not strictly true since there are such nuclear reactions as the (n,2n) reaction
can serve as secondary
in beryllium (capture of neutron producing two neutrons) which

neutron sources.
REACTOR ANALYSIS
16
[CHAP.
will escape from the system, and the remainder will be captured by vari
Thus absorption or escape is the only eventuality open to a
ous nuclei.
neutron.
If
we define
fa = fraction of neutrons in the reactor which are eventually absorbed

(by either fuel or nonfuel nuclei)
/, = fraction of neutrons in the reactor which eventually escape
// = fraction of neutrons in the reactor which cause fission in the fuel
nuclei
fy = fraction of neutrons
which experience radiative capture by fuel
nuclei
=
fraction
of neutrons which experience radiative capture by nonfo
fuel nuclei
then of the available n neutrons, nfa are absorbed and nf. escape; there
fore,
1
where, by definition,
/. + /.
(1.20)
/.-//+/+/.
(1-21)

Of the
nfa neutrons absorbed, nff cause fission, and from these fissions,
neutrons
are emitted.
If the population n is to remain at steady
vnff
state, we must require that
n = nfa + nf. = vnff
(1.22)
and it follows that
/.+/.
This relation may
be
rewritten with the aid of Eq. (1.21); thus,
/,+//('/.+/.
For convenience,
(1.23)
(1M>
we define the ratio
=
-*
a
(1.25)
T,
This quantity is strictly a function of the nuclear characteristics of the

Equation (1.24) may be written in terms of a by making a direct
fuel.
substitution from (1.25). The result is
v
a-p-P-T^M
(1.26)*
The importance of this relation rests with the fact that for
a given fuel,
the average number of neutrons per fission, must be at least as large
os 1 + a if the chain reaction is to be self-sustaining.1
Thus, the number
v,
Without any external
reactions.
sources,
i.e.,
no sources
of neutrons other than fission
SEC.
INTRODUCTION
1.2]
of neutrons per fission must exceed

which accounts for the nonproductive
by the o term) and the captures in
escapes from the reactor [the last term
17
unity by precisely the amount

captures in the fuel (represented
the nonfuel components and the
It should
on the right in (1.26)].
be noted that the fraction of neutrons which escapes from the reactor is
At best, / = 0, a situation which
dependent on the size of the system.
Clearly, even for an infinite
is possible only in an infinite geometry.
system, some losses (radiative capture) still occur.
Much of the attention in reactor analysis is devoted to the calculation
A knowledge of these quanti
of the fractions /a, / etc., defined above.
distributions
ties forms the basis for determining the neutron-density
will
later
derive
suitable
for
procedures
computing
We
in the reactor.
these quantities, not only as a function of the spatial coordinates, but

also in terms of other independent variables of interest, such as the

neutron energies.
Some mention must be made of the frequent use of the phrase "on the
As in the kinetic theory of gases, we are dealing in the
average."
problems of reactor physics with very large numbers of particles which
are moving sufficiently rapidly so that the elapsed time between collisions
For example, the entire lifetime
is very short on the laboratory scale.
of a neutron in a reactor from birth in a fission process to final death in
absorption or escape may be of the order of 10-i to 10~3 sec. The proper
ties of the assembly which are usually measured are bulk
which depend on the occurrence of huge numbers of nuclear
thus, the actual fate of a given neutron is of little interest.
really seek in every case are the various possible life histories
properties
collisions ;
What
we
a neutron
may have and the probability of each such life history ; then
we may take
an average over these probabilities, to find some macroscopic property
of the system which is of interest.
Inasmuch as these problems involve
large numbers of particles and time intervals long in comparison with
their average free flight time (time between collisions), the deviations
in the behavior of the assembly from the average expected behavior
will be very small. So long as these conditions hold, we may replace
the actual assembly by a model in which all particles behave according to
This approach is used over and over again in
some average pattern.
reactor analysis; consequently, the phrase "on the average" is heavily
worked and becomes such a burden that it is often omitted as understood.
1.2
a. Classification
Nuclear Reactors
of Nuclear Reactors.
From the viewpoint of the
reactor physicist, nuclear reactors may be classified according to two

criteria.
These are: (1) the geometric arrangement of the nuclear fuel
and moderator materials, and (2) the average kinetic energy of the neu
18
REACTOR
ANALYSIS
[CHAP.
In terms of the first

trons which cause the bulk of the fission reactions.
criterion, nuclear reactors are separated into two categories, namely:
(a) homogeneous reactors, and (b) heterogeneous reactors.
The principal nuclear constituents in the core of a homogeneous reactor
The fuel-bearing substance will appear, in gen
are intimately mixed.
eral, as some molecular compound, dissolved in a suitable solvent to
form a solution, suspended in a fluid carrier to form a slurry, or fused
at some appropriate temperature (with or without a carrier) to form a
The purpose and operating conditions of the reactor

complex determine which of these possibilities is the most attractive for a
homogeneous fluid.
given engineering application.

Heterogeneous reactors are characterized by the geometric separation
In systems of this type, the fuel may
of fuel and moderator materials.
in
the
form
of
metal
appear
(or ceramic) plates, rods, or lumps which are
distributed throughout the moderator according to some prescribed

lattice configuration.
Heterogeneous reactors necessarily contain mod

erator material, in contrast to homogeneous systems which may not.

However, the mere physical separation of the fuel and moderator mate
rials does not establish a reactor as belonging to the heterogeneous
This point is perhaps better illustrated by the extreme exam
category.
ple of a reactor core which consists of "tissue-thin" foils of fuel material
In so far as the neutron

closely packed in a medium of moderator.
population is affected, this configuration is equivalent to a homogeneous
mixture of fuel and moderator.
The neutrons do not " see " the geometric
discontinuities in passing from fuel to moderator to fuel, etc., because of
This degree of fineness is
the "fineness" of the physical structure.
determined jointly by the various dimensions of the system and by con
siderations of the average energy of the neutron population and the
material composition.
The basis for the choice between a heterogeneous and a homogeneous
reactor is determined in large part by the available fuel enrichments,
heterogeneous arrangements being most attractive (if not the only
The now classic example
possibility) for low-enrichment fuel materials.
of this application was the choice of a heterogeneous configuration for
The famous University of Chicago reactor
the first nuclear reactors.
and
operated by Enrico Fermi's group, the X-10 graphite
(CP-1) designed
reactor at the Oak Ridge National Laboratory, and the Hanford reactors
In each case the selection of the hetero
were all heterogeneous systems.
geneous configuration was dictated by the fact that the then-available
quantities of enriched U235 (the only nuclear fuel found in nature) were
insufficient to charge a reactor, and it was necessary to employ natural
uranium which contains only 0.714 per cent Us". Homogeneous systems
utilizing mixtures of various moderators and natural uranium, however,
cannot support self-sustaining chain reactions because of the high neu
SEC.
INTRODUCTION
1.2]
19
The only exception is

tron-absorption properties of these materials.
heavy water (D20). The other good moderators, such as graphite, Be,
BeO, and H20, require enriched fuels, or if natural uranium is to be used,
it must be arranged in a heterogeneous configuration.
(The "lumping"
of the fuel gives it an advantage over the moderator in competing for
the available neutrons and thereby improves the possibilities for sustain
In the early days of reactor technology, only
ing the chain reaction.)
graphite and water were easily available, and of the two, graphite had
the lesser affinity for neutrons and was therefore selected for the earliest
designs.
The energy spectrum of the neutron population in a reactor is another
This description, when coupled
useful criterion for classifying reactors.
with the specification of the geometric arrangement of fuel and moder
ator, gives a reasonably complete picture of the reactor physics of the
Now it was noted in the preceding section that fission neutrons
system.

are born into the system with energies of a few Mev, but are subsequently
slowed to thermal speeds by a succession of elastic- and inelastic-scatter
ing collisions with the surrounding nuclei. Thus neutrons are distributed
If one were
over the entire energy scale from thermal to several Mev.
to plot the density of neutrons as a function of energy, the resulting

distribution would reveal that in each system the bulk of the neutron
population occupies a relatively narrow band. It is convenient, there
fore, to speak in terms of an "average energy" of the neutron population.
It was also mentioned previously that the fission-reaction rate, as well
as the rates of the other neutron-induced reactions, is a direct function
Thus, if the neutron-energy spectrum is known,
of the neutron energy.
the total fission rate in the reactor can be computed as a function of
neutron energy, and an "average energy of fission" can be derived from
this function. It is customary to classify reactor types in terms of this
For this purpose the entire energy scale is
average fission energy.
If the average fission energy is greater than
divided into three domains.
100 Kev, the system is called a fast reactor; if the average fission energy
lies between 100 Kev and thermal energy, the reactor is designated inter
mediate; and if the bulk of the fissions are caused by thermal neutrons,
In some areas of reactor technology,

the reactor is designated thermal.
the expression epilhermal has become popular for describing reactors in
which the average fission energy is in the electron-volt range.
typical energy spectrums for the neutron population
in the three principal reactor types. The ordinate in these curves gives
the relative density of neutrons as a function of their kinetic energy.
The intent here is to present some illustrative sketches which will give
Other than indicat
some idea of the nature of the distribution function.
ing the general shape of these curves, the only feature of interest is the
relative "sharpness" of the spectrum in the thermal reactors.
Figure
1.4 shows
REACTOR ANALYSIS
20
[CHAP.
The classification of the neutron-energy spectrum of a given reactor

is determined principally by the neutron-moderating
materials which
If the nuclear masses of the nonfuel components of the
it contains.
reactor are relatively low, then the neutron spectrum will correspond to
that of a thermal reactor (cf. Fig. 1.4c); if they are large, a "fast"
spectrum will result (cf. Fig. 1.4a). The spectrums of intermediate
reactors may be due to a number of nuclear characteristics, the presence
of nuclear masses of moderate magnitudes being one cause.
(a)
(b)
(c)
Neutron energy
Fig.
1.4
Neutron-energy

reactors.
spectrum in (a) fast,
(6)
intermediate,
and (c) thermal
b. Applications of Nuclear Reactors.

Since the successful operation
of the first reactor by Fermi's group at Chicago in 1942, nuclear reactors
have been utilized for a number of important applications.
These can
under the following three headings: (1) power production,
(2) research, and (3) breeding and converting.
The first of these applications, power production, may be further
be grouped
according to three specific functions: central-station power,

package power, and mobile power.
Central-station-power applications
of nuclear reactors refers to the production of power for large municipal
classified
or industrial areas.
Package-power applications, however, refer to power
In this classification, one might include
production for limited facilities.
power plants for areas in unusual climatic conditions, advanced bases,
or small, isolated establishments.
The application of nuclear reactors
to power mobile units can conceivably include any device which is
designed for terrestrial operation or, for that matter, space craft as well.
The possibilities in this area of application are as yet hardly known.
The application of nuclear reactors for research purposes includes the
production of high-intensity neutron and radiation fields and the produc
tion of radioisotopes.
The use of radiation and neutron fields for experi
mental research in physical and biological sciences is well established,
and the construction of nuclear reactors for use in university and research
laboratories is already under way.
The production of radioisotopes has,
likewise, become an important application of nuclear reactors, no doubt
stimulated by the considerable success of radiotracer techniques.
SEC.
INTRODUCTION
1.2]
21
The third, and the earliest, application of nuclear reactors was for the
The first of these devices were the
production of fissionable materials.
Hanford reactors which were built for the production of plutonium by
Nuclear reactors which
the conversion of the U238 in natural uranium.1
material are generally classified as breeders or con
has become the custom to distinguish between these two
reactor types by designating as breeders those reactors which produce
fissionable material of the same kind as the actual nuclear fuel which
produce
verters.
fissionable
It
on the other hand, are reactors that pro

duce fissionable material different from the reactor fuel which maintains
The best-known example of a converter is the Han
the chain reaction.
which
ford type reactor
produces Pu239 from U238, using the U23s in the
Converters,
is
is,
natural uranium as the fuel. The important fact in any case is that some
fissionable material is produced from the fertile material added to the
The U238 in the Hanford reactors is such a material, as is Th232
reactor.
In either of these fertile materials, the nuclear
yields
(which
U233).
is
involved
the
capture of neutrons by the nucleus and the subse
process
Although breeders and converters
quent formation of fissionable nuclei.
provide only a portion of the fissionable material required in present
reactor technology, they are, nevertheless, an important source. Another
application of these reactors which is being given increasing attention in
recent times is the combined production of fissionable material and power.
The idea of a power source which is capable of producing its own fuel
of course, highly attractive from the economic standpoint. Although
many such devices have been proposed, their practicability
not yet
entirely established.
Many questions in both nuclear physics and
economics must be answered before such applications can become engi
neering realities.
In
Sec. 1.2a, we discussed the classification of reactors according to

their nuclear characteristics.
There are important correlations between
these characteristics
and the operational purpose of the reactor.

This
in
due
part to engineering considerations and in part
is
is
interdependence
to the nuclear properties of the reactor constituents.
A simple illustra
tion of this point
found in the case of the thermal reactors.
Great
pains are taken in the design of these systems to obtain efficient moderator
materials
reason
to slow the fast

thermal spectrum
fission
to thermal energy.
The
that the fission characteristics
neutrons
is
is
desired
of the principal fuels are superior in the thermal-energy range. As will
be shown later, this results in
minimum requirement in fuel mass.
would be important in a design situation which
limited by the scarcity of nuclear fuels.
We have here another reason
for the selection of the thermal-reactor type for the earliest designs.
The radiative capture of a neutron in U"8 yields
is
This consideration

they consume.
Pu"'
[cf.
Eq.
(1.17)].
22
REACTOR
1.3
ANALYSIS
[CHAP.
The General Reactor Problem
of Neutron Population. A complete understanding of

of a reactor requires a detailed description of its
In its most general form, such a description must
neutron population.
give the distribution of the neutrons in space, energy, direction of motion,
a. Description
the nuclear behavior
and time.
A study of the various nuclear features of
a reactor
usually
begins with a general statement consisting of a detailed neutron count

which is summed up in an integrodifferential equation of the Boltzmann
These equations are conveniently written in terms of the neutrontype.
density function, and the statement of the equation gives a balance
relation between the various nuclear processes which can affect the
neutron population.
Generally speaking, the dependence of the neutron density upon all

seven independent variables implied above is customarily indicated in
the initial, very broad statement of the problem.
Very seldom is such
For example, the time-dependent behavior of the neutron population is

of interest only in regard to the nuclear stability and control of the
reactor, and in many cases an analysis of the steady-state problem is
Likewise, the directional dependence of the neutron
entirely adequate.
density (i.e., the specification of the direction of motion of the neutrons
being studied) is of little interest except in regions of the reactor system
close to discontinuities in the material composition (boundaries, etc.).
Only when a complete description including all seven variables is
required is it necessary to solve the Boltzmann equation in all its generali
ties. Frequently, simplifying assumptions and limiting conditions can
be imposed which reduce the integrodifferential
equation to more tract
able form.
Thus much of the subject of reactor analysis is devoted to
the development and the application of simplified analytical models which
define, within the limits of engineering needs, the nuclear characteristics
of the reactor complex.
b. The Problems of Reactor Physics.
It is convenient to list the
types of nuclear-reactor problems under the following general
categories: (1) critical-mass calculations, (2) neutron-density and fission-
various
confronted.
It
is,
energy distributions, (3) control effectiveness, (4) reactor stability, (5)

fuel burnup and production of reactor poisons, (6) nuclear accidents and
The computation of the critical mass1 for a given
special problems.
reactor system is one of the first tasks with which the reactor analyst is

In most situations, attention is

generality required in actual practice.
confined to certain specific aspects of the neutron problem, and less
precise descriptions involving only a few variables will often suffice.
likewise,
one of the easier calculations
to perform.
The nuclear fuel mass required to sustain a chain reaction in a reactor configuration
of specified composition.
SEC.
1.3]
INTRODUCTION
23
Frequently coupled with this problem
is the determination of the opti

which yields a minimum fuel mass.
Reason
able estimates1 for preliminary studies can be made with relatively little
effort, and many crude analytical models are available for this purpose.

mum nuclear configuration
Accurate estimates require more elegant methods or the use of critical

experiments. Although precise mass figures per se are only infrequently
required in modern practice, this information is usually available in every
reactor study as the by-product of solutions to more essential problems
involving neutron-density distributions. As a practical matter, rela
tively large discrepancies in mass estimates can be readily accommodated
with the increased availability of high-enrichment fuel samples.
Neutron-density and fission-energy distributions in a reactor can also
be obtained with sufficient accuracy for many needs by the use of rela
tively simple models. Again, more exact information requires elaborate
mathematical analysis which is usually carried out with the aid of fast
Occasionally associated with these
computing machines or simulators.
of the fuel-loading pattern which will
problems is the determination
yield a spatially uniform production of power throughout the reactor.
Except in the case of relatively simple core geometries, nonuniform fuelThe
loading problems can be treated only by numerical methods.
distribution of neutron density and power production are closely related
A knowledge of these
and of paramount interest to the reactor analyst.
distributions determines the design of the cooling system and provides a
Once the density of neutrons
detailed count of the neutron population.
known at every point in
all the important

interactions
are easily computed.
Such information
neutron-nucleus
yields the rate of fuel consumption, the production of radiation, the rate
is
a reactor, the reaction rates of
of fuel breeding and conversion, and the flow of neutrons out from the
This last factor determines, in part, the shielding
surfaces.
requirements and the intensity of neutrons beam which may be available
for research purposes.
The problems of control effectiveness deal with those considerations
reactor
which enter into the design and the operation of the nuclear elements
which control the neutron population.
The mechanical configuration
forms;
of these elements may take many
the determining factor is
usually design and operating convenience.
The customary device is a
rod or some other cylindrical unit which contains a material that is a
By some appropriate motion of the unit, this
strong neutron absorber.
material is brought into a region of the reactor which contains a high
By either withdrawing or inserting this neutron
density of neutrons.
sink, the neutron balance can be quickly altered and the population
caused to rise or fall, thereby accelerating the chain reaction or shutting
it off. Other control devices that have also been successful include the
1 In the order of 10 to 15 per cent accuracy.
REACTOR ANALYSIS
24
[chap.
use of burnout poisons and movable reflectors.

The use of burnout poisons
be better included with the group of reactor-physics
for control would
problems dealing with long-time nonstationary processes, such as fuel

Control devices which utilize movable reflectors rely upon
burnup.
the control of the neutron population through changes in the rate of
neutron loss by escapes from the reactor.
The problem of estimating
accurately the effects of such devices on the neutron-density distribution
is difficult; however, if the effect of the device is not "too large" in the
nuclear sense, perturbation methods may be useful for such calculations.
Otherwise, multiregion analyses may be required which in some instances
involve the solution of the Boltzmann equation.
Here again critical
experiments are helpful in checking the analysis.
Although most reactors contain some sort of external control device,

at least to meet emergencies and make planned shutdowns, such devices
Many reactors in operation
are not required for all types of reactors.
today contain some degree of inherent nuclear stability. Reactor designs

which incorporate liquid coolants are frequently of this type.

In such
reactors, any disturbance (or control command for that matter) which
momentarily changes the operating condition of the system (e.g., opening
the throttle on the turbine of a power reactor) will cause some fluctua
tions in the neutron population and power production, but these quickly
The prediction
decay and the system returns to steady-state operation.
of the thermal response of the reactor to these disturbances belongs to a
All these involve questions of dynamic
related class of problems.
response and stability, and in the majority of the cases of engineering
Although some insight into

they are nonlinear in character.
the physical features of a given situation can be gained from linearized
to the problem, the nonlinear solutions are ultimately
approximations
Much of this work is carried
required for a complete understanding.
interest
out with the aid of simulators and high-speed computing machines.

Another class of time-dependent problems of concern to the reactor
physicist are questions on fuel burnup, poison production and burnup,
These problems differ from those on reactor
breeding ratio, and the like.
stability in that they involve time scales measured in hours (or years)
in contrast to stability problems which are concerned with fractions of a
Reactor-analysis problems, such as the determination of critical
second.
distributions, are based on the steady-state

mass and neutron-density
condition
of
the
reactor.
The day-to-day operation of the
operating
reactor at steady state involves, however, long-time changes in the fuel
circulating-fuel reactors, the fuel is
As
introduced into the reactor according to some predetermined cycle.
the fuel is consumed,1 some gradual adjustments can be made by means
concentration.
1 The
Except in the
case of
absorption of a neutron by a fuel nucleus "destroys" the nucleus whether it
fissions or not.
SEC.
1.3]
INTRODUCTION
25
of the controls (shim rods) which will introduce additional fuel into the
reactor and thereby maintain the chain reaction.
The rate of fuel consumption and the amount of control adjustment
available
analyst.
for this purpose are among the problems which confront the
Closely related to the fuel-burnup problem is the production
of fission fragments which are strong neutron sinks.

These nuclei are
called poisons and are serious competitors for the neutrons in the chain
The production of these nuclear species and their concentra
reaction.
tions at all times during the operating life of the reactor are determining
factors in the design of the control elements and the fuel-loading require
Another set of problems which are similar in nature to those of
ments.
fuel burnup and poison production is the production of fissionable mate
rial from fertile nuclei. The time scales involved are comparable, and
the same analytical methods may be applied in each case.
In addition to the class of conventional reactor-analysis problems men
tioned above, one frequently encounters special problems which arise
from emergency or unusual operating conditions of the reactor.
It is
to include also in this set the analytical problems dealing
with special loading configurations designed to accommodate experiments.
Such situations as the placing of test specimens and measuring devices in

convenient
or near the reactor proper require special attention.

c. Methods of Attack and Analytical Models.
In the present discus
sion we summarize the principal analytical models and tools which have
It is believed that this
proved useful for treating reactor problems.
information will serve best as an outline to locate and orient the various
ideas and methods within the over-all structure of the subject.
The general field of problems described above, except in some special
areas, may be treated by the well-known methods and analytical models
of mathematical
It has already been noted that the most
physics.
general description of the neutron population usually starts with a
neutron-balance relation of the Boltzmann type.
The Boltzmann equa
tion was developed in connection with the study of nonuniform gas
mixtures, and the application to the neutron problem represents a con
siderable simplification of the general gas problem.
(Whereas in gas
problems all the particles are in motion, in reactor problems only the
neutrons are in motion.1)
The fundamental equation of reactor phys
ics, then, is already a familiar one from the kinetic theory.
Further,
many of the most useful neutron models obtained from approximations
to the Boltzmann equation reduce to familiar forms, such as the heatconduction, Helmholtz, and telegraphist's equations.
These simplifica
tions result from the elimination of various independent variables in the
1 This
is not strictly true because the nuclei themselves are in motion, but in many
problems the neutron speeds are very great, and nuclear motion can be neglected in
comparison.
REACTOK ANALYSIS
26
[CHAP.
is,
general formulation and are usually introduced by virtue of certain

limitations imposed on the physical situation.
Perhaps the most elementary, and limited, model in reactor analysis
is the so-called infinite medium.
The essential feature of this model
is
of course, that the geometric confines of the real problem are ignored and
attention
focused on other aspects of the situation.
The infiniteis
especially useful for treating reactor problems which

deal with phenomena well inside the physical boundaries of the system.
Also, chain reactions in systems that are large in a nuclear sense can be
medium model
An especially important problem in reactor phys

the neutron-energy spectrum, may be studied by means of this model.
handled in this way.

ics,
is
is
is
it
is
to its spatial dependence.

third analytical model which combines the most important features
of the two already described
the continuous slowing-down diffusion
model, better known as the Fermi age model.
Fermi age theory offers
is
first-order approximation to the Boltzmann equation.

In this model,
the neutron population
described in terms of two independent variables,
The directional dependence of the neutron density
energy and position.
is
eliminated by assuming that the velocity vectors of neutrons in regions

far from boundaries are nearly uniformly distributed, i.e., isotropic.
The
is
is
application of this assumption reduces the Boltzmann equation to par

tial differential equation of the heat-conduction type.
Fermi age theory
most effective for analyzing reactors which contain a moderator of rela
tively large mass (anything heavier than deuterium).
It most useful
for obtaining fuel-mass calculations and neutron-density distributions.
As we have already mentioned, reactors are usually designed to
A

is
it
is
The question to be answered here

the determination of the density
distribution of neutrons as a function of kinetic energy from the instant
neutron appears as
fission particle down through the slowing-down
captured.
process until
second model which has great usefulness in treating reactor problems
the one-velocity model.
We noted in Sec. 1.2 that the energy dis
tribution (spectrum) of the neutrons in the chain has
sharply peaked
form (cf. Fig. 1.4).
As first approximation, one can treat the neutrons
in the reactor as though they all possess the same average energy defined
by the shape of the spectrum.
As crude as the approximation may seem
at first,
has proved to be extremely effective for treating a large variety
of reactor problems.
It especially useful for studying thermal reactors.
By eliminating the general energy dependence through the assignment
of a single energy (or the corresponding one velocity) to all the neutrons
in the chain, the Boltzmann equation for the multiplying medium1 can be
reduced ultimately to the Helmholtz (stationary-wave) equation; this
frequently called the neutron-diffusion equation. The
approximation
application of this model limits the description of the neutron population
system containing fissionable material.
SEC.
INTRODUCTION
1.3]
include
Thus, the simplest configurations,
a reflector.
the solution
27
at best, require
problem in order to
of a two-region
(core and reflector)

As a rule, engineering require
the reactor system considerably, and the analyst is
obtain the neutron-density distributions.

ments complicate
frequently confronted with three- and four-region boundary-value-type

Both the elementary one-velocity diffusion model and the
problems.
Fermi age theory may be applied to these systems. Diffusion theory
is the most easily applied of the two and can be extremely useful in
handling multiregion problems, provided that the one-velocity approxi
mation
If
gives an adequate description of the neutron-energy spectrum.

diffusion theory is inadequate, and the energy distribution must be
taken into account, the Fermi age model may still be used; however, the
analysis of multiregion systems which include the neutron-energy spec
The
trum usually requires the application of numerical methods.
so-called mulligroup methods have been developed for this purpose.
Many of these methods are based on the continuous slowing-down model
One very useful model which has been devised for
Various forms
treating multiregion reactors is the two-group model.
of this model have been used for analyzing reactors of both the thermal
of Fermi age theory.
The essential feature of the model is the description of the

The bulk of the
neutron population by means of two energy groups.
neutrons which cause most of the fissions is treated as one group, and
neutrons of all other energies are treated as a second group.
The par
ticular merit of this method is that reasonably good estimates of the

and fast type.
of the reactor can be obtained by hand calculation.

More involved multigroup (three or more) methods must be handled on
nuclear properties
large computing machines.

Another type of nonhomogeneity
encountered
in reactor analysis
is
The multiregion configura

that associated with heterogeneous reactors.
tions mentioned above are on a relatively large scale, and each region in
In a
(e.g., the core, the reflector) is a homogeneous medium.
heterogeneous reactor, the core consists of a lattice of many small cells,
each of which contains both moderator and fuel.
The moderator and
the fuel exist as separate regions; however, the scale of the nonhomo
An
geneity is much smaller than in the general multiregion problem.
analysis of this type of reactor requires a detailed study of the nuclear
itself
characteristics
of the individual cells.
In particular, the analyst must
determine the space and the energy distribution of neutrons in these cells.
From this information, he can readily predict the over-all nuclear char
For many purposes, simple diffusion theory

acteristics of the reactor.
and the Fermi age model will give adequate descriptions of these distribu
tions in the cells.
More accurate analyses require the application of
approximations to the Boltzmann
experiments are also used in this connection.
the models which have been mentioned can be altered to include
semiempirical
equation.
All
models or higher-order
Exponential
REACTOR ANALYSIS
28
[CHAP.
the time dependence of the neutron-density distributions.

Thus, each
of these models can be applied equally well to nonstationary reactorphysics problems as to steady-state problems.
However, the differential
equations which include the time dependence can be easily solved for
only
simple reactor geometries.

More complex configurations
involve
two
or
more
and
regions),
systems which include energy
(which
as
are
more difficult to treat.
dependence
well,
Again, one must resort
to numerical or graphical methods for analyzing these systems.
Timea few
problems which include the coupling between the reactorpower output and the neutron population will, in general, involve a
nonseparable relationship between the time and the space variables.
However, in many instances, the important features of the problem can be
determined by using simple physical models which do separate out the
Such solutions will not yield an accurate
space and the time variables.

dependent
description of the neutron physics, but they can give valuable insight
into the basic characteristics of the system in question.
The analytical treatment of reactor-physics problems is derived in
main part from the fundamental Boltzmann equation.
These models
of this equation to various degrees of accuracy.
are approximations
Which model is selected for a particular reactor problem is determined
by the accuracy required and the nature of the phenomenon to be
When the problems deal with the bulk, or integral, neutron
analyzed.
characteristics of the reactor, relatively crude models can usually be
Nuclear-fuel-mass
calculations fall in this category.
On the
applied.
This
other hand, more detailed studies require more elegant methods.
relatively
of
the
small
dimensions
which
are
is a consequence
involved
In these situations, many of the simplifying assump
in these problems.
tions about the neutron-density distribution and the composition of the
media (such as isotropy) are no longer valid, and more precise descrip
Problems which involve regions close to material
tions must be used.
boundaries and heterogeneous cell calculations are problems of this type.
Accurate descriptions of the physical situation in these problems can be
This is not to say, however, that
given only by the Boltzmann equation.
Again, cruder
the Boltzmann equation must be solved in all such cases.
models have been devised which can be applied even to these problems,
and previous experience can always reduce appreciably the work required
One of the purposes of the present work
to obtain adequate solutions.
is to explain some of these approximations which have proved to be use
ful and to demonstrate their application to typical situations.
CHAPTER
PROBABILITY CONCEPTS AND NUCLEAR

CROSS SECTIONS
2.1
Neutron-Nucleus Interactions
Neutron Life Histories.
of the problems of
reactor physics require
instance a description of the average
behavior of the neutrons in the reactor.
The average behavior is
obtained through an understanding of the probabilities of the various
a.
Detailed
analyses
in each
interactions which can occur.

From a knowledge of
these probabilities can be obtained a complete description of the neutrondensity distributions and of the reaction rates of all neutron-induced

neutron-nucleus
The nuclear reaction rates determine

processes throughout the reactor.
the competition between the fuel and other nuclear components of the
reactor for the free neutrons and thereby establish the characteristics
If the fuel can compete favorably, the chain will
of the chain reaction.
and the neutron population either will be maintained
the competition is unfavorable, the neutron popula
tion will eventually die off and all nuclear reactions come to a halt.
As an example of a typical life history of a neutron, consider the
be self-supporting
or will increase.
trajectory of
If
born from fission in a reactor which consists of a

On the average, the neutron will be
single medium (i.e., no reflector).
ejected from the point of fission with a relatively high kinetic energy
a neutron
It will move outward

(about 2 Mev) and with a random direction.
from this point in a straight-line path until it encounters a nucleus or
the boundaries of the system.
passes through
Throughout reactor
analysis, we consider the region outside the medium of interest as being
void of material so that there is no chance for back-scattering of the
Thus, a neutron which passes through the boundaries of the
If, on the other hand, the neutron encounters a
reactor is lost for good.
nucleus, by which we mean if the neutron passes sufficiently close to a
nucleus that the nuclear forces interact,2 then a reaction will occur which
neutrons.1
will either remove the neutron or alter its energy and direction.
1 In reality, this will seldom be the
case.
In most situations, the reactor will be
surrounded by the atmosphere, and some neutrons will always be scattered back to
the reactor by collisions with air molecules.
But, these events are negligibly few in
number and can be ignored in so far as their effects on the chain reaction are concerned.
* The range of nuclear forces is in the order of
10" cm.
29
ANALYSIS
REACTOR
30
[CHAP.
For the present, we will not concern ourselves with the details of the
various collisions which can occur. It will suffice to distinguish between
If the neutron is scattered on its first collision,
scatterings and captures.
Again,
it will lose some energy and will move on in a new direction.
will
is
that
this
new
direction
allow
the
neutron
to
possibility
there
a
reach the boundary of the reactor without further collisions and so escape
from the system. On the other hand, if a collision occurs, there is once
Fig.
Spatial trajectory of a neutron.
2.1

more a probability that the neutron will be absorbed or that it will be

If
scattered, with subsequent loss of energy and change in direction.
this second collision occurs and it too is a scattering, then once again
there is a possibility that the neutron will be lost by escape or absorption.
This scattering process will continue until the neutron is eventually lost
to the system.
For any single neutron, the spatial trajectory will be
some complicated series of broken straight-line paths which extend from
its point of birth to the point at which it is eventually absorbed or leaves
the boundary of the reactor (refer to Fig. 2.1).
Number of scatterings
Fio.
2.2
Neutron trajectory in energy-space.
The trajectory of the neutron in terms of its kinetic energy is deter

Thus,
mined from its speed over each straight-line portion of its path.
the trajectory in energy-space is a staircase function, as illustrated in
Fig.
2.2.
Note that each scattering collision causes a jump change in the
neutron energy.
As we have already mentioned, the history of any one
neutron is of little importance to the over-all neutron problem.
We are
concerned only with the average behavior of the neutrons, and this can be
sec. 2.2]
PROBABILITY
CONCEPTS,
CROSS SECTIONS
31
described in terms of the probabilities of the various nuclear events and

the probability for escape at each instant during the life of a neutron.
One of the goals of reactor analysis, then, is the calculation of the prob
ability of a neutron-nucleus interaction of every possible type.

b. Types of Interactions.
Nuclear collisions may be classified under
two main categories, absorptions and scatterings.
It will be convenient
for the purposes of reactor analysis to define absorptions in the following
way:
Absorption: Includes all neutron-nucleus reactions whose end products
do not include a neutron and the original nucleus.
The following processes
are specifically included in the absorption interaction:
1. Capture of a neutron: all neutron absorptions which do not produce
fission, for example, radiative capture (n,y) and charged-particle produc
ing reactions such as (n,a) and (n,p), etc.
Fission
The various scattering interactions
2.
are included
under the following

general heading:
Scattering: All neutron-nucleus reactions which produce a neutron and
a nucleus that differ from the reactants only in their energy content
(kinetic and/or internal). Scattering reactions are of two types:
Elastic scattering: The products of the reaction are a neutron and the
struck nucleus in its original internal-energy state.
2. Inelastic scattering: The products are a neutron and the struck
1.
nucleus in a higher internal-energy state.

For the present discussion, all the neutron-nucleus reactions defined
Thus, when
above will be included under the general heading collision.
we talk about the probability of a nuclear collision, we imply any one
of the four principal reactions: capture, fission, elastic scattering, or
inelastic scattering.
2.2
a. Average
nucleus
Probability of Collision
The calculation of the number of neutronthat occur per unit volume per unit time at each
Behavior.
collisions
space point of a medium supporting a neutron gas is based on a knowledge

of the various interaction probabilities between the neutrons and the
nuclei of the medium.
These probabilities depend upon the varieties
and densities of the nuclei present, as well as upon the relative speed of
the neutrons.1
A typical cubic centimeter of a reactor may contain 10" nuclei and
107
neutrons.
1 The
In any calculation
involving such enormous
numbers
nuclei are generally in motion, and the relative speed of the neutrons is given
by the magnitude of the vector difference between the velocity of the neutron and
that of the nucleus.
32
REACTOR ANALYSIS
[chap.
of particles, it is neither possible nor desirable to follow the detailed

behavior of each particle and to obtain in this way the exact number of
Instead, it is entirely
collisions at each space point and instant of time.
sufficient to concentrate attention upon the average behavior of the par
As usual in
ticles and to calculate the expected number of collisions.
statistical problems, the large number of events considered guarantees
that deviations from the expected behavior will be slight for the individual
The exact details of the life history of a neutron wandering

particle.
through a medium are not important, since attention is confined to the
This average behavior is obtained by consideration
average behavior.
of the functions describing the probabilities of the various nuclear col
The present section is devoted
lisions1 that the neutron experiences.
to the construction of these probability functions.
As an initial approach to the basic probabili
b. A Physical Approach.

ties for neutron-nucleus interactions, consider the following situation:

A number of neutrons na, all of a given speed, are incident normally
upon a unit area of surface of a uniform
semi-infinite medium (see Fig. 2.3).
Each
neutron will penetrate some distance into
the medium before suffering its first colli
For the present, the

type of nuclear interaction which occurs
V/s Semi-Infinite
n0 neutronsper unitarea
^'
medium
(and unit time)
will be left unspecified.
The problem is to
Fig. 2.3 Collimated beam of
find the distribution in space of these first
neutrons incident upon the face
collisions.
Let the number of neutrons
of a semi-infinite medium.
surviving as far as distance x without col
liding with a nucleus be defined as n{x). Now we assume that the fraction
of the neutrons which suffer their first collision in a small interval dx is a
If we denote the
constant, independent of the position of the interval.
number of neutrons making their first collision in dx by dn(x), then the
mathematical statement of our assumption may be written
sion with a nucleus.
1 dn(x)
dx n(x)
= 2
(2.1)
where 2 is a constant appropriate for the given medium, the particular

neutron-nucleus interaction, and the neutron speed.
The solution to Eq. (2.1)
is easily seen to be
n(x) = ntfrZx
(2.2)
under the boundary condition

n(0) = no
1 The
(2.3)
frequency of neutron-neutron collisions is some six orders of magnitude less

than neutron-nucleus collisions and, therefore, entirely ignored in reactor calculations.
SEC.
PROBABILITY
2.2]
CROSS SECTIONS
CONCEPTS,
33
The physical content of Eq. (2.2), that the number of uncollided neutrons
falls off exponentially with distance, may be subjected to experimental
verification directly, and the fact that the form of (2.2) is found to fit
the data very well is evidence for the validity of the assumption expressed
byEq. (2.1).
The result (2.2) may also
The number of
of two factors; thus
be stated as the product
neutrons
the number
of neutrons
surviving to
starting
distance x
at x = 0
the probability that

a neutron
will
not
suffer a collision
|_in traveling distance
(2.4)
x_
If
this equation is compared to (2.2), we arrive at one of the basic prob

ability functions for neutron-nucleus interactions:
probability that
a neutron
w(x) = have an uncollided

length x

The related probability that
a neutron
flight of =
will suffer
a first collision
reaching distance x is called the distribution function for

and is given by
probability that
a neutron
Further,
- w(x)
the
before
first collision
will have suffered its
Fix) = first collision before distance

=
(2.5)
e~Xx
e~Zl
(2.6)
frequency function for first collision, i.e., the probability that

the neutron have its first collision in unit distance about x, may be
obtained from the distribution function; thus
the
fix)
probability that a neutron have its first collision

in unit distance about x
^-Fix)
ax
= Xe-Z*
(2.7)
In the sections which immediately follow (c and d) these functions are

again derived in a more complete treatment which exhibits more clearly
the underlying hypothesis (the general form of the distribution and
functions is shown in Fig. 2.4). The physical content of
these probability functions is discussed in Sec. 2.3.
c. The "As Good As New" Hypothesis.
There is an alternative for
frequency
mulation of the problem of calculating the probability of collision which

may be obtained by asking the following question : If a neutron is observed
moving through a medium at some speed v, what is the probability that
it suffers a nuclear collision before it has traveled a distance x? This
No detailed description is
question is stated in very general terms.
given of the physical
condition,
composition,
or geometric form of the
REACTOR ANALYSIS
34
[CHAP.
medium, nor of the type of collision which is to occur.

Although each of
these factors is pertinent and together they determine a particular situa
tion, the general question is still meaningful and the approach generally
We will discover shortly that the answer to our question may
written in terms of certain parameters which are related only to the
Thus,
physical properties of the medium and the speed of the neutron.
if one asks what is the probability that a 100-ev neutron will be absorbed
by a U235 nucleus within a distance of 5 cm, the answer can be expressed
in terms of the density and certain nuclear characteristics of the Um
The specific question is concerned with a par
nuclei in the medium.
ticular physical situation and involves detailed information. The general
problem inquires into the analytical description of the phenomenon and
applicable.
be
examines the conclusions predicted by the model.

The essential feature of the analytical model we select for calculating
the collision probabilities is the "as good as new" hypothesis which is
is
is,
along its path before it encounters a nucleus is now measured from the
the neutron
That
as good as new and can be considered
position x.
The mathematical statement of this
to have originated at point x.
F(x)
probability that
neutron make its first

collision before reaching
given point x
(2.8)
<
PU
X)
physical principle may be written in terms of the probability function1

P{ < x), which we define as
is
is
is
where
a random variable and F(x)
the corresponding distribution
function. For this calculation, we require also the conditional probabil
defined as
ity function. The conditional probability
between x and x +
it
x),
occurs
(2.9)
= occurs after x(
>
y\ >
x)
< x
P(x <
probability that given the collision
y\t
reaches
before
collision
it
makes
[i.e.,
H. Cramer, "Mathematical Methods of Statistics,"

ton University Press, Princeton, N.J., 1951.
chaps.
y)
P(
<
>
x)
<
P(x
<
Or,
Eq. (2.8)].
first observed at
is
it
if
is
The "as good as new" hypothesis can be stated with the aid of these
known to have survived to x, then the
Thus,
the neutron
moves an additional
probability that the first collision occurs before
neutron
identically the probability that
distance
[i.e., Eq. (2.9)]
functions.

defined in the following way: If a neutron moving through a medium is

known to have traveled a distance x along its path without suffering a
Thus, if a neutron is as yet uncollided
collision, then it is as good as new.
at position x, the probability that it will travel an additional distance y
13,
(2.10)*
14,
and 15, Prince
sec.
PROBABILITY
2.2]
This equation is the mathematical

hypothesis
ity functions.
35
of the "as good as new"

we construct the collision probabil
statement
from which
and is the basis
CROSS SECTIONS
CONCEPTS,
d. Distribution Function for Collision.
The distribution function for

collision based on the relation (2.10) may be derived by applying the
properties of a conditional probability, namely,
P(*
P(x < t < x + |{ > x)
<
*f
V>
* >
*>
(2.11)
The function P(x < < x + y, { > x) is the joint probability that the
collision occurs in the interval (x, x + y) and that the collision point is
If we substitute this relation into (2.10), we obtain
beyond x.
Pit
< V) P(Z > x) =
P(x
<
< x + y,
> x) =
P(x
<
< x
y)
(2.12)
The last equality follows from the fact that P(x

implies P(x < $ < x + y).

terms of the corresponding
definition (2.8) and write1
F(y){l
It
is convenient
distribution
- F{x)]
This relation is in the form of

through by y, we obtain
we suppose that,
for
each x,
= l-
y)-
[F{x + y)
Fix) has
We apply the
functions.
a difference
i F(y)[l - F(x)\
If
F(x +
< { < x + y, f > x)

to write this equation in
F(x)
(2.13)
equation.
If
- Fix)]
a derivative,
we divide
(2.14)
the right-hand side
of (2.14) tends to F'(x) as y tends to zero; the left-hand side must there
fore also converge.
Thus we find
S[l
- Fix)]
F'ix)
(2.15)
where the constant 2 is given by
lim
v o
^
y
= F'(0) = S
As noted, this limit must exist, and since it does exist, F(0)
the identification (2.16).
The solution to Eq. (2.15) is
Fix)
(2.16)
= 0, hence
- e-s*
which agrees with our previous result (2.6).

As we have already noted, the function F(x) is properly called the
distribution function for the first collision of a neutron along its path.
The derivative of this function, which is called the frequency function,

1
Ibid., p. 166.
REACTOR ANALYSIS
36
[CHAr. 2
is usually of somewhat greater physical interest.

This function has been
given the symbol f(x), and a formal definition was made in (2.7).
We summarize the first collision distribution function and frequency

function below:
=
F(x)
- e-s*
2e-Sl
(2.17)
fjf(w)dw
(2.18)
f(x)
Note also that

F(x)
f*^f(w)dw

The general form of these functions is shown in Fig. 2.4 where, for
convenience, the parameter 2 is taken as unity. The origin of the x axis
is the point at which the uncollided
neutron is first observed. The func
tions F(x) and f(x) are identically
zero for all x < 0. The distribution
function F(x) approaches unity in
the limit as x becomes arbitrarily
large, i.e., for very large x, the prob
o
ability that the neutron has made
Fig. 2.4 Distribution

and
functions for first collision.
frequency
a collision
is nearly
certainty.
Note, also, that the probability that

the neutron makes its first collision per unit path of travel is greatest near
its point of origin.
Likewise, for any position xo < Xi
f(x0) dx >
f(xi) dx
(2.19)
It
is easy to show that f(x) is indeed the probability of collision per unit
path of travel around x by means of the following computation: The
probability that the first collision per unit path traveled occurs around x
is given by
< x + Ax)
Ax
axô
< < x + Ax) = F(x + Ax)
- F(x)
< + A*)
- Fix)
lim P(x
But,
P(x
<
Therefore,
Um
P(s<f
AX
Al-0
2.3
Hm
AI-.0
Fix + Ax)
Ax
Macroscopic Cross Section
Number of Collisions. The parameter 2 introduced in

intimately related to the nuclear properties of the struck
In addition, it has a
nucleus and the relative speed of the neutron.
physical significance associated with the concept of a mean free path.
This relationship is easily demonstrated by computing the average dis
tance to the point of the first collision, i.e., the first moment of the fre
a. Expected
Eq. (2.1)
is
SEC.
PROBABILITY
2.3]
In the usual way,
quency function /(x).

"
Jfm
x =
CROSS SECTIONS
CONCEPTS,
dx = 2
xf(x)
JO
37
we write1 [using (2.17)]

xe~Zx
dx = ^ =
2i
(2.21)*
Thus the parameter 2 = 1/X, the reciprocal of the average distance to

collision (mean free path) ; or, saying it in another way
2 = average number of collisions per unit path of travel
(2.22)*
A physical argument which displays the basis for this definition of 2

be obtained from the simple experiment
According to Eq. (2.1)
may
\_dn
dx n
= 2 =
discussed in Sec. 2.26.
fraction of first collisions in dx about x
where n(x) is the density of uncollided
neutrons at x.
Or, rewriting

(2.23)
Since the derivative at the left of this relation gives the change in the
neutron density n(x) over the distance dx, it must correspond also to
the number of collisions which occur in dx. Thus a verbal statement
of (2.23) might be as follows
Number of
number of
number of collisions
collisions in
uncollided
per neutron per
traversing dx}
neutrons at x
unit path
(2.24)
If
this equation is compared to (2.23), we immediately identify 2 as the

quantity defined in (2.22).
b. Radioactive Decay.
By introducing appropriate changes in ter
minology, the analysis given above for the collision processes applies
directly to the problem of radioactive decay. This is possible since the
physical principle which served as a starting point for the analysis of the
collision problem has an exact analogue in radioactive decay ; namely, if a
nucleus has not suffered radioactive decay by a given time t, it is "as
good as new." For this calculation, we define A, the decay constant, as
the number of decays per unit time.
This quantity is analogous to the
parameter 2. On this basis all the analysis from Eqs. (2.8) through
(2.18) applies directly. If we call the frequency function for decay g(t),
then
g(t) dt =
=
Ae-A< dt
probability of decay between time I and
The distribution function G(t)

1 Ibid.,
p. 174.
is obtained by integrating
g(t).
dt
(2.25)
REACTOR ANALYSIS
e-A<
probability that nucleus decays before time
t,
is
the number of undecayed

iV(0)
then
the number at
- G(t)]
= JV(0)[1
N(t)
nuclei present at
If
g{d)dd
[CHAP.
(2.26)
G(t) =
J'_m
38
and N(t)
= iV(0)e-A'
(2.27)*
We have used here the function

v
probability that nucleus has

decayed by time
not
given by the derivative N'(t),
dN
dt
AN(t)
dN
dt
(2.29)
is
present at time
to
ta,
is
it
is
A,
a
a is
is
in accord with our previous definition of A.

which
radioactive decay process,
quantity frequently used in describing
the half-life of the nucleus.
to
the
constant
decay
which
related
and
given concentration of
takes
defined as the time
The half-life
radioactive (unstable) nucleus to diminish to half its initial value by
If we call the half-life
then the number of undecayed nuclei
decay.
N{t0)
N(0)c-*"
(2.30)
ta
0;
is
0)
thus,
As before, N(0) denotes the number of nuclei present at =
=
of
at which the number
nuclei
the time instant (measured from
We require, therefore, that
exactly half the initial amount.
present
is
obtained
iN(0)
(2.31)
by using this relation in Eq. (2.30).
result
is
The half-life
t0
N(U)
The
is

time
which decay per unit
is
Thus, the fraction of the nuclei present at time
The decay per unit time
is
q.
(2.32)*
The half-life, decay scheme, and energy of radiation (or of the emitted
particle) for some radioisotopes of interest to reactor technology are
The various known decay schemes for these nuclei
listed in Table 2.1.
are given in the third column.
The symbol a denotes radioactive decay
of the unstable isotope by the emission of an a particle; /3~, emission of a
negative electron; y, emission of electromagnetic radiation; K, orbital
SEC.
PROBABILITY
2.3]
Table
Isotope
6B
Half-life
CONCEPTS,
CROSS SECTIONS
Radioactive Isotopes
2.1
Decay
Energy,
scheme
Mev
Decay
Energy,
scheme
Mev
43 days
12
0.022 sec
Half-life
Isotope
pr
1.85
0.50
1.6
.C"
,N
5100 years
/s-
7.35 sec
0.15
6JTe135
3.5, 10
..I1"
<2 min
6.7 hr
p~
1.40, 1.00,
0.47
8o
28 sec
6.2, 6.7
(3-
2.9, 4.5
i,Al
14.8 hr
0-
>.K
1.8
.7Co
years
10.7 min
10

250 days
1.7
0~
3.01
1.8
hPiii1"
47 hr
/?-
1.9
,oTh!
0.93
0.25
0-
1.35
1010
or
3.98
0.32
(")7
1.11
Pu
108
4.56, 4.40
0.162
a
(e-)y
24,110
years
electron capture; /3+, positive

version of a gamma ray.1
The preceding analysis
8.91
years
0.056, 1.5
K(/3+)
1.6
1.1
years
1.54
1.39
0.247
r
r
hr
.oNd1"
1.38, 2.76
,oZn
(")7
P~
1.39
7
2.30 min
9.2 hr
1.6
i.Na"
.Xe
electron emission;
and
e~,
5.15
0.42, 0.20
internal con
of the radioactive-decay
process is entirely
general and applies to all the decay schemes mentioned above.
Some
of the isotopes listed in Table 2.1 may decay by the emission of more
than one type of particle.
For example, Al28 decays by either /3~ or
y emission.
The half-lives of nuclei that possess multiple decay schemes
are based on the combined probabilities of the various possible events.
Thus, in the case of Al28, the 2.30-min half-life means that a given sample
of this material will be reduced to one-half its initial concentration in
2.30 min by either
emission or y emission.
Although it was not specifically stated, the decay process selected for
the above derivation implied a single decay scheme. The analysis is
more general than this, however, and can be changed to account for the
possibility of additional modes of decay. To continue our example,
consider the case of Al28.
Let Gp(t) denote the probability that an Al28
nucleus will decay before time
that it will decay before time

[1
6,(0111
- Gy(t)]
_
=
t
t
f}~ emission and Gy(t) the probability

y
emission.
Then [cf. Eq. (2.28)]
by
by
probability that nucleus has

before time
by either
f}~
not decayed
or y emission
(2.33)
1A
detailed description of these processes can be found in any of the standard texts
on nuclear physics.
An introductory treatment is given by F. Bitter, "Nuclear
Physics." chap. 4, Addison-Wesley Publishing Company, Reading, Mass., 1950.
REACTOR ANALYSIS
40
[CHAP.
This statement follows from the assumption
that the two events 7 and

emission are independent, and, therefore, the joint probability is given
If \g and AT denote the
by the product of the individual probabilities.
decay constants for these two processes, then by (2.26)
[1
- G0(t)][l - Gy(t)]
= e-<Ao+V
(2.34)
The number of Al28 nuclei which survive to time t (i.e., have not decayed
by either 7 or /3_ emission) is the product of this joint probability and the
initial population iv*Ai(0); thus
=
NM(t)
It
half-life of the joint events
is easy to show that the corresponding

=
vrrr
+
Afi
This general approach can also
(2.35)
ArA1(0)e-<A+V (2.35)
is
<2-36)
Ay
be applied to nuclei which decay
by three
or more modes.

2.4
Macroscopic Cross Sections and Physical Properties
a. Microscopic Cross Section.

As presented in Eq. (2.22) the param
eter 2 is the expected number of neutron-nucleus
collisions per unit
In
of
traveled.
this
section
we shall attempt to relate this
length
path
parameter, from this definition, to the basic properties of the medium

As a first, admittedly crude,
through which the neutron is passing.
model of the physical situation, introduced primarily for its simplicity
and to give some feel for the physical situation, consider the nuclei as
spheres of radius R. Then, if the neutron passes within a distance R
of the center of the nucleus, a collision is sure to result; otherwise, there
Thus in traveling a unit distance along its path the neu
is no collision.
tron will collide with any nucleus whose center falls within a distance 7?
If N is defined as the
measured perpendicularly from the path line.
number of nuclei per unit volume in the medium, then along unit length
of path a total of ttR2N nuclei fall within the required range for collision.
Thus on the basis of this simple model
*
2 =
Thus 2
is
(tR*)N
"
TfQ
jf
(2.37)
directly proportional to the nuclear density and to the target
area presented by a given nucleus.

Of course, the picture of nuclei as hard "billiard balls" is not really
allowable ; a more valid calculation might proceed as follows :
Define a probability:
probability per unit time that

h(r)
dr = between radial distances r and
a neutron
found
r + dr from
a given
nucleus will suffer a collision with that nucleus
(2.38)
PROBABILITY
sec. 2.4]
CONCEPTS,
CROSS SECTIONS
41
as the neutron moves an infinitesimal distance ds, it spends a

time ds/v at a distance r from all nuclei in the spherical shell r to r + dr
Thus in this formulation
whose center is at the neutron (see Fig. 2.5).
the parameter 2 is given by
Then,
--
21
In
--"1
-*(r)*
JW
(2.39)*
the last step in (2.39) the integral of the weighted probability func
tion over all space is designated by the symbol a. Equation (2.37) may
also be written in this form with a particular choice for the function h(r).
Thus, quite generally, the expected number of collisions per unit
distance traveled 2 is proportional to the nuclear density through a
proportionality constant which has

the units of length squared and plays
This proportionality
per nucleus.
constant will be designated by the
symbol a and referred to as the mi
croscopic cross section for neutron-nu
cleus collision throughout this work.
Correspondingly, the parameter 2
will
be referred to as the macroscopic
cross section, although
its dimensions
are, as shown by Eq. (2.21), those of

The microscopic
reciprocal length.
cross
for nuclei
sections
veniently
measured
in
are con
barns, with
barn = 10-24 cm2; from the result

Fio. 2.5 Spherical shell dr about r.
of Eq. (2.37) a nucleus with a a of 1
barn would be expected to have a radius of the order of 6 X 10-13 cm,
1
which agrees roughly with various methods of estimating nuclear radii.

However, the hard-sphere model is not really valid, as may be seen from
the fact that for a given nucleus the value of a for interaction with a neu
speed
may be a few barns, while the value of a for the same
nucleus, but for interaction with a neutron of speed

may be several thousand barns.
ent from
b. Cross Sections for Neutron Reactions.
v',
tron of
only slightly differ
v,

the role of an effective target area
In our general treatment
of the collision problem (Sec. 2.2), we were not concerned with the phys
ical process which took place between the neutron and the nucleus at
Instead, we confined our attention to determin
the instant of collision.
Thus our initial examination

ing the probability that a collision occurs.
of the collision problem has not been limited to any particular type of
reaction, and the results obtained are generally applicable to interactions
For the problems of reactor
between many kinds of elementary particles.
REACTOR ANALYSIS
42
[CHAP.
physics, the interaction phenomena of primary importance are the

neutron-nucleus reactions.
We should like then to consider the applica
tion of the results of Sec. 2.2 to the description of the various types of
neutron-induced reactions.
It has become
the practice in the field of reactor technology to represent

types of neutron reactions by a fixed set of nomenclature.
will be convenient for our present needs to introduce this set here.
particular type of reaction will be described by the cross-section
the various
It
symbols given above, distinguished by an appropriate

most commonly used quantities are:
S =
27 =
=
2/
N<ra =
N<7y
Nof
subscript.
cross section
macroscopic absorption
= macroscopic radiative-capture cross section

= macroscopic fission cross section (by convention,
absorption includes radiative capture and fission

when dealing with nuclear fuels)
=
=
Na,
2,
macroscopic elastic-scattering cross section
2in = NaiB = macroscopic

The
(2.40)*
cross section
inelastic-scattering
The application of the probability concept to the description of these

We will demonstrate
processes is the purpose of the present discussion.
the basic idea involved by considering an example.
Suppose that we
have a neutron moving through a medium which consists of a single
nuclear species and that this medium has only absorption and elasticscattering cross sections; i.e., the only possible neutron-nucleus reactions
are capture
It
and elastic scattering.
is meaningful
then to ask the
following question: What is the probability that the neutron will travel
a distance x before it is either absorbed or scattered?
We can answer
this question by means of a calculation much like the one we used for the
radioactive-decay problem of Al28.
The two processes elastic scattering and absorption
and their respective distribution functions are
F,(x)
- e--
Fa(x) =
are independent,
e-1-*
(2.41)
These functions give the probability that the particular event denoted
by the subscript occurs before the neutron reaches x. Thus, the prob
ability that the neutron remains uncollided by the time it reaches x is
[cf. Eqs. (2.28) and (2.33)]
[1
- F.(x)][l -
Fa(x)}
- F.(x)
e-*
We have defined the product at the left as some function

This notation is used for convenience.
Note that
F0(x)
(2.42)
1
F(x).
- e-
= probability that neutron is

either absorbed or elastically
scattered before it reaches x
(2.43)
SEC.
2.4]
PROBABILITY
CONCEPTS,
CROSS SECTIONS
43
The corresponding frequency function fo(x) is
/.(*)
with
2 = 2
Soe-^
(2.44)
We conclude from
these results that, in a situation

of
neutron
reactions,
which allows two types
the joint probability that
either of the events occur may be described by means of a combined cross
This cross section, called 2 in the present instant, is simply
section.
This idea is easily
the sum of the cross sections of the two reactions.
+ 2..
extended to situations which involve more than two types of collisions.

In general, if a nucleus has neutron-reaction cross sections for n different
types of collisions, then the sum 2,
2( =
2, +
22
2B
(2.45)*

gives the total number of collisions of all n types which the neutron
Any partial sum, e.g., 2i + 2S,
experiences per unit path of travel.
gives the number of collisions of types 1 and 2 per unit path of travel.
Macroscopic cross sections, then, are additive. This operation, how
ever, offers some conceptual difficulties in connection with the absorption
process. The idea of a single neutron being absorbed more than once
appears to be in contradiction to the definition of the absorption process.
In dealing
And, indeed it is. But, this is not what Eq. (2.45) implies.
with the absorption process (and with any of the others, for that matter),
it is more meaningful to talk about the mean free path to absorption
A physical interpretation
Xa, which is the reciprocal of 2 [cf. Eq. (2.21)].
of (2.45) is better made, then, in terms of the corresponding mean free
Thus, we write
paths.
X*
y'
-j
-+-++Xi
X2
(2.46)
The total mean free path is the average distance along its path that a
neutron travels before any one of the n events occurs.
Note that, if the
mean free path to any one of these individual events approaches zero
(i.e., the probability of the event occurring increases to unity), the total
mean free path also goes to zero.
The functions F0(x) and/(x) derived above may be used to describe the
neutron population and distribution of collisions in the example con
If n(0) neutrons leave the origin of the x axis, moving in the
sidered.
positive x direction, then the number of neutrons which reach x without
collision is [cf. Eq. (2.2)]
n(x) = n(0)e~*-
(2.47)
which follows directly from the relation (2.42).

We can also compute the
total number of first scattering and absorption collisions A(x) that occur
REACTOR ANALYSIS
44
[CHAP.
Thus
in unit distance around x by applying (2.44).

Ao(x) = n(0)2e-z>x
(2.48)
The number of the individual events which occur at

Scatterings:
As(x) = n(0)2.erx*
Absorptions:
Aa(x) = n(0)Se_I!r
a;
is
(2.4ft)
Note that the neutron population and the density of events all decrease
exponentially from the origin.
It has become the practice in dealing with combined events to define
The
a suitable set of symbols for some of the important reactions.
standard practice is to call the total cross section 2, of a given material the
Thus
sum of the cross sections of all the possible types.
2, = 2 + 2. + 2in
(2.50)*
In
process is unimportant,
situations wherein the inelastic-scattering
As previously remarked
the usual definition is to write 2< = 2 + 2,.
is
cross
section
given
by
[Eqs. (2.40)], the absorption

2 = 2T
+ 2/
(2.51)*
In
the case of nonfuel materials (2/ = 0), the absorption cross section
is just the radiative-capture cross section.
We found in Sec. 2.4a that the
c. Measurement of Cross Sections.
macroscopic cross section 2 could be described in terms of the nuclear
and the microscopic cross section a. Thus, given the density

of material and either cross section, the other is immediately known.
However, as yet we have given no explanation of how the number a
With our present imperfect knowledge of
(or 2) is to be obtained.
density
dynamics, it is not possible to calculate from first principles

neutron cross sections with better accuracy than roughly a factor of 2.
Another approach to the calculation of cross sections is the phenomenoFrom a knowledge of experimentally determined values of
logical one.
to compute a for other reactions in the same nucleus.
<r, one attempts
Such calculations are generally of greater accuracy than those based on
nuclear
first principles.
Thus when high accuracy
is required,
we
must rely
principally upon experimental measurements.

Unfortunately, this approach is not free of major obstacles either.
Except in the case of the total cross section [Eq. (2.50)], cross-section
measurements are difficult to perform and require extensive instrumenta
This is especially true in situations wherein one
tion for precision.
attempts to measure the cross sections for the individual processes listed
in Eqs. (2.40).
However, considerable effort has gone into cross-sectionmeasurement programs, and a large body of data is gradually being
accumulated.
sec. 2.4]
PROBABILITY
CROSS SECTIONS
CONCEPTS,
45
Although the basic problems of reactor physics are not generally con
cerned with the analysis of cross-section experiments, there are some
special areas of reactor analysis which involve the same techniques and
analytical models. The analysis of radiation experiments in research
reactors and criticality studies are examples of such problems.
There is
one cross-section experiment, however, which is extremely simple to
study from the analytical point of view and which will serve very well to
demonstrate some of the ideas presented in this section.
This problem
has the additional merit that it gives some further insight into the
The experiment in question is the
physical concept of a cross section.
measurement of the total cross section of a given material.
The essential features of this experiment are as follows: A collimated
beam of monoenergetic neutrons is allowed to strike a slab of the material
of interest.
If n(0) is the number of neutrons crossing a unit area of the
slab surface per unit time at x = 0 (see Fig. 2.6), the number of uncollided
neutrons crossing a unit area per unit time at dis
~~
*
tance x inside the slab is [cf. Eq. (2.47)]

n(x) = n(Q)e-z-z
Suppose that a neutron detector is placed on the

opposite side from the face exposed to the inci
dent beam and that this detector registers a
count only when a neutron reaches it by a path
normal to the slab surface.

are scattered
Thus, neutrons that
about
Fig.
J*
2.6
Collimated neubeam inident
*
inside the slab and even

tually emerge at face x = a traveling in some random direction are not
counted.
It is possible that some of these neutrons might still leave
the surface with a trajectory normal to the surface and at a point directly
opposite the detector (point A in Fig. 2.6) ; however, these are few in
number, and most of them can be eliminated by using very thin slabs
The number of neutrons which survive
(i.e., by requiring that a2( <K 1).
the passage through the slab without being scattered or absorbed is
n(a) = n(0)e-s'a
(2.52)
The numbers n(0) and a are known, and n(a) is measured by the detector.
Thus, one can compute S,
S, =
knowledge
N<rt =
of the nuclear density
- In
a
n{a)
of the slab material
(2.53)
yields
the
total cross section at.

Experiments of this type are most useful for measuring the total
As already noted, cross-section meas
cross section of a given material.
microscopic
REACTOH
46
ANALYSIS
[CHAP.
urements for specific neutron reactions are much more difficult to make
and will not be discussed here.
Extensive compilations of such data
have been made, nevertheless, for many materials of interest to reactor
These
technology and are now available in several standard sources.1
data for the various elements are customarily presented in the form of
Also available in
cross-section curves as a function of neutron energy.
The thermalmany sources are the thermal-neutron cross sections.
neutron cross sections refer to the cross section for neutron-nucleus inter
action in a system wherein the kinetic energy of the neutron corresponds
to the equilibrium temperature of the material.
This reference tempera
ture is customarily taken as 20C (0.025 ev).
Table
Material
Density,
g/cm*
H
D
<r,
0.330
0 .46 mb
Li
0.534
1.85
Be
Thermal
2.2
B
C
N
71
10
mb
755
1.60
3.20mb
1 .88
<0.2
Na
Al
0.971
2.70
0.87
7.86
mb
38.0
Neutron Cross Sections"

Material
Mo
7.0
Cd
1.4
In
Xe1"
7.0
4.0
4.8
10
4.2
0.505
0.230
4.0
1 .97
1 .5
1.4
The corresponding neutron
Sm>"
Th
U"
U"
U"8
I'll"9
Pu"1
H20
Fe
Co
2.53
37.8
11 .0
8.9
7.0
D,0
Ni
8.9
4.6
17.5
BeO
Density
10.2
8.65
7.28
a.
2.5
2550
2.2
190
2 .7
7.7
11.3
18.7
18.7
18.7
10
50,000
7
12.6
585
687
2.75
1065
533
10
8.2
9.6
580
750
1100
1.0
1.10
3.01
0.66
0 .93 mb
9
mb
108
14.3
10.0
All
cross sections are given in barns except where noted as millibarns (mb).
Refers to thermal energy (20C or 2,200 m/sec).
The thermal-absorption,
elastic-scattering, and
speed is 2,200 m/sec.
fission cross sections for some common reactor materials are listed in
Table 2.2. The density of the material at 20C is also included.
The
millibarn notation (mb) used in the table is defined
as one-thousandth
of a barn.
d. Calculation of Macroscopic Cross Sections.
In order to determine
the various neutron-nucleus reaction rates, we require the appropriate
Since these probabilities are related to the
probabilities for collision.
collision cross sections 2 (or the mean free path X), the reaction rates
may be computed from a knowledge of the microscopic
cross sections a
for example, Donald J. Hughes and John A. Harvey, in U.S. Atomic Energy
Commission, "Neutron Cross Sections," McGraw-Hill Book Company, Inc., New
1 See,
York,
1955.
sec. 2.4]
and
PROBABILITY
the nuclear
relation :
densities
nuclei
vol
CONCEPTS,
N.
CROSS SECTIONS
The densities
47
are obtained
from
the
/ mass\ / mole\ / nuclei\

\ vol / \mass / \ mole /
By the usual definitions:

p =
= mass/ volume
= nuclei/mole = Avogadro's number (0.6023
= mass/mole = atomic (or molecular) weight
density
We use these definitions
in the expression for
N
The macroscopic cross section may
1024)
N and obtain
&
be
written in terms of
(2.54)*
these
quantities
2 =

Thus, 2
is immediately
(2.55)
determined once the density of the material
is
specified.
The magnitudes
of the numbers involved in these relations may be

demonstrated by means of a sample calculation.
As an example, con
sider BeO.
The pertinent data for computing the macroscopic cross
sections are: p = 3.01 g/cm3, A = 25, a = 0.6023 X 1024, a, = 10 barns,
= 9 mb.
These yield N = 7.26 X 10" nuclei/cm3, 2. = 0.726 cm"1,
<ra
and 2 = 0.000654 cm"1,
e. Mixtures of Nuclear Species.
The usual reactor configuration con
sists of various homogeneous regions composed of mixtures of different
nuclear species.
Thus, the general problem is to compute the reaction
This
rates for media which contain more than one type of nucleus.
calculation is easily carried out when the media are homogeneous.
The
methods derived in the preceding portions of this section for mononuclear

media apply directly to the uniform mixture.
Suppose that the medium through which the neutron is moving con
sists of a mixture of nuclear species such that
number of nuclei of the tth species present per unit volume

i = 1, 2, . . . , k
of mixture
The frequency function for collision is still of the form given in Eq. (2.17),
but the cross section 2 is now to be computed from the properties and
densities of all nuclei present.
Consider A; fictitious media, each of which
consists solely of one of the nuclear species of the final mixture.
Let the
nuclear density of the medium composed of the ith nuclear type be the
Then, for each of these
same as the density of that type in the mixture.
ANALYSIS
REACTOR
48
[CHAP.
media, we can define a distribution function

.
_
=
probability that
such that
with
a neutron collides
a nucleus of type
before reaching x
e~sli)x
(2.57)
2W> is to be computed
from the properties of this ith
Now
medium, using the known <r(i) and the density JV. from Eq. (2.54).
imagine these various media to be combined to give the resultant mix
ture.
Furthermore, let us assume that collisions with the various nuclear
The parameter
is
some function
distance x without
with each nuclear type.
must avoid collision
Fm(x), where
2<
[1
k
- Fk(x)]
(2.58)
e-zt,, = e-2""'1
2'"'
- F1(x))[l - F,(x)]
f]
(2.59)
i-i
The last equality follows from our usual definition of a. The frequency
function fm(x) for collision in the medium may be computed from the
distribution function Fm(x) by means of the result (2.58).
=
F'm(x)
= 2'-'e-sl"1*
Cross
Sections
of Compounds
AV(1>
The mean free path to first collision in the mixture
(2.60)
is
/(*)
f.
AV
and Nuclei in Bound States.
(2.61)
The
of the cross sections for molecules may be obtained from

knowledge of the microscopic cross sections of the individual atoms in the
structure.
a straightforward one for capture-type
The procedure
calculation
is
is
reactions; the capture cross section of the compound

given by the sum
of the cross sections for all the atoms present.
Scattering collisions,
however, are much more complicated phenomena.
scattering reaction
between a neutron and a nucleus bound to
molecular structure differs
greatly from the reaction between the two corresponding particles in the
free state.
The chemical bonds between the struck nucleus and the rest
a

with
to survive
Fm(x) = probability of survival to x in the mixture
k [1
Thus, the joint probability
neutron
is
If
it
is
given by
Ft(x).
collision in the mixture,
is,
species are independent events, that

the presence of other nuclear
species in no manner affects the interaction between
given nuclear
neutron.
Then the probability in the mixture that neutron
type and
reach distance
without suffering
collision with a nucleus of type
sec.
2.4]
PROBABILITY
CONCEPTS,
CROSS SECTIONS
49
of the molecule give it an increased rigidity; consequently, the neutron

responds as though it had collided with a much larger mass than that
of the target nucleus.
The nature of the collision
is determined
in main part by the
speed of
ev, the target nucleus
responds as a free particle because at these energies the chemical binding
forces are negligible in comparison to the collision forces, and so the bonds
are disrupted.
At lower neutron energies, internal vibrational states
the neutron.
At neutron energies in
excess
of
of the molecule can be excited, and the neutron will experience inelastic
collision with the molecule.
However, there exists some neutron speed
below which the lowest vibrational mode of the molecule cannot be
excited, and the target nucleus responds to the collision with effectively
the mass of the molecule.
In addition to these considerations, the

scattering from molecules is further complicated by interference effects

due to the presence of multiple scattering centers.
Some attempts have
been made to evaluate these various factors, and we will consider them
later in connection with the study of the scattering process proper.
PROBLEMS
2.1
Monoenergetic neutrons arc being emitted isotropically from a point source
in an infinite medium at a rate qo neutrons/sec.
If the nuclei of the medium have an
absorption cross section <ra for neutrons of the given energy, a negligible scattering
cross section, and a nuclear density Ar, find the number of neutron absorptions occur
ring per cubic centimeter per second at each point of the infinite medium.
2.2
Suppose that in the same medium as in Prob. 2.1 neutrons arc emitted per
pendicularly from a line source at a rate qt neutrons/sec per centimeter length of line.
Find the number of absorptions per cubic centimeter per second at each point in the
infinite medium.
2.3
Suppose that the neutron source in the infinite medium of Prob. 2.1 is a per
pendicular-emitting plane with a strength qt neutrons/cm* of source area per second.
Find the number of absorptions per cubic centimeter per second for all points in the
medium.
2.4
Suppose that the neutron source in the infinite medium of Prob. 2.1 is a plane
which emits neutrons such that from each side of a small area dS, q3 dS neutrons are
emitted per second with an angular distribution given by
q(0) de dS = 2q, dS sin 0 cos 0 d0
where 0 is measured from the normal to the plane (0 having the range 0 *-/2 for
each side). Show that the number of absorptions per cubic centimeter per second for
all points in the medium is given by
CHAPTER
MULTIPLICATION
CONSTANT AND NEUTRON FLUX
3.1
Collision Density
The importance of the nuclear reaction rates (the number of neutronnucleus collisions per unit volume per unit time) to the problems of
reactor physics was discussed in the preceding chapters.
The purpose

of the present chapter is to establish the relationship between the various

reaction rates and the neutron population and the nuclear characteristics
of the reactor materials.
The ideas developed in this connection will
in
be used to examine,
greater detail than was done in Chap. 1, the
for a self-sustaining chain reaction.

The expected number of collisions per unit volume per unit time can
requirements
be expressed in terms of the neutron density and the appropriate cross

In order to introduce the
section for the type of reaction in question.
necessary concepts for this calculation, let us examine the following
Consider a medium which is supporting a neutron
simplified situation.
kinetic energy. Thus, all the neutrons in this system

have the same speed, but their directions of motion are randomly dis
tributed. We should like to compute the average number of neutronnucleus collisions which occur in time dt in the element of volume dt
gas of a specific
some space point1 r in this medium.

expressed in terms of the following factors:
around
Expected
number of
collisions
in time
dt in
volume dt
be
expected
number of
distance
number of
collisions
traveled
per neutron
by neutron
in time dt
neutrons
in volume
per unit
distance traveled
But, by our previous definition
This quantity may
[cf.
(3.1)
dt
Eq. (2.22)]
expected number of collisions
unit path of travel
by neutrons of
speed
per
(3.2)
The symbol t denotes a set of three coordinates to locate a point in space.

50
SEC.
If
MULTIPLICATION
3.2]
CONSTANT
AND
NEUTRON FLUX
51
we define, also,
,
n(X'V't}
_~
F(T,V,t)
number of neutrons of speed v per unit volume around r

at time t
expected number of collisions by neutrons of speed v per
unit volume around r and per unit time at time t
(3.3)
then, (3.1) may be written

F(r,v,t) di dt =
F(r,v,t)
or
2(i>) v dt
= 2(t>)
n(r,v,t) dt
n(r,t;,<)
(3.4)
The quantity F(r,v,t) is called the collision density for the nuclear reac
tion described by the cross section 2(y). No subscript has been assigned
to 2 so as to keep the derivation general.
Equation (3.4) shows clearly
the dependence of the reaction rate (or collision density) on the neutron
This relationship was stated previously in the intro
density n(r,v,t).
duction to Chap. 1.
In dealing with specific types of reactions, we add the appropriate

subscripts to the cross section according to the notation of (2.40).
Thus,

for example,
24(y)
2<,(i;) v
n(i,v,t)
total number of collisions of all types which occur

= between nuclei and neutrons of speed v per unit
volume around r per unit time at t
n(i,v,t)
which occur between

number of absorptions
= nuclei and neutrons of speed v per unit volume
around r per unit time at t
(3.5)
It is important to remember that all these expressions for the

various collision densities are statistical quantities and should be taken
As with all statistical results,
to give the expected number of collisions.
and so on.
the actual number of collisions will vary around the expected number
with a variation which will depend upon the neutron and nuclear densities.
The macroscopic cross sections 2(y) may be written in terms of the
nuclear densities in the usual way [cf. Eq. (2.39)]. In general, these
densities will be spatially variant functions and so should be properly
written in the form N(r). The resulting equation for the collision density
is then
F(T,v,t) = <r(y) N(t) v n(i,v,t)
(3.6)
Note that we have retained the specification of the neutron speed in

defining the microscopic cross section o(v). This is in keeping with our
recognition of the fact that the neutron cross sections are in general a
function of the neutron speed (cf. Sec. 2.4a).
3.2
a. Infinite-medium
Model.
Neutron Balance
The actual
reactor may be described as a continuous
situation in an operating
cycle of neutron production
REACTOR ANALYSIS
52
[CHAP.
The principal nuclear events which constitute this

and destruction.
cycle are the fission, scattering, and nonproductive capture reactions
which take place throughout the reactor.
At virtually every point
core, high-energy neutrons are continually appearing
Likewise, neutrons of all energies are being continually
removed from the general population at all points of the reactor system
because of capture reactions.
Neutrons are also continually lost by the
of the reactor
from fission.
net flow outward
through
the boundaries
interval between birth and death,
of the system.
a neutron
In the time
wanders about the reactor
collisions which cause it to deposit by

decrements its initial kinetic energy from fission at the points of collision.
If we were to examine a small volume of the reactor, we would find
and experiences many scattering
that the "microscopic" picture would be much the same as the over-all
picture of the reactor system, with neutrons of all energies continually
Thus, at any given time
streaming in and out of the little volume.
instant, there would be present in this volume neutrons of all energies,
corresponding to different stages in their life history. Some of these

neutrons would subsequently be absorbed in the volume element.

Others
might be scattered about for a while before they finally pass out through
the boundaries.
The neutrons which escape would be leaving with
energies less than that with which they entered.1 If the volume element
we selected were in the core, the situation would be further complicated
As in the case of neu
by the continual appearance of fission neutrons.
trons which arrive in this volume by passing through its boundaries
(transport), the fission neutrons might disappear by absorption while

still in the volume, or they might be scattered about before leaving.
In
any event, we realize that the physical situation in any part of the reactor
is extremely complex.
For the present, we will not attempt a detailed description of all the
processes that occur in an actual system, but instead will limit our study
of the reactor system to a considerably simplified model.
The model
admittedly an unrealistic one; however, it will serve well to

display some of the basic features of all reactors. The reactor model we
propose has the following characteristics:
we select is
(1)
(2)
(3)
(4)
The reactor medium contains fissionable material and is

infinite in extent and homogeneous and isotropic in composi
tion
All fission neutrons appear at the same
speed v
(3.7)
Neutrons do not lose speed (energy) in scattering collisions

The reactor material has cross sections for only absorption
and elastic-scattering
1 These
reactions
remarks do not apply to thermal neutrons, as will be explained later, but

only to fast neutrons.
SEC.
3.2]
MULTIPLICATION
CONSTANT
AND
NEUTRON FLUX
53
The first of these assumptions
is imposed in order to remove the spatial

In an infinite homogeneous medium,
dependence of the neutron density.
no neutrons can be lost by leakage (escape) , and all regions of this medium
are identical.
Thus, there are no special space points in this system (all
points are alike), and consequently there can be no net change in the
We recognize
neutron population in going from one point to another.
then that the neutron-density function n will not include spatial coordi
nates in its argument [cf. Eq. (3.3)]. Assumption (2) further restricts

the system to neutrons of a fixed speed.

This is known as the oneThus, in our definition of the neutron density n, we can
velocity model.
omit the speed variable v, it being understood that all the neutrons in
Although we specified a similar
this medium have the same speed.
in
of
situation
the discussion
Sec. 3.1, it was convenient to retain the
independent variable v in that instance in order to emphasize the func
Assumption (3) is intro
tional dependence of the quantities involved.
duced so as to simplify the problem and allow attention to be focused on
other aspects of the physics.
The energy changes resulting from scatter
will
be discussed in detail in Chap. 4.
ing collisions
Assumption (4)
is again a simplification, but it does not impose any great limitation on the
As in the case of the neutron density n, the
usefulness of the results.
cross sections will refer to neutrons of speed v, but the symbols will not
include the argument v for the reason given above.
b. Neutron Density in Infinite Medium.
Even though
the system
described above is a highly fictitious one, we can nevertheless acquire
from it a considerable insight into the nature of a chain-reacting complex.
We approach this objective by asking the fundamental question of all
reactor problems: What is the distribution of neutrons in the system?
Because of the nature of the limitations we have imposed, this distribu
tion will be expressed in terms of the only remaining variable, the time t.
Thus any balance condition we write for the neutron population must
On the basis of the assumptions (3.7),
be in terms of its time dependence.
all the materials are uniformly distributed throughout the space.
Fur
thermore, the only sources of neutrons are the fission reactions, and the
No neutrons can be lost by escape
only sinks, the absorption reactions.
of
this
boundaries
reactor.
The neutron-balance condition
through the
for this system must
be expressed then in terms of the time rate of change
of these competing
is as follows:
processes.
Number of
neutrons gained
per unit volume
|_per unit time
The verbal statement of this condition
number of
number of
neutrons produced
neutrons lost by
from fission per
absorption
unit volume per

unit time
unit volume per

unit time
per
REACTOR ANALYSIS
54
[chap.
This statement may
be written in terms of the concepts developed in

n(t) denotes the number of neutrons per unit volume at
time t, then by introducing the appropriate collision densities for absorp
tion and fission [see Eq. (3.5)], we have
Sec.
3.1.
If
T n{t)
v2fvn(t)
- 2avn(t)
(3.8)
where 2/ = NFaf = macroscopic-fission cross section of fuel

cross sec
20 = X'/' + 2j,0) = total macroscopic-absorption
tion of reactor material
and
Nf
2
No
2^0)
= neutrons per fission

= fuel nuclei per unit volume
= nonfuel nuclei per unit volume
= macroscopic-absorption
cross section of fuel
= macroscopic-absorption
cross section of
(3.10)
all nonfuel
materials
JSVi1'
2^2)
2^*"

sa>
+"
2 =
_|_
In general, if there are M such materials

NttvM
(3.11)
Eq.
The symbols JVi, Nt, . . . are the

(2.59).
nuclear densities of the nonfuel components and the <x(*\ a, . . . their
cross sections.
respective microscopic-absorption
Two points are to be noted from Eq. (3.8) : (1) No mention is made
of loss or gain by migration across the boundaries of the unit volume.
In our idealized model, no such losses or gains occur on the average
since each neutron lost by wandering out is immediately replaced by a
which
follows
from
neutron gained from wandering in [cf. assumption (1), Eq. (3.7)].

(2)
No mention is made of the scattering cross section 2,; these reactions
have no effect on the neutron population in the present model.
Both
of these facts result from the assumption of an infinite medium and the
In the more general case of a
consequent uniform neutron density.
finite reactor, the scattering cross section will affect the neutron balance,
and migration (or transport) losses and gains must be considered.
Suppose now we ask for the physical condition which must hold if a
neutron density, once set up in this infinite medium, is to persist for all
time, i.e., we ask that
(3.12)
It
follows from Eq. (3.8) that

(3.13)*
SEC.
3.3]
MULTIPLICATION
CONSTANT
AND
NEUTRON FLUX
55
Equation (3.13) is one of a class

of conditions called the criticality condition; a reactor which satisfies such a
condition is said to be critical.
If condition (3.13) is not satisfied for the infinite reactor, then (3.12)
As an
does not apply, and the neutron density must change in time.
example, suppose that in some way a uniform density no is established
= 0).
(so long as we have not selected n
until time
then for all t >
in the infinite medium and maintained
t 0.
At
= 0
all
0, Eq. (3.8) applies.

nonfission sources are removed;
The solution of this equation gives the neutron density n(t) for all subse
We write (3.8) in the form
quent time.
dt
n(t) = (vZ,
- Xa)vn(t)
The solution to this equation

n(t) = noe''2'-2-"'
(3.14)
is
(3.15)*

where we have applied the initial condi

tion n(0) = n0. Note that if v2f = 20,
Fio.
3. 1
Time dependence of neu
tron density in infinite medium.
n(t) = n0 for all t > 0; this is the condi

tion we examined above. But, if v2f > 2, the neutron population rises
exponentially with time, and if v~Lt < 2a, it decays exponentially.
These
three possible situations are shown graphically in Fig. 3.1.
3.3
Infinite-medium Criticality Problem
a. Neutron Chain and Multiplication Constant.

The criterion for a
self-sustaining chain reaction in a system containing fissionable material
is that, on the average, the number of neutrons produced per unit volume
per unit time be equal to the number of neutrons lost per unit volume
From the viewpoint of the
per unit time [cf. Eqs. (3.12) and (3.13)].
behavior of the neutrons, the cycle may be imagined to be as follows:
neutrons are produced per unit volume per unit time.

If these n0
neutrons are produced as the result of fission reactions, they will, in
general, have kinetic energies of about 2 Mev.
Each of these n0 neutrons
n0
moves through the system and makes collisions with the various species
of nuclei present; the results of these collisions will be: (1) change in
direction of motion of the neutron, (2) loss in kinetic energy of the neu
tron, (3) disappearance of the neutron by absorption in the nucleus, and

(4) loss of the neutron by passage from a final scattering collision across
the boundaries of the system.
Each of the original n0 neutrons will have a different history, but, on
the average, a certain fraction of them will disappear as a result of
processes (3) and (4). Only those absorbed in nuclei have a chance to
produce additional neutrons; a certain fraction of the absorbing nuclei
REACTOR ANALYSIS
56
[CHAP.
will be those of the moderator, the coolant, and the structural materials,
all of which have cross sections for nonproductive capture of neutrons.
Only the fraction of neutrons absorbed in the fissionable fuel can be
Furthermore, not all captures in fissionable materials result
productive.
in fission and, hence, more neutrons, because each fissionable element has
also a cross section for radiative capture.
In any event, a certain fraction
of the no will cause fission, and these fissions in turn will yield, say,
If the reactor system is self-sustaining, the ni neutrons
rai neutrons.
of the second generation will be identical in number with the no neutrons
of the first generation; thus, the population is maintained.
The ratio of the neutron population in two successive generations,
such as n0 and n\, is a useful number for describing the characteristics
This ratio is called the multiplication constant
of the chain reaction.
of the reactor and is formally defined as
,
ni
n0
ratio of number of neutrons of any

number in preceding generation
As noted above, the system

one generation to
(3.16)*
is at steady state when
fc
1.
In this
case,
itself, and the neutron population remains

generations of neutrons increase in
the
subsequent
1,
population; if fc < 1, the subsequent generations decrease in population.
The case fc = 1 corresponds exactly to the condition given by Eq. (3.13),
each generation
fixed.
If fc >
reproduces
This is easily demonstrated for the infinitei.e., the criticality condition.

medium reactor model defined by the assumptions (3.7).
In our treatment of the infinite homogeneous reactor model, we argued
that the neutron density was
spatially invariant function.
Thus, if
in any one volume element

of this infinite system, these conditions will apply to all such elements
Let us write, then, the quantities involved in our
throughout the space.
definition of the multiplication constant in terms of the nuclear events
we determine the neutron-balance
conditions
which occur in a unit volume of the reactor.

write Eq. (3.16) in the form
^
It
is convenient
here to
number of neutrons produced per unit volume per unit time

number of neutrons introduced per unit volume per unit time
(3.17)
According to the definition of the collision density, the neutron production

These neutrons are the
per unit volume per unit time is given by k2/tm>.
generation, of which there must have been
That 2anv is
exactly 20nv produced per unit volume per unit time.
indeed the number of neutrons introduced follows from the fact that all
Although very few
neutrons must eventually be absorbed somewhere.
offspring
of the previous
of the neutrons introduced into the unit volume are actually absorbed
there, every one of the neutrons which leave this volume element to be
SEC.
MULTIPLICATION
3.3]
NEUTRON FLUX
AND
CONSTANT
57
absorbed elsewhere is replaced by one which arrives from another region

It follows, then, that
of the reactor.
- fS -
(3I8)*
is
is,
Thus, the multiplication constant for the infinite-medium reactor is given

When k = 1, we have a critical system, and the
by the ratio v2//2a.
Equation (3.18)
neutron population maintains itself.
therefore, con
sistent with (3.13).
Equation (3.18)
frequently written in an alternative form which
better shows the essential features of the reactor.
Let us write the total
Then
absorption cross section 2 according to the definition (3.9).
"Z/
+ 2!,'
by the absorption cross sec
multiply numerator and denominator

of
tion
the fuel 2j,'',) we obtain
= v2f
2<"
2/>
2/ + 2<f
vNF<r,
Ntfn 2/)
20>
is
77
t\
is
is
is
?;
t)
is
The ratio (vaf/aa)F, which depends only on the nuclear properties of

the fissionable material,
and
usually denoted by the symbol
to
some extent a measure of the value of fuel in
reactor.
Of course, the
value of
varies with the relative speed (energy) of the neutron, but
for many reactors there
one neutron speed which
of predominant
importance (cf. Sec. 1.2a), and the value of at that speed
an appropri
ate criterion for the value of a fuel.
The quantity may also be defined
in terms of the ratio a defined previously [cf. Eq. (1.25)].
In terms of the
fuel cross sections,
^_
(3.20)
follows that
*
and
it
a =
ff7T^>
The quantity
if".
(3"21)*
is
is
it
is
is
usually denoted by the symbol

Zi"/[2^ + 2]
called the thermal utilization;'1
dependent upon the concentra
tion of the fuel and
measure of the competition for neutrons between
and
is
if
in the reactor.
Note that
If this notation
used in
1.
No
nuclei
0,
if
>
0;
likewise,
0,
and the nonfuel
the fuel nuclei
r,f
where
(3.19), the multiplication constant may be written

(3.22)*
^^
(3-23)*
This expression was first used in conjunction with "thermal reactors";

cross sections would refer directly to the property of thermal neutrons.
1

we
If
thus, the
REACTOR ANALYSIS
58
[CHAP.
It
was remarked in Sec. 3.26 that the scattering cross section did not
play a role in the analysis of the infinite-medium model, and indeed
This is entirely a consequence
the result (3.22) is independent of 2,.
of the assumptions of an infinite medium and no change of speed in
scattering.
For more general finite multivelocity reactors, other expressions for

the multiplication constant will be obtained; these must take into account
loss by escape from the system and absorption and production of neu
trons over the entire range of energies.
b. Critical Fuel Concentration.
The criticality equation1 (3.13) gives
the condition which must be satisfied if an infinite homogeneous reactor
is to operate at steady state, i.e., be self-sustaining.
This equation is a
condition on the material composition
If
the nonfuel
coolant, etc.) are
of the reactor.
of the reactor (moderator, structure,

specified, Eq. (3.13) defines the critical fuel concentration for that system.
We will show later that critical equations in general also involve the geo
metric proportions of the reactor.
For the infinite-medium problem,
the
factor2
in
critical
equation which depends on the reactor
the
however,
components
1, and the resulting expression depends only on

involving the composition.
The infinite-medium one-velocity model has some useful applications
in reactor analysis.
One application is the estimation (in first approxi
mation) of the critical fuel concentration of large, highly thermal reactor
A sample calculation using this model is presented in Sec. 3.4.
systems.
In the present section, we derive the basic relations for determining the
For simplicity in notation, we assume that the
critical concentration.
reactor core contains only nuclear fuel and moderator and that these
If the fuel and
components are intimately and homogeneously mixed.
moderator are not homogeneously mixed, and this will be the case in
many reactor designs, one can still introduce a suitable homogenization.
The usual procedure is to assume that the various reactor materials,
which in reality appear as distinct and separate regions, are mixed
Thus, if
uniformly and spread throughout the entire reactor volume.
the region of interest is the core, then one assumes that the constituents
Consequently, each of
are spread uniformly through the core volume.
The
the constituent materials of the core appears at a reduced density.
the
ratio
of
volume
of
by
is
the
the
actual
given
density-reduction factor
substance (as it is to be found in the reactor) to the total volume of
therefore, maintained in the
the core.
The total volume of the core
similar procedure may be applied
homogenized version of the system.
These will be
to the other regions of the reactor, such as the reflector.
dimensions is identically
is,
1.
See also Eq. (3.22) with

The neutron nonleakage probability.
discussed later.
1

quantities
MULTIPLICATION
SEC. 3.3]
AND
CONSTANT
NEUTRON FLUX
59
We compute the critical fuel concentration NF from Eq. (3.22). If

the symbol M denotes the moderator components, then the appropriate
form of (3.22) for a homogenized fuel-moderator system is
The solution for Nr
is
-*"
nJN+n^
(3-24)
(3-25)
This equation gives
one relation between the two unknowns N.v and Nr.

may be obtained by requiring that the total volume
of the system be conserved. This condition may be written
second relation
-I"
(3.26)
where the volume fractions are given by

Vi
and
V{ =
actual volume
reactor
-y
of component
(3-27)
t as
it is to appear in
total volume of system under consideration
(3.28)
Equation (3.26) may be stated in terms of the nuclear concentrations.

For this purpose, we use (2.54) in the form
=
Ni
Pa
~
(3.29)*
is the density of component i after it has been dis

persed uniformly through the region V. According to our previous
definition, this is simply the normal density p< of material i times its volume
In this expression, P,
fraction; thus,
Pi
The normal density
temperature
p<
= PiVi
(3.30)*
is the actual density
and pressure of interest.

Vi
It
of material
follows that
at the reactor
(3.31)
The substitution of this relation into (3.26) yields
pf
pm
This is the second equation between NM and NF.
The simultaneous.
REACTOR ANALYSIS
60
[CHAP.
solution with (3.25) gives
pmQ
Am
[(ij
s
A*
m =
- l)m/sp
1]
(3.33)
*P
l)m/sp +
o-,(A"
Pa/
,(')
(3.34)
PF
Application to Safety Storage Calculation

Chemical-processing Plant
of
a
3.4
Ar[(v ~ l)m/P +
1.
N*
where
PfGL
(ij
Nr
In
is,
is
the present section, we carry out sample calculation for the safety
The calculation
chemical-processing plant.
storage requirements in
based on the infinite-medium one-velocity model.
Our consideration of nuclear reactors, thus far, has included only the
of course, only
principal features of the reactor proper. The reactor
and the
operation
The
complex.
one element of an extensive engineering
of this device require many additional component systems.
*
COOLING
REACTOR
CHEMICAL
PROCESSING
DILUTION
To reactor -
Fig.
3.2
Chemical processing
New fuel
elements
cycle.
is
determined by the
Precisely what supporting components are required
Reactors designed for indus
type and the design purpose of the reactor.
trial or military application are primarily used for the production of power
These applications are frequently made on
or fissionable materials.
is
Under
large scale; consequently, the supporting facilities are also large.
such circumstances, the preparation and supply of the reactor fuel ele
This
particularly true
mammoth undertaking.
ments can become
in the case of reactors which utilize fissionable materials in solid form.
Such large quantities of solid fuel elements can be handled only in large
chemical-processing
plants.
is
problem
When dealing with large amounts of fissionable material,
which inevitably arises
the criticality consideration of the system.
important to the chemical-processing cycle for two rea
The problem
is
sons: (1) the safe storage of the fissionable material at each stage of the
process, and (2) the optimization of the batch size handled by the process.
Both of these factors can enter into the design considerations of the
The principal
number of intermediate stages.
process at any one of
stages in the chemical processing of solid fuel elements are shown in the
a

Spent fuel elements
maintenance
SEC. 3.4]
MULTIPLICATION
CONSTANT
AND
NEUTRON FLUX
61
block diagram of Fig. 3.2. The first stage is denoted cooling and cus
tomarily involves the storage of the spent fuel elements in a suitable
vault or canal wherein the radioactivity of the fission fragments present
in the elements is allowed to decay to tolerable limits. The second stage,
here called dilution, is the preparation of the fuel elements for processing
(e.g., dissolving by some suitable acid solution). The final stage involves

Fuel cluster
Core
A-A
Spent slugs
Fio.
3.3
Core and fuel-element configuration.
the reconcentration of the unburned fissionable material present in the

element and the preparation of new elements.
At each stage of the process, the geometric configuration of the batch
must be safe.
We will illustrate the treatment of the criticality problem
in this connection by carrying out a sample calculation for one stage
of the process.
For this calculation, we select the cooling stage and apply
the one-velocity infinite-medium model to determine the criticality of the
proposed cooling batch size for a representative reactor configuration.
Let us assume that the reactor in question is a heterogeneous system
The
of slightly enriched natural uranium slugs imbedded in moderator.
slugs are cylindrical in shape and are loaded into the core through vertical
REACTOR ANALYSIS
62
channels so as to form clusters (see Fig. 3.3).

elements are given below :
[CHAP.
The design details of these
Slug length
Slug diameter
Slugs per fuel-element cluster
Enrichment (% U")
Total number of slugs
12
in.
0 . 4 in.
500
3
150,000
It is proposed that the reactor be loaded by adding new slugs (a complete

vertical layer at a time) at the top and withdrawing spent slugs (a layer
This operation would yield 25,000 slugs
at a time) from the bottom.
per layer, these to be loaded into storage crates in batches of 500 for
The storage crate is a rectangular box 9.2 by
cooling in a water canal.
8.8 by 12 in., with ^-in. -thick steel-plate sides (no top or bottom).
This
We should like to know if the loaded crate is
crate can hold 506 slugs.
critical when immersed in water.
We compute the multiplication constant for this system by assuming
Thus, we require the macro
that it is a homogenized infinite medium.

scopic cross sections of a system consisting of enriched slugs, steel, and

water.
The volume fractions of these constituents may be computed
from the cross-sectional area of the crate:
area of crate occupied by 506 slugs = 410 cm2
Cross-sectional area of crate occupied by steel = 56.4 cm*
Cross-sectional area of crate occupied by water = 55.7 cm*
Cross-sectional
The densities of the various materials in the homogenized system are com
puted from Eq. (3.30), using pF = 18.7, p.*., = 7.8, and pHl0 = 1 g/cm8
(cf. Table 2.2).
Pf
pfVf
/V,
(18.7)(H#) = 14.7 g/cm'

P,o = 0.107
= 0.843
For the microscopic cross sections (at 0.025 ev), we

= barns
<r = 2.75
a<a26)
barns
= 687
2.75
ff(...n
= 2.53
a<a26)
a?
use
= 687
= 0.66
where the superscript (28) refers to U238 and (25) to U236.

We will use
these cross sections directly in our calculations; this is not a correct
procedure, but it will suffice for the present purposes.1 In order to
simplify the calculation of the macroscopic cross sections, we introduce
the following conservative assumptions: (1) Spent fuel elements contain
original concentration of U2". (2) Fission fragments (which are strong
neutron absorbers) in the spent slugs may be ignored.
1 These
data should be corrected for the energy distribution of thermal neutrons

which are not all clustered about 0.025 ev, but are spread out over a wide range (refer
to Chap. 4).
SEC. 3.5]
MULTIPLICATION
CONSTANT
NEUTRON FLUX
AND
63
For the 3 per cent enriched uranium slug, we obtain the absorption
cross sections [cf. Eq. (3.29)]
Nf&m
= (14.7)(0.6023)(687)(0.03)
= (0.843)(0.6023)(2.53)
5o
= (0107)(0.6023)(0.66)
2(26>
/ [cf.
vf
2<,t,el)
is,
And, the multiplication constant
n ocrj
2|H!0>
by (3.22),
= (2.08) (0.860) = 1.79
is
is
It important to note that this result based on a very elementary

model of a chain-reacting system and should not be taken
priori as
representative value for the infinite-medium
multiplication constant.
Some of these limitations were mentioned in the discussion following
Better estimates for the infinite-medium multiplication con
stant than that given by Eq. (3.22) are developed in Chap.
4.
Eq. (3.19).
3.5
Neutron Flux
is
statement of the neutron-balance condition involves the collision

densities of the various possible neutron-nucleus
reactions.
The col
lision densities, in turn, may be described in terms of the neutron densities
and the appropriate cross sections [cf. Eq. (3.5)].
Although the neutron
the fundamental
density n
quantity which describes the neutron
is
related to the neutron
density
through
the definition
(v
is
it
population,
frequently more convenient in reactor calculations to
work with another function, called the neutron flux. The neutron flux
= neutron
speed)
= nv
(3.35)*
the neutron density

function of space, speed, and time,
given as
then the flux
we use the definition (3.3), the
similarly specified; thus,
flux
is
if
is
is
If
4>(T,v,t)
= n(i,v,t)v
(3.36)
Note that the flux

refers to neutrons of a particular speed v. More
correctly, the function n, and therefore
must be given for an interval

(3.23)]
^5
2<28)
?m
cm_,
V(26)
-|-
These give for the thermal utilization

^
^
_^^
^^
(14.7)(0.60BK2.75)(0.97)
S<n, =
REACTOR ANALYSIS
64
[CHAP.
Thus, the formal definition of the density n(T,v,t) is
of speed dv.
n(i,v,t) dv dr =
number of neutrons in volume dr around space

point r at time t which have speeds between v and
v
dv
(3.37)
The corresponding definition of the neutron flux

4>(r,v,t)
dv dr =
is
total flux of neutrons in volume dr around r at

time t which have speeds between v and v + dv
The quantities n(r,v,t) and
(3.38)
(r, v,t),
therefore, are given as functions per

unit speed around v and per unit volume around r. By the expression
"around v," we mean speeds which differ from v ever so slightly so that,
if the deviation is
dv, then as dv > 0, the speeds
in question approach v.
is customary to call
the total track length per unit volume per unit
time.
The basis for this terminology is evident from the definition (3.35).
We write this relation in an expanded form
It

(T,v,t)
dv
di
= vn(i,v,t) dv dr
or
(t,v,t)
distance
number of
traveled
neutrons
per unit
per neutron
per unit
volume per
time
unit
speed
speed
interval J
rvoiume]
total distance (track length) traveled by all

v in
volume dr per unit time
dv dr = neutrons which have speeds in dv around
This definition
(3.39)
(3.40)*
abbreviated to: total track length per unit

Whenever this form is used, the complete state
is frequently
volume per unit speed.
ment (3.40) is implied.

It is sometimes more convenient to describe the neutron population in
terms of the neutron kinetic energy instead of the speed.
In this event,
the neutron density and the flux are defined per unit of energy.
In order
to show the relationship between the speed and the energy definitions,
let us consider the function
n(r, E,t) dE dr =
of neutrons in dr around r at time

which have kinetic energies in dE about E
number
This function gives the neutron density in the energy space E.
(3.41)
It
can
be directly related to the density in the speed space [given by (3.37)]

by selecting the speed interval dv which corresponds to the energy interval
dE. Clearly, if we specify that dv is to correspond to dE, then the num
ber of neutrons per unit volume around r must be the same whether we
describe it in terms of the energy variable
or in terms of the speed
SEC.
MULTIPLICATION
3.5]
variable v. Thus, if
(cf. Fig. 3.4)
AND
CONSTANT
dE
n(i,v,t)
n(r,,<) = n(t,v,t)
or
65
function (3.37), then we require that
we use the
n{r,E,t)
NEUTRON FLUX
dv
(3.42)*
(3.43)
This statement tells us that if the neutron density is known in terms of
E or v the expression in terms of the other is obtained by using the

transformation (3.43). The differential ratio1 dv/dE is computed from
It follows
the relation E = imv2, where m is the mass of the neutron.
that
either
n()
. "00
mv
_ n(V2lA)
= n[v(E)]
VWm
y/2Em
(3
u)

The
where we have omitted the variables r and t for the sake of clarity.
corresponding expression for the flux per unit energy range is obtained
Fig.
3.4
Equivalent
intervals in speed and energy spaces.
a similar calculation.
The fundamental relation between the expres
sion for the flux in speed space and energy space is derived from (3.35)
by
using (3.42)
If
v(E) n(i,E,t) dE
we call <p(r,E,t) the flux per
n(r,v,t) dv
(3.45)
unit energy range, where
v(T,E,t)
then
(r,E,t) dE
v(E) n(T,E,t)
= (t>(r,v,t) dv
(3.46)
(3.47)*
This equation corresponds to the relation (3.42) and states that the total
track length per unit volume per unit time around r at time t is conserved
in going from speed space to energy space.
If we introduce the Jacobian
dv/dE, it is easily shown that (again omitting r and t)
,<A')-4^g^-:m (v^)
\\ m J
It has
become customary to distinguish
cms.
between the flux per unit speed
1 The Jacobian for the transformation v E.

See, for example, I. S. Sokolnikoff
and E. S. Sokolnikoff, "Higher Mathematics for Engineers and Physicists," pp.
183-185, McGraw-Hill Book Company, Inc., New York, 1941.
REACTOR ANALYSIS
66
[CHAP.
and the flux per unit energy solely by the argument following the flux
Thus, if 4> denotes flux and n denotes density, then
symbol.
^(
4>(E)
total track length per unit volume of neutrons

with speeds in unit interval around v
total track length per unit volume of neuv(E) n(E) trons with energies in unit interval around
( \
(3.49)
The concepts introduced

characteristics
above may be extended to describe integral

of the neutron population.
The function (r,E,t), which
gives a detailed description, can be used to construct a quantity which

refers to the "bulk" properties of the system.
Thus, if we wish to
determine the total track length $ of neutrons in a system of volume V
which have energies in dE about E, we perform the following integration :
*(,<)
The total number
31
dE
Jr 4>(i,E,t)
=
(3.51)
Likewise we can compute the total track length

energies in volume V from the relation
(0
(3.50)
of such neutrons is computed from the relation
3l(#,0
di dE
jE *(E,t) dE
fg dE
jy
of neutrons
4>{r,E,t) dr
of all
(3.52)
The application of these functions to the calculation of neutron popula

tions in reactors will be considered later.
The flux concept is useful in describing the collision densities in reactor
The general expression for the collision density was developed
systems.
in Sec. 3.1 with the result (in energy notation)
F(x,E,t)
In
2(E) v(E) n(T,E,t)
(3.53)
terms of the neutron flux, this expression gives
F(t,E,t) = Z(E)
=
4>(t,E,t)
total number of collisions per unit volume around r

and per unit time at time t due to neutrons in unit
energy range around E
(3.54)*
Thus, for absorption collisions
S a\.(E)
(t
<p\ , -, t)
dF>
total number of absorptions per unit volume

arounc^ r an(* Per UIU^ time at time t due to
with energy between E and
neutrons
E + dE
(3.55)
SEC. 3.5]
MULTIPLICATION
CONSTANT
AND
NEUTRON FLUX
67
and so on. This relation should be compared to (3.5). The form (3.54)
for the collision density is especially useful in writing neutron-balance
conditions such as (3.8). In the present notation, this equation takes
the form
I ~at (<)
*2/*(0
- 2.0(0
where we have applied the definition (t) = vn(t) from (3.35).

The neutron flux has a directly useful physical meaning when con
sidering systems in which all the neutrons have the same energy and
move in the same direction, i.e., a beam of monoenergetic neutrons.
Suppose that there are n neutrons in each unit volume of the beam and
Now consider a unit area which is
that v is the speed of the neutrons.
to the beam and compute the number of neutrons passing
through that area per unit time. Clearly, this is exactly the number of
neutrons in a volume whose cross section is unit area and whose length
is numerically the speed v, since all neutrons in this volume will be able
Then, in the
to pass through the unit area in unit time (see Fig. 3.5).
perpendicular
3.5
Collimatcd
beam of neutrons of speed v.

4>,
Fig.
is
of
beam, the number of neutrons crossing unit area in unit time is nv =

the neutron flux.
This
true only in the case
monoenergetic beam.
In the more general case wherein neutrons are moving in all directions,
the neutron flux is not the number of neutrons crossing
unit area in
unit time. For example, in the case of a unit area placed in a gas of
neutrons whose directions of motion are isotropically distributed, the
number of neutrons crossing the unit area in unit time in any one direc
is
is
This point will be demonstrated later.

tion
\4>.
We conclude the present chapter with a few remarks on the units
of measure for neutron flux and collision density.
It
customary
if,
if
to specify neutron densities in neutrons/cm3; thus,

the neutron speed
given in cm/sec, the flux has the units neutrons/cm2/ sec.
Likewise
as
the custom, the neutron microscopic cross sections are given in
square centimeters and the nuclear densities in nuclei/cm3, then the
is
is
are given in cm-1 and the collision densities

macroscopic cross sections
in collisions/cm3/sec.
We can obtain some idea of the magnitudes of
2

Unit area
ANALYSIS
REACTOH
68
[CHAP.
these quantities by considering some representative numbers.

if we use the data of the sample computation in Sec. 2Ad,
Thus,
= 7.26 X 1022 nuclei/cm3

S. = 0.726 cm-1
2 = 0.000654 cm-1
(a typical value),
then for a flux of 1013 neutrons/cm2/sec

which neutrons are absorbed in BeO is
Xa
= 6.54
109
the rate a
absorptions/cm3/sec
These cross sections refer to neutrons of 0.025 ev energy or 2,200 m/sec

The density of these neutrons in the BeO is
speed.
n =
= 4.54
107
neutrons/cm3

PROBLEMS
Consider an infinite reactor composed of a uniform mixture of beryllium and
5.1
The atomic ratio of isotopes in natural uranium is U"8:Um =
natural uranium.
993:7. Assume that the one-velocity model applies, and use the microscopic cross
sections and densities of Table 2.2.
o. Compute the atomic concentration of U*" required to make the reactor critical.
6. Show that, if it is assumed that the average density of the U-Be mixture can be
approximated by the density of pure Be, then, in this case, the error in the Um con
centration is of the order of 1 per cent.
c. Find the mean number of scattering collisions made by a neutron during its
lifetime.
3.2
It is proposed that a reactor of pure UF be constructed, the fluorine to provide
the moderation and the U"' in the uranium to serve as the nuclear fuel. A first
estimate of the critical fuel concentration for such a system can be obtained from the
one-velocity infinite-medium model.
a. If the uranium in the UF8 is composed of U"5 and U"*, determine the atomic
Use the data
ratio of U*" to U"* required to make the infinite medium critical.
t)
2.08
(U")
fo=9mb
for fluorine
The uranium data are given in Table 2.2.

multiplication constant for UF6 with natural
6. What is the infinite-medium
uranium (cf. Prob. 3.1)? Note: A much more realistic calculation would include the
absorption of neutrons during the slowing-down process by the F and U"*. These
nuclei have large 2 for high-energy neutrons, and the omission of this effect here gives
an underestimate of the fuel requirements (cf. Chap. 4).
CHAPTER
SLOWING-DOWN
4.1
PROCESS IN THE INFINITE MEDIUM
Mechanism of Energy Loss by Scattering Collisions
The importance of the neutron-energy spec

Reactions.
the
nuclear
features of a reactor was mentioned in
determining
in
trum
Sec. 1.2. The spectrum itself is determined partly by the nuclear reac
tions which can occur and partly by the geometry and size of the system.
In this chapter, attention is limited to the influence of the nuclear

a. Scattering
phenomena.
All three types of reactions enter into the determination of the spec
trum. The fission reaction determines, of course, the average energy
The scattering and absorption
of the neutrons borne into the reactor.
reactions together determine the detailed shape of the spectrum.
The
largest influence is due to the scattering interactions which establish the
The effects of the absorptions are some
over-all shape of the spectrum.
what secondary, and in most reactor situations these tend primarily to
Thus in the present

distort the spectrum produced by the scatterings.
treatment of the subject we begin with a study of the spectrum arising
from scattering phenomena alone and then extend and modify this
analysis to include the deviations which arise when one includes also the
effects of the absorptions.
Both elastic- and inelastic-scattering reactions have a significant effect
upon the neutron-energy distribution since each process is an important
mechanism of energy degradation over some portion of the energy scale.
The inelastic collisions are important when the neutron kinetic energy
Below the Kev range, the elastic collisions
is in the Kev range or above.
mechanism.
Before attempting an analysis of the
primary
the
are
spectrum it will be of interest to examine the essential
features of these two scattering reactions.
It was mentioned pre
process.
Consider first the inelastic-scattering
collision occurs, a compound
viously that when an inelastic-scattering
nucleus is formed, and this nucleus appears in an excited nuclear state.
neutron-energy
The energy of excitation is acquired from the kinetic energy of the

incident neutron and the binding energy associated with the addition
In the case of light nuclei, the
of this particle to the target nucleus.
excited energy states of the compound nucleus are far removed from the
69
REACTOR ANALYSIS
70
[CHAP.
ground state, and these states are spaced widely apart (in the order of
1 Mev).
If the struck nucleus is of large mass, however, the first excited
state is closer to the ground state (10 to 100 Kev), and the adjacent states
are much more closely packed1 than in light nuclei.
These general properties of light and heavy nuclei are illustrated in
The energy E* indicated in the figure denotes the excited state

of the compound nucleus formed from the absorption of the incident
It is evident from the figure that the
neutron by the target nucleus.
with
its
nucleus,
heavy
closely packed energy levels, has by far the greater
number of accessible decay channels for returning to the ground state.
One of these may involve the emission of a neutron which leaves the
nucleus with some residual energy (e.g., the state E\ in Fig. 4.16). This
The light nucleus
decay channel corresponds to an inelastic collision.
has fewer channels for decay to the ground state, for a given excitation
Fig.
4.1.
energy
E*, and, furthermore,
appreciably
possesses
threshold
energies
(the separation energy between the first
excited state and the ground state) . For
JL*_P_
higher

example, in the case of U2", the threshold

energy is about 40 Kev, and the level spac
ing, in the order of several Kev. For light
Ground state
Fig.
e0
4.1
Nuclear energy levels:
(a) low mass number; (6) high mass
number
nuclei, such as Li and Be, however, the

thresholds and level spacings are as much
as 2 and 3
-
Mev.
Because of this wider

.
.,
spacing, fewer energy states are available

to the compound nucleus, and the oppor
tunities for decay with neutron emission (i.e., by inelastic scattering)

the inelastic-scattering
cross sec
Consequently,
are greatly reduced.
tions of light nuclei are relatively small, and this process is not an impor
tant mechanism for energy degradation in media which contain appreci
able amounts of these materials.
relatively large inelastic-scattering
On the other hand, heavy nuclei have

cross sections and can therefore serve
neutron-energy moderators so long as the neutron energy

is above the threshold value.
Thus inelastic collisions are an important
mechanism for energy degradation only in the case of the heavy nuclei
as effective
Although these nuclei

range.
could cause some slowing down of neutrons, it is clear that they could
not be satisfactory as moderator material because of the threshold-energy
and, even then, only in the high-energy
limitation.
The elastic-scattering
process, however,
suffers no such restriction.
Regardless of the energy of the incident neutron, an elastic-scattering

The extent
collision can cause some reduction in the neutron energy.
1 See,
J. M. Blatt and V. F. Weisskopf, "Theoretical

John Wiley & Sons, Inc., New York, 1952.
for example,
pp. 365-374,
Nuclear Physics,"
SEC.
4.1]
BLOWING-DOWN
PROCESS
IN THE INFINITE MEDIUM
71
of the energy change does depend, however, on the initial energy of the
As will be shown shortly,
neutron and on the mass of the struck nucleus.
the light nuclei are the most efficient moderators from the viewpoint of
the elastic-collision process.
It is clear from the above discussion that a reactor which is designed
to utilize the superior low-energy properties of its nuclear constituents
Thermal reactors
must contain large amounts of moderating material.
In accord with the remarks above, these
are systems of this type.
reactors contain materials of low nuclear mass which slow down the high-

At the higher
energy fission neutrons by means of elastic collisions.
neutron energies (above threshold) the inelastic scatterings with the
heavy nuclei in the structural components of the reactor, and with the
Although these con
fuel itself, can assist in this slowing-down process.
tributions are generally small, they can be significant in some instances.1
Inelastic scatterings are most important in systems which contain large
amounts of high-nuclear-mass materials, such as iron, liquid metals, and
Reactors which utilize materials of this type tend to have
uranium.
Sometimes this property is deliberately designed into
fast spectrums.
In
such
systems good moderator materials are kept to a
reactor.
the
minimum in order to reduce the slowing down of neutrons as much as
In this way the average neutron energy is increased so as

possible.
better to exploit the superior nuclear properties of the fuel in the higher
The energy spectrum of these systems is determined by
energy range.2
their inelastic-scattering characteristics.
The greater part of the present work is directed toward the study
of thermal reactors which require efficient moderator materials as prin
The low-mass-num
cipal ingredients of the core (and/or the reflector).
Although their inelasticber substances are best suited to this purpose.
scattering cross sections are small, their elastic-scattering cross sections
are relatively large, and, moreover, the energy loss per collision is sub
Thus the elastic-scattering process is the mechanism of energy
stantial.
The analysis which

degradation of primary interest to the present study.
follows begins with an inquiry into the details of the collision mechanics.
In this treatment we examine the nuclear reactor with the aid of an
infinite-medium model which describes the neutron-density distributions
This model is the immedi
in terms of the kinetic energy of the neutrons.
ate extension of the one-velocity approximation introduced in Chap. 3.
In the present analysis we discard the one-velocity restriction and attempt
to describe the trajectory of the neutron in energy space as it is slowed
down by elastic-scattering
1
In
collisions with the nuclei of the medium.
In
reactors and in reactors containing highly enriched fuel.

conditions for breeding and converting, for example, are improved with
For U2" this ratio is noticeably
decreasing values of the ratio a = <ry/a/ in the fuel.
reduced in the Kev range from its value around thermal.
heterogeneous
1 The
HEACTOK ANALYSIS
72
[CHAP.
order to simplify the calculation we assume that the medium is homo

Furthermore, the
geneous and isotropic as well as infinite in extent.
neutron sources (to be consistent) are taken to be uniformly distributed
throughout all space.
These sources are assumed to emit neutrons
isotropically in direction, with a distribution of speeds (energies) to be
We introduce these restrictions in order to elim
specified in each case.
inate the spatial dependence of the neutron density and thereby allow us
to concentrate attention on the energy distribution of the neutrons in the
system.
Because of the assumption of isotropy of medium and source emission,
the number of neutrons moving in a given direction per unit solid angle
Thus for systems
per unit volume will be the same for all directions.
at steady state, the energy distribution alone will suffice to describe the
neutron population.
Consequently, for this calculation, we can use the
functions for the neutron density and flux given by (3.49)
(E)
dE
v(E) n(E) dE
total track length per unit volume per unit time of neutrons with kinetic energies between E and E + dE

The material of the infinite medium

sections 2(.fc'), 2,(2?), and 2/(2?), and
We define
per fission.
is characterized
v
by neutron cross
is the average number of neutrons
number of elastic-scattering collisions between

2 rp\ = exPecêd
neutrons of energy E and nuclei per unit distance of travel
ME)
= 2T(,')
(4.1)
(4.2)
+ Xf(E)
in the notation of (2.40).

We shall also use the total cross section 2<(/?)
according to the definition (2.50).
b. Motion in Laboratory and Center-of-mass Systems of Coordinates.
If a neutron collides with a stationary nucleus (described in the same
frame of motion), the neutron will emerge from the collision with a
reduced kinetic energy.
The nucleus in turn acquires a momentum
increment which will result in its displacement (recoil) from its lattice
position or an increased amplitude of vibration about its lattice site.
These remarks apply, of course, to nuclear materials in solid (crystalline)
form.
If the medium of the struck nucleus is in the liquid or gaseous
will recoil from the impact, moving off with some

velocity prescribed by the details of the collision.
If the struck nucleus
is a part of a molecular structure, the situation is further complicated,
and the details of the event are determined principally by the relative
For the present calculation, we limit the problem
speed of the neutron.
state,
the nucleus
to the consideration of a collision between a neutron and a free nucleus.

This situation describes in first approximation the collision process
between a neutron and any nucleus for neutron energies in the electronvolt range or higher. The problem of predicting the energy loss of the
8EC.
4.1]
SLOWING-DOWN PROCESS
neutron for these situations may

niques of the
be treated
73
by the purely classical tech
"billiard-ball" type of collision.
Consider, then, the collision between a neutron of mass m and a nucleus

We describe the motion of these particles in two reference
of mass M.
frames, namely, the laboratory and the center-of-mass coordinate sys
tems.
The velocities1 of these particles in the laboratory system (L) are

v0 =
Vo =
velocity
velocity
Vi = velocity
V*i = velocity
of neutron before collision

of nucleus before collision
of neutron after collision
of nucleus after collision
For the center-of-mass system (C) we
W for the
and nucleus velocities, respectively.
The subscripts 0 and 1
apply as before.
The immediate objective is to determine the relationship between the
initial and final energies of the neutron, i.e., the energy loss suffered by
For this
the neutron of mass m in colliding with a nucleus of mass M.
use the symbols w and
neutron

calculation we require the conservation laws for momentum and energy.

The conservation of (vector) momentum in the (L) system yields
mv0
From the kinetic-energy
+ MV0
= mv,.
MVi
(4.3)
balance we have
imvl + MVl
lmv\ + \MW\
(4.4)
where v! = v v. In carrying out the proposed calculation it is con

venient to transform these equations to a coordinate system in which the
center of mass of the neutron-nucleus system is at rest, i.e., the (C) sys
The velocity of the center of mass2 c is denned by
tem.
(M +
m)c = mv0 + AfV0 = mvi +
il/Vi
(4.5)
and V0 are specified, c may be computed directly

from (4.5). This expression for c may be used to relate the velocities
It follows that the velocities in (C) before
in the (L) and (C) systems.
collision are
Since the velocities
v0
w0 = v0
and after collision

Wl
c =
-ry
m
- - TX
(v0
Vo)
(4.6)
(v,
- VO
W.-Vi-c-jÔTt-TO
1
1
Boldface type denotes vector quantities.

The vector c is defined in the (L) system.
(4.7)
REACTOR ANALYSIS
74
[CHAP.
Thus the velocity vectors of the neutron
and nucleus in system (C)

are oppositely directed both before and after collision (cf. Fig. 4.2).
If
we compute the momentum in (C) before and after collision, using rela
tions (4.6) and (4.7),
it is
total
that
= mw,
+ MW0
mwo
so the
seen
MWi
= 0
(4.8)
of the system is zero in the center-of-mass
momentum
In the same way it is easily shown

system before and after collision.
the
in
that
kinetic energy
The substitution of the
(C) is also conserved.
definitions (4.6) and (4.7) into (4.4) yields
imw* +
pfW
*h
i

Wl =
w0
Wi
Velocities
niass system.
4.2
(4.9)
Z
-^w,
(4-10)
The substitution of these relations into
ff
Fig.
p/W?
But, we have from (4.8) that
imvfl +
(4.9) yields
in eenter-of-
w\ = w\
W\
W\
(4.11)*
Thus the magnitudes of the velocities

where |W| = W and |w| = w.
of the neutron and nucleus in the center-of-mass system do not change
during collision.
c. Energy Loss from Stationary
Nuclei.
When the neutron kinetic
energy is at least as large as 1 ev, the speed of the nucleus V0 (due to

thermal motion) is negligible in comparison to the speed of the neutron ;>0.
Thus a very good approximation, which may be introduced when dealing
with the slowing-down process, is the assumption that the nucleus is
This assumption
initially at rest in the laboratory system of coordinates.
simplifies the analysis somewhat ; however, it may introduce appreciable
error in the study of collisions between low-speed neutrons (in the order
of thermal energy ~0.02 ev) and nuclei.
The speed of the nuclei in
these situations is comparable to the speed of thermal neutrons, and the
approximation is therefore poor. For the present we will not consider
collisions by neutrons of electron-volt (or lower) energies and will limit
our calculation to the portion of the slowing-down process above the
The properties of these low-energy or thermal neu

electron-volt range.
trons will be examined later.
Consider, then, the special case in which the struck nucleus is initially
at rest, that is, V0 = 0. The velocity of the center of mass for this
situation
is [see
Eq. (4.5)]
C
: J>
+ M
Vo
(4.12)
SLO WING-DOWN PROCESS
SEC. 4.1]
and the various velocity vectors in (L) are shown in Fig.

purpose is to compute the neutron energy after collision,
The quantity v\ is obtained from the rela
tion
v\
75
Our
4.3.
namely, %mv\.
[see (4.7)]
= (Wl
= w?
+ c)s
c2
+ 2w!
(4.13)
But from (4.6), (4.11), and (4.12)

=
w? = w
we have
(^TT/)2v
(4.14)
The dot product which appears on the
Center of mass
Fio.
4.3
Fio. 4.4 Coordinate

velocity vectors.
Collision with stationary nucleus.
system for
This result may
2^
be
m)
A/)
(-^)2
(415)
The application of (4.14) and (4.15)
(^X/)2
vl
* (jT?
vl
u*
where we have denoted cos

to (4.13) gives the relation
(iff
r\
/3
Wl
'
C
/S.
right-hand side of (4.13) may be expressed in terms of the scattering angle

in the (C) system. This quantity is denned in terms of the coordinate
The z axis of this system is selected in the direction
system of Fig. 4.4.
Note that v0, Vi, and Wi are coplanar.
of the vector Vo.
It is evident
from the figure that the dot product Wi c = W\C cos
The substitution
of (4.14) and (4.12) into this expression yields
<

written in terms of the kinetic energies.
2mMvfa
(m
+ M)2
(4.16)
Let
/i0 = imvl = kinetic energy of neutron in (L) system before

collision
=
=
kinetic energy of neutron in (L) system after
Ei
imv\
collision
(4.17)
If
(4.16) is multiplied through

applied, the result is
E*
(A +
[chap. 4
ANALYSIS
REACTOR
76
by
1)2
m and
(A* +
the definitions
(4.17) are
2^ + !)
(4-!8)*
Thus, in the laboratory system, the kinetic energy

where A = M/m.
of the neutron after the collision is determined by the kinetic energy
before collision if the scattering angle 0 in the center-of-mass system is
It
known.
is clear
that,
as varies from 0 to ir, the final neutron energy
varies from a maximum value of E0, the initial energy, to the minimum
value
EmiB
= E0
a =
where
(x^pi)2~
= aE
|Y
(4.20)*
In order to determine
Function for Elastic Collision.
d. Frequency
(4-19)
it in another way,
the frequency function for the angle

describes the angular distribution of neutrons
This function, call
neutron of given speed and a nucleus
emerging from a collision between
will depend upon the energy of the
particular type. In general,
of
incident neutron and the nature of the struck nucleus. Its specific form
/S.
it
f,
it
we require
either from experiment or from the application of some

It usually convenient to
theoretical model of the nucleus.
determined
with the
or theory.
express this function analytically as an infinite series in

coefficients of the series to be determined from experiment
Thus we define a function such that
suitable
is
is
= fraction of all scattering collisions which

angles in (C) whose cosines lie between
=
result in scattering
and
+ drj
tj
dr}
f(jj)
AnPn(.v)dv
(4.21)
Pifo)
Po(u) =
where Pn(y) are the Legendre polynomials1

etc.
is
/3.
it
i?
has been used in the definition of since

this quantity
which enters into the energy relation (4.18) rather than
Likewise,
in
carried
out
terms
of
the
Legendre
polynomials in
the expansion
The argument
is

the distribution of final energies between the limits E0 and Emln [see
Eq. (4.19)], there must be some additional information as to the distribu
Or putting
tion of the scattering angles in the center-of-mass system.
'See, for example, F. B. Hildebrand, "Advanced Calculus for Engineers,"

Prentice-Hall, Inc., Englewood Cliffs, N.J., 1949.
171-181,
pp.
SEC.
SLOWING-DOWN
4.1]
PROCESS
77
order to accommodate the physical fact that the scattering process is

uniform in azimuth about the initial direction v0 of the neutron.1 Thus
This vari
the argument i) alone is sufficient to describe the scattering.
able takes on all values in the interval ( 1,1), the very interval over
which the functions P are denned.
If the function f(7j) is specified, the coefficients A are readily com
Thus we determine the coefficient A, by
puted by the usual methods.
multiplying both sides of (4.21) with the polynomial P,(rj) and integrating
over the interval (1,1):
P.M
^
Kv) * =
A,
Pn(v) P.(v) dv =
A.
or
2s
2A,
2s
(v) Kv) dv
(4.22)
case of the function f corresponds to isotropic scattering in

In this case, the cosine of the scattering
the center-of-mass system.
Thus we take
angle tj is uniformly distributed between 1 and 1.
=
And from (4.22)
0 for all n > 0.
An
A0 =
i f^PoWCdr,
where C, a constant, denotes the value of f for

this case. We obtain C from the requirement
on the frequency function, namely,
Kv)=i
scattering
follows that
-1<V (4.23)*
(4.23)*
Fig. 4.5 Scattering angle

that, even though the
in laboratory system.
isotropic in the center-of-mass
not isotropic in the laboratory system.
To show this we com
$u
should
d"
be noted
is is
It
and
/_',
= C
= A0 =
Therefore
dr>
it
/_". K")
This point
discussed
Wtf
(4.24)
in Sees. 7.2b and 7.3e.
(^J
2V
wlV)2]i =
Vl
(^j
(c
Vi = [w\ sin2
is
where
Mo
it
is
is
d0
do
it
system,
in the laboratory system. The
pute the cosine of the scattering angle
shown in Fig. 4.4.
An alternate view
angle
The
given in Fig. 4.5.
cosine of 80, call
mo, may be computed from the geometry of Fig. 4.5 by
using the relation

The simplest
(4.25)
KEACTOK ANALYSIS
78
[CHAP.
But, from (4.14),

Wi
Wo
(4.26)
The substitution of (4.25) and (4.26) into (4.24) yields

1 + Ar,
+
2VA +
(A2
Mo
(4.27)*
1)1
This expression relates the cosine of the scattering angle in the laboratory
system to the cosine of the scattering angle in the center-of-mass system.
In order that a scattering collision be isotropic in any reference frame,
it is necessary that the average value of the cosine of the scattering angle
This condition is already met for the (C) system by the func
be zero.
To demonstrate this fact

selected according to (4.23).
Now, the
we compute the average value of ij over the interval (1,1).
average value of any function of ij, call it s(tj), may be denned by
f
(4.28)
M.(l) f(l)
M0O7),
*>=!/_, (a|
(4.29)
on the other hand,
1)
dr>
is
mo
1*1
Ja
=
f0) *>
/_\
/_!,
rj
i
=
*J
and find
The average value of
/_i
7.
as expected.
f(l)
(4"30)*
is
is,
it
is
Thus, the average value of mo

where we have used (4.23) and (4.27).
not zero, and therefore, the scattering in the laboratory system cannot
assumed to be so in the center-of-mass
be isotropic even though
Equation (4.30) shows that the scattering in (L) tends to be
system.
the neutron tends to continue moving in its original

"forward," that
This effect
greatest for the light nuclei (A small), but as
Only in the limit for infinite nuclear mass
increases mo decreases.
For the heavier
does mo *
and the scattering becomes isotropic in (L).
is
direction.
0,
is
is
mo
it
can be assumed to good

very nearly zero, and
isotropic.
approximation that the scattering in (L)
The assumption of isotropic scattering in the center-of-mass system
gives
reasonably accurate description of the elastic-scattering process.
nuclei, however,
It
serves especially well in the thermal-to-Kev range of neutron energies

breaks down in the thermal region where chemical binding
however,
also poor in the highenergies become important. This assumption
is
it
The angular distribution of highly energetic

neutrons scattered from stationary nuclei
not isotropic in the centerof-mass system; the distributions show peaks in both the forward and
energy range
(>Kev).
is

Thus we compute
/_\
tional form of
SEC.
4.1]
79
A specific nonisotropic distribution of some prac

backward directions.
In
tical importance in reactor calculations is the so-called p scattering.
scattering reactions of this type, the distribution may be written in the
form
fWrfô
i(l
-1
ar,) dr,
<
= 0
i<-i
i?
<
>
(4.31)
is
0,
is
if
is
is
if
is
\a\
<
and
the cosine of the scattering angle in (C).
As compared to the frequency function for isotropic scattering (4.23),
seen to have an added term %ar\.
This situation
the above function
described physically by saying that there
component in the
Note that a > the
scattering, and this component has magnitude a.
predominantly forward, and
predominantly
<
scattering
where 0 <
backward.
in terms of the neutron
energy.
The appropriate function
is
is
The frequency function for elastic scattering from stationary nuclei

It more con
has been described in terms of the scattering angles.
venient for some calculations to work with frequency function described
obtained
is
is
is
dE = fraction1 of all scattering collisions which result in

in range
to
(4.32)
<
final kinetic energies of neutrons

< Ea
+ dE, where aE0
t)(E;Eo)
is
tj)
it
is
if
it
rj
is
The energy E0 denotes the initial energy of the scattered neutrons, and
a
Now, to each final energy E there
an associated
given by (4.20).
to
each
increment
scattering angle cos-1
Furthermore,
[cf. (4.18)].
Thus
change in tj, call
we
d-q, there corresponds a change dE in E.
scattered
relate Eton and dE to dr\, then the probability that a neutron
into
final energy range in dE about
[see (4.32)] must be exactly the
be scattered into di\ about 17. In other words, we
probability that
require that the probability that a certain event occur be the same
used to describe it.
Thus
regardless of the independent variable {E or
we require that
\KE;Eo) dE\
algebraic signs of
dE and
signs are introduced

dr\; note
that
(4.33)*
|f() d\
The absolute magnitude
so as to account
positive increment
for the
dE corre-
We use the notation \)(E;Eii) in order to keep in mind that the initial energy of the
scattered neutron was 0.
1

If
the frequency
directly from the energy-angle relationship (4.18).
function
specified in terms of either angle or energy, the corresponding
readily computed by the usual trans
function in the other variable
formation of variable.
demonstrated in
The procedure to be followed
We define the frequency function in energy by
the following calculation.
ANALYSIS
REACTOR
80
[CHAP.
Now, both frequency functions

sponds to a negative increment dy in r\.
f and f) are denned as positive quantities
(cf. Fig. 2.4), as are the prob
abilities t)(E;E0) dE and f(7/) dri; therefore,
ME;Eo)
drj
Kv)
(4.34)
dE
The transformation is carried out by computing the Jacobian
dy
dE
from
(4.18).
dy
dE
(4.35)
2AE0
Note that in the present calculation E, the final neutron energy, cor
If this result is used in Eq. (4.34), the
responds to Ei of Eq. (4.18).
frequency function in energy for the case of isotropic scattering in the
center-of-mass system is
(A +
t)(E;E0) dE =
\YdE

iAE0
dE
=
(1
- a)E0
aEo <
< Eo
(4.36)
Thus t)(E;E0) does not depend upon E, so that the kinetic energy of the
scattered neutron falls with equal probability into any unit energy inter
val in the range aE0 < E < E0- It is of interest to compute also the
distribution function for this process. If we define
&(E)
probability
then by Eq. (2.18)

=
that final
neutron
energy is less than
- a)E
de
&(*;) d( =
aE, (1
(1
- aE9
- a)EB
(4.37)
(4.38)
where the lower limit of the integral is given by Emin according to (4.19).
Incident neutrons of energy E0 cannot emerge from an elastic-scattering
collision with kinetic energies <a,0; thus &(E) m 0 for E < aE0.
The
maximum
energy which a scattered
From (4.38) ^>(Eg)

energy E0.
tion of a distribution function.
neutron can possess is its initial

= 1, as is to be expected from the defini
The distribution function and frequency

function in energy for isotropic scattering in (C) are shown graphically
in Fig. 4.6.
The analytical expressions for these functions are summa
rized below.
Energy range
i)(E;E0)
0
aEo
sE
E
^ Eo
> Eo
(1
- c)Eo
0
E
(1
aEo
- a)Eo
1
(4.39)
SEC.
4.1]
81
(1^) .
average energy of the scattered neutron after collision
UE;E0) dE
(^n2)
(o,
is
kw E
The
a,
a-
/;.
The energy lost by the neutron is given by the difference &E = E0 E.

This is the energy acquired by the nucleus and dissipated as heat in the
immediate vicinity of the collision as the recoiling nucleus is brought to
rest.
The average loss in energy per collision may be computed by means
of the frequency function $(E;Eq) [see (4.36)].
(4-41)
Fig. 4.6 Frequency

mass system.
and distribution
functions for isotropic scattering in center-of-
medium of beryllium (A = 9). For this nucleus, a = 0.64

= 0.82.
After the first collision the neutron energy will
will be
lie somewhere between 0.64 and
Mev. On the average
we assume that the neutron energy after each collision
0.82 Mev.
the average value given by (4.40), then after the second collision
will
it
If
it
a) a
is
released in
and ^(1 +
is
be
El
(l^j
E\
(ly Eo
= 0.67
Mev
is,
logarithm of the energy, that

the average value of In E0 In E{^
We compute the number
using the frequency function (4.39).
{.
is
and E3 = 0.55, EA = 0.449, and so on. On the average, each elastic

collision with Be reduces the neutron energy by 18 per cent.
e. Lethargy Variable.
quantity of greater interest in reactor analy
sis than the average loss in energy per collision
the average loss in the
,

Some idea of the magnitude of these changes in energy can be obtained

1-Mev neutron which
from the following sample computation Consider
REACTOR
82
ANALYSIS
[CHAP.
where a is given by (4.20).

The utility of the quantity rests with the
fact that it is independent of the initial energy of the neutron.
A graph
of this function in terms of the mass number A is given in Fig. 4.7.
The average logarithmic energy loss may also be obtained by another

In this connection we introduce the lethargy, which we define
approach.
in the following way:
w
= lethargy of a neutron of kinetic energy

In

The number
E*
(4.43)
E+ is a reference energy (selected
7 8 910
4.7
30
20
Mass number
Fio.
for the convenience of any
40 50 60
80 100
The average logarithmic energy loss
200
given problem) which corresponds to the zero of the lethargy.

Usually,
E* is taken to be some very high energy (say 10 Mev) such that prac
tically all neutrons of interest have energies below E+; thus, it is not
necessary to work with negative values of the lethargy.
It should be
noted from (4.43) that
du =
dE
(4.44)*
SEC.
SLO WING-DOWN PROCESS
4.1]
83
From the definitions of and u, it is clear that is the average gain in the
lethargy of a neutron per collision.
From (4.42) we have
= In
Eq
-g
= In
E*
- In ^,
E+
= u
(4.45)
w0
where u corresponds to energy E, and w0 to Ea.

In a typical reactor situation, neutrons appear from fission at energies
which are very high compared to the thermal energy of the medium.
Consequently, we can think of all fission neutrons as appearing at some

energy E0
The history of these neutrons may
Eth (thermal energy).
be described in the following way: (1) Neutrons appear at high energy
E0.
(2) Each neutron makes a first collision with a nucleus of the
medium, and this collision may be either an elastic scattering or an
absorption (in this discussion, only scattering collisions are considered,

since we are interested in only the energy changes of the neutron).
(3) After the first collision, the neutron will have an energy 2?i some
where between aE0 and E0.
(4) The neutron makes a second collision,
after which it will have energy E2 between aE^ and E\. (5) The col
lisions continue, and at each collision the neutron loses some kinetic
energy.
(6) The neutron finally attains some low energy at which its
speed is comparable to the speeds of the nuclei in thermal motion in the
medium.1
Neglecting absorption losses and thermal motion, each neutron travels

from energy E0 down to energy zero, or from zero lethargy up to infinite
While each neutron traverses this lethargy range by a differ
lethargy.
ent path (i.e., with collisions at different points on the lethargy scale),
it is possible, nevertheless, to compute some average quantities during
the slowing-down process.
For example, one quantity which is of fre
quent interest in reactor analysis is the number of collisions made by a
neutron in traversing a lethargy interval ux to w2, where 0 < Wi < w2 < 00 .
S. M. Dancoff2 has shown that the average gain in lethargy Am by a
neutron after n collisions with nuclei of mass A is given by
Aw = n
(4.46)
defined by (4.42).
This is not properly the average number of
collisions to traverse an interval Aw; however, we will use the form of this
result and assume that
with
M2
Ui
average number of collisions

passing from Mi to
made by a neutron
in
w2
An analysis of the behavior of the neutrons in a medium wherein the nuclei are in
thermal motion is given in Sec. 4.7.
' One form of the distribution function for collision from which this result was
obtained is given by E. U. Condon and G. Breit, Phys. Rev., 49, 22<J (1936).
1
84
REACTOR
ANALYSIS
[CHAP.
We justify this seemingly crude assumption by arguing that the expres

sion (4.47) is identical in form with the exact solution which has been
obtained in the case of hydrogen
U%
U\ =
= 1), namely,
average number of collisions made by a neutron

from U\ to it2 in a medium of hydrogen
in passing

Needless to say, similar calculations for the general case (A > 1) are
difficult to carry out and have led to complicated expressions for which
However, it is
simple approximating formulas have not yet been found.
reasonable to assume that, if u2 tti is small compared to In (1/a) (the
maximum jump in lethargy per collision), Eq. (4.47) should hold, at least
approximately.
(Here we are considering single-element media only.)
that
this
condition
corresponds to the situation in hydrogen wherein
Note
In (1/a) is infinite and any (u2 Ui) is small in comparison.
For media
in which A > 1, and w2 U\ ~ln (1/a), Eq. (4.47) is commonly used
even though it is at best only an approximation for which the limits of
validity are not known.
Since the results of using (4.47) over the extended
lethargy range have generally been good, we shall accept the approxima
tion in this work; thus, for all A and for any lethargy interval i * u2,
however large, we will take (u2 tti)/f to be the average number of
collisions in traversing the interval (2 Ui) and recognize that the
approximation is particularly good for
M2
U\
(4.48)
We conclude this section with a calculation of the frequency function

in terms of the lethargy variable for the case of isotropic scattering in
the center-of-mass system.
The computation is carried out by trans
forming the energy function i)(E;E0) given in (4.39) to the lethargy space
by means of a relation of the type (4.33).
We define
probability that a neutron of lethargy u0 when scat
with a lethargy in da about u
tered emerges
where
Mo
= In
p-'
(4.49)
Then
(4.50)*
The corresponding distribution function (u;uo)
is
(4.51)
SEC.
It
4.2]
85
is important to note that this result gives the probability that a
neutron scattered at itn emerges

with lethargy <u, that is [cf. Eq.
(2-8)],
(w;uo)
P(w
< u)
where w is a random variable.
Uol
(4.52)
u =
In;
Fur
thermore [see Eq. (4.39)]
- (;uo) = l_a n $()

p ( uo)
(4.53)*
u = oo
E = 0
This relation simply states that the

Fig. 4.8 Final lethargy and energy of
probability that a neutron of leth
scattered neutron.
argy M0 (or energy E0) emerge with
lethargy > u is identically the probability that the neutron emerge with
energy <E. This relationship between the distribution functions in leth
argy and energy is evident in Fig. 4.8.

4.2
Slowing-down Density in Pure Scatterer
Neutron Flux. In this section we consider the problem of finding

the energy distribution of neutrons slowing down in an infinite, homo
geneous, isotropic medium.
We have already noted that, if the neutron
sources in this medium are distributed uniformly in space and time, the
neutron population may be described by means of the neutron flux or
a.
These
given in terms of the neutron energy alone.
functions may also be described in terms of the neutron speed, or as we
will shortly demonstrate, in terms of the lethargy.
Which one of these
density
functions
independent variables, lethargy, speed, or energy, is selected for a specific

calculation is entirely a matter of convenience.
In any case, if the neu
tron population is described in terms of one of these three variables, the
function in terms of the other two is easily obtained by a

For example, in the case of the energy and
transformation of variable.
the
fluxes
speed variables,
(or densities) satisfy a relation of the type
Thus we write
(3.47).
appropriate
\*{E) dE\
where
(E)
(v)
\(v)
dv\
dE = total track length of neutrons per unit volume

per unit time with energies in dE about E
dv =
(4.54)
..
total track length of neutrons per unit volume

per unit time which have speeds in dv about v
As previously mentioned, the symbol

denotes neutron flux and the
argument following indicates the variable in terms of which it is defined.
REACTOR ANALYSIS
86
Properly, the functions
(E)
[CHAP.
(v)
should be given distinct symbols,

but this practice has been discarded here in the interests of simplicity in
Equation (4.54) states that, if the intervals dE and dv cor
notation.
respond, then the neutron flux described by either coordinate system
(E or v) must be identical.
A similar relationship may
and
be
written in terms of the lethargy.
If
we
define
.
total track length of neutrons per unit volume per

unit time which have lethargies in du about u
then,
\4>(u) du\
\(E)
dE\
(4.56)
(4.57)
The neutron densities may also be related by equations of this type

Throughout this work we will use all three independent
[see Eq. (3.42)].
variables E, u, and v freely and transfer from one to the other as the need
arises. For completeness we list below the various collision densities
in terms of the new variable u:
Z,(w)
2/(m)
{u)
{u)
(u)
(elastic
number of \
=
<
du
neutrondu
nucleus
reacôns Per um^

/ volume per unit time
> due to neutrons with
(4.58)
letnarS>es m du
I
J about u
/ fission
\
It
That
is,
is important to recognize that, whereas the neutron fluxes (or densities)

are related by expressions of the type (4.54) and (4.57) which require a
specification of the interval, the
cross sections are identified directly.
2()
Typical
neutron cross-section
JK
S(u)
(4.59)*
.c
be specified,
r
lhe cross sections, on

'
4.9
Fig.
curve.
= 2(y) =
The distinction arises from the defi

nition of these two types of func
tions.
The fluxes and densities are
denned per unit energy (or lethargy,
or speed) and therefore require that
the width of the interval in question
,
the other hand, specify the interac

neutrons of a particular energy.
and
nuclei
characteristics
between
tion
Figure 4.9 shows a typical cross-section curve given as a function of
Note that all three cross sections c{Ex), <r(i), and
neutron energy.
tr(ui) refer to the same point on the curve which may be specified by
either Ei, Vi, or Ui.
It was remarked in Sec. 2.1a
b. Continuous Slowing-down Model.
discontinuous function.
that the neutron trajectory in energy space
is
a

I(tt)
du
SEC.
4.2]
87
If
the neutron energy were to be given as a function of its position along

its path in actual space, the resulting function would be of the form
Thus, the " slowing-down track" is a step function
shown in Fig. 4.10.
which has a unique form for a given neutron history. Three such
Although no two can be alike, most
histories are sketched in the figure.
of the neutron trajectories for a given medium follow closely a general
In fact, it is entirely meaningful to talk about an "average"
form.
Whatever the
trajectory or an average neutron slowing-down track.
form of this function, it will describe the average behavior of a neutron
in the slowing-down process. As in the case of the neutron density, the
function for every neutron is not required (or
specific slowing-down
For
of reactor analysis, a function which describes
the
purposes
desired).
the average behavior is sufficient.

Urrtz;
Distance
Fig.
4.10
0(source energy)
traveled
by neutron
Slowing-down
track.
Distance
Fia.
traveled
by neutron
Slowing-down
track
medium of heavy nuclear mass.
4.11
in
The actual discontinuous function of the neutron-energy track illus

trated in Fig. 4.10 is very difficult to handle in the study of the slowingConsequently, much attention has been devoted to find
down process.
ing some analytically continuous function which will serve as a good
average over all the possible step functions of Fig. 4.10 and thereby
From a physical view
represent the average behavior of the neutrons.
point such a continuous function is easily imagined in the case of heavy
nuclei.
For these nuclei the maximum jump in energy that the neutron
can experience is small relative to the energy of the colliding neutron.
For example, in the case of deuterium (A = 2), the maximum fractional
jump in energy is
for helium (.A = 4), fa, and for lithium (A = 7), ^
Thus the maximum allowable jump decreases rapidly
[cf. Eq. (4.40)].
with increasing nuclear mass. For the heavier nuclei, then, a continuous
function could describe to good accuracy the slowing-down track of the
In the case of the light nuclei, the concept of
neutron (cf. Fig. 4.11).
a continuous function is not so useful, and a more accurate description
which accounts for the discrete jumps in neutron energy must be made.
For the present we will confine our attention to a study of the slowingdown
process which utilizes the concept of a continuous slowing-down

In this model, the energy-loss process is a continuous one, with
changes in neutron energy due to collisions occurring in arbitrarily small
model.
REACTOR
88
ANALYSIS
[chap. 4
To describe this process we introduce

(differential) decrements.
slowing-down density q(E) and define
density at E
= number of neutrons per unit volume per unit time whose
energies are changed from some value above E to some
value below E
the
q(E) = slowing-down
(4.60)
Thus this function gives the rate at which neutrons are slowing down
It should be observed that this concept is
past some given energy E.
similar to that of a flow (current) in a fluid.
One may think of the
function q as giving the flow rate of neutrons down the energy scale.
In the fluid analogy, q could be used to specify a continuity condition
on the neutron population in terms of the factors which affect the slowingdown process.
This we propose to do shortly.
The slowing-down density may also be defined in terms of the neutron
As in the case of the neutron cross sections [cf.
speed or lethargy.
Eq. (4.59)],
= q(v) = q(u)
q(E)
It
is
is
v,
is,
the slowing-down rate

the same on the energy, speed, or lethargy
the measurement
made at corresponding values of E,
or u.
if
scales
important to note that the concept of a slowing-down density

not limited to the continuous slowing-down model.
This function has a
much wider application, and
will be used later in connection with the
analysis of the slowing-down process which allows for the step-function
behavior illustrated in Fig. 4.10 (cf. Sec. 4.2c).
We discuss now a simple model which will adequately display several
basic features of the slowing-down process.
Consider an infinite medium
into which neutrons of energy E0 are introduced continuously in time
and uniformly in space at a rate q0 neutrons per unit volume per unit
Eventually
time.
achieved in this medium so that the
steady state
number of neutrons in any given energy interval AE in
given volume
remains constant in time.
Let us further idealize the medium by assum
ing that the nuclei are pure scatterers, i.e., Za(/) = 0. Then each neu
tron of energy E0 which appears in the medium moves off from its point
of origin until
makes a first scattering collision.
Since the distribution
of sources
uniform, the distribution of first scatterings will also be
is
it
is
it
is
is
is
is
uniform.
In this model the slowing-down density
different from zero for all
< E9, but
identically zero for all
energies
> Ea [since by (4.39)
no neutrons are speeded up by scattering collisions, and no neutrons
All neutrons introduced into the medium
are introduced at
> E0].

That
(4.61)
at Eo slow down to some energy below E0; thus, in steady state

q{E0)
q0
(4.62)
SEC.
4.2]
SLOWING-DOWN
PROCESS
89
Also, in the steady-state condition, since there are no losses by absorp

tion or leakage in this model, the number of neutrons slowing down into
a given energy interval must be equal to the number slowing down out
of that interval; i.e., neutrons do not accumulate at any one energy over
a period of time.
Thus, for E < E0,
q(E + dE) = q(E)
or
^ q(E)
q(E)
and
If
we
(4.63)
= constant
apply the initial condition (4.62),

q(E)

= 0
(4.64)
q0
The slowing-down density in this model, then, is constant in energy.

One of the important problems to be considered in this chapter is the
determination of the relationship between the flux and the slowing-down
As an introductory treatment
density for various physical situations.
of this problem we carry out the following calculation, namely, compute
the number of scattering collisions which occur per unit volume per unit
time, using first the concept of the neutron flux and second the concept
of the slowing-down density.
We write the scattering-collision density
in terms of the flux by means of Eq. (4.58).
S (u)
(u)
du
numDer f scatterings per unit volume per unit

time due to neutrons whose lethargies lie in du about u
**a^
This quantity may also
expressed in terms of the slowing-down
be
density q(u). According to (4.47), du/ is the average number of scatter

But, q(u) neutrons per unit
ings made by a neutron in traversing du.
volume per unit time slow past lethargy u; therefore,
.
du
If
total number of scatterings per unit volume per unit

time by neutrons with lethargies in du about u
these two expressions are equated,
it follows that
q(u) = {2,(tt)
(u)
(4.66)
between q{u) and (u) applies in an infinite medium

of pure scatterer.
It will be our purpose later to develop suitable
expressions of this type which can be used in finite systems containing
This relationship
materials.
neutron-absorbing
It is of interest to compare the expression (4.65) with the correspond
The symbol denotes
ing relation given in terms of the neutron energy.
in
the average gain
lethargy per collision; its counterpart in energy space
is
(1
a)E/2
[see
Eq. (4.40)].
i \ du
?(")
-j
Thus if dE corresponds to du, then

2q(E) dE
(1
- a)E
(4.67)
90
REACTOR ANALYSIS
[chap. 4
This statement follows directly from the physical fact that the number
of collisions which occur in traversing dE must be identically the num
ber which occur in traversing du if dE implies du. But, by (4.61),
q(u) = q(E), and by (4.44), du = dE/E; therefore,
q(u) du
and
it follows that
2q(u) du
1
a
(4.68)
which appears to be in contradiction to the definition of given by

In fact, the above result is the
(4.42); however, this is not the case.
limiting form of for values of a
1.0
approaching unity, i.e., A a> (see
Fig. 4 . 1 2) . Thus the scattering-col
0.8
lision density may be computed by
either of the two

0.6
forms given in
(4.67) provided that a is near unity.
Note that if this condition is to be
OA
satisfied,
l-a-"
condition (4.48)
as
well, then clearly du* 0; i.e., (4.67)

applies if Au is small and the nuclear
0.2
and
mass large,
Density by ColliIntervals.
sion
We turn our atten
tion now to a more realistic treat
c. Scattering
0.2
Fig.
4.12
0.4
0.6
Approximation
0.8
for
l.O
.
ment of the slowing-down process

which takes into account the discontinuous form of the neutron trajectory
in energy space.
We carry out the analysis by considering the collisions
at each stage of the slowing-down process and determine the distribution
in energy of the neutrons for the physical situation in question.1
For
the present calculation we assume that the scattering collisions are
isotropic in the center-of-mass system of coordinates and recognize that
our results will be of questionable value, therefore, when applied to
neutrons in the thermal range, or to very fast neutrons (see Sec. 4. Id).
A situation of fairly general interest in reactor analysis is one in which
neutrons of some relatively high energy E0 (which may be in the order of
1 Mev) are introduced
into a uniform infinite medium and subsequently
slow down by elastic collisions.
Suppose, as a particular case, that the
medium consists of a single nuclear species and that the following set of
assumptions apply: (1) q0 neutrons of energy E0 appear per unit volume

per unit time from uniformly distributed sources throughout the medium.
1 From our knowledge of the scattering phenomenon as defined by the frequency
function.
SEC.
4.2]
91
(3) The system is at

(2) All nuclei in the medium are stationary.
steady state.
It must be observed that in this system a finite number of neutrons
at any given instant have exactly energy E0; thus the neutron density
n{E) has a singularity at E = E0. The mathematical difficulties inher
ent in this single-source energy model may be overcome by regarding the

function n(E) as defined up to, but not including, Ea; then the condition
at E = Eo may be used as an initial condition (in energy) on the slowing-

down density q(E).
aE'
Fig.
4.13
First collision interval.
We begin the calculation by considering the distribution of neutrons

whose energies fall within the first collision interval, namely, in the inter
val aE0 to E0. Our purpose is to determine the neutron balance and the
slowing-down density at any energy E for aE0 < E < 0- Consider
first the slowing-down density q{E).
The rate at which neutrons are
E is determined by the collision density at all

Thus we compute the number of col
energies E' > E (see Fig. 4.13).
in
E'
lisions
dE' about
which yield neutrons of energy < E. Only these
The reasoning here is clear
neutrons can contribute directly to q(E).
slowing down past energy
if it is recalled that q(E) represents a "flow rate" in energy space. Thus,

if a collision around E' produces a neutron of energy greater than E,
this neutron will not be observed passing energy E and therefore cannot
be counted in the density q(E).
The number of scattering collisions per unit volume per unit time in
dE' about E' is 2,(') (E') dE'. Of this total number a certain fraction
REACTOR ANALYSIS
92
[CHAP.
will produce neutrons with energies <E. But this number is precisely
the probability that a neutron which suffers a collision at energy E'
will emerge with some energy <E. We obtain this from the distribu
tion function (2?) [see Eq. (4.39)]; thus the fraction of neutrons scat
tered to energies <
is
P( <
E)
(f
(4.69)
The total number of neutrons scattered from dE' which emerge with
is
'1
4>{E')
"f
X.(E')
f.f
energy less than, or equal to,
dtf'
(4.70)
In order to obtain the total contribution to q(E) from all such intervals
E to E0. To be correct, the

should extend to E/a because this
the highest energy which
How
neutron can possess and still "reach"
by
single scattering.
ever, the integration need extend only to Eo, for by definition there are
no neutrons in this system with energies greater than E0.
dE' we integrate the above result from
For the pure scattering medium considered here, there are

first scatterings per unit volume per unit time.
Of this total
fraction (E aE0)/(l a)E0 yield neutrons of energy <E.
number of neutrons from first collisions which can contribute
therefore
(1
qo(E
exactly go
number, a
The total
to q(E)
is
is
The contributions to q(E) so far considered are due to scattering of

neutrons of energies E' >
but less than EQ. As previously mentioned,
=
the flux
not defined at
E0, and the contributions to q{E) from the
first collisions must be taken into account by a separate computation.
- aE0)
- a)E0
(4.71)
We obtain the total contribution to q(E) by performing the integration

described above and adding the contribution from the first scatterings;
thus, for the first collision interval
+
m- j>*W)[^]^ ^f>
(4.72,
is,
<
This expression gives the slowing-down density q(E) for all energies
in the interval aE0 <
E0, that
> aE0.
(4.72) applies for all
< aE0, the source cannot contribute directly to q(E) and the
When
omitted; thus,
5o term
q(E)
[B/a2.(E')4>(E')\
^L]dE'
for
is

is
integration
< aE0
(4.73)
SEC.
4.2]
Note that the integration now extends to
B/u
93
due to the fact that
E/a < E0, and the flux is therefore defined over the entire interval.
The above results involve two unknowns 4>(E) and q(E). In order to
obtain a complete definition of the neutron distribution in energy,
This can be obtained from the neutronConsider, therefore, an energy interval dE about E
The number of neutrons disappearing from this interval
we require a second relation.

balance condition.
(see
Fig. 4.13).

(due to scatterings which change the neutron energy to a value which

falls outside the interval) must be balanced by the number scattered
The rate
into dE from energies above E, in the steady-state condition.
at which neutrons are removed from dE is 2.(E) (E) dE. The rate at
which neutrons enter dE is given by the number of collisions which occur
at energies above E times the probability that these scatterings "throw"
the neutron into the interval dE. We compute this number by consider
ing some interval dE' about E', where E' > E. As before, the scattering
The fraction of these collisions which
rate in dE' is X,(E') 4>(E') dE'.
yield neutrons that have energy in dE about E is just the frequency
function ]){E;E') [see Eq. (4.39)]; this follows from the definition (4.32).
If E < aEo, the neutron source cannot contribute directly to the neutron
population in dE; thus, the neutron balance in dE about E is given by
Z.(E)
4>(E)
dE
fE/a2.(E')
JE
(E') \n
\_\\
dE
- o-)E
dE'
for
<
aE,
(4.74)
E > aE0, then the contributions from the source

must be included, and the neutron balance is
On the other hand, if

at
A'o
Z.(E)
HE) dE
pME')
HE')
y^jp]
dE' +
(r^fo
(4.75)
The results (4.74) and (4.75) define the neutron flux (E) for all values
The solution of these integral equations when substituted
of E < Ea.
into the expressions (4.72) and (4.73) yields the appropriate forms for
q(E) over the entire range 0 < E < E0.
We demonstrate the procedure for solving these equations by carrying
out the calculation for the first two collision intervals, namely, aE0 Ea
For convenience call the scattering density 2,< in the
and a2E0 aE0.
first interval Fi{E) = X,(E) i{E). Then (4.75) may be written
/m
F1(E)
94
REACTOR ANALYSIS
This is
[CHAP.
first-order differential equation in Fi{E) ; its solution is

F1(E)
F1(E0)
(j?)
(4.77)
We obtain fi(o) by taking the limit of (4.76)

lim FÊ)
E-*E,
?0
jT
U
)#0
E -*
as
E0
= Fi(Eo)
Substitution into (4.77) yields the scattering density in the first collision
interval (aE0 < E < E0)
=
me) me)
For the second collision interval
â)-Eo($y
l/(l-a)
(4.78)*
write
we use (4.74) and
. urn wn - Jef" 'J^g

- a)L + JaE,
- a)E
('"w^w
Fm
(1
(1

Note that the integration over the interval E to El a has been separated
into two parts so as to accommodate the
change in the functional form of 4>{E) as the
proceeds from the second collision
integration
otE0
E
a2E0
Fig.
4.14
interval to the first (see Fig. 4.14). As in the

case of the first interval above, we transform
this integral equation into a differential equation ; the result is
--
mF,{E)
Second collision
interval.
where we have used the solution
Ft(E)
ME) ME)
(i
(1
(4.78).
a)E
- a)2()1/(I
goy'
(1
(4
K
)+l
80)}
The solution to (4.80) is
(r^ - (irj +
[1
"
al,(1-)]|
(481)*
We have used here the limiting value

E-aE,
(1
a)haaini
a>
This computational procedure may be continued indefinitely to obtain

the collision densities F3(E), Ft(E), . . . and the corresponding fluxes
Although the calculation is somewhat laborious,
4>3(E), *(E)
the method is straightforward.
SEC.
4.2]
d. Asymptotic
densities Fi and
Fi{aEo)
It
Solution.
F 2 at E
~
(1
95
is of interest
to compare the collision

The appropriate expressions are
= aE0.
- a)E^-<,
Fl(aEa)
(1
- a)E^'^
and a comparison of these values shows that

=
Fi(aE0)
- a1"1-']
Fi(o#0)[l
Evidently the collision density F(E) is discontinuous
(E)
at
= aE0; there
must also be discontinuous at aE0.

If the procedure
mentioned above is extended to the third collision interval and beyond,
one finds that F(E) is continuous at E = a2E0, but F'(a2E0) is discon
tinuous.
At E = as0, both F(a*E0) and F'(a3E0) are continuous, but
fore, the flux
F"(a3E0) is discontinuous, and so on; at energy

first appears in the (n l)st derivative of F(E).

the discontinuity
as the neutrons
progress down the energy scale, the collision densities of any two adjoin
ing collision intervals, say F(E) and Fn-\(E), which describe the energy
distribution of the neutrons, approach the same analytical form, and dis
similarities between the two functions recede to higher and higher deriva
tives.
As the ratio E/E0 * 0, F(E) approaches a limiting form.
This
limit can
a"2?0,
Thus
from the solution of the nth interval.

lision density for this interval is
be obtained
tJK -
JB
(1
Ot)h
+ ("
Ja'-'Eo
(1
F(E)
|.
(4.82)
Ct)E
But, as we have already observed, for large values of n, F(E)

thus approximately
A solution to this equation is
The col
~ F_i(i?)
F{E)
(4.84)
where C is some constant.

The symbol F(E) = 2,(5) {E) denotes the
limiting form of the collision density.
The constant C may be obtained
for the prescribed condition on the slowing-down density q(E). The
appropriate equation for q(E) for neutron energies far removed from the
If we use (4.84) for the collision
source energy E0 is given in (4.73).
density in this equation, then
qiE)
_~ fE/"F{E')(E JE
(1
aE')
a)E'
dE>
[*/- (E
]E
(1
- aE') dE' _
- a){ET
The last equality follows from the fact that no neutrons can
~ qo
(4-85)
be lost in a
medium of pure scattering material, and the number of neutrons slowing
REACTOR
96
[CHAP.
must be the same for all E, namely,

The solution of (4.85) for the constant yields C =
down past energy

value.
ANALYSIS
F(E)
or, if we use q(E) =
energies below
solution
(E)
the source
thus,
<7o/;
||
(4.86)
q0,
q(E)
The asymptotic
Z.()
50,
= a2E0,
&.(E)
4>(E)E
(4.87)*
(4.87) is a very good approximation for all

and for many purposes it will be acceptable
80
1
70
60

50
1
\
40
, Asymptotic
o
30
20
\\
1
1
N
1
^
10
0.2
0.1
n-
O0
4. 15
of source
4th
a*
0.3
3rd
4-
, Collision density
F(E)Ea/qa
1
1
1
1
Jgttl
Fig.
form
0.4
0.5
' 0.6
2nd interval
0.7
0.8
0.9
1.0
1st interval
Collision density in pure scattering medium of mass A

neutrons.)
i0 = energy
Hydrogen is a special case; since a = 0 for hydrogen, the

up to E0.
However, for deuterium
entire energy range lies between 7J0 and aE0.
a =
and if E0 is chosen as 2 Mev (a representative energy for neutrons
from fission), a20 25,000 ev.

Thus the most important part of the
slowing-down range even in thermal reactors with deuterium lies in the
For other substances,
region in which (4.87) is a good approximation.
the situation is even more favorable.
Figure 4.15 shows the collision density
scattering
material
of mass A
= 7.
F(E) E0/qo in
This function
is
a medium of pure
plotted in terms
SEC.
4.2]
PROCESS
BLOWING-DOWN
97
of the energy ratio t E/E0, where E0 is the energy of the source neu
The discontinuous form of the collision density is clearly dis
trons.
It is of interest to note that,
played in the first two collision intervals.
in the third interval and beyond, F{E) "smooths out" and rapidly
approaches the asymptotic solution E0/E% (given by the broken-line
In the preceding discussion of the asymptotic solution, it was
curve).
remarked that this form could be used as an approximate description of
F(E) even in the first and second intervals, especially for the heavier
The relatively good match of the asymptotic form to the actual

function F(E) in this energy range is well demonstrated in Fig. 4.15.
We close this section with a brief discussion of the solutions for the
slowing-down density and the collision density in terms of the lethargy
variable u. The set of equations which correspond to the results (4.72)
through (4.75) are readily obtained by introducing the distribution func
tion &)(u;u') and the frequency function g(u;u') from Eqs. (4.50) and
nuclei.
It can be
(4.51) and making a suitable change of variable from E to u.
shown that the slowing-down density q{u) is given by [see also (4.53)]

q(u) =
OC
fU2.(u')
Jut
4>{u'){e"'-"
- a) du' +
~
go(e"^
1
"}
Of
u < o + In
=
j /
"
- a) du'
"
,
S.(m') 4>(w')(eu'-u
u > uo + In
(4.88)
As a general rule when dealing

corresponds to the energy E0.
neutrons, it is convenient to define u in
Thus, if we take u = In E0/E, then m0 = 0.
terms of the source energy.
The collision density F (u) when written in this notation takes the form
where
with
u0
a source of monoenergetic
F(u) =
"
! / F(u')
a Jo
^
1
e'- du'
= T-^ I
, F(u')
1 a /-ini
eu'
-"
du'
u < In
u > In
(4.89)
integral equations (4.88) and (4.89) for q(u) and F(u) may be
solved by the same procedure as used in the case of the energy variable.
It is easy to show, for example, that the collision density in the first
These
interval Fi(u)
is given
by
Fl(tt)
and so on for the other
intervals.
q*_a
It
is of interest
(4.90)
to determine
the
REACTOR ANALYSIS
98
[chap. 4
asymptotic solution for F(u) in the lethargy range far removed from the
source at u = 0. In this range we use the form
1
1
fCn-l)
a yM-in
In
du'
"
/
l-aj(n_i)inl
e-'- du'
Fn(u')
(4.91)
"
F,(u')
/
i-yu_ini
.
e'- du'
in the nth collision interval. The approximate form used at the right
follows from the same arguments as used in the calculation of F(E).
The solution to Eq. (4.91) is
L4
F,(u>
II
by direct substitu
Thus in the asymptotic range,
as can be verified
1.3
tion.
Ail'"
the slowing-down density q(u) and

the flux (u) are related by
1.2
1.1
q(u) = 2,(u) *(u)
(4.93)
l"a
Fig. 4.16 Collision density F(u) in pure
= 2.
scattering medium of mass
is
3lno
1ns
Inj
if
This result states that,

the col
lision density F{E) in
medium
varies like 1/E, then F(u)
a con
stant [cf. Eqs. (4.86) and (4.92)].
Figure 4.16 shows F(u)/q0 in a medium of pure scattering material of
= 2.
mass
It should be noted that, although there are relatively wide
variations in the function F(u) through the first two collision intervals,
the function rapidly "smooths out" and approaches the asymptotic solu
tion l/.
1.0
4.3
Slowing Down in Hydrogen
Collision Density. In the preceding section we examined the slow

In
ing-down process in an infinite medium of pure scattering material.
all problems of practical interest to reactor physics, however, the neu
trons slow down in media which contain some neutron-absorbing
mate
rials.
The amount of absorber present depends, of course, on the
reactor region in question.
The reflector, for example, usually consists
of materials of relatively small absorption cross section.
In such regions,
the results based on the model of
pure scattering medium would give
Even here, however, the
first approximation to the actual conditions.
In the reactor core, the
of questionable value.
pure scattering model
situation
still more unfavorable for the use of this model because of the
presence of large amounts of fuel materials which have large absorption
a
a.
is
is

(4.92)
Collision dansl
*.<) =
= 1.38.
LB
cross sections.
SEC.
4.3]
SLOWING-DOWN
PROCESS
99
A typical history of a neutron slowing down in a medium which con

tains both absorbing and scattering materials was described in Sec. 2.1a.
An important feature of this process was that at every collision between
a neutron and a nucleus there was a finite probability that the neutron
The probability that a neutron is absorbed upon
could be absorbed.
making a collision is 20()/2,(), and that it be scattered, ME) /ME)Thus at each stage in the slowing-down process, the neutron can be
The physical consequence
removed from the slowing-down "stream."
of this fact is that the slowing-down density q(E) will not be constant
The process is further complicated if the medium contains
in energy.
In this case when a neutron is removed from the
fissionable material.

slowing-down stream because of an absorption collision, other neutrons

may appear from fission sources. Thus in developing the neutronbalance relations for the slowing-down process in an absorbing medium,
additional terms must be introduced which take into account the absorp
tion losses and fission sources [see Eqs. (4.74) and (4.75)].
As a first step in deriving the expressions for the neutron density in an
absorbing medium we consider the relatively simple case of an infinite
The
medium of pure hydrogen, for which a rigorous solution is possible.
model for this calculation is defined by the following set of assumptions :
(1) infinite medium of pure hydrogen (A = I); (2) 2a() j 0; (3) go
neutrons of energy E0 appear per unit volume per unit time throughout
all points of the medium; and (4) the hydrogen nuclei are stationary.
We begin the analysis with the calculation of the total collision density
F(E) defined by
F(E)
- ME)
4>{E)
The appropriate neutron-balance

seen to be [cf. Eq. (4.75)]
ME)
*{E)
[z.(E) + ME)]
condition
ME') (E')
(E)
for a medium of hydrogen is
rr,
+p
E <E0
(4.94)
The term at the left gives the rate at which neutrons are removed, by
both absorption and scattering collisions, from the unit energy interval
around E.
The integral, as in Eq. (4.75), gives the rate at which neu
trons enter unit energy interval about E by being scattered down from
Note that the frequency function for scattering in
energies above E.
hydrogen takes the simple form 1/E' if E' is the neutron energy before
collision.
Because a = 0 for hydrogen, a neutron can be scattered from
Consequently, the neutron source can
Et> to any energy 0 < E < Ea.
contribute directly to the neutron population at every energy from Ea
down to zero. Thus the term qo/Eo appears in the balance relation for
This form for the source term is only an approximation,
all E < E0.
Actually,
all neutrons that appear from the source do not
however.
REACTOR ANALYSIS
100
[chap.
enter into the slo wing-down process; some source neutrons are absorbed
at the first collision, and therefore the go factor should be corrected to
account for these losses.
It would be better, then, to use the form
Number of neutrons
slowing down from
E0 per unit volume
per unit time
..
J^.
q{E)
q{Eo)
Z,(g0)
f"
[za(E0) + S.(^)
(4.95)
But in many
cases, ^,a{Ea)/Z,{E0)
and
<5C 1,
q(E0)
?0
will often be taken as an approximation.

Equation (4.94) may be solved by reducing it to
Thus, if we differentiate (4.94), we obtain
a differential
equation.
--mi-m
f(E)
-^^jrj
F(E)=F(E0)exp[JE
But from (4.94), F(E0)
W) -g exp [ - ^
= q0/E0; therefore,
b. Nonabsorption Probability for Hydrogen.

The slowing-down den
be
the
result
The appropriate
computed using
sity q(E) may
(4.96).
integral equation is [see Eq. (4.72)]
q(E)
J*
X.(E')
4>(E')
The identity follows from (4.94).

of F(E) given in (4.9G) ; thus
qB(E)
= go exp
d~ +
^-
EF(E)
This result may
f*'
be
(4.97)
written in terms
- qoPn(E;E0)
(4.98)*
probability that they are

exponential
not absorbed
between
and E0.
Thus the
identified as the
in (4.98), which we denote by pH(E;E0),
nonabsorption probability:
is
verbal
where the subscript H has been added to denote hydrogen.
statement of this result might be expressed as follows: The number
of neutrons which succeed in slowing down from E0 as far as
given
multiplied
the
by
starting
of
neutrons
at
(namely,
by the number
g0)
E0
is

The solution to this differential equation is
SEC.
4.3]
101
(4.99)*
2,(E')E'
= probability that a neutron will slow down from E0
to E without being absorbed (in a medium of pure
hydrogen)
If
This function may also be written in terms of the lethargy.

define u = In EB/E, then it follows from (4.99) that
p(M;0) = exp
\- ]o -^y-J
we
(4-100)
which gives the probability that a neutron of lethargy zero released

in a medium of pure hydrogen slows down to lethargy u without being

absorbed.
It can be demonstrated by means of a crude probability argument
that these results (4.99) and (4.100) are indeed the nonabsorption prob
abilities defined above. For convenience, we will carry out the calcula
tion in terms of the lethargy variable.
Suppose that the interval 0 to u
(which corresponds to the interval E0 to E) is divided into

Ai which are sufficiently small so that the
cross sections may be given one particular value
[2(m,),2,(w,), . . .] over each subinterval (see
Fig. 4.17). Then let p, denote the probability
that a neutron will pass through Am,- without
being absorbed.
But, each p, must be an inde
pendent probability;1
Ph(u;0)
[Jp-
i-i
m-i
(4.101)
the probability that a neutron is

absorbed in traversing Am^, then by definition
we call
ith interval
-n
(4-102)
T
-f-"w-i
pj1
Pi =
subintervals
_o_n
therefore,
If
./V
Fio.
4.17
Lethargy sub
intervals.
The absorption probability p? may be given as the product of the prob

ability that the neutron will make a collision in Am; and the probability
that this one collision will be an absorption.
Now it was shown in
Sec. 4.1e that Am is the average number of collisions made by a neutron
with hydrogen in traversing the interval Am. Thus if the subintervals
Aw; are very small, then the probability of more than one collision in each
interval becomes negligibly small. In the limit as the Aw, > 0, Am, gives
the probability that the neutron will make a collision in traversing this
subinterval. The probability that a particular collision at m, is an absorp
tion is simply 20(m,)/2,(m,) ; therefore, p? = AM,2a(Mi)/2((M,), and
1 The
probability of collision in any subinterval

history of the neutron.
Am cannot depend
upon the past
REACTOR ANALYSIS
102
[CHAP.
2/
AT
limit
4'
-4-Sr]
2("i)
Aw,
may be replaced by
lnpH(U;0)
Jq
>
0,
then the summation

an integral over the variable u, and we obtain
If
we pass to the
i-1
and note that for small Aui
^]
['-
ft
pH(M,0)
(4-103)
We compute the logarithm of
I1[1--1^J
i-l
Ph<.-0)
-^y-
called the resonance
escape
probabilities.
These
may be applied to more general media; an approximate

discussed in the next section.
4.4
functions
treatment
is
and p(u;0)
Slowing-down Density with Absorption
We turn now to
treatment of the more gen
in
down
medium
which contains absorbing
eral problem
For the present we consider a medium of a single nuclear
nuclei.
later analysis the problem of slowing down in
species and leave for
In this more general treatment both a and
mixture of nuclear materials.
of slowing
Integral Equations.
a.
q0(e~"
du' +
/_ln
r^
a
f"
u < In
S.(m') *(')("'-"
a)
2,(u') (u')(e"'-"
Jo
r-1
a
q(u) =
is
is
2a(E) are different from zero. As before, we assume that the nuclei
are stationary and that the neutron source (g0 neutrons per unit volume
monoenergetic (source neutrons have lethargy u0).
per unit time)
identical in form
The integral equation for the slowing-down density
=
If we take for convenience, Wo
then
to the relation (4.88).

is
in agreement with Eq. (4.100).

The functions Ph(E;E0) and pH(w;0) are defined for pure hydrogen,
and these are special cases of a more general pair of functions p(E;E0)
which
- a)du'
(a)
(4.104)
u > In
(6)
SEC.
4.4]
103
The collision
density in an absorbing medium F(u) differs from the

expression (4.89) by the addition of a term 2(w) (u) to the left-hand
side. Thus,
F(u) =
S(w)
*(u)
["
Jo
S.(m')
(u')
'- du' +
u < In
"
a
(a)
(4.105)
/u-lni
S.(m') (m') e"'-' cfu'

w

> In
(6)
The term at the left gives the rate at which neutrons are removed from
the unit lethargy interval about u by both scatterings and absorptions.
The terms at the right give the rate at which neutrons are added to this
interval by scattering in from above.
b. Solution in Asymptotic Range.
A general solution of Eqs. (4.104)
and (4.105) has not been found; however, an approximate solution may
be obtained which is valid for small absorption.
This approximation
is based on the assumption that the asymptotic solution applies for all
values of u > 0. We select then Eq. (4.104b) for the slo wing-down
This is one equation in the two unknowns q(u) and (u). As
density.
a second relation we introduce
^q(u)
-Z(M) *()
(4.106)
with the initial condition q(0) = q0. This relation is entirely satisfac
tory from physical requirements inasmuch as the only mechanism for the
loss of neutrons from the slowing-down stream in an infinite medium
is the absorption collision.
However, even with this simplification the
set of equations (4.1046) and (4.106) cannot be solved rigorously.
There
fore we introduce the next approximation, which is to regard the absorp
tion as small so that we have nearly the solution for the pure scattering
case.
Under these restrictions there can be no rapid changes in (u);
such changes could arise only from the high rate of neutron removals
due to the presence of strong absorbers.
Thus we assume that the
scattering density 2(u) (u) changes only slightly over each collision
interval u
In
to
w.
By making this assumption
it becomes pos
sible to evaluate the integral of Eq. (4.1046).

Accordingly we expand
in
a Taylor series about the point u and use the first two
2,(u') {u')
terms; then
2.(u') *(')
~ S.(m)
*() + (u'
u)
[Z.(u) *(u)]
(4.107)
REACTOR ANALYSIS
104
[CHAP.
The substitution of this expression into the integral of

upon integration,
q(u)
~ fS.(u) *()
7=1-
where
As
a first
"
and
2.(u) (u)
use this estimate of the
q(u)
This equation
The result is
(4-108)
(4.109)*
^ g()
derivative [2,(u)
2,(u) >()
(u)]'
[2.(u)
(")]
in (4.108); thus,
- 7 ^ ?()
may be solved by using the relation

q(u) = [2.(u) + 72.()]

*(")]
- a)
ln*
2(1
yields,
(4.1046)
approximation to (4.108)
g()
We
- t{ [2.()
(4.106) for q'(u).
{u)
(4.110)
which relates the slowing-down density q(u) to the flux (u) in a medium
of light absorption.
One of these functions can be eliminated by using
this relation in (4.106). If we solve for q(u), we obtain the expression
It
should be noted that as
^J
- Jq
q(u) = 3(0) exp
(4.111)
and 7 approach unity, and this

approximate
expression for the
0 both
slowing-down
density
approaches
the result for pure hydrogen
[see
Eq. (4.98)].
Equations
(4.110) and (4.111)

are applicable only when the term
72a(u) is small compared to f2,(u).
and 7 differ by
quantities of the order of 25 to 50
per cent of their values (see Fig.
Thus, although
4.18), it is often acceptable to as

sume that
q(u)
Fig.
4.18
Comparison of
and y.
7, so that
[2(u) + Z.()]
= 24()
()
(u)
(4.112)*
even though poor, is being made in a correction

in this work, it will be assumed that
term.
the flux and slowing-down density are related by Eq. (4.112); however,
since this approximation,
For much of the material
SEC.
4.4]
105
it should be borne in mind that (4.112) is valid only if the quantity

2,(u) (u) is slowly varying over any range of length In 1/ot in lethargy
and that the use of (4.110) in place of (4.112) may give better results.
Probability.
c. Resonance-escape
The nonabsorption probability for
a system containing some absorbing material is given by the exponential
factor in Eq. (4.111). In general for any interval u0 to u\, we define the
nonabsorption probability, or as it is commonly called the "resonanceescape probability," p(mi;mo) by
(4.113)*
This nomenclature has
[if
been adopted because the principal contribution

to the integral in (4.113) would
(4.113) were applicable in those
regions] come from the energy range of the absorption cross-section curve
Neutron energy
(a)
Fio. 4.19 Resonances in absorption

broad resonance.
cross
section,
(a)
Resonance structure;
(6)
is
is
which included the resonance peaks (see Fig. 4.19). Since the absorption
very large in the region of resonance peak, the require
cross section
ment that the scattering collision density be a slowly varying function
over a collision interval (In 1/a) will not be satisfied, and separate con
sideration must be given to the problem of evaluating the nonabsorption
probability in regions which include an absorption resonance. It would
appear, then, that the term "resonance-escape probability"
misap
plied to p(ui;u0) however, in many practical cases (4.113) may be used to
good approximation even in the presence of absorption reso
obtain
medium
nances, since the macroscopic absorption cross section of
a
particularly
large
consisting of several nuclear species may not exhibit
Morevalue at
resonance of any one of the nuclear types present.
a

a.(E)
REACTOR ANALYSIS
106
[CHAP.
over, the mere existence of a resonance in the absorption cross-section

curve does not invalidate, a priori, the applicability of Eq. (4.113).
The usefulness of this relation for estimating the resonance-escape
probability in an absorbing medium depends upon the "size" of the
The size of a resonance is determined by the maximum

resonances.
value of (Ta(E) at the resonance peak and by the width of the resonance.
The width of a resonance r is customarily defined as the distance across

(in energy units) the resonance at a height of one-half the maximum value
On the basis of a physical argument,
of the tr() curve (see Fig. 4.19a).
it is clear that, if the resonance is high and broad,1 many neutrons will be
captured in this energy range (see Fig. 4.196).
Consequently, the neu
tron density, and therefore the flux, will be highly depleted in this vicin
ity, and the assumption on which the result (4.111) is based would be
Thus (4.113) could not be expected to give a good estimate of
violated.
the resonance-escape probability in such a system.
On the other hand,
if the absorption resonances are "sharp," (4.113) may still be applicable.
The sharpness of a resonance is determined by the ratio A,/rf, where
Ai?i denotes the average loss in the energy of a neutron per collision
around the resonance energy 2J<. The resonance is sharp if this ratio is
By this definition the resonance at energy
much greater than unity.
Et in Fig. 4.19a is sharp inasmuch as A^3/r3 y> 1. The resonance at
E4, however, is broad because in this instance AEt/Tt ~ 1.
Equation (4.113) is applicable to systems which exhibit sharp reso
nances due to the fact that in these situations a large fraction of the
neutron population will never "see" some of the resonances.
So long
as the average jump in energy per collision at the resonance energy is
much larger than the resonance width, there is a strong likelihood that
the neutron will be scattered from some energy above the resonance to
some value below without ever occupying the energy band in which the
Thus many of the neutrons in the
large resonance cross section appears.
system will avoid capture by never having been exposed to the resonances
and thereby enjoy a greater opportunity to reach any specified energy
level below the source energy.
However, if the resonances are broad
in comparison to the average energy jump in the vicinity of the resonance
energy, no neutrons will succeed in jumping clear across the resonance in
one collision, and all neutrons will spend some time in the resonance
As a result, the probability for capture increases, and the
energy band.
resonance-escape probability for a system which possesses broad reso
nances will be significantly less than one for a system which has sharp
resonances.
are not meant to imply, however, that the result
(4.113) for the resonance-escape probability is rigorously correct for sys
tems which have strong but sharp resonances.
The intention here is to
These arguments
And the concentration of absorber large, i.e., when 2a is of the order of S,.
SEC.
4.4]
107
offer some justification for the use of this formula in systems which con
In this connection it is appropri
tain appreciable amounts of absorber.
ate to note that (4.113) takes the rigorously correct form in the limit as
Furthermore,
1, i.e., in a system of pure hydrogen [cf. Eq. (4.99)].
{
we know that the presence of a high sharp resonance in the cross-section
curve does not act as an impenetrable barrier to the slowing-down process.
Many neutrons will get through such a barrier in a single collision.
Equation (4.1 13) shows that, although the appearance of such a resonance
will decrease the resonance-escape probability (the integrand increases
1,000
c
h -1} \h
\
100
-i
'
'
\f\*
1
\ Tl ff
10
0.1
0.01
0.1
1.0
10
10*
10*
10*
10*
10*
10T
106
10'
(a)
1,000
n
A
--
loo
10
0.1
0.01
f
0.1
1.0
10
102
10J
10*
105
Energy, ev.
(b)
Fig.
4.20
Cross-section curve for a nuclear fuel:
(f>) fission cross section for U"s.
(a)
total cross section for
U"s;
limited by the presence of the 2(

term in the denominator of the integrand.
This result exhibits, then,
some of the features that would be required of such a function on simple
physical grounds.
Equation (4.113) should be most useful in analyzing
media with light absorption and may be extended to systems which have
as Z increases), the rate of decrease
is
sharp resonances.
It was pointed out previously that the preceding analysis was inappli
cable to media which contained heavy absorbers (2 ~S.), especially
large, broad resonances appeared in the absorption cross-section curves.
if

Energy, ev.
Unfortunately these are the materials of greatest interest to reactor tech

Reactor cores invariably possess relatively high concentrations
nology.
REACTOR ANALYSIS
108
[CHAP.
of fissionable material; these materials are characterized by absorption

cross sections which exhibit many large resonances (see, for example,
Thus, although some preliminary estimates can be made for
Fig. 4.20).
such systems with the aid of Eq. (4.113), other methods must be utilized
for accurate determinations of the resonance-escape probability in these
The principal difficulty in analyzing media with large absorption
cases.
that the assumption of a slowly varying collision density cannot be

We found that in pure scattering media the collision density
applied.
This is not the case, however, in absorbing media. As
varied as \/E.
the macroscopic absorption cross section increases, the collision density
departs further and further from the 1/E behavior, even in the absence of
If resonances appear, the collision density is further dis
resonances.
torted and significant dips appear in the vicinity of the resonances (see
Fig. 4.21).

is
Em
Energy
Fig.
4.21
e0
Distortions in collision density due to resonance
absorptions.
These features of the problem complicate the calculation of the neu

tron flux, and the integral equations for highly absorbing media cannot
be treated by the elementary methods used above.
In these instances
Thus much of
it is more practical to resort to experimental methods.
the information on the resonance-escape probability of the various mix
tures of nuclear fuels and moderators is presently available in experi
mental form.1 The resonance-escape probabilities in these media are
usually obtained by measuring the slowing-down density as a function of

energy for a given neutron-source distribution.
The ratios of these data
to the known source strength yield the resonance-escape probability [see,
for example, Eq. (4.111)].
An advantage which this method offers over a purely analytical
approach is that the details of the resonance structure need not be known
in order to determine the resonance-escape probability.
In view of the
1
A detailed discussion
of this subject is given in Sees. 10. le and 10.3.
SEC. 4.4]
109
considerable difficulties involved in making cross-section measurements in

the resonance range, we have here another argument for abandoning the
analytical approach in favor of the experimental.
d. Collision Density by Monte Carlo Method.
The analyses of the
preceding sections showed the inherent difficulties associated with the
computation of the neutron distributions in a slowing-down medium
Analytical solutions were obtained only
containing absorbing materials.
for the special case of a medium of small absorption cross section. Excep
tions to this restriction were allowed for media which possessed large
absorption
nances.
cross sections over short energy intervals, i.e., sharp reso

Even in these somewhat limited cases, it was necessary to
introduce several assumptions and approximations into the analysis in

The solutions obtained were limited,
order to obtain useful results.
moreover, to the asymptotic range of neutron energies [see Eqs. (4.110)

and (4.111)].
Two problems of importance to reactor physics which were not treated
in these analyses are the slowing-down distribution of neutrons in the
energy range close to the energy of the source neutrons and the general
These calculations
slowing-down problem in strongly absorbing media.
cannot be carried out by elementary analytical methods.
The experi
mental approach mentioned in connection with the calculation of the
resonance-escape probability (see Sec. 4.4c) is an alternate method for
treating these problems.

Unfortunately, in many instances it is as diffi
cult to perform good experiments as it is to carry out more exact ana
lytical studies.
A computational method which is eminently suited to the analysis of
the slowing-down distribution in strong absorbers and in regions close to
the source energy is the M onte Carlo procedure. 1 This method is essen
tially a statistical compilation.
The average behavior of a system of
many particles (or events) is predicted from a knowledge of a limited
number of individual histories obtained by a Monte Carlo calculation.
In some applications (e.g., the resonance-escape probability) the Monte
Carlo calculations can be performed "by hand." For many reactoranalysis
problems,
however,
extremely
large numbers of histories are

required (greater than 104), and these can be handled only by means of
It is not within the scope of this work
high-speed computing machines.
to examine the details of the method; however, we will have frequent
occasion to refer to results which have been obtained from machine
calculations using this method.
The Monte Carlo method has been successfully applied to a variety of
An application of interest to the present

reactor-physics
problems.
is
the
of
the
analysis
slowing-down problem in an absorbing medium
topic
1 See, for example, Herman Kahn, "Applications of Monte Carlo,"
Rand Corpora
tion, Santa Monica, Calif., AECU-3259, Apr. 19, 1954.
REACTOR
110
ANALYSIS
[CHAP. 4
A part of this treatment

carried out by Coveyou, Bate, and Osborn.1
was devoted to the calculation of the neutron-scattering collision densi
ties within the first two collision intervals below the energy of the source
The physical model on which these computations were based
was defined by the following set of assumptions: (1) neutrons slow down
in an infinite, homogeneous medium from a monoenergetic source of
strength q0 neutrons per unit volume per unit time, (2) the nuclei of the
medium are unbound and (3) are distributed in energy according to a
Maxwell-Boltzmann distribution, (4) the scattering cross section is inde
neutrons.
pendent of energy, and (5) the absorption cross section is inversely pro
portional to the relative speed between the neutron and the nucleus.
These assumptions are similar to the set used in Sec. 4.2c. The principal
is
(if
difference appears in 3 and 5. Assumption 5 is a convenient specification

of the absorption cross section; however, it does have some physical sig
nificance since in general ca{v) ~ 1/v
one omits the contributions from
an attempt to describe the thermal
Assumption
the resonances).
motion of the nuclei in the medium; further remarks on this phenomenon
it
is
is
is
this decision
scattering
is
if
if
is
the event
an absorption, a new neutron
introduced at the source
the event
a scattering,
speed and the procedure repeated; (4)
decision must be made regarding the speed of the target particle and the
relative angle of approach of the neutron to the target.
Having made
(3)
is
it
then a simple matter, on the assumption of isotropic

in the center-of-mass system, to select the new speed of the
(5) these steps are repeated until an absorption collision occurs,
is
it
neutron
thus terminating a history. By tracing many such histories
possible
to compute the average number of the various events (such as scatterings
and absorptions) per energy interval for all energies per neutron intro
Thus the collision densities, neutron flux, and
duced into the system.
the slowing-down density can be obtained.
Note that the resonance-
R. R. Coveyou, R. R. Bate, and R. K. Osborn,
/.
Nuclear Energy,
2,
is
is
probability
an especially easy quantity to compute by this
the number of neutrons which survive to
method; clearly, p(E)
divided by the total number of histories examined.
specific calculation for the scattering collision density was carried
out for the first two collision intervals using a slowing-down medium of
escape
153-167,
These decisions are made by

random numbers.
(1956).
*

is
will be deferred until Sec. 4.7.

The application of the Monte Carlo technique to the present problem
may be briefly outlined in terms of the following steps: (1)
neutron
introduced into the system with preassigned source speed (2) decision
then made as to the nature of the first event to be experienced by the
to be absorbed or scattered, and by what species;2
neutron, i.e.,
suitable procedure which involves the selection of
SEC. 4.4]
111
A = 9. The results of this calculation are presented in

The ordinate is the function ~L,{x) <p(x)/q0 where
mass number
Fig. 4.22.
Sv
=
Vt
Vt
m
/:U-7'xA*
(4.114)
v is the neutron speed, m the neutron mass, Tn the moderator

tem
perature, k the Boltzmann constant,1 and 8 = 16 (a convenient scale fac
tor). Thus the collision density is given as a function of a speed ratio.
The source condition for this computation was taken as xo = 512. The
and
open circles in the figure represent the points computed by the Monte
The discontinuity in the collision density at the lower
Carlo method.
400
Fig.
420
460
440
480
Scattering collision density in
4.22
500
1 /v
J 520
absorber.
interval2 x = x\ is clearly discernible (cf. Fig.

The solid-line curve is the solution for the collision density in the
end of the first collision

a
This function
1/t absorber.
9o
(1
where
Tt a 293K,
- a)x,
fxV
\ZTN)
\N.)
\x0)
\x
is
+ KiV* -a)/(l-a)
+ k5
(4.115)
first interval for
fx0
4.15).
(4.116)
a.
is
the nuclear
is
density of absorbers, and N. the

nuclear density of scatterers.
It should be noted that the parameter
=
measure of the absorption strength, and
corresponds to the case of
pure scatterer; whence,
and
9o
It
)/(l-o)
{x)
(1
Z.(-r)
(4.117)
a)j.-0
is
is
is
easily shown that this function

the direct transformation3 of expres
sion (4.78) when the variable
transformed to the variable x.
The
= 2-6,
results for the Monte Carlo calculation were carried out for
which corresponds to a medium with relatively low absorption.

X 10-,6erg/K
= 8.6275
lO"5
is
Jo, where a
given by Eq. (4.20).
By using a relation of the type (3.43).
Xi =
a'
>Jfc = 1.3804

380
ev/K.
The
REACTOR ANALYSIS
112
[CHAP. 4
method is not limited, however, by the value of k, and the flux distribu
tion in more highly absorbing media can be obtained as readily as in the
It is seen from the figure that the Monte Carlo data showpresent case.
some scatter; this is due to the poor statistics in the high-energy range.
The essential features of the collision density in this range are well
exhibited, nevertheless.
4.5
a. Asymptotic
slowing
Mixtures of Nuclear Species
Equations.
down of neutrons
from a monoenergetic
consider the
source in an infinite
we
uniform mixture of several nuclear species.

The
calculations which follow apply to the asymptotic range (well below the
We noticed in our previous calculations
energies of the source neutrons).
for mononuclear media that this energy range was the most important
for the analysis of thermal reactors.
Therefore, even though we ignore
the details of the neutron distributions in the vicinity of the source
energy, the results we obtain will be of some immediate value for reactor
medium
which
In the present section
is a
calculations.
The appropriate
equations for describing the neutron distributions in

a medium of several nuclear species are obtained from a generalization of
Eqs. (4.104) and (4.105). The only change which is required in these
results involves the integral term which gives the rate at which neutrons
are slowed down into the lethargy range of interest from collisions above.
For the mixture, this term must include the contributions from collisions
with all the various nuclear types present; it

scattering integral for each nuclear species i.
N species)
is necessary, then, to add a

define (i = 1,2,
,
...
If we
Si" = macroscopic scattering cross section of
iih
nuclear species
where At = nuclear mass of ith nuclear species

then it follows that for the asymptotic
range
u > In
q(u) =
2,(m) 4>(u) =
2>(u')
(') (e"'-u
- 0 du'
2i'>(w') 0(w') e"'- du'
(4.118)
(4.119)
Equation (4.118) is a statement of the physical fact that each neutron

which slows down past lethargy u has made its last scattering collision
with some particular nuclear type at some lethargy u' < u. The prob
SEC. 4.5]
113
ability that these collisions increase the lethargy of the neutron to a value
above u must be calculated separately for each nuclear type; finally, the
combined contribution from all scattering collisions must be summed up
with the length of the scattering interval (In l/a<) varying from one
Similarly, Eq. (4.119) is a balance condition for the
nucleus to another.
u.
lethargy
unit interval around
In the special case wherein all nuclei
b. Mixture of Pure Scatterers.
in the mixture are pure scatterers, Eqs. (4.118) and (4.119) have a simple
The integral equation (4.119), with
solution.
[so
Si" n 0
has the solution
are
where the
C,
2<()*(u)
= C,
set of constants.
2,
that 2,
iThe
t-i
1,2,
use of
Z]
,N
this result in the integrand
of (4.118) gives
N
i-1
"'
If
we define the average
fc2>(u)
(u)
(4.120)
AT
for a mixture by
N
(4.121)
Eq. (4.120) may
be
written
q(u) = ?(")2,(w)
*() =
(4.122)
<l(0)
is
is
is
The last equality follows from the physical requirement that, in the
absence of absorbers, q(u) = q(0).
mixture of pure scatterers
treated, then, as a single scatterer, with the quantity
replaced by the
This definition
average defined in (4.121).
plausible from
physical
standpoint since
effective in slowing down neu
given nuclear species
trons only in proportion to the fraction of the scattering collisions which
It should be noted that, since 2i')(u) lethargy
are due to that species.
is
mixture of pure scatterers
is
The neutron flux for

from (4.122).
is
tends to be less
dependent, (w) must be also; but the dependence of
Thus
particular
sensitive because of its functional form.
usually
chosen to represent the entire lethargy range of interest.
a

JV
obtained
directly
ANALYSIS
REACTOR
114
[CHAP. 4
9(0)
\{u) 2. (it)
Note that if one assumes that all the scattering cross sections have the
same functional dependence, then
|(u) =
= constant
(4.124)
and we obtain for the flux
this
case of
independent of lethargy;
of course, the assumption
that all
then
.V
are
(4.126)
The simplest
is,
2.(it)
[2<()
yields
i=
*()
fcZ>(tO
q(u)
If
the derivative term in (4.128)

mation
is
y, ;m
where
-^
ln2
The substitution of this expression into the integrand
[2'()
*()]
(4.127)
of Eq. (4.118)
*()])
a<
(4.128)
(4.129)
neglected, we obtain as
~ 2() *(u) + ('
*(')
u)
2<"(')
use the approximation
first approxi
We
use this result in (4.128) to
obtain
9() =

is
also a constant (u) = 0O = </(0)/2.

The flux
In the presence of absorption, Eqs. (4.118)
c. Mixture of Absorbers.
and (4.119) do not have simple solutions, and even the limitation that
2 <K 2, for all the nuclear species present does not lead easily to a solu
If we introduce this assumption, then the variation in the scatter
tion.
ing-collision density over a collision interval should be small, and we can
>-l
f,2''>(u)
*(M)
"
7"(u)
^9(M)
(4130)
SEC. 4.5]
[Z() (u)]' f,7.
y(u) = <~
where
(4.131)
[S*(u) *()]'
As in the calculation for the mononuclear

{u) are related by
=
^q(u)
115
case, we assume
that q(u) and
-2(u) *(u)
and substitute this relation into (4.130).
(4.106)
Thus
q(u) =
fc2>()
*() +
2.() *()
7(m)
=
where we have used (4.121).
[{(u) 2.(u) + y(u) 2(u)] 0(a)
Again the approximation
is often used
1()~7(M)
so
q(u) = |(w) 2,(w)
that
(4.132)
that
(4.133)
()
(4.134)*
This relationship between q(u) and 4>(w), which seems to be related to the
accurate equations only through several very limiting assumptions, has
been used in many calculations on actual reactors with very good results.
It will be generally assumed throughout this work.
An explicit expression for the slowing-down
substituting (4.134) into (4.106). The solution
q(u) = ?(0) exp
wherein
absorbing
the exponential
medium.
[L
density
is
obtained
by
is
- P
(4.135)*
Jo {(') 2,(u')J
gives the resonance-escape
probability in an
It should be recognized that if it is desired to use the relation

an actual computation, instead of the simpler form (4.134), it is
to carry out an iterative procedure.
This complication arises
definition of y(u) which involves the neutron flux [see Eq.
(4.132) in
necessary
from the
(4.131)].
This difficulty may be overcome by following the procedure outlined

below: As a first approximation use Eq. (4.134), and call the correspond
ing expression for the flux i(u); thus,
Mu)
*(M)
*(0)
exP[-
f MO*q
(4.136,
from (4.121).
This expression for the flux may
now be used to compute the first estimate 7i(u).
We obtain from (4.131)
where
%(u)
is computed
Yi()
W _
[2<<>(u)
second estimate of
From (4.132),
gn(")
Z.(u) + 7i() Z.(u)
,.A
g(0)
Jo
X exp
f(0 2.(uO
f.(')2.(')J
g() 2.(u) + 7i(w) S.(u)

^
for
(4.137)
011
be used
[CHAP. 4
*,()]' f.T*
[2<*(u)
i-l
This relation may
ANALYSIS
<.
REACTOR
{,
116
In view of the numerous assumptions imposed on the preceding analysis,

the refinements suggested above are somewhat questionable and the con
siderable labor required in performing the iteration seems hardly justified.
is
is,
d. Moderating Ratio.
It was shown in Sec. 4.1 that the effectiveness
of a material as a slowing-down medium was determined in part by its
value of
the average gain in lethargy per collision.
Another factor
which can affect this characteristic of a substance
of course, its macro
scopic scattering cross section 2,. The product of these two parameters
gives an indication of the moderating properties of
This
material.
quantity
sometimes called the moderating power:
= moderating
2a
power
(4.139)
is
Thus the larger the moderating power of material, the greater its effec
tiveness as a slowing-down medium.
This number, however,
not a
complete measure of the slowing-down effectiveness.
better quantity
the moderating ratio, which takes into account the adverse effect of
The definition of this quantity
high-absorption cross section.
is
is
'
= moderating
ratio
(4.140)
The influence of this parameter on the slowing-down density may be

demonstrated by means of
simple example.
Consider two uniform infinite media of dissimilar composition.
It
is

The substitution of this result into (4.131) yields the second estimate
7u(w), which in turn gives m(u), and so on. This procedure should
converge quite rapidly. As pointed out previously, however, for most
practical systems of interest, Eq. (4.134) should be entirely adequate.
immaterial whether these media consist of mononuclear materials or are

uniform mixtures; the following arguments apply in either case. Let q0
monoenergetic neutrons per unit volume per unit time be introduced into
SEC. 4.5]
BLOWING-DOWN
each of these media, and let u

source neutrons.
and (4.135)] we
PROCESS
*()
to the energy of these
0 correspond
From our previous analyses

know that
117
[see Eqs.
(4.111), (4.113),
9.Pl(;o)
= q0pi(u;0)
q2(u)
where the resonance-escape probabilities p<(u;0) are denned
- exp r
(
<w
p.(;0)
["WW
du'~\
Jq
....
(4.142)
For convenience we define an average value of the integrand such that

p,(w;0) = exp
If
we call m, = (2,/20),-,
g2(u)
(4.143)
= e-u(l/m,-l/m,)
(4 144)
for comparing the moderating effective

Clearly if mi > m2, then qi(u) > qi{u), that
more neutrons succeed in slowing down to lethargy u in medium
than
in
Note that the parameters wi, are essentially the moderating ratios
a useful criterion
2.
is,
ness of two materials.
defined by (4.140), since for slight absorption
<5C
2 and
we have
|2.>
2a,
Thus (4.144) states that qi(u)
2.
if
> q2(u)
the moderating ratio of medium
Good moderating materials, then,
greater than that of medium
It should be recognized
possess large
(small A), large 2,, and small 2.
that, in using the moderating ratio as a criterion for moderating effective
necessary to establish unambiguously the cross sections which
ness,
enter into the calculation.
It
evident from (4.143) that the proper
is
is
it
is
suitably averaged
cross sections to use are

of interest.
set over the lethargy
In many instances, however,
range
is
a rough estimate
sufficient,
has become customary to compare moderating media in terms of
these properties at thermal energy (0.025 ev).
The moderating ratios at
and
it
thermal energy of some efficient moderators are listed in Table 4.1.

Table
4.1
Material
Moderating
Ratios
Moderating ratio
H,0
164
D20
13,200
Be
144

then from (4.141)

9'(M)
This ratio gives
236
[CHAP. 4
REACTOR ANALYSIS
118
4.6
Infinite Homogeneous
Multiplying Media
Arbitrary Source Distribution. Our considerations of the slowingdown problem so far have been limited to infinite media with monoa.
energetic sources. In an actual reactor the neutron sources are in main

part the fission reactions which take place; these source neutrons appear
in the system according to a particular distribution in energy, the fission
A somewhat more realistic analysis of the slowing-down prob
spectrum.
lem can be obtained by taking this source distribution into account.
We
begin the calculation by setting up the following general problem: (1)
neutrons appear in the medium isotropically at the rate S(u), where

S(u) du is the number of neutrons which are born per unit volume per
unit time with lethargies in du about u; (2) the nuclei of the medium are
The flux
stationary; and (3) the medium is isotropic and homogeneous.
density are assumed to be related by the equations
and the slowing-down
^ q(u)
q(u) =
-20() +() + S(u)
(4.145)
&,(u) *()
(4.112)
any change in q(u) must be due to the combined effect of the removals
S(uJ
Fig.
4.23
Source
function S(it).
Equation (4.112) is taken from the

and of the source S(u).
analysis of Sec. 4.4b and therefore applies to the asymptotic range and to
media with slight absorption.
These relations replace the more general
form (4.104).
Equations (4.112) and (4.145) are to be used with the
2(m)
4>{u)
initial conditions
5(0) = 0
5(0) = 0
(4.146)
which imply that the lethargy scale is so selected that the origin is well
above all neutron sources (see Fig. 4.23) ; thus no neutrons appear in the
system at energies above the value which corresponds to u = 0. The
solution to (4.112) and (4.145) is readily shown to be
9() =
/"

Equation (4.145) is an immediate extension of (4.106) and states that
S(u'Mu;u') du'
(4.147)*
SEC. 4.6]
BLOWING-DOWN
where
p(u;u>)
PROCESS
exp
119
[-
(4.148)
The physical statement of (4.147) is that the number of neutrons slowing

past lethargy ?i (in an infinite medium, per unit volume per unit time) is
given by the number of neutrons produced at u', S(u') du', multiplied by
the probability p(u;u') that a neutron at u' will reach u without being
absorbed and the product integrated over all u' < u [for which S(u') has
a nonzero value].
A special case of (4.147) is the medium

b. Monoenergetic Source.
with a monoenergetic source; thus, suppose that all neutrons enter the
medium with lethargy
Then, we take
u = 0 (corresponding
to some fixed energy E0).
S(u) = q5(u)
(4.149)
Note that the total number of

where 5(u) is the Dirac delta function.1
neutrons which appear at all lethargies per unit volume per unit time is
du =
jo"
SM
S(") du =
(4.150)
(') p(u;u') du' =
q0
exp
jf"
qB
9(u) =
jf"
is
is
obtained by the direct substi

The slowing-down density in this system
tution of (4.149) into the general result (4.147). The solution
(4.151)
previously obtained [cf. Eq. (4.111)].

Fission Source.
As a second special case of Eq. (4.147), suppose
uniform distribution of
that the infinite medium in question contains
fissionable material, so that the nuclear characteristics of the composition
are now
2,(w), Xf(u), 2(w), and Z,(w).
Suppose further that, as
result of each fission,
neutrons appear at the point of the fission reaction
and that these neutrons are distributed in lethargy according to the func
tion j(m), where
as
Thus j(w)
i(u) du =
f~
{,
c.
(4.152)
is
is
the normalized fission spectrum in lethargy.

Note that
|(u) du gives the fraction of neutrons which appear from fission with
lethargies in du about it.
It sometimes more convenient to work with
the corresponding
1

/_".
given by the integral
function in energy space z(E), which may also
be
H. Margenau and G. M. Murphy, "The Mathematics of Physics and Chemistry,"

Company, Inc., Princeton, N.J., 1943.
p. 325, D. Van Nostrand
REACTOR ANALYSIS
120
normalized;
[chap. 4
The fission spectrum
dE =
i(u) du
(4.153)
z{E)
f~
is fQ"
thus,
obtained from experiment.

Figure 4.24 shows
data along with an analytical approximation.
some experimental
One of the initial assumptions (2) was that the nuclei were stationary
thus tbe neutrons may slow down all the way to u =
The source
1.000
750
500
Fig.
=
E, Mev
energy
Time of flight data X 0.204

Photographic-plate data
Fission spectrum in energy for U*"; indicated formula

good up to
Mev.
[From L. Cranberg, G. Frye, N. Nereson, and L. Rosen, Phys. Rev.,
is
4.24
9
3
}
Neutron
103(3): 662 -670 (1956).]
Jo"
therefore
S(u)
is
function for this system
The slowing-down density may
2,0')
4>{u')
be obtained
du'
(4.154)
by using this expression in
2,(u") *(") du"
j(M') p(u;u') du'
(4.155)
to be the only source of neutrons, and

the neutron flux in the infinite medium
to be in the steady state, we
require that
is
if
is
the fissionable material
fQU
q(u) =
f0"
(4.147).
If
du
Jo"
statement
fQ"
S(u) du
The equivalent mathematical
produced from fissions per

\unit volume per unit time
is
volume per unit time/

(Number
per unit
introduced
/number of neutrons
of neutrons
fo"

250
2/(w')
(u')
du'
(4.156)
If
PROCESS
BLOWING-DOWN
per unit volume
we introduce one neutron
du
Jo S(u)
1
f"
1,
121
per unit time,
then
and
j(tt) du
f* 2,(u')
{u')
du' =
fQ"
SEC. 4.6]
2,(u')
du'
4>(u')
(4.157)
The equality at the right states that the system

(4. 152).
to be at steady state then exactly one neutron per unit volume per unit
time must be produced as result of all the fissions at all lethargies in that
volume.
The application of this result to the relation for q(u) given in
a
is
if
by virtue of
(4.155) yields
/"
()
J(u') p(u;u') du'
(4.158)
and for the flux we have [from (4.112)]
f"
Jo/
jzm
f2<(w)
*("') p(u;u') du'
(4.159)
('
"
Jo
For convenience
^7^r
f'lW ^)u')
f2,(u) Jo
(4-161)
fi(w)
dw~]
j(m') exp
y(u) du
Jo"
(4.160)
Ku) s W&j
|fe)
f*
Then (4.160) gives

=
du'
we define1
y{u)
du'
(4.162)*
and
is
condition,
Condition (4.162)
requirement
is
no thermal group.
it
tion, there
is
is
the condition for the steady-state operation of a multiplying

medium with only fission sources of neutrons and slowing down allowed
over the infinite lethargy range; i.e., in the terminology of the next sec
This
called the criticality
on the nuclear characteristics
of the
is
it
multiplying medium.
We will have more to say of this later.
The fission spectrum z(E), foi the particular fuel Um, has maximum
Mev. For many purposes
adequate to group all
value at about
neutrons from fission at some one energy in the neighborhood of the posi
The same can be done with j(u). If for

in the vicinity of the maximum point (see
convenience we choose u =
approximated by monoenergetic
Fig. 4.25), then the fission spectrum
source at u =
be confused with
j(u) = *(u)
Not to
0.
is
tion of the maximum point.

The expression for the steady-state condition

d. Criticality Equation.
(4.157) may be written in terms of the result (4.159); thus
from (4.109) or f(u) from (4.131).
(4.163)
ANALYSIS
REACTOR
122
[CHAP. 4
y(u) exp
y(u) p(w;0) du
Jq"
Jq"
- fjp(u') du'] du
(4.164)*
Since the cross sections which appear in (4.162) and (4.164) are generally
obtained from experiment, simple analytical representations usually are
not available for these quantities, and the evaluation of the integrals must
be done numerically.
physical interpretation of the

integrals in (4.164) may be demon
strated by writing the integrands in
the form
Fig.
Fission spectrum in lethargy.
4.25
That
fraction of
collisions
at u which
successfully
|_are fissions_
reaches lethargy
|_w
_time at u
neutron at u
probability that
Number of neutrons
produced per unit
volume per unit
average
average
number
number
collisions
neutrons
_in du at u_ _per fission_
y(u) du
j(w') p(u;u') du'
f*
fQ"
km
is
is
There
an alternative way of regarding (4.160) which may be more
given generation
instructive. The ratio of the number of neutrons in
to the number in the preceding generation
the multiplication constant
[see Eq. (3.16)]; then
(4.165)
is
is
where kx
defined as the multiplication constant of the infinite medium.
If the composition of the medium specified, then all the cross sections
are known [and therefore y(u) and /3(w)] and the integrals may be evalu
is
if
is
is
y{u) du
Jo"
|Q"
r.
j(u') p(u;u') du'
is
supercritical.
greater than unity, the reactor
Alternatively, whatever the value of the right-hand side, some number
may be found such that
value
ve
unity, then the system

the right-hand side of (4.165)
subcritical and
its
its value
less than unity, the system
is
critical
if
If
ated.
is

is
is,
the number of neutrons pro

duced per unit volume per unit time (which must be unity)
the sum
over all lethargy intervals du of the product
(4.166)
SEC. 4.6]
123
i.e., for any composition of the infinite medium, there exists some (ficti
tious) average number of neutrons per fission which would make the sys
tem critical.
This value is given the symbol vc and is to be compared

with c, the experimentally determined number of neutrons from fission.
It follows from (4.165) and (4.166) that
(4.167)*
e.
Slowing Down by Average Jumps.
The actual slowing-down proc

consists of a series of discrete jumps in its
ess experienced by a neutron
kinetic energy; thus the neutron trajectory in energy space is a discon

tinuous function with a random distribution of "steps" (see Fig. 4.10).
The magnitude of these steps (or jumps) is given by the frequency func
tion for the elastic-scattering collision [see Eqs. (4.39) and (4.50)]. In
each of the preceding analyses an attempt was made to take this behavior
Only in the case of a hydrogenous medium, however, was
into account.
it possible to obtain exact solutions.

An alternative analysis of the slowing-down process, which may
be more
attractive in an intuitive sense, treats this process as a sequence of

"average jumps." In this treatment the actual frequency function is
replaced by a function which allows the scattered neutron to acquire
the average gain in lethargy per collision { [or lose the average
in
change
energy AE per collision; see Eq. (4.40)]. The neutron trajec
tory in lethargy space, then, is a step function with a series of jumps, all
of magnitude .
This approach to the slowing-down problem is at best a
first approximation ; however, it does have some merit in that it retains a
exactly
number of the essential features of the physical phenomenon, although in

crude form.
Moreover, it allows a solution of the exact equations [see
Eqs. (4.105)].
The average-jump collision process described above requires that the
initial and final lethargies of a scattered neutron (w' and u, respectively)
be related by the equation
u = u' +
The frequency function for

average jumps is defined
l(u;u') du
a slowing-down
probability that
(4.168)
process consisting
of these
neutron of lethargy u' when

scattered emerges from the collision with lethargy in
du about u
= S[u'
- {u
{)]
du
(4.169)
where 5 is again the Dirac delta function. The physical statement of this
function is that the scattered neutron will emerge from the collision with
lethargy in du about u only if u differs from u' by exactly the amount .
REACTOR ANALYSIS
124
[CHAP. 4
As an example of the application of these concepts, consider the analysis

of the slowing-down problem in an infinite, homogeneous multiplying
Let the distribution of
medium which is at steady state, i.e., critical.
fission neutrons be given by the normalized spectrum j(m) defined in
Eq. (4.152), and let 2(u), 2,(w), 2((m), , and 2,(u) denote the nuclear
constants of the medium.
The total collision density for this system is
given by
(u)
du =
2,(u')
4>(u')
I(u;w') du' du
/"
2,(u)
+ "i(")
2/(u') 0(u') du' du
(4.170)
2/(m')
}()
fo"
2.(u')
+ 2,(
(')
du'
(4.157)
with the definition (4.169), then

S[u'
4>{u')
)*(
- (
used in (4.170), along
2,(u) *() = j(u) +
{)]
du'
(4.171)
is
0,
a difference equation1 in the
If
we
unknown function
=
<
for all u
define the origin of the lethargy scale such that j(u)
then the appropriate initial condition to be used with (4.171)
is
This relation
f~
{)
this relation
(4.172)
<K0)
- we
*<
Thus, for example,
if
The solution to the difference equation may be carried out in steps by

starting with the known value of the flux at the origin (4.172). In
denotes the flux at the lower limit of the nth collision
general,
interval, then
[.<ae
?fjP]
(*')
can be
and so on. An approximate solution to the difference equation

Thus we expand
obtained in the case of
slowly varying flux.
" Mathematical
Theodore v. Karman and Maurice A. Biot,
Methods in Engineering," chap. XI, McGraw-Hill Book Company, Inc., New York,
1

If
is
is
The first integral gives the rate at which neutrons are scattered into the
interval du about w, and the second, the rate at which they appear from
fissions.
The system described by (4.170)
at steady state; therefore,
the criticality condition must be satisfied:
See, for example,
1940.
SEC. 4.7]
*(u .- 8
*()
- '()
125
(4.175)
and use the first two terms on the condition that {<'() <3C {u). The
substitution of this expression into (4.171) reduces the difference equation
The result is (2, is taken to be inde
to a first-order differential equation.
pendent of lethargy)
2,(u)
()
as j(u) + 2.*(u)
{2.
'()
(4.176)
which has the solution
It
is seen that this result is in good agreement with the relation (4.159).
Note that the resonance-escape probability, given by the exponential of
(4.177), is defined in terms of the scattering cross section, whereas the

form in (4.159) involves the total cross section. The discrepancy here is
not significant however, inasmuch as (4.159) applies to media with slight
This is also the case in the present calculation ; otherwise we
absorption.
could not have assumed that the flux was a slowly varying function over a
single collision interval.
4.7
Thermal -group Cross Sections
In all the preceding

a. Thermal Motion of Neutrons and Nuclei.
analyses dealing with the interaction between neutrons and nuclei it was
assumed that the nuclei were stationary. Actually, the nuclei in any
material at a temperature above absolute zero will be in thermal motion.
It is known from the kinetic theory1 that the average energy associated
with the translational motion of such particles is given by f kT (where k is

Boltzmann's constant and T is the absolute temperature of the medium).
So long as the neutron speeds are much greater than those of the nuclei
(i.e., the neutron energy is much greater than %kT), there is no important
effect of the nuclear motion, and this is the case for materials at ordinary
temperatures (20 to 1000C) and neutrons with energies of a few electron
However, when neutron speeds are comparable to
volts or greater.
nuclear speeds, the nuclear motion cannot be ignored.
Nuclear motion becomes important, then, when neutron energies are
in the order of kT {kT = 0.025 ev for T = 20C).
The effects of the
nuclear motion are significant in connection with two different aspects of
the interaction phenomenon, namely, the frequency function for scatter
ing collision and the specification of neutron cross sections for neutron
It will be recalled that the frequency functions for
energies around kT.
for example, J. E. Mayer and M. G. Mayer,
1-17, John Wiley & Sons, Inc., New York, 1948.
1 See,
"Statistical Mechanics,"
pp-
REACTOR ANALYSIS
126
[chap. 4
collision used in the previous calculations [see Eqs. (4.39) and (4.50)] were
defined for stationary nuclei; these functions do not apply, therefore, to
scattering collisions by thermal neutrons (neutrons of energy around kT).
In fact, a scattering collision experienced by a thermal neutron will result
in an increase in its kinetic energy in a large fraction of the collisions.
Because of the thermal motion of the nuclei, some energy can be trans
ferred from the nucleus to the neutron, thereby increasing the neutron
This is in contrast to collisions between neutrons and stationary
speed.
nuclei which always result in a decrease in the energy of the neutron (and
an increase in the energy of the nucleus).
Elastic-scattering collisions between thermal neutrons and nuclei do

not, on the average, raise or lower the kinetic energy of the neutron.
Once a neutron has been slowed to the thermal-energy range, it wanders
through the medium from one scattering collision to another, sometimes

gaining energy, sometimes losing energy, until it is finally absorbed (or

The effect of the
escapes from the medium if the system is finite).
nuclear motion on the slowing-down process is to decrease the average
logarithmic energy loss (decrease ) as the kinetic energy of the neutron

decreases toward kT; thus, the neutron actually makes more collisions in
a small energy interval (a few kT) above E = kT than would be pre
dicted by a theory which treated as a constant in energy.
The use of a
then, underestimates the number of absorptions in a small
energy interval at low energies, just where the absorption cross sections
are usually high; this may, in part, be compensated for by the fact that
constant
the number of fissions in the same interval is also underestimated.

The second effect of the nuclear motion appears in the definition of the
neutron cross sections for thermal neutrons and the specification of the
In the case of
neutron and nuclear densities as a function of energy.
media with stationary nuclei the cross sections could be readily defined
inasmuch as the relative speed between neutron and nucleus was the
absolute speed of the neutron.
When the nuclei are in motion, however,
the neutron cross sections must be expressed in terms of the relative speed
Thus in computing collision densi
between the two moving particles.
ties for thermal neutrons and nuclei it is necessary to use the relative
speed, and this requires a complete description of the motion of both
neutrons
and nuclei.
In general, these motions are extremely compli
cated, and detailed descriptions are difficult to specify and even more
difficult to apply.
The behavior of neutrons at the lower end of the energy scale is much
Because of
different, then, from the behavior in the high-energy range.
the basic difference in the nature of the two phenomena, slowing-down
and thermal motion, it is convenient to separate the neutron trajectory
in energy space into two parts.
The first part deals with the slowing-
down process from the source energy down to the thermal range, and the
SEC. 4.7]
127
second, with the behavior of the neutrons around E = kT.

The slo wingdown process has been studied in detail in the preceding sections, and
various models for treating this portion of the neutron trajectory have
The problem which remains, and to which we now

address ourselves, is the selection of suitable functions for describing the
energy distribution of neutrons and nuclei in the thermal-energy range.
been developed.
Given these distributions, expressions for the nuclear-reaction rates can

be readily obtained.
To a fair approximation the physical situation in the thermal-energy
range may be regarded as that of two gases (neutrons and nuclei) in
thermal equilibrium. On the basis of this picture it is possible to derive
by the methods of the kinetic theory the functional form of the distribu
tion functions for both the neutrons and nuclei.
The present derivation
of these functions follows the treatment of Chapman and Cowling.1
Consider, then, an infinite medium consisting of neutrons and nuclei in
thermal equilibrium, and for convenience let us assume that only one
nuclear species is present.
Further, let us assume that only scattering
reactions occur between the neutrons and nuclei so that no neutrons are
lost from the system by absorption.
Let the distribution of neutrons
fracôn f neutrons in unit volume whose velocities
lie between v and v + dv
fraction of nuclei in unit volume whose velocities lie
at? cm /W
=
{ '
between V and V + dV
V
n(v) dv
(4.178)
The derivation of the functions m and 2J? is developed from the neutronbalance relation for all reactions which can affect the density of neutrons
in a differential element dv. For this purpose we require two additional
functions, namely,
g
= v
V = relative velocity between neutron and nucleus

scattering cross section for a
neutron of velocity v and a nucleus of velocity V which will rotate relative velocity vec
tor p through angle u
=
with p =
V|,o>) dto = microscopic

<r(|v
|p|
collision
g'.
is
is
typical scattering collision in the laboratory

shown in Fig. 4.26.
The unprimed quantities
system of coordinates
refer to the velocities before collision, the primed, after collision.
As
indicated in the definition above for a, u the angle between the relative
velocity vector before collision
and the relative velocity vector after
The vector diagram for
Sydney Chapman and T. G. Cowling, "The Mathematical Theory of Non

uniform Gases," 2d ed., pp. 374-377, Cambridge University Press, New York, 1952.
1

and the nuclei be given by the functions
REACTOR
128
We begin the neutron-balance
ANALYSIS
[CHAP. 4
by computing first the total
calculation
number of neutrons scattered out of the differential element dv. Since

the speeds of the nuclei are comparable to those of the neutrons, it is
necessary to compute the scattering densities in terms of the relative
Thus for each nuclear

speed between each pair of colliding particles.
velocity V there is a corresponding relative speed p and a cross section
In order to compute the total scatterings out of dv we integrate
a(p,u).
the product of m(v) dv and
over all V and
p<t9J?
u>;
thus,
Number of neutrons
scattered out of element =
dv per unit time
Jv
m(y) dv
po{p,u>)
9D?(V)
dV da
(4.180)
In order to complete the neutron-balance
relation for steady state, we

require also the rate at which neutrons are scattered into the element dv.
For this purpose we introduce the density of nuclei
of any velocity
before collision 3K(V) and the
density of neutrons of some velocity v' before colli

sion m(v').
It is important to recognize that the
velocity v' is not entirely arbitrary since given V,

a scattering angle
and a final velocity v there
exists only one velocity v' which the neutron may
possess if the neutron is to emerge from the collision with velocity in dv about v.
Or saying it in

',
-v,
',
Fio.
Velocity
4.26
vectors before and after
collision in a laboratory
system of coordinates.
another way, given V,

and v, v' is completely
defined
This mayJ be seen from the figure.
Thus
"
to compute the total number of scatterings into dv
we sum over all possible
and
to'
v' =
and select each v' such that
v'(v.W)
(4.181)
causes the emergent neutron to appear with velocity in dv about v.

sum is given by the integral
J[V',<o'
m[v'(v,V>')] dv' p'a(pV) SW(V) dV'
d*'
This
(4.182)
Note that
a(p>')
do>'
angle
Also
microscopic scattering cross section for a neutron

of velocity v' and a nucleus of velocity
which
=
...
.
,
.
x .
,
will rotate relative velocity vector p through
o>'
= <r(pV)
. . , .
(4.183)
(4.184)
which states the physical fact that the scattering cross section to rotate
the vector p (corresponding to v and V) through angle to into p' must be
the same as that which rotates p' (due to v' and V) through w' into p.
In the present generalized notation, u = u'.
SEC. 4.7]
The next step is to transform the integral (4.182) by

Thus we write (4.182) in the form
tion variables.
im/imjirj
129
a change of integra
number
scati >\
/
, . ,
,
/f
.. ..
m(v') dvp<r(p,u) 3fl(V) dVdw = .
Jv,a
tered into dv per unit time
of neutrons
,OE.
(4.185)
Note that we still denote velocity vectors before collision by the symbols
v' and V; however, the integration variables which correspond to the
have been changed from the primed symbols to

thus, dv' dv, dV' dV, and do>' do>. Also, observe
coordinates of v' and

the unprimed;
that
dv' dV' du' = dv dV
In transforming from (4.182) to (4.185)

= p'; i.e., the relative
do>
(4. 186)
we have used the fact
that
and nucleus remains

collision.
The proof of this
unchanged through an elastic-scattering
statement follows directly from the results obtained in Sec. 4.1b. We
note first that in terms of the velocities in the center-of-mass system (C),
p
speed
between neutron
the relative velocities may be written [using (4.6) and (4.7)]

= w
-W
e' =
w'
(4.187)
If we
square each of these expressions, subtract p- from p'", and apply the
results (4.10) and (4.11), it follows immediately that
P
= p'
(4.188)
At steady state the expressions (4.180) and (4.185) must be equal.

Since the integration variables are the same we may write the balance
condition in the form
[m(v') 9D?(V)
- m(v) 2H(V)]p
A physically acceptable solution to this equation

m(v') 9tt(V)
The next step in the analysis

that
/(v)
If
(4.189)
is
m(v) 9fl(V)
is to define two new
= In m(v)
we use these in (4.190),
dv dV dw = 0
(4.190)
functions
F(V) = In 9)i(V)
/ and
such
(4.191)
it follows that
/(v') + F(V') ^ /(v) + F(V)
(4.192)
which states that there exists some scalar function of the velocities before
The only scalar
and after collision which is unchanged by the collision.
invariant in the present system is the total kinetic energy of the colliding
Thus the functions / and F may be immediately identified

with the kinetic energies of the neutron and nucleus, respectively.
particles.
REACTOR ANALYSIS
130
Therefore,
[CHAP. 4
we set
/(v) =
F(V) = a(p/F2)
a(mt>J)
(4.193)
where a is some constant, m the neutron mass, and M the nuclear mass.
Finally, we substitute these expressions into (4.191) and obtain
m(v) = 6e*
The coefficients
and
2K(V)
Be^>
(4.194)
B are simply normalization factors and are easily
computed by setting the integrals of m and 2JJ over v and V, respectively,

The determination of the quantity a which appears in
equal to unity.
the above equations is accomplished with the aid of certain thermo
relations.
The computation is straightforward but of no par
ticular interest to the present study.
Without further comment, then,
we note that it can be shown that1
(4.195)
is
the Boltzmann constant and

the temperature of the
this result in (4. 194) and carry out the normalization
If we use
of the functions m and
50{,
we
obtain
(set)
is
which are the well-known Maxwell-Boltzmann

distributions. There are
two essential properties of these functions which are of particular interest
to the present development.
(1) The temperature parameter which
appears in each of the exponentials
the same quantity for both the dis
is
is
a is
tribution of the neutrons and of the nuclei.

(2) These functions are of
Gaussian shape.
When the functions (4.196) are plotted against speed,
they will exhibit peaks around the value of the speed corresponding to
the energy kT, and with increasing values of temperature these peaks will
shift to higher values of the speed.
Thus higher temperatures correspond
to higher average speeds for both neutron and nuclei.
The fact that the same parameter
appears in both distributions given
in (4.196)
a direct consequence of the assumption that the medium in
We know, however, that many
question
purely scattering material.
media of practical importance to reactor technology possess appreciable
This
absorption cross sections.
especially true, of course, for media
In these situations the results
comprising the core of thermal reactors.
valid,
are
not
and
generally
some
modification
must be made in
(4.196)
1

where
medium.
is
a m
j,
dynamic
Mayer and Mayer, op. cit., pp. 8-14.
SEC. 4.7]
131
the definitions of the distribution functions which accounts for the effects
of the absorption losses on the neutron spectrum.
The principal effect
of the absorption collisions is to depress the neutron population over the
energy interval wherein the absorption cross section is relatively large.
Since for most materials the microscopic absorption cross section varies
more or less as 1/v, this means that the neutron densities are increasingly
reduced, from the pure scattering media values, as one progresses down
the energy (speed) scale. As a result, the neutron distribution m is dis
torted and tends to peak at higher values of the neutron speed correspond
ing to the range wherein the absorption cross section is relatively small.
This shift in the peak, and distortion in the shape, of the neutrondistribution function is called hardening of the spectrum, and it is the more
These trends are illustrated in
pronounced, the greater the value of 2.
Fig. 4.27.

Purely scattering
0
Neutron
Fig.
energy
medium
Distortion in thermal-neutron spectrum due to the presence of absorbing

materials ('normalized so that each system contains the same total number of neutrons).
4.27
One way in which this shift in the neutron density may be taken into
account is by introducing an effective neutron temperature T.
Thus we
define some fictitious temperature Tn which, when used in the function
m of (4.196), gives a new Gaussian that yields a better fit to the distorted
density distribution arising from the presence of the absorber.

The
determination of this parameter is deferred until a later section.
For the
present, all we require is that such a number can be specified for a given
system.
It
The results (4.196) may now be written in terms of this parameter.

will be convenient for subsequent treatments to introduce also a tem
perature Tn which describes the thermal state of the nuclei in the system.
Thus a suitable set of functions for the neutron and nuclear densities in
an absorbing medium may be given by
m(v) dv =
fraction of neutrons in thermal range whose

ties lie between v and v + dv
veloci
(4.197)*
REACTOR ANALYSIS
132
cm/m
fraction of nuclei whose

+ dy
nr
2K(V)dV-v
It
[CHAP. 4
velocities lie between
V and
that these distributions are defined in terms of

velocity space, which is not to be confused with speed space.1
The dis
tributions in velocity space are more detailed than those in speed space
inasmuch as they specify not only the neutron speed but also its direction
of motion.
The transformation from the velocity function to the cor
responding speed function is readily obtained by integration.
Thus in
the case of the neutron distribution (4.197) we compute the correspond
ing function in speed space by integrating (4.197) over all directions of
should
motion2
If
a.
. .
be noted
mWdv =
we define
fraction of neutrons whose

v
speeds lie between
and
dv
..
(4-199)
then

m(v) dv =
I m(v) dv
Ja
v2
dv
I m(v) da
Ja
= iirv*
Also, if
,
n,h
(^ftJ
e~mvmkT" dv
(4-20)
is the total number of thermal neutrons per unit volume, then
, ,
N
n(v) dv = nt m(v) dv =
of neutrons per
unit volume
y
,
,
.
,.
whose speeds he between v and v + dv
number
(4.201)
b. Definition of Thermal-group Cross Section.

The Maxwell-Boltzmann distributions (4.197) and (4.198) may be used to construct suitable
average cross sections for the neutrons in the thermal-energy
range.
These average cross sections will be useful in subsequent analyses wherein
the entire thermal neutron population is represented by the one-velocity
If 2th denotes some average cross section and th the total track
model.
length per unit time (in some specified volume) of the thermal group, then
a convenient definition of the group cross section is given by
2th, =
ntb
m(v) dv
<r"(\v
- V|)|v -
V|7v~
9K(V) dV
(4.202)
where N is the number of nuclei per unit volume, nth the number of
thermal neutrons per unit volume, and 2Ji and m are given by (4.197) and
The symbol <r(|v V|), which represents the neutron-nucleus
(4.198).
1
Ibid., p. 14.
' The symbol Q denotes a unit vector in actual space (i.e., a
direction) and can be
specified by giving two angles, a co-latitude i? and an azimuth ^; therefore, a = Q(^,^),
and the differential solid angle (i.e., area on surface of unit sphere) is da m sin t? dd d\f>.
SEC. 4.7]
133
interaction phenomenon of interest, is a function only of the relative

It is not, in general, the cross sec
speed between the colliding particles.
tion measured by the experimenter in the laboratory. The relationship
between the measured a and the quantity which appears in the above
equation is easily exhibited upon introducing two additional definitions
and rearranging the form (4.202).
Let
Sib = ffthjv
4>*
(4.203)
= n.h
then, (4.202) reduces to

a*
= J~
m(v) dv
- V|)|v -
<r-(|v
m(v) dv
(4.204)
V| 2TC(V)
dV
m(v) dv
(4.205)
The V integral in this expression is in fact the quantity obtained by the

Thus we identify the reaction rate v<r(v) as the integral
experimenter.
- V|)|v - V| 2N(V) dV
/v<r-(|v
(4.206)*
The physical interpretation of this relation is discussed in the section

which follows.
If we use this definition in (4.205), we obtain
cr'"
/ iv (v) m(v) dv
= J-^-
(4.207)*
m(v) dv
This result may
m(v) dv =
yV(!>) dv
Jt iv(v)
JQ"
be written in an alternative form by separating the integra

tion over velocity space into an integration over speed space and an
integration over all directions of neutron motion SI. Then, for example,
m(v) da =
iv(v) m(v) dv
(4.208)
is
Jo
iv (v)
v
^th
If
where we have used the definition (4.200) for the neutron distribution in
similar operation
speed space m(t>).
performed on the denomi
nator, then (4.207) reduces to
m(v) dv
m(v) dv
(4.209)*
We find then that the computation of the thermal-neutron group cross

may be carried out by integrating the measured cross-section
data weighted by the neutron distribution in speed space (4.200) over the
> .
<
It would appear that the details of the computation
range
introduce some ambiguity from a physical viewpoint since the integration
<
sections

va(v) =
134
REACTOR
ANALYSIS
[CHAP. 4

Conceivably, a thermal neutron

extends over the entire speed range.
speeds
acquire
extremely
large
by
suffering successive scattering
could
This
collisions each of which caused an increase in its kinetic energy.
situation might raise some question as to how this neutron is to be classi
From a prac
fied, i.e., is it a thermal neutron or is it a "fast" neutron.
tical standpoint, however, this causes no complication inasmuch as the
number of such high-speed neutrons is negligibly small due to the Gaus
sian form of the function m(v).
c. Reaction Rates for Neutrons and Nonstationary Nuclei.
We have
already noted that the cross sections measured in the laboratory are
the quantities a(w), and these are related to the interaction functions
ax(\v
V|) through the integral equation (4.206). Actually, the precise
forms of the functions <s" are not generally known.
The only information
about a specific reaction rate which is obtained from a cross-section experi
ment is the integral quantity <r(v).
In fact, the experiment itself consists
in performing the integration indicated in (4.206). This may be more
easily recognized if one recalls the experimental procedure involved in
A brief description of this pro
making a cross-section measurement.
cedure was given in Sec. 2.4c.
Such measurements are performed for
neutrons of various speeds (energies) by irradiating a sample of the mate
rial in question with a monoenergetic beam. For each neutron speed,
the experimenter obtains a measurement of the beam intensity, with and
without the specimen. The difference between these intensities, which
is usually given in terms of the reaction rate R(v), where
R(v) = va(v) =
fa"(\v
- V|)|v - V| 2K(V)
dV
(4.210)
Thus neutrons of a par

is attributed to the reactions in the specimen.
ticular speed (established by the beam) are allowed to interact with the
These nuclei, however, are in a random motion
nuclei of the sample.
the
by the temperature of the material.
Consequently
of
consists,
ultimately,
the
interaction
between
experiment
measuring
neutrons of a given speed and nuclei of all speeds.
The integrated effect
determined
of all these interactions is the quantity <r(v), and the graphic representa
tion of these functions is the cross-section curves to which we have made
Finally, it should be pointed out that Eq. (4.206) is a
frequent reference.
statement (based on the model of a two-body collision) of
the physical process involved in performing the experiment.
We have noted that the reaction rate R given by (4.210) is dependent
mathematical
upon the nuclear composition
of the specimen and upon the neutron
There is a third factor, however, which can affect this function

In the present
significantly, namely, the temperature of the material.
speed.
model, the influence of the temperature upon the reaction rate is seen
through the distribution function 9)i(V), which appears in (4.210). Thus,
SEC. 4.7]
SLOWING-DOWN
PROCESS
135
properly, R should be written as R{v;TN); we will make this distinction

later when this effect is examined in more detail.
For the present it will
suffice to point out the physical significance of this dependence.
It is customary when making cross-section measurements to perform
However, thermalthe experiment at standard temperature (20C).
group cross-section data will be required, in general, at temperatures
different from this value, depending upon the operating conditions of the
Thus, if an accurate calculation of the <rth is required,
reactor in question.
This means, then, that a

be taken into account.
procedure must be developed for determining the reaction rate R at the
temperature of interest from the available data which has been obtained
at some known reference temperature.
The temperature-corrected rate
R may then be used in (4.209) to compute the corresponding ath. A suit
this difference must

able procedure for generating these data from the laboratory measure
ments is discussed in Sec. 4.7e.
d. Effective Neutron Temperature. The neutron distributions around
thermal energy in both velocity and speed space involved an effective
neutron temperature Tn. This number appears as a parameter in the
functions (4.197) and (4.200) and establishes the position and shape of
the Maxwell-Boltzmann distribution which has been assumed to describe
the thermal-neutron
The calculation of this quantity has
population.
received considerable attention from various investigators.
For the
present discussion we draw from the work of Coveyou, Bate, and Osborn
previously cited in connection with the calculation of collision densities
near source energy (see Sec. 4.4d).
The work of Coveyou, Bate, and Osborn was aimed at determining the
effect of moderator temperature upon the neutron flux in an infinite
capturing medium.
The calculations were carried out for a uniform
medium
of unbound
nuclei of constant scattering
cross section and
l/v
absorption cross section. It was assumed that the nuclei were dis
tributed according to the Maxwell-Boltzmann relation (4.198) and that
neutrons were introduced
into the system from a monoenergetic source.
The details of the slowing-down process and the scattering around thermal
A
energy were determined with the aid of the Monte Carlo method.
brief description of the mechanics involved in the use of this method was
given in Sec. 4.4d. Again, we will not enter into a discussion of the
method but will confine our attention to the use of the results which were
obtained.
The Monte Carlo calculations carried out for the infinite-medium model
above were used to determine the distribution of neutrons
around thermal energy. These distributions were obtained by tracing
The density of neutrons
the individual life histories of many neutrons.
described
collision densities at each energy were computed from

Some of the
the statistics compiled from these individual histories.
and the various
REACTOR ANALYSIS
136
results obtained
[CHAP. 4
by this method are presented in Fig. 4.28.
As in the
collision-density calculation described in Sec. 4.4d, these results are given

in terms of the speed ratio x = 8v/vT, where v is the neutron speed, S a
scale factor, and vT = (SkTs/m)^
The curves shown in Fig. 4.28 give
the neutron flux 4>(x) = n(x)x for media of various mass number A and
cross section (the parameter k is proportional to 2a, see Sec.
The solid-line curves in the figure are the neutron fluxes obtained
4.4d).
from the Monte Carlo calculation.
The broken-line curves are the Maxabsorption
well-Boltzmann fluxes that would be obtained in infinite media of the

same mass number and temperature but which contained no sources or
sinks.
The systems described by the two curves in each of the three cases
=2"*
A =16
i.t
/
/

>
l
/
/
/
1
1
|
1
1
/
1
5 10
100
1,000
1 2
5 10
100
\
1,000
1 2
5 10
100
1,000
Fig.
4.28
Neutron flux in the thermal range.
presented contain the same total number of neutrons per unit volume;
thus, in the thermal range, the distortion of the solid-line curves relative
to the broken-line curves is a measure of the effect of the (1/tf) absorber
on the neutron speed distribution.
A useful measure of the flux distortion is provided by the concept of
We define this temperature as
the effective neutron temperature Tn.
that number which when used in (4.197) gives the best least-squares fit
of a Maxwell-Boltzmann distribution to the computed flux (solid-line
The ratio of the effective neutron
curve) in the range1 0 < x < 35.
temperature to the moderator temperature TK/TN is indicated in the
The ratio has been omitted from the last
figure for the first two cases.
=
case (A
9, k
2~') because the flux was so severely distorted from a
1 The
the
width of this range was selected fairly arbitrarily.

(l/E) region of the flux.
The intent was to exclude
8EC. 4.7]
137
Maxwell-Boltzmann distribution that the concept of an effective neutron

temperature was meaningless.
It was found that the linear relation
Tn = TV(1 + 1.11.4k)
(4.211)*
This
correlated fairly well1 all the cases reported in the reference paper.
a crude interpolative device for estimating the

parameter Tn from a knowledge of the moderator temperature TN, its
mass number A, and its absorption properties k. This result may be
relation provides then
A more accurate estimate of

(4.197) and (4.200).
the neutron distribution than can be obtained from this adjusted Max
well-Boltzmann relation would require a detailed calculation of the
used in the relations
Monte Carlo type or a solution of the Boltzmann equation.

In many prac
e. Temperature Corrections to Cross-section Data.
tical cases of interest the operating temperature of the reactor will be
above the temperature
measured.
estimating
at which
For these situations it
the available
cross-section
data were
is possible to develop a procedure for
the reaction rates at the higher temperature
from the meas
of a given speed with a medium at two different temperatures.

This is
followed by an approximate treatment which is valid when the two tem
peratures of interest differ only slightly.
We begin by stating the integral relation for the reaction rate in the
more general notation previously mentioned; thus, in place of (4.210), we
introduce
R(p;Ts)
where
/v
ffi is given by (4.198).
- V|)|v - V| 2R(V) dV
(4.212)*
For convenience in notation
we define also
<r(|v
< s Sfsji
S(p) = pa"{p)
then with
p m[\9\
|v
- V|
(4-213)
(4.214)
(4.212) takes the form

R(v;6N) =
=
f S(P)
g(v
e, Os) dp
(4lrb/PS(p)eHT"e'V49Vrfe
^
6'N
In deriving the relationship bet ween the reaction rate at one tempera
ture R{v;df/) with that at a higher temperature R(v;6'N), where
> Bn,
5
To within
per cent, for the cases considered. This relation may be used for
systems wherein .4k < 0.2.
This relationship was pointed out to the authors by R. R. Coveyou.
*

In the analysis which follows2 we

ured data at the lower temperature.
derive first a general relationship between the reaction rates for neutrons
ANALYSIS
HEACTOR
138
[CHAP. 4
we will work with the Fourier transforms of these functions rather than
with the actual functions.
For this treatment we require the three-
Fourier transform of R(v;6n) with respect to the variable v.

We define the one-dimensional Fourier transform of any function f(x) by
dimensional
the relation
fix) e*dx
${/(*)}
where
is the transform
If
variable.
(4.216)
s is the three-dimensional
form variable, where the components of
s are given
trans
by
s =
then the three-dimensional
(4.217)
transform of
R is
t R(v;BN)("dv
ff{fl(t>;M) = <R(s;M =
(4.218)
The substitution of R from (4.215) into this relation yields
If
Wift L S{p) dg
exp
"
ib
|v
- e|J + ts
v]
dv
(4.219)
we define the components of the velocity vectors v and

V v(x)2/,2)
by
9(a,b,c)
(4.220)
6)2
aY +
"
*fx]
(x
- cy +
**]
["
{z
wN
v =
18 '
e|2
*w]
"
{y
|v
- ib
then the exponent in (4.219) may be expanded into the form
Now, the x, y, and

integrations implied in dv are identical,
matically. Consider, therefore, the x integration.
a)1
+ ix
(4-221)
mathe
dx
-(4,^)1
exp
(x
/_>p[-4i
is
it
If
is
which
readily obtained by completing the square in the exponent and
this operation
repeated
applying the definition of the error integral.
for the and integrations,
follows immediately that the v integral of
y
p|s
is
dv = (4x0*)' e'e-*"
(4.
-~ -
/exp
(4.219) reduces to
|v

^s'0n)
222)
BLOWING-DOWN PROCESS
SEC. 4.7]
139
The substitution of this result into (4.219) yields

=
Tou /
-S(p) *-d9
(4.223)*
is,
It is of interest to note that the above integral is in fact the Fourier trans
form of pa"{p), the reaction rate corresponding to the fundamental inter
action function.
The procedure outlined above may be repeated for the reaction rate at
the higher temperature R(v;8'N). The Fourier transform of this function
obviously,
Si
If
S(P)e^dg
(4.224)
we compare this result with (4.223), we find that

= <R(;M e-''(-''"-'">
<R(a;0's)
(4.225)*
)
,
ff-M/tt)l
If
we apply this operation
J"
is
obtain the relationship between the actual functions, we require only an

inversion of Eq. (4.225).
The inversion operation for the Fourier transform
given the symbol
S-1
where, for example,
K& e~illdt
JF-1{(R(s;M e
(4.226)
to (4.225), we obtain
(in three dimensions)
R(v;8'N)
(4.227)
is
is
The right-hand side of this expression involves the

duct of two Fourier
transforms.
The first of these (R(s;9jv)
of course the transform of
R(v;8n).
The second
the transform of the Gaussian function in v
+ g)/4(^
- 0n)V
expI-q'
n'
[4r(ft
- Ml)
i
[-(s2 +
y*
* fexp
f'X^
- M]
demonstrated
z2
and
s2
This result
f2.
>;2
x2
y2
if
where
v1
(2x)
(4.228)
is
space, i.e.,
easily
one notes that
SF-'{e-'l
=^j"
e-o* dt
(4.229)
Since the relation (4.227)

given as the product of two known trans
forms, the inversion may be carried out with the aid of the convolution
is

Thus we have obtained

simple relationship between the Fourier trans
forms of the reaction rates at two different temperatures; in order to
REACTOR ANALYSIS
140
theorem:1
[CHAP. 4
In terms of the present definition of the Fourier transform,
this theorem states that
j"
h{&) m
g{t) h{&
=
e-^df
f"
H (x
G(x')
- x') dx'
We now apply
d'N
the function (4.213).
RWs)
where 9Ji
is
-'{911(8,
9m)
8H)\
- (2^|
=
e'N
then by (4.228),
(4-231)
(4.232)
BN)
The convolution theorem yields, then,
R(pM
9K(v
e,
e'N
d'N
where g() and h() are the transforms of G(x) and H(x).
this operation to (4.227).
Let
3K(*,
(4.230)
eN) d9
R(vv"*>
UîT'
Jg
14^(0^
flTii
dN)\*
(4.233)*
dP
This equation relates the reaction rate (cross section) of
a material at the
to
its
cross
section
at
lower
temperature
TK. As in all
T'N
the preceding analyses,
assumed that the nuclei are distributed
The above equa
according to the Maxwell-Boltzmann relation (4.198).
tion may be used then to compute the cross-section curves for a reactor
is
it
temperature
is
if
operating at any temperature T'N from the known cross-section data of

the reactor material which have been determined at some temperature
Note that
TN.
T'N = TN then Eq. (4.233) reduces to an identity [see
In general, the indicated integration must be carried out
also (4.225)].
in detail.
There one special case, however, which leads to an especially
If the measured cross-section curve varies as 1/w, which

easily
the case for many absorbers in the low-energy range, then
shown that the reaction rate
independent of the moderator tempera
some constant,
ture.
For example,
we take aa{v) = c0/v, where c0
if
is
is
is
it
is
simple result.
then from (4.233)

=
|[^rryj
wherein we have used the fact that

in a system in which the absorption
9)?
\/v,
<4-234)
normalized
N. Sneddon, "Fourier Transforms,"

See, for example,
Book Company, Inc., New York, 1951.
I.
d* =
Thus
function.
the absorption reaction rate is
is
a
m vaa(v;e'N)
is
R{vfi's)

and the application of (4.213) gives
pp. 23-27, McGraw-Hill
BLOWING-DOWN
SEC. 4.7]
PROCESS
141
independent of speed and temperature and can be determined from any

single point on the cross-section curve.
Another interesting property of Eq. (4.233) is that it is identical in
form, mathematically, to the integral solution of the heat-conduction
Thus the reaction-rate function R(v;8\) must also satisfy the

differential equation
equation.1
V'RivM
~ R(v;8N)
(4.235)
where it is now understood that the Laplacian operator V2 refers to the

v space, the coordinates of which are x, y, and z. A further analogy of
some interest is the fact that the reaction rate R has the same mathe
matical properties as the slowing-down function q{x,r), which is developed

in Chap. 6 (in particular, compare with the Fermi age model of Sec. 6.2a).
Finally, it should be pointed out that the general result (4.233) may be
used to compute the temperature coefficient (i.e., temperature depend
This has considerable
ence) of a Breit-Wigner resonance (see Sec. G.5g).
practical significance from the standpoint of reactor control.

In the event that the reactor system of interest is to operate at a tem
perature slightly above the temperature at which the cross-section data
were obtained, it is possible to derive a simple relationship between
R(v;8n) and R(v;8'N), which does not require the evaluation of the
Let us consider therefore the situation wherein
integral in (4.233).
=
In this case we can generate a solution
8s + A0jv and A6n/ 8n <3C 1.
0'N
for R{v;8'N) by making a Taylor series expansion about 8N. Thus,
R(v;8'N)
R(v;8N) +
MN
00 N
fl(y;0.v)
We compute the derivative of R with respect to

Therefore, in general,
R(v,e'N)
But, R
is a spherically
R(v;6'N)
8^
(4.236)
by utilizing (4.235).
R(v;8N) + A8N V-R(v;8N)
symmetric function in velocity space; hence,
R(v;8N) + A6N
l-
^ [t>i2(t>;M]}
(4.237)
where the derivative term may be obtained from the cross-section data at
the temperature corresponding to 0.v. In actual practice,2 the applicafor example, H. S. Carslaw and J. C. Jaeger, "Conduction of Heat in Solids,"
Oxford University Press, New York, 1948.
1 A Monte Carlo reactor code is available for carrying out this computation.
See
R. R. Coveyou and W. E. Kinney, "Neutron Reaction Rate Codes,'' Neutron Physics
Division Annual Progress Report for Period Ending Sept. 1, 1959, Oak Ridge
National Laboratory, ORNL-2842, pp. 144-145, Nov. 9, 1959.
1 See,
chap.
XIII,
[chap. 4
REACTOR ANALYSIS
142
tion of (4.237) may offer some computational difficulties, especially if the

cross-section curve is a rapidly varying function of v, since the indicated
operation involves the determination of the curvature of <r(v). Thus the
integral relation (4.233), which involves a "smoothing" type mathe
matical operation, may be more satisfactory from this standpoint.
f. Thermal-group Cross Section for l/v Absorber. As has been fre
quently noted, many materials of practical importance to reactor tech
The calculation
nology are l/v absorbers in the thermal-energy range.
absorption cross sections for materials of this
of the thermal-group
type is therefore of special interest; moreover, it is fortuitous that, in
these cases, the general relation (4.209) or (4.207) may be evaluated
analytically.
The appropriate form for the thermal-group absorption cross section in
the notation of Eq. (4.209) is given by
cr.M
we take for the present case
(4.239)
If
defined by (4.200).
(4.238)
m(v) dv
m(v) dv
(4-240)
v
m(v) dv
Co
<
f"
then (4.238) reduces to
m(v) dv
is
normalized function.
The denominator of this result
since m
clearly the expression for the average value of the speed (v) corresponding
to the Maxwell-Boltzmann distribution.
That
(4.241)
carried out, the expression for
is
m(v) dv =
the indicated integration
is
If
fo"
is
is
a
found to
be
*

where m(v)
is
/"
vaa{v) m(v) dv
v
/
Jo
T.
We evaluate the constant

aa
c0
2(^Y
(4.242)
in (4.240) in terms of a measured value of

(for example, the values given in Table
in terms of the neutron energy E.
at some reference temperature
2.2).
First we write
(4.239)
(4.243)
SEC. 4.7]
If we denote by Tt the reference energy at which the

ured, then for E = kTt we have
7.(kT.)
aa
has been meas
co^-j
143
(4.244)
which provides a relation for computing c0 in terms of known constants.

The substitution of this expression for c0, along with v from (4.242), into
a.(*r,)
Note, also, that
(^J
2?
c.(kTn)
(4.245)*
No*
(4.246)
that the absorption cross section of
substance
is
Thermal-group Cross Section for Non-l/v Absorber.

a
g.
(j
(4.240) yields
not
In the
case
\/v, the thermal-
group cross section may be computed from the general relation (4.209).
This calculation requires, of course,
detailed knowledge of the cross-
is
the thermal-group properties.
For
these estimates the constant of pro
portionality
Co [see Eq. (4.239)] may

be obtained from the more generally
available 2,200 m/sec value of the

cross section1 [see (4.244)].
Detailed cross-section curves in
the low-energy range are presently
available for a number of the more

important reactor materials (see,
for example, Fig. 4.29), and the gen
Fig. 4.29 Absorption cross section of

eral scheme described in Sec. 4.7b
U".
may be used to compute the ther
As rule the integration indicated in (4.209) must be
mal-group values.
carried out numerically however, in a few instances, notably in the case
of the fissionable materials U233, U23s, and Pu239, some of the preliminary
computation has already been performed.
Thus for these fuels data are
available which permit one to compute the thermal-group cross sections
by applying
suitable correction factor to the values obtained directly
from the cross-section curves. One definition of such
factor has been
See, for example, Donald J. Hughes and John A. Harvey, in U.S. Atomic Energy
Commission, "Neutron Cross Sections," McGraw-Hill Book Company, Inc., New
1

is
is
it
In the absence of such information,

section curve in the thermal range.
frequently assumed in practice that the absorption (or fission) cross
section
1/v, and the result (4.245)
used directly as first estimate of
York,
1955.
REACTOR
144
given by Harvey and Sanders,

the Greek symbol f).
ANALYSIS
[chap. 4
the so-called
factor (here represented by
This quantity is defined so that
a(kTn) F(Tn)
(4.247)
where a(kTn) denotes the value to be read off from the cross-section curve
at the neutron energy corresponding to the temperature Tn.
The func
tion f (Tn) denned by (4.247) may be written in the form
KTn)
ffth
(r/4)ff(*T)
(/4)ff(fc7i0
The application of
(4.241)
a(v) m(v)v dv
[*m(v)vdv
yields
jtmhm
jo
a{v)
miv)v dv
(4-249)

which is in the form used by Harvey and Sanders.

The argument Tn has
been indicated in the correction factor f in recognition of the fact that
m(y) is a function of T, the effective neutron temperature [refer to Eq.
This correction accounts, then, for the non-l/v part of the cross
(4.21 1)].
section and is based on a Maxwell-Boltzmann distribution for the tem
Curves of this function have been computed for the
perature T.
principal fuels and are available in Hughes and Harvey2 and in the refer
ence paper.
The data for U233, U235, and Pu239 have been reproduced in
Fig. 4.30.
An alternative formulation of
behavior is given by Westcott.3
f factor given above by
ff(T) =
Westcott defines also
a suitable
This
factor for non-1 /v

factor; it is related to the
correction
is the g
,(^)^(^)J
(4.250)
which may be applied to the

effective thermal cross section in order to account for the lower portion
of the l/E tail from the slowing-down distribution.
As mentioned pre
viously, the total neutron population may be represented by a thermal
group, described by a Maxwell-Boltzmann distribution, plus a slowingdown portion, approximated by a l/E distribution.
In practice, the use
of these models usually yields an abrupt intersection somewhere in the
' J. A. Harvey and
Energy," ser. 1, vol.
'Ibid.
J.
1,
a second correction
E. Sanders, in R. A. Charpie et al. (eds.), "Progress in Nuclear

Pergamon Press, New York, 1956.
' C. H. Westcott, "Effective Cross Section Values for Well-moderated Thermal

Reactor Spectra," Atomic Energy of Canada Limited, Chalk River, Ontario, CRRP680,
Jan.
25, 1957.
SEC. 4.7]
145
kT range where the l/E portion can be significantly larger than the "fast
tail" of the thermal group (refer to Sec. 4.8). Values of the l/E correc
tion factor have been obtained for several important elements and are
tabulated in the reference report.
The purpose of the present discussion
h. Scattering from Molecules.

is to present in a very elementary way some of the essential features of

The scattering
the scattering process between neutrons and molecules.
com
in
is
an
extremely
of neutrons from bound nuclei (nuclei
molecules)
plex interaction phenomenon, and a detailed study of the physics involved
The primary interest in this process
is beyond the scope of this work.
from the viewpoint of the reactor physicist is the determination of slowneutron cross sections of various compounds of practical importance in
O.990 I
200
400
600
200
400
600
200
400
Effective
Fig.
4.30
neutron
temperature
L_
600
T, K,of MaxwellBoltzmann Distribution
factors for fission and absorption cross sections.
reactor design.
Since these data are usually best obtained from crosssection measurements using the actual materials, the analytical treatment
of the process is of little immediate concern to the reactor physicist.
However, a discussion of the principal phenomena involved in this inter
action may be of some value in achieving an accurate interpretation of
Moreover, the process itself is of considerable
the experimental data.
interest to the nuclear physicist since the diffraction of slow neutrons by
important research tool.

Much of the theoretical
work on this problem has been carried out in this connection.1
The difficulty in describing the scattering of neutrons by molecules
arises from the fact that such interactions are the combined result of
gases has become an
for example, E. Fermi, Ricerca set'., 7, 13 (1936); W. Lamb, Phys. Rev., 66,
R. G. Sacks and E. Teller, Phys. Rev., 60, 18 (1941); A. Messiah, Phys.
Rev., 84, 204 (1951); A. C. Zemach and R. J. Glauber, Phys. Rev., 101, 118 (1956),
and 102, 129 (1956).
1 See,
190 (1939);
146
REACTOR ANALYSIS
[chap. 4
and molecular effects. At high neutron energies (above the

electron-volt range) the molecular effects are not important in these
collisions, and the process is effectively a nuclear one. Below the elec
nuclear
tron-volt range, however, and especially in the thermal range, molecular

binding forces are comparable in magnitude to the "impact" forces from
the incident neutron, and molecular phenomena contribute appreciably
to the over-all collision process. Thus, the scattering cross section of
molecules for thermal neutrons is determined by several different kinds of
The principal phenomena involved are: (1) the
particle interactions.
nuclear scattering process per se, (2) the natural frequencies of the chem
ical bonds which hold the target nucleus in the molecule and the various
(3) the rotational energy states of the
molecule, and (4) the interference effects from the multiple scattering
centers (nuclei) in the molecule.
vibrational states of the molecule,
The extent to which each of these factors influences the interaction

by the neutron energy. At very high energies (between
determined
is
1

and 100 ev), the scattering process is essentially a nuclear phenomenon.

In these cases, the neutron energy is so great that the average recoil
energy transferred to the struck nucleus1 is sufficient to rupture the chem
ical bonds, and the nucleus behaves as if it were unbound (an atomic gas).
As the neutron energy is reduced to the 0.1-ev range, items 2 and 3
become important.
In this range the excitation energy of the highest
internal quantum states of the molecule are comparable to the translational energy of the neutron.
Depending upon the temperature of the
medium, a certain number of these rotational and vibrational states of
the molecule will already be excited, and when a collision occurs, some
fraction of the neutron energy is imparted to the molecule.
This energy
transfer may result in the transition of the molecule to higher rotational
In some
states and/ or the excitation of various vibrational modes.
cases, depending on the neutron energy, the initial internal-energy state
of the molecule, and the mass of the struck nucleus, the energy transfer
may be in the opposite direction, and the neutron may emerge from the
These considerations
collision with a speed greater than its initial value.
become increasingly important as the neutron energy approaches the
thermal range.
Finally, there are the interference effects (4) noted
These are generally very complex, and arise when the neutron
wavelength is comparable to atomic spacing.
The influence of these several factors on neutron-nucleus interactions is
above.
well displayed in the case of neutron scattering from hydrogen nuclei

bound in molecules.
This process has been studied in detail and experi
mental measurements of the scattering cross section of hydrogen in the
water molecule show that there is a substantial change in the cross section
1
Clearly, the energy
the struck nucleus.
transferred depends
on the neutron energy and the mass of
SEC. 4.7]
147
from the high- to the low-energy range. The cross-section data are shown
in Fig. 4.31.
The steady rise in a. from the 20.36-barn value in the elec
tron-volt range to around 80 barns at 0.004 ev is due to the combined

effects of the various phenomena mentioned above.

Blatt and Weisskopf1 give an analytical treatment of the effect of
They show that for
chemical binding on the scattering cross section.
10
'
'
0.01
0.02
0.04 0.06
1
0.1
'
0.2
0.4
0.6
1
1.0
2.0
4.0
6.0
10
Energy, ev
Flo.
4.31
'
Scattering cross section of hydrogen measured in water.
'
very slow neutrons the scattering cross section (o-) B of a hydrogen nucleus
bound to a molecule of mass Am (in units of neutron mass) is given by
WB
4(rrx:y
where (a,) F is the cross section of the hydrogen nucleus in the free (atomic)
The generalization of this result to the case of a bound nucleus of
state.
mass A is
<
It is interesting to note that, for many molecules, the factor involving Am

will be very near unity and the principal contribution to the chemical
binding effect stems from the factor involving the mass of the struck
nucleus.
Even then, the effect of the chemical bond is of importance
primarily in connection with the very light nuclei.
1
Blatt and Weisskopf,
op. cit., pp. 71-80.
148
REACTOR
4.8
ANALYSIS
[chap. 4
Four -factor Formula
Thermal Group.
The analysis of Sec. 4.6 dealt with the energy
distribution of neutrons in an infinite multiplying medium wherein neu
trons could slow down to zero energy.
We found in Sec. 4.7, however,
that in all media at temperatures above absolute zero the neutron popula
tion at the very low energies acquired a peaked distribution around some
energy of the order of kT which was determined by the thermal motion
(temperature) of the moderating material.
We showed also that this
distribution could be well approximated by a Maxwell-Boltzmann dis
tribution if the medium was not too highly absorbing.
In fact, we were
able to specify an effective neutron temperature Tn for this purpose which
a.
was dependent upon the actual moderator temperature and its absorption
properties [see Eq. (4.211)]. This temperature could be used to define the
neutron distribution [see Eqs. (4.197) and (4.20O)] in the thermal range.
In the present section we combine these results, the neutron distribu
tion in the high-energy range according to
Sec. 4.6 and the
distribution
This gives a complete

range according to Sec. 4.7.
description of the neutron population in energy space from the source
energy down to, and including, the thermal range. We will use this
picture to carry out a more realistic analysis of the neutron physics in an

in the low-energy
infinite multiplying medium.

For this calculation we propose the follow
ing model: (1) Medium is infinite, isotropic, and homogeneous.
(2)
System is at steady state.
(3) Neutrons appear in the system accord
ing to some source distribution S(u).
(4) These source neutrons
slow down to the thermal range as if they were in a medium in which
motion of the nuclei.

(5) At some energy E =
process ends, and all neutrons (not captured while
slowing down) enter a one- velocity diffusion process; they wander through
the medium at this energy and suffer no change in speed as a result of
scattering collisions.
(6) These thermal neutrons are eventually lost by
there was no thermal
Eth, the slowing-down
absorption.
Thus our model replaces the actual neutron distribution by a slowingThe thermal group consists of
down distribution plus a thermal group.
neutrons of one speed, this speed to be defined by the effective neutron
The actual thermal distribution, which was approxi
temperature T.
mated by the Maxwell-Boltzmann relation (4.200), is replaced then by a
"spike" or "group" at E = Elb (see Fig. 4.32). The selection of this
thermal energy EA is discussed shortly.
The mathematical statement of the model described above is (cf. Sec.
4.6a)
q(u) = -S0(m) *() + S(u)

q(u) = 2((u)
(u)
0 < u < uth
(4.145)
(4.112)
SEC. 4.8]
where q(0) = 0 and
wlh
use
Si"th
For the thermal group we
to Eth.
corresponds
lim q(u)
149
m g(u(h)
(4.252)
which follows from the physical requirement that, if the system is to be at

steady state, the rate at which neutrons enter the thermal group q(uih)

fa)
(b)
I
/The rmal
n(E)
Fig.
4.32
Distribution
group
Slowing-down
distribution
So,
E0
of thermal neutrons: (a) actual distribution;
must be balanced by the rate at which they are removed

above relations we have used the definitions
2^ = absorption cross section of thermal
th = ntbvth = thermal neutron flux
nth
i>th
= number of thermal neutrons per

= speed of thermal neutrons
(6) model.
2J,h<th.
In the
group
unit volume
The absorption cross section for the thermal group is computed by means
of Eq. (4.238). As already demonstrated, in the event that the absorber
has a 1/y cross section, the integral yields an especially simple form; this
result is given in Eqs. (4.245) and (4.246).
The absorption cross section
REACTOR ANALYSIS
150
[CHAP. 4
of many materials has this l/v dependence, fortunately, and these rela
tions may be used directly to obtain the 2*\
An appropriate definition of the thermal flux which is consistent with
the thermal-cross section formula is given by [cf. (4.204)]
th
J0
n(v)v
dv = ntbv
(4.253)
where the average speed v is defined by (4.242).

According to this
definition the appropriate value for the speed of the thermal group is
= v; however, the most probable speed v0 is sometimes used to describe
vtk
In terms of the Maxwell-Boltzmann dis
the speed of thermal neutrons.
tribution,
vo
is found to
be1
vo
= v(kTn) =
(2*y
(4.254)
This number represents the position of the maximum point in the Max
well-Boltzmann distribution m(v) [see Eq. (4.200)]. Note that t>0 cor
responds to the energy kTn so that we immediately identify Eth as

E*
imvl = kTn
(4.255)*
It
must be recognized that this model, which treats the neutron dis
tribution by two separate regions (one for the high-energy range and one
for the thermal group), only approximately represents the actual distribu
tion, since the slowing-down process just above the thermal range is incor
rectly represented. We have also neglected certain crystalline effects

which may play an important role in the low-energy range; neutrons may
scatter from several nuclei simultaneously in what amounts to a diffrac
This effect changes the effective scattering cross sec
tion phenomenon.
tion (see Sec. 4.7h) and lowers the average logarithmic energy loss per
collision.
The solution to Eqs. (4.145) and (4.112) was previously obtained and
If we compute q(ulh) using this relation and sub
is given in Eq. (4.147).
into
we obtain for the thermal flux
(4.252),
the
result
stitute
=
S(u)
p(Mlh;M) du
(4.256)
The neutron flux in the range 0 <

where p(uth;u) is given by (4.148).
from
using
u < Mth is computed
(4.147).
(4.112)
*(M) =
1 See
f" SW
V(u;W) du'
(4.257)
D. K. Holmes and It. V. Meghreblian, "Notes on Reactor Analysis," part

54-7-88, Oak Ridge National Laboratory, Aug. 2, 1955.
ORNL CF
II,
SEC. 4.8]
If
all the neutrons are introduced at lethargy u
= 0,
151
Eqs. (4.256) and
(4.257) reduce to
**
^p(tW))-^
(4.258)
PM
2,(u)
(4.259)
&t(u)
where So = number of neutrons introduced per unit volume\

per unit time at u = 0
=
p(Mti>;0) = resonance-escape probability from
Pth
u = 0 to thermal energy
p(u)
p(u;0)
= resonance-escape
m = 0 to u
(4.260)
probability from
It is
should be noted that Eqs. (4.258) and (4.259) are not accurate.
with
zero
and
with
emerge
that
the
neutrons
appear
lethargy
imagined
<So
directions of motion isotropically distributed about the point of birth.

It
As these neutrons move away from this point, there are certain prob
abilities per unit distance of travel that they will make both absorption
If the neutrons are to contribute to the slowand scattering collisions.
ing-down density for lethargies u > 0, their first collisions must be scatter
ing collisions; thus in reality, only the fraction 2,(0)/2,(0) of the <S0 neu
trons should be counted as a source for the slowing-down process. We
will continue to neglect this effect {first-flight losses from sources), since
It
the absorption cross sections are often very small at higher energies.
is understood, then, that at any time it is felt to be important the source
strength may be reduced by the ratio 2,/2<.
b. Criticality Equation. Suppose that the infinite homogeneous
medium contains fissionable material and that the only sources of neu
Then, as above, if the system is
trons present are the fission reactions.
to operate in steady state, the number of neutrons produced from fissions
must equal the number of neutrons introduced into the medium, namely,
S(u). For this case the source function has the form
5(u)
[/o",h
2,(') *(') dv! +

Z?*h]
(4.261)
The fission spectrum j(w) has the shape shown in Fig. 4.25, but for the
present calculation we set the origin of the u axis at the far left so that
Note that the integration in (4.261) extends
j(0) may be taken as zero.
The last term 2}h<,h
up to but does not include the thermal group.
accounts for the thermal fissions.
We choose to introduce one neutron per unit volume per unit time by
requiring that
|oUth
S(u) du =
(4.262)
if we integrate (4.261) from u
Therefore,
du
[fQUlh
[CHAP. 4
= 0 to ulk, we obtain
Zf(u') *(u') du' + 2}"*,h]
is a normalized function [see Eq. (4.152)]; hence
[|o"th
2,(u')
4,(u') du'
(4.263)
2}^th]
We expand this expression by introducing the solutions

(4.257) and noting that in this case S(u) = j(u).
1
r
v I
/"uth
r
I j(m')
y(u) du
(4.256) and
p(u;u') du'
S
I
in Jo
h(u) p(uth;u)
du]
But,
ANALYSIS
REACTOR
152
(4.264)
The above relation

the criticality condition for the infinite homogeneous reactor, using the
asymptotic solution for the slowing-down distribution and the concept
of
the system
Equation (4.264) must be satisfied
thermal group.
j(w') p(w:u') du'

vth
+ ^4
,lh
y(u) du
JoI
h(u) p(uth;u) du
into (4.167),
j(m') p{u;u') du'

uth
th
\fast fissions/
(4.266)
number of
number of
of neutrons
per fission
average number
ka =
h(u)p(ulh;u)du\
convenient to write this statement in the form
It
is
iI
Jo I r
(it) du
Jo
vc
(4.265)
uti.
km
Or, by substituting this expression for
i J0
Jo
(4.167)
Vc
from the equation
vc
uth
vc
if
K
We compute
is
the system
not critical, the multi
Now,
plication constant kx may be computed analogously to the reasoning
Thus we employ again the concept of a fictitious
leading to Eq. (4.166).
neutron yield per fission vc and the relation
to operate in steady state.

is
if
is
is
where the function y(u)
\thermal fissions/
(4.267)
where we use the expression fast fissions to denote the number of fissions
SEC. 4.8]
153
per unit volume per unit time produced by neutrons with energies above
Etik, and thermal fissions to denote those produced by neutrons in the
thermal group represented by the energy E = E^. Thus we may write
^
If
fast fissions + thermal fissions

thermal fissions
thermal
fissions
(4.268)
we define t the fast fission factor by
total fissions
(4.269)*
thermal fissions
then (4.266) takes the form
)
This result may

average resonance
be
written in
p(ulh;u) du
(4.270)*
introducing an
for neutrons released according to
a more compact manner by
escape probability
pth
the fission spectrum; thus,

fQUtb
i(u) ?(*;) du
(4.271)
and Eq. (4.270) reduces to
*.
(4-272)
This relation for computing
kx is useful only if e may be calculated or

estimated on bases other than those given here. An especially simple
In these
situation occurs in the case of thermal reactors (see Sec. 1.2).
~
1 and for many purposes may be taken as equal to unity.
systems, t
The deviation of e from unity may also be estimated, without introducing
How
any large error, on the basis of known values for similar reactors.
ever, in general (4.270) is to be used in calculating kx.
If it is a good approximation that all neutrons from fission are intro
duced at one energy (chosen at some appropriate energy near the peak
of the fission spectrum), then by setting the zero of lethargy to correspond
to this energy,
J()
= ()
The expression for the multiplication constant
(4.163)
in the infinite medium
(4.266) becomes
y(u) p(u) du +
v.:
(4.273)
or in the form of (4.272),
(4.274)
REACTOR ANALYSIS
154
[CHAP. 4
Following the treatment of Eqs. (3.19) and (3.22), we write
where the subscript

[Eq. (3.23)]
V and
By our usual definitions for
refers to the fuel.
(4.275)*
>?/Pth
This equation
is referred to as the "four-factor formula" and is the exten

sion of (3.22) to the mullivelocity reactor.
Again it must be mentioned
that (4.275) is useful only if can be easily estimated; in general (4.273)
or (4.266) is to be applied.
To summarize, then, we use the relation
/""th
/
y(u)du f l(u') p(u;u') du' +
y*
1/
Sa
2th
-L
/"""'
^Jo
j(u) p(utll;w) du
(4276)
Jo
case of the general fission spectrum j(u).

If on
(4.163) is an acceptable approximation for j(u), then
(4.276) reduces to
for the fast effect in the
f"th 7() P() du + 2fp*/Z
z.-p /S*
(4-277)
The four-factor formula (4.275) may be given a simple physical inter

First we note that in the case of a critical system kK = 1, so
that
1 = vfp**
(4.278)
pretation.
is,
This relation may be used to follow the neutron production cycle at steady
If we begin the cycle by introducing one thermal neutron into the
state.
multiplying medium, then a fraction (SJ,h) p/2? is captured by the fission
able material.
Of these, a fraction S}b/(S^h)f causes fission. Thus the
single thermal neutron produces /2}h/(2h)? fissions.
However, if we
consider that in general the fuel material will also possess some nonzero
fission cross section in the fast range, then for each thermal fission there
will be e thermal fissions plus fast fissions. The total number of fissions
in the system
Each fission produces neutrons
therefore, e/2}h/(2J,h)jr.
Thus qtf fast neutrons appear for each thermal neutron
on the average.
added to the system.
These fast neutrons are eventually slowed down
to thermal energy by the various scattering collisions with nuclei.
How
only
certain fraction survive the entire slowing-down process since
ever,
some neutrons are removed from the system by fast captures.
The
therefore,
fraction reaching thermal
fast
neutrons
are
eventu
p,;
yfep,,,
the system
ally thermalized.
Clearly,
to remain at steady state,
is
if
is

the other hand
77/ep,h
SEC. 4.8]
1,
so as to reproduce the
155
original thermal neutron which started
the cycle; thence, (4.278).

As an illustration of the use
c. Application of Four-factor Formula.
of the four-factor formula (4.275), consider an infinite medium which
consists of a homogeneous mixture of two elementary substances, one
The problem
acting as the moderator and the other as the fuel material.
is to find the multiplication constant of the medium, given the concentra
tions of the fuel and moderator, the neutron cross sections of each as a
function of energy, and the neutron production properties of the fuel.
In many cases of practical interest, particularly in thermal reactors, the
fuel concentration will be very small in comparison to the moderator con
centration. Under these circumstances the volume-fraction effects dis
cussed in Sec. 3.3 may be neglected and the moderator concentration Nm
may be taken to be that of the pure moderator material [see Eqs. (3.33)].
In the calcuTations which follow, we will assume that this is a valid
assumption
for the system in question.
the fission spectrum
Further, we will assume that

by the monoenergetic
may be well approximated

source (4.163).
The appropriate formula for kx in the present case is seen to be (4.273).

From the form of this equation it is evident that both the fast and thermal
fissions must be calculated; the computation of the fast fissions involves
integrals which are usually performed numerically over the cross sections
and which are not readily presentable in this discussion.
However, it
should be observed that in many cases the fraction of the fissions caused
by thermal neutrons increases as the fuel concentration decreases; there
fore, it is possible to find a range of fuel concentrations for which the fast
fissions are negligible (although the multiplication constant may not be
Let us con
greater than unity for this range of fuel concentrations).
sider, then, the ratio of fast fissions to thermal fissions as the fuel concen
tration is decreased. Now,
Fast fissions
Thermal fissions
_ jo*
+(>
du
2/0th
fj? ggO
() *u
<r}hti,
where the fluxes are obtained from (4.112) and (4.252).
These may be
written
M-lH,
For
<4-280)
(at some high energy) both <(w) and th

increase as the fuel concentration is lowered.
With decreasing fuel con
centration, the slowing-down density will be increased at every value of
u; however, the greatest increase will be in q(utU) since q(uih) records the
a given source strength
cumulative effect from the entire fast range.
Further, and more impor
REACTOR
156
ANALYSIS
[CHAP. 4
is,
tant, is the fact that both 2((u) and 2J,h decrease with decreasing fuel con
Again, this gives an increase in both the fast and thermal
centration.
fluxes; moreover, the thermal flux will in most cases gain relative to the
fast flux since the fuel-absorption cross section will usually be a larger
part of the thermal-absorption cross section than that of the total fast
cross section.
(Note that in the fast range the total cross section is due
The combined effect of these
primarily to the scattering cross section.)
variations is to decrease the ratio (4.279) as N? decreases.
Now consider a moderator-fuel mixture in which the ratio Np/NM is so
low that the fast fissions may be completely neglected, that
in Eq.
(4.275), e~l.
Then
*.^H/Pih
The quantity
(4.281)
is
is
is
fixed once the fuel

selected (for U236, the value 2.08).
For this case the thermal utilization
0.002 0.003
Fig.
0.004 0.005
Nr/Nu
Infinite-medium
multipli4.33
cation as a function of the atomic ratio
(4.282)
The ratio of the thermal-absorption

cross sections of carbon and
example,
Um, for
is
0.001
0.0047
of fuel to moderator.
The calculation of the resonance-escape probability involves

gral over the fast cross sections; thus
jo
exp
05,(u)
is
low (high mass of

high and the
may be valid to make the following approxi
Because the absorption cross section

the fuel nuclei) for the fuel,
mation for all energies:
an inte
(4-283)
2,(u)
is
pth =
it

NFc,
_ UNM*iM)(u)
Further, the scattering cross section of the moderator will often have so
little variation over the fast range that some representative value may be
taken over the entire range; then, Eq. (4.283) may be approximated by
^"STO* P^)du]
(4.284)
SLOWING-DOWN
SEC. 4.8]
PROCESS
which appears in this expression is called the
The integral
157
resonance
integral:
futh
=
o-(w) du
/
Resonance integral =
[E
jEt),
JO
aa{E)
(4.285)
Cj
Values of this integral have been obtained for various fuel materials and
are given in the literature.1
Equation (4.281) may be written in terms of these results; thus,
=
<
i1
% W\X exp [-
dfe* r^,(u)
du]
*-
(4.286)
t.v,
is
1)
(t
it
and the resonance

Then, with given values for r\, WaM)/<x(aF)]lh,
calcula
integral, A-., may be computed as function of N p/N M. Such
Of course,
curve of the type shown in Fig. 4.33.
tion might yield
would be necessary to determine whether the assumption of no fast fissions
acceptable up to concentrations
as high as
N F/N M
= 0.005.
>
-1
<
0<a
<
-1
i(l+av)
f(i) =0
<
is
Let the frequency function for the cosine of the scattering angle in the center4.1
of-mass system be f(ij), as given below, for the case in which
particle of mass A/i
initially at rest in the laboratory system:
collides with particle of mass Mi which
Ml
M\
2Af,A/s
(A/, + A/,)1
- A/A'
- cos
a In a
a
the re
1,
b In E0/E, and express

Compute
sults in terms of {o, where
=
is
E,
is
As compared to the frequency function for the case of isotropic scattering in the
This situation
called
scat
center-of-mass system, ((?;) has an added term âr).
Note that the relation between the neutron energies before and after collision
tering.
component;
unaffected however by the addition of the
respectively,
Eo and
that is,
/A/,
U/,
Collision in center-of-mass system
MtJ
0 (1
1,
1,
J.
r-=-
a)i.o
L. Dresner, The Effective

Am, Nuclear Soc.,
08-79 (l'J56).
See, for example,
Th-232,
dE
>
Et
aEo <
j.
= -7=
f)
(';) dE =
is
If
4.2
neutron of kinetic energy Eo emerging from a scatter
ing collision with energy in dE about
given by

PROBLEMS
< Eo
< E0
Resonance Integrals of U-238 and
REACTOR ANALYSIS
158
where a = [{A
I) /(A +
1)]* and
[CHAP. 4
= nuclear mass number, show that the average
kinetic energy of the neutron after two collisions is
It
is possible to show a more general relation of this type, namely,
Calculate the average number of neutron-nucleus collisions needed to reduce

of a neutron from 1 Mev to 1 ev on three different bases.
Treat all
neutrons as if they suffered their average variable change at each collision on the three
Take the value 0.1 for the a of the medium in
scales energy, lethargy, and speed.
which the neutrons are slowing down.
(Note that the average speed change per
collision can be calculated from the frequency function given in Prob. 4.2.)
4.4
Consider the slowing down of neutrons in an infinite, nonabsorbing homogene
Use the vari
ous medium of a single nuclear species with thermal motion neglected.
able w = Et>/E, where Eo is the kinetic energy with which the neutrons are introduced.
o. Find the distribution function and the frequency function for w after a single
scattering collision (assume that the initial value of w = u)).
b. Find the average gain in w per collision.
c. By considering the balance of neutrons in an interval dw, find the (asymptotic)
density F(w), where F(w) dw =
integral equation for the scattering-collision
4.3

the kinetic energy
2,(u>)
(w)
dw.
the slowing-down density q(w) as an integral over the scatteringcollision density.

4.6
Neutrons of energy Eo are produced uniformly throughout an infinite medium.
Subsequently, they slow down by elastic collisions until they reach Eth and there
The medium is characterized by the macro
enter a one-velocity diffusion process.
scopic cross sections 2 2, 2/, and 2f, which are all inversely proportional to the
neutron speed at all speeds.
Find the fraction of fissions caused by thermal neutrons.
Assume the following slowing-down density and flux relation: q(E) = 2i(i?) (E)E.
4.6
Consider a homogeneous infinite medium for which the macroscopic scatter
ing and absorption cross sections may be taken to be constants and for which the
probability is given by
resonance-escape
d. Compute
appropriate average gain in lethargy per collision.

for the slowing-down density g(u) at some u much larger
than the source value in the two cases: (1) all neutrons are introduced with lethargy Ui,
and (2) neutrons are introduced uniformly in lethargy over the interval 0 u* (in the first approximation Ui ^ u* /2, for
where
a.
is the
Find
2u./|2,
the expression
1).
Consider an infinite medium of uniformly distributed fission sources and

Neutrons appear at all energies from these sources and sub
moderating material.
A good approximation to the fission spec
sequently slow down by elastic collision.
trum is given by the normalized function
4.7
z(E) dE
Jge-E/B. dE
El~
159
Determine the corresponding function j(u) du in terms of the lethargy variable

u and
un = In E, /Em.
h({) df in terms of f, where f = u
b. Obtain an approximation for A(f) by expanding its exponent in a Taylor series
about its minimum point fo (use only the first three terms of the series).
c. Compute and plot the exact and approximate expressions for the function h({),
using E, = 10 Mev and E = 1 Mev.
d. If 2(u)/i2i(u) = 0 = constant, find the slowing-down density q(u), using the
approximate solution for }(;*) based on the results of part b above.
e. Show that this result (part d) varies like the slowing-down density due to a
monoenergetic source at uo = fo + um of the same total strength1 when u is far removed
from tto.
4.8
Consider a homogeneous infinite medium consisting of a mixture of a fuel
material and a moderator material into which neutrons are introduced with a high
energy o- Suppose that each of the macroscopic cross sections 2a(E), 2,(E), and
2/(25) has the following energy dependence:
a.
shE, /E, and
2(E)
(E)E
Assume also that q(E) = 2,(E)
p(E;Ea)
E,
= 2(0' = constant
= 2< = constant
= 2th = constant
= exp
<
Elh <
E < Et
E < E,
= 2?,h
and
fB,2JE') dE'l
JE t2,(E')E' J
Define
*
i
7i =
2?
f2,(<)
"th"
= 0, 1, or
according to the above notation and 7th/0th = 2'//2^.

expression for the multiplication constant of this medium in
terms of the parameters: /30, /Si, 70, 71, 7th/0th, E0, Eu and
b. Compute the numerical value of the multiplication constant using
where
a.
Find
the general
v
2.46
E0 = 10 ev
Am =
Ei
12
= 10 ev
AF
Elh
= 235
= 0.025 ev
and an atomic ratio of moderator to fuel of Nm/Np = 500.

microscopic cross sections (in barns) to be as follows:
Fuel, F
Take
Moderator,
the
Energy range,
ev
10^
c. What percentage
1
1.0
10'
0.025
0.025
10
10.
650
0.5
2.0
550
a.
a.
5.0
5.0
5.0
0.001
0.003
0.010
of the fissions is caused by thermal neutrons?
Note that the integral of the approximation for j(u) is not unity.
5.0
5.0
5.0
average
CHAPTER
DIFFUSION THEORY: THE HOMOGENEOUS
ONE-VELOCITY REACTOR
5.1
a.
One-velocity Diffusion Equation
One-velocity Model.
The most general problem
reactor analysis is the determination

inhomogeneous
anisotropic
medium
of interest
in
of the neutron density in a finite,

as a function of space, neutron
velocity, and time coordinates.

For the present, the general problem
will be avoided, and attention will be confined to uniform, homogeneous,
and isotropic media in steady state.
Even with this idealization, the
problem of computing the neutron flux at all space points and velocities is
sufficiently difficult that it is worth while to lead up to the general prob
lem through a series of simpler cases.
The first of these we have already
considered in the analysis of Chap. 4 which dealt with the energy depend
ence of the neutron flux.
In this case, the spatial dependence was
removed by considering only homogeneous infinite media.
The direc
tional dependence was removed by limiting the description to an inte
grated flux (v) which specified only the neutron speed.1
It is the purpose of the present chapter to study the second model in
the series.
This model limits the description of the neutron population
to its dependence on the spatial coordinates.
Thus for the time being we
remove the energy dependence entirely by assuming that all the neutrons
in a system (including neutrons from sources) have the same speed.
Later, in Chap. 6, we will consider a third, more complex, model2 which
gives a description of the neutron population in terms of the neutron
energy and the spatial coordinates simultaneously.
Further generaliza
tion is achieved in Chap. 7 wherein the flux is denned in terms of not only
spatial coordinates and energy but also the direction of motion of the
neutrons
(the angular
dependence of the flux).
This is the complete
transport-theory formulation.
The basis of the following analysis, then, is the one-velocity approxima
This model was introduced in Chap. 3, but our attention there was
tion.
confined to the infinite medium.
In the present treatment we will
1 The calculation of the total
track length <p(v) took into account the contributions
of all neutrons of speed t>irrespective of their direction of motion.
1 The Fermi age model.
160
DIFFUSION THEORY
SEC. 5.1]
161
examine in detail the spatial distribution of neutrons in finite systems of

various shapes; moreover, we will consider both multiplying and nonmultiplying media. It is recognized that the one-velocity model, like
" multivelocity " model of Chap. 4, gives a very
the infinite-medium
limited picture of the neutron distribution, inasmuch as each of these
Thus while
models focuses attention on only one feature of the problem.
in
each
in
itself,
very
is
realistic
a simpli
presents,
of
these
models
neither
fied form, concepts and general results which remain valid in the more
general case.
The immediate purpose of this calcu

b. Neutron -balance Equation.
lation is the formulation of a neutron balance (or continuity) relation,
equation (3.8), which accounts for all
analogous to the infinite-medium
the neutrons in an elementary volume of the system in question.
The
present development of the neutron-balance condition, however, is not
mathematically rigorous and appeals in main part to physical intuition.
Even though the analysis is relatively crude and does not bring out clearly
involved, it is expected
and extent of the approximations
A rigorous
nevertheless to be instructive as an introductory calculation.
derivation of the result obtained below can be developed only from the
more accurate statement of the transport equation.
We will use this
approach later in Chap. 7 in connection with the general transport-theory

the nature
methods.
The model we use for the present calculation has the following features :
(1) the system is a homogeneous and isotropic medium of finite extent;
(2) all neutron sources and sinks are uniformly distributed throughout
Thus all neutrons in
the system; and (3) the one-velocity model applies.
the system, including those which appear from sources, have the same
We specify the neutron population in terms of the density
speed
v.
n(r,<),
where
.
.
'
We will also
.
'
i\
number of neutrons of speed v per unit

volume around space point r at time t
use the neutron
flux
4>{r,t)
total track length of neutrons of speed v per unit

volume per unit time around space point r at time t
By the usual definition

*(r,t)
n(r,0
(5.3)
Note that these functions refer to neutrons of a single speed v according to

our model.
In all the analyses of this chapter the speed v will be implied
in our definitions of n and 0, and therefore the number v will not appear as
an argument of these functions.
Also note that the argument r denotes a
general position vector in some appropriate coordinate system set up in
REACTOR ANALYSIS
1G2
[CHAP. 5
The position r will be specified by means of one, two, or

the medium.
three space coordinates as required for a specific calculation.
We now generalize the balance condition (3.8) to the present case. For
the finite medium, the number of neutrons lost from a given volume ele
is,
ment in unit time by passage across the boundaries of the element will
That
there
not, in general, be compensated by a corresponding gain.
will be in general net "leakage" (gain or loss) from the volume element.
An appropriate term must be introduced into the neutron-balance rela

tion to account for these net losses or gains through the boundaries.
balance condition for the differential volume ele
verbal statement of
,
,
,n
.
number of neutrons absorbed
[net number of neutrons

escaping from dr per
unit time
If
[in di per unit time

we introduce the absorption
cross section
for neutrons of
20
(5.4)
speed
r
I
number of neutrons pro[duced in dr per unit time
Jr
rate of increase of the

neutron density in dr about
[Time
ment dr which includes such a term is:
and
volume per unit time around
|i
the mathematical statement of Eq. (5.4)
at
n(r,<)
di
S(r,t) dr
produced per unit

at time
t
number of neutrons of speed
i-i
is
c/
24>(r,0 dr
2{r,t) dr
(5.6)
The function ?(r,<) dr denotes the net losses from the volume element dr
due to transport of neutrons through the boundaries.
lost per unit
net number of neutrons of speed
volume around point per unit time at time due to
passage through boundaries of the volume
v
sa
%(r,t)
(5.7)
is
The primary interest in this chapter

the derivation of solutions to
In each
for
of
Eq. (5.6)
various situations
importance to reactor physics.
case we require also a suitable
set of boundary
conditions
and an ana
is
lytical expression for the leakage term 8(r,<)- The remainder of the
devoted to the derivation of useful relations for defining
present section
these quantities.
Fick's Law Approximation
is
for Net Neutron Current.

The calcula
somewhat involved, and
tion of the leakage term $(r,t) in Eq. (5.6)
will
of
this
not
be
considered until we have
quantity
estimate
proper
For our present
introduced the machinery of the transport equation.
will suffice to use instead
convenient approximation which
needs
c.
it

the source function S(r,t), where
DIFFUSION THEORY
SEC. 5.1]
163
What is really desired in making an estimate of

has given good results.
the leakage is a knowledge of the net current of neutrons at any point,
since, if this were available, it would be a simple matter to sum up the
contributions from the net outward currents across all boundaries of the
In order to introduce the approximating

volume element in question.
a
simplified
for
the
case is used.
current,
expression
Consider then a portion of a medium in which the neutron density is a
function of only one space variable
x and is independent
of time (i.e.,
assume that steady state has been

Suppose a test area dA
reached).
is placed in the medium at the
;0;
te
that its
surface is normal to the direction of
the axis (see Fig. 5.1). At any instation x and
dA
f*
I-
so oriented
F[Q
Differential arca dA

stant, neutrons are passing through

this test surface from both sides and at all angles. We describe the net
effect of these passages by means of the net current which we define
net number of neutrons which pass through test

area êr Un^ ^me' w'*n neutrons moving positively
(x) dA =
in x counted as positive and those moving nega
tively in x counted
g.
as negative
The approximating form for this function is chosen by the analogy of the
In the case of par
over-all neutron movement to a diffusion process.1
ticles wandering through a medium in such a manner that they suffer
many collisions in traveling over distances of interest, and emerge from
each collision with a nearly isotropic distribution in direction, it has been
found that the assumption that the net current be proportional to the
negative gradient of the particle concentration leads to a correct descrip
tion of many aspects of the process. In the one-dimensional neutron
problem under consideration, then, we might describe the behavior by
the assumption
Jx(x)
- Dv
n(x)
(5.9)
This is the form of Fick's law for the diffusion of gases.

The form of the
proportionality constant in (5.9) is chosen for convenience in introducing
the neutron flux; thus,
Jx(x)
-D^(x)
1 P. M. Morse and H. Feshbach, "Methods of Theoretical

Physics," Part I,
McGraw-Hill Book Company, Inc., New York, 1953.
(5.10)
p. 173,
REACTOR ANALYSIS
164
[CHAP. 5
= vn. Note that the use of (5.9) or (5.10) implies that in a region
in which the neutron density is spatially constant there is no net current,
since
i.e.,
Jx(x)
= 0
for n(x)
= n0, a constant
(5.11)
This result is expected physically, since as many neutrons move through

an element of area dA from the right as from the left.
It must be emphasized that (5. 10) is not rigorously correct for the neu
tron case, and it will be inapplicable to some cases of practical interest.
On the other hand, there are many situations in which it may be used
In any event, the neutron application is intuitively appeal
profitably.
because
of
the analogous situation in gases; however, in the gas there
ing
is a pressure, due to molecule-molecule collisions, which is higher in the
more densely populated region, and this pressure causes a net driving
In the case of neutrons there are
force toward the low-density regions.
virtually no neutron-neutron collisions; the drift from the high-density

regions to the low-density regions is a purely statistical effect (which is

also present in the gas case) and is due to the fact that there are more
neutrons available for the movement from high density to low than for
the reverse movement.
The constant of proportionality D which appears in (5.9) and (5.10) is
called the diffusion coefficient.
In general it will be a function of the
nuclear properties of the medium (and therefore a spatially dependent
If we
function in nonhomogeneous systems) and of the neutron speed.
define the net neutron current in the units "neutrons per unit area per
unit time," it follows that the diffusion

This definition differs from the
length.
In the gas problems
diffusion problems.
"length squared divided by time." Thus
D has the units of

one customarily used in gasthe coefficient 3) has the units
if the current of gas particles is
Evidently the
problem, 5) = vD.
coefficient
defined analogously to the neutron

particle speed v is absorbed into the definition of the proportionality con
stant in the gas problem.
A detailed development of the concept of neutron current given in
Chap.
shows that
(5.12)
where X,, is called the transport mean free path.

In the usual way, we
define a macroscopic cross section to correspond to X,r; thus
2.Mo
(5.13)
where /no is given by (4.30).

This expression for 2tr is derived in Chap. 7.
It is of interest to note that in moderator-type materials wherein Zt ~ 2,,
DIFFUSION THEORY
SEC. 5.1]
2* ~ 2,(1
mo);
165
or, in terms of the mean free paths,

Xtr
~
1
> X.
(5.14)
AK>
which tells us that the transport mean free path is at least as large as the
scattering mean free path and increases with increasing values of /i0, the
Since
average cosine of the scattering angle in the laboratory system.
find
that
the
large values of mo correspond to small scattering angles, we
flow of neutrons is aided by forwardness in the scattering process. Thus
one would expect, in very general terms, that low-nuclear-mass materials
would support larger currents for a given neutron flux (density) gradient
[see (4.30)].
Some values' of the diffusion coefficient for various materials of interest

to reactor technology are listed in Table 5.1.
Table
5.1
Thermal
Density,
Material

Diffusion Lengths
g/cmJ
Water
Heavy water*
1.00
1.10
and
Diffusion Coefficients"
Diffusion
length L,
cm
2.85
171
0.16% H20
116
1 .85
2.69
1.60
GBF
20.8
29.0
54.4
52.0
AGOT
Diffusion
coefficient
D, cm
Absorption
cross section,
0.16
0.66 barns
0.800
0.883
0.477
0.300
0.93 mb
9 mb
1 .043
4 .4
1.040
4.8 mb
mb
data were obtained from U.S. Atomic Energy Commission, "Reactor Hand
1.5.4, p. 487, Document AECD-3645, March, 1955.
should be noted that L for D20 is a very sensitive function of the H20 content.
These
book," vol. I, Table

1
It
The concept of net current introduced above may be used to obtain an

analytical expression for the leakage term 2 from an elementary volume in
the medium in question.
Consider the differential volume dx = dx dy dz
shown in Fig. 5.2 which has faces normal to the x, y, and z axes.
If we
=
assume as before that n
n(x), then clearly there will be no net leakage
of neutrons across the faces dx dy and dx dz by the argument preceding
Eq. (5.11).
Net loss of
neutrons in
unit time
from di
Thus,
net current in posi
tive x direction
through face
\_dy dz at x + dx
net current in posi
tive
x direction
through face
(5.15)
dy dz a t x
1 Further discussion of these quantities will be given in Sec. 5.5 in connection with
the diffusion-length concept.
REACTOR ANALYSIS
166
Or, in differential
[chap. 5
notation
Net loss of neutrons

,
per unit volume
per unit time
dy dz
J +1JNJ,
dx)
T
z{x
T/st.
ay dz
Jz(x)
\-L z
,
(5-16)
In terms of the flux [using (5.10) and assuming that D is

Net loss of neutrons
per unit volume
per unit time
a constant]
n D-j2(x)
d2
(5.17)
dx1
Equation (5.16) shows that, in order that there be a net leakage different
from zero, there must exist a spatial variation of the net neutron current.

Fig.
5.2
Differential volume di = dx dy dz.
In the general problem the neutron density would
function of all
three space coordinates; thus, for a Cartesian system, we would define a
In this situation there would be a net
density function n = n(x,y,z).
of
As
movement
neutrons along each of the three coordinate axes.
before, we would compute a net current in the x direction /i(r) ; but now
there will also be, in general, components in the y and z directions.
These
If Jy{t) and
we compute in the same way as we did the x component.
be a
J,(i)
denote these components, and dr is a differential volume at r, then

the functions Jx, Ju, and Jz represent the flow rate of neutrons through
In dealing with these quantities
the three sets of faces of this volume.
it is convenient to combine them into
a single vector
function
J(r)
(see
DIFFUSION THEORY
SEC. 5.1]
Thus the
Fig. 5.3).
vector net current
J(r)
167
J is defined
Jx(r)i + J,(r)j + J,(r)k
The symbols i, j, and k denote unit vectors in the x, y, and

This relation may
respectively.
also be written in terms of the neu
tron flux <(r) by writing the com
ponents
(5.18)
z directions,
Jz, Jy, and Jz in the form
of (5.10); then
J(r)
+
-D
DV(i)
k(r)
(5.19)
We compute the leakage 8 from

the unit volume at r for this gen

eral case wherein n = n(r) by ap

plying the operation
(5.17) to
of
of
each
the three sets
faces of the
volume.
In this way
we
Fig.
obtain
5.3
Net neutron current.
the net leakage per unit volume at r per unit time

8 =
This expression
V-
is the generalization
= -DV24>
(5.20)
of the form (5.17) to a neutron flux
which is a function of all three space coordinates.

We will use this
analytical expression for the net leakage in (5.6). The neutron-balance
equation is then written
1 d
v
dt
*(r,0
= ,S(r,0
where we have used also (5.3).

tion reduces to
-Z>V2tf>(r)
- So*(r,0
DV*4>(r,t)
(5.21)*
In the steady-state condition, this equa

+Za<Kr)
S(r)
(5.22)*
The time-dependent function (r,l) is here replaced by the steady-state

function (r). The remainder of this chapter is concerned with the solu
tion of Eqs. (5.21) and (5.22) for various specifications of the source S
and of the geometric configuration of the medium in which the onevelocity neutrons are diffusing.
d. Neutron Partial Currents.
In much of the analysis which follows
it will be convenient to separate any given component of the net current
According to our original definition [see Eq. (5.8)] the
into two parts.
net current at any point gives the net flow rate of neutrons through a
unit area; thus, Jx is the excess of neutrons moving to the right (in the
HEACTOR
168
ANALYSIS
[CHAP. 5
of positive x) over those moving to the left.

An appropriate
description of Jx, then, would be in terms of these two currents, one
direction
toward positive x, the other toward negative
J.(r)
where
^ j+(t)
Therefore we define
x.
- j_(r)
(5.23)
Jz(t) is the net current toward positive x at space point
functions j+ and
j-
The
r.
are called the partial currents along the x axis at r.
Specifically,
number of neutrons which pass per unit time from
=
left to right (irrespective of the incident angle)
j+(r)
through a unit area normal to x axis at r
It
(5.24)
important to note that every neutron that passes through the unit
area is counted if its velocity vector has a component in the positive
x direction, regardless of the angle which its trajectory makes with this
area. The partial current j-(r) defines the movement toward negative x
in an analogous way.
The two functions
and /_ account for all the
neutrons which pass through unit area perpendicular to the x axis, and
Jx the measure of the net movement toward positive x. We now com
pute the partial current j_(r).
Consider
differential area dA in an infinite medium supporting a
Let the center of dA be the origin of
one-velocity neutron flux.
Car
tesian coordinate system such that the
axis
normal to dA.
Define
also
meas
polar coordinate system such that the co-latitude angle
ured from the positive
axis and the azimuth angle
from the positive
axis (see Fig. 5.4). Then, by definition, j-(0) dA
the number of neu
trons which pass through dA per unit time having made their last scatter
We can com
ing collision in the upper half-space in which x
positive.
pute this number by summing all the contributions from such collisions in
the half-space.
Consider then an elementary volume dx at space point
= r(r,0,^) (see Fig. 5.4).
There are 2<(r) dr scattering collisions per
Of the neutrons leaving dr from these collisions,
unit time in this volume.
directed toward dA.
We will estimate this fraction
certain fraction
on the assumption that the neutrons scattered in dr emerge with an iso
tropic distribution in angle; i.e., for the medium in question elastic
scatterings are isotropic in the laboratory system.
It
difficult to improve upon the isotropic assumption with the
exactly at this point that the
present method of approach, and
In
lost.
information about the angular distribution of the neutrons
(L),
not
in
and
the
neutrons
scattering
isotropic
the
neutron
general,
they impinged upon dr
scattered in di will emerge isotropically only
Thus,
net transport of neutrons in the region
there
isotropically.
surely invalid, and the
of dr, the assumption of isotropic scattering
6
is
\f/
is
is
is
is
is
is
a
if
if
is
is
is
it
is

is
,;'+
is
error made cannot be estimated by the theory (in this form).
Suppose,
DIFFUSION THEORY
SEC. 5.1]
169
however, that we accept it for the purpose of completing the calculation

and correct the theory later by using the results of more exact models.

We compute then the number of neutrons directed toward dA on the

The
assumption that they emerge from dr uniformly in 4ir solid angle.
probability that a neutron leaving dr isotropically is headed toward dA is
simply the projected area of dA visible at di divided by the surface area
of a sphere of radius r. The projected area of dA is seen to be cos 0 dA ;
Not all the neutrons headed for dA actually reach dA, however.
Some
of these neutrons suffer further collisions along the path and are thereby
removed from the stream either by being absorbed or by being scattered
into a new direction which does not pass through dA. We estimate this
number on the assumption that the principal source of removals is
Thus we are
scattering collisions; i.e., we neglect absorption losses.
implying that the medium in question is only slightly absorbing.
We will
show later that this is in fact the only condition under which diffusion
theory yields a good approximation of the spatial distribution of neu
If we assume that scattering collisions are the only mechanism by
trons.
which neutrons can be removed from the group directed toward dA, then
by the concepts developed in Sec. 2.2 we have [cf. Eq. (2.17)]
P({ > r) =
Thus
- P(f < r)
- - e---r)
(1
tr**
(5.25)
the probability that a neutron headed for dA traverses the

distance r without suffering a scattering collision along the way.
We
multiply this result with cos 6 dA/4wr2 to obtain the fraction of neutrons
e~s,r is
170
REACTOR ANALYSIS
leaving dr which reach dA
If
[chap. 5
we define a current
number of neutrons which pass through
j-(Q;r,6,t) di dA = dA per unit time due to last scattering
(5.26)
collisions in dt about r(r,0,^)

then for n = cos
and dr =
dr dfi
r2
we have1
d\f/,
turz'^.4>{rM) dr dA
j_(0;r,M,^) dt dA =
(5.27)
4bt2
The integral of this quantity over all r yields the contribution to the
current through dA from all neutrons which made their last scattering
collision somewhere in the differential solid angle dp df (see Fig. 5.5).
We define a second function
number of neutrons which pass through
'
(0-'
\b) d ** d\l>
dA
êr
Un^ ^me having made their last

collision somewhere in the solid
angle dp

In
2g\
scattering
and obtain
j-{0uijt)
dA
d/x
# dA
df
about the direction (nrf)
>L*!W^
dAjj tr^rjijf)
dr
(5.29)
order to proceed we must have some additional information as to the

functional form of over the positive x
Since we are dealing with
situation, such specific infor
mation is not available; however, the
presence of the exponential factor e-2,r
half-space.
a general
sm$d$d\l/
in (5.29) allows us to make an approxi

mation of the integral which does not
require
It
so detailed
a knowledge
of
d>.
is evident that because of the expo

nential in the integrand, the principal
contribution to the integral comes from
Fig.
(,#)
is sufficiently
expanding
ment
the
Scatterings from direction
5.5
is
region
of
few
lengths about the origin, i.e., about the

elementary area dA . Thus, if the flux
slowly varying, approximate results may be obtained by
4> in
(Note that this require

with the previous assumption that the medium is
Let us use, then,
a Taylor series about the origin.
consistent
slightly absorbing.)
,f>(r)
= 0(0)
+ r
V*(0) + |r2V2(0) +
For the present calculation the first two terms will suffice.
1 The
mean-free-path
parameters r,
p,
and
(5.30)
The first term
indicate the source of the current in question.
DIFFUSION THEORY
SEC. 5.1]
171
V(0)
r|V0(O)| cos
(5.31)
/3,
of (5.30) is the flux at dA ; the second term is the first derivative correction
for the displacement r. If the angle between the vectors r and V> is
then
designated
Fig. 5.4). The substitution of the first two terms of (5.30) into
(5.29) along with the expression (5.31) yields
fi
(see
dn dyf/Zs
dA
dA
dn dip
|V(0)| cos
*(0) +
4x
cos
re-*\ dr
g"
[(0)c~z-r + |V>(0)|
></;
4-K
(5.32)
2.
is
x)
\p
The integral of this quantity over all

the contribution to the current
due to all neutrons which made their
at dA (in direction of negative
last scattering collision in the conical shell dp
We define then
(see Fig. 5.6).
of neutrons
dA
(5.33)
nave ma{je their last

scattering collision in
the shell d\i about
y.
'
J-K(0-
'j-{0)v,
j*
by
cos
4>y
we denote
and
angles of the two vectors
the co-latitude and azimuth angles
sin
sin
cos
(\f>v
yp)
= cos
cos
8Z
Fig. 5.4), then

+
V (see
in terms of the direction

8X
V.
cos
Fig. 5.6 Scatterings from

conical shell dp.
(5.34)
M|o2'co80d^]
and
If
We can write
of
^[2x*(O)
8Z
rfM
/3
i_(0;M)
*)
and
(5.35)
dfl
ym
M^cosj,
d,
;_(0;M)
Substitution into (5.34) yields

A
{5M)
Equation (5.36) gives the current through unit area normal to the x axis
(located at the origin) due to all last scatterings in a cone of width dp
about the x axis. This result will be very useful for computing the cur
azimuthally symmetric.
rent in systems in which the neutron density
is

number
passing through dA
Per um* time which
REACTOR ANALYSIS
172
A final integration of (5.36) over all
[chap. 5
in the upper half-space gives the
j_(0;M) d
0(0) +
JT1
j_(0)
|V0(O)| cos
0X
partial current through unit area in the direction of negative
x.
Thus,
(5.37)
Fio.
5.7
Coordinate system for negative x half-space.
definition of the direction angles of the vector quantities

Thus, we take
involved.
J+(0;r,,*) dA = Z.*(r,M,*)e-
1^
be taken in the
dA
(5.38)
4irrJ
y.
is
where the absolute magnitude sign

introduced in order to account for
the fact that, throughout the negative
takes on the values
half-space,
to
The contributions from the solid angle dfi d\j/ are
(see Fig. 5.7).
1
(5.39)
dn
0(0) +
|V0(O)|m cos
6.
Im|
j+(0;n) d/i
and from the conical shell dp

(5.40)
*(0)
rfM
|V0(O)| cos
t>2.
cos
0,^0(0)
s.
ex
0(0)
|M|
j+(0) =
Finally

is
The computation of the partial current j+{0) may be carried out in the
same way.
The procedure
straightforward; however, some care must
(5.41)
DIFFUSION THEORY
SEC. 5.1]
173
These expressions for the partial currents
(5.37) and (5.41) may be

used to compute the net current according to the statement (5.23).
Thus, for an arbitrary space point r in an infinite medium, we obtain
32.
9X
(5.42)
makes with the positive x axis.

6X is the angle which the vector
the flux is a function of only one space coordinate, say x, the angle
where
If
cos
|V<(r)|
Bz = 0,
and
(x);
<(r)
= 4>'{x), and
moreover, |V0(r)|
/,(r)
Jx(x)
- r
4>'{x)
(5.43)
A comparison of this result with (5.10) shows that the elementary model
used in the calculation of (5.37) and (5.41) yields an expression for the
net current which differs from that obtained by a more elaborate theory
through
the constant of proportionality.
Whereas the transport theory gives for the
diffusion coefficient [see Eqs. (5.12) and
(5.13)]
3(2,
2,/io)
(5.44)
diffusion theory gives
(5.45)
32.
Fiq.
5.8
Gradient of neutron
Thus although diffusion theory is able to pre
flux.
dict the functional form of the net current,
it fails to give a correct estimate of the diffusion coefficient.
However,
the estimate (5.45) is fairly good for media of large mass number and small
absorption cross section; if both 20/2, and 1/ A are much less than unity,
then by (5.44)
D
32.(1
- 2/3A) J_
32.
1
no)
32,(1
(5.46)
In all subsequent work we will use the relation (5.44) for calculating the
diffusion coefficient.
The result (5.42) for the net current can be used to construct the
vector net current at point r.
If 8 and 02 denote the angles which
makes with the y and z axes, respectively (see Fig. 5.8), then
Jx(r)
J(i)
J,(t)
-DjV</)(r)|
-D|V(r)| cosfl,
-D|V(/>(r)i cosfl,
cos
8X
(5.47)
REACTOR ANALYSIS
174
[chap. 5
where we have used (5.44) for the constant of proportionality.

expressions (5.47) may be combined according to (5.18)
J(r)
J,(t)i
J(t)j + 7,(r)k
Z>|V^(r)|(i cos
cos 0
+ k cos
D; thus in place of (5.37) and (5.41) we will
If
<(r)
j+(T)
=j*(r)
j_(r)
be
DV(i)
written in terms
use
- ^|V*(r)|cos*.
i +(r) + | |V*(r)| cos
(5.48)
ex
0(x), we have
j+(x) =
j_(s)
It
0Z)
The partial currents may also
as is to be expected.
of
- *'(*)
i *(x)
(5.49)
*'(x)
is of interest to note that if there is no net current at a given space

= 0, then
|V
point, i.e., if
+ j
6Z
The
(i)\
j_(r)
which states that,
if
(5.50)
i4>(i)
are moving through a unit area

at r along a given direction as are moving in the opposite direction, then
the neutron flow rate is >/4.
This case was previously mentioned in
as many neutrons
Sec. 3.5.
For the remainder of this chapter we will use the following set of equa
tions in discussing the one-velocity approximation:
-DV*(t)
n
J(r)
+ 2,(r)
=
3Str
3(2t
(a)
(6)
- 22./3A)
1
We will require in addition to these equat ions

which define the specific problem.
tions is given below.
S(r)
-Z)V(r)
a set of
M
W
(5.51)*
boundary conditions
A general discussion of these condi
A complete state
General Statement of Boundary Conditions.
in the application
which
to
be
used
are
ment of the boundary conditions
e.
of the one-velocity
model is not possible on the basis of the physical con

These considerations1 must be deferred until we
cepts so far developed.
have a more detailed description of the neutron flux than presently avail
It is possible nevertheless to establish an appropriate set of
able.
A detailed description of the boundary conditions to be used in the one-velocity
model involves the methods and results of transport theory (see Chap. 7).
1
DIFFUSION THEORY
SEC. 5.1]
175
boundary conditions from the machinery of the diffusion theory which,

although lacking in details, will not be inconsistent with the results of
more accurate models.
On the basis of physical arguments we must require that the neutron

flux satisfy the following conditions:
(1) The neutron flux must be finite and nonnegative in all
regions of a medium except possibly with certain source
(5.52)
distributions
(2)
At an interface between two media with different diffusion

properties, the density (or flux) of neutrons moving in any
given direction must be a continuous function through the
(5.53)
interface
(3) At
interface (the outermost boundary of a sys

tem) there must be no neutrons returning to the medium
from any direction in the vacuum
a vacuum
(5.54)

These conditions are perfectly general and must be met, at least in approx
imate form, by whatever function is specified for the neutron-density dis
tribution. The neutron flux defined by the elementary diffusion theory
cannot meet all these conditions.
The function <(r) is much too coarse a
description of the neutron population in that it gives the track length per
unit volume per unit time of all neutrons of a single speed.
Conditions
such as (2) above require, however, a much more detailed statement, and
be replaced by less exacting requirements.
Before inquiring
the
into
nature of these approximations, let us first examine the implica
tions of the three conditions as stated.
so must
Consider condition (1) : it is clear from this requirement that all regions
of a medium which are in the vicinity of an elementary source (such as a
point source) are to be excluded.
Equation (5.51a) does not apply then
in the neighborhood of a singularity; everywhere else, however, the flux
Condition (2) requires that at an

<(r) is defined and must be bounded.
interface between two dissimilar media the density of neutrons moving in
a given direction must be identical on either side of the interface.
This
statement implies, of course, that there are no neutron sources or sinks at
the interface.
Thus (2) states that neutrons
boundaries.
Condition (3) is a special case
a neutron has left the outer boundary of the
the system.
This requirement is stated in
are not to "pile up" at such

of (2) and implies that, once
system, it will not return1 to
detail, namely, that no neu

trons return from any given direction.
That conditions (2) and (3) cannot be satisfied by the function <(r) is
entirely clear; <p(r) gives the total track length, whereas what is needed
is the track length (or density) of neutrons which are traveling in a par1 This is
of course true only if the outer boundary of the system is a simply con
nected and nonreentrant surface.
REACTOR ANALYSIS
176
[CHAP. 5
Such a function which specifies the angular depend

only from the transport equation.
ticular direction.
ence of the neutron flux can be obtained
(2) and (3) are far too detailed

and must be replaced by suitable approximations which deal with the
bulk behavior of the neutron population rather than with the detailed
The approximations to the
behavior of specific directional components.
For our present needs, then, conditions
exact boundary
(5.51) are:
conditions
which
we will use in connection
with Eqs.
The neutron flux must be finite and nonnegative in all

regions in which the diffusion equation (5.51a) applies
(2') The partial currents must be continuous through an
interface between two dissimilar media
(1)
(3)
At
(5.55)
(5.56)
a vacuum interface, the neutron flux must satisfy one of
the following two conditions : (a) The partial current from

the vacuum must be zero, or (b) the function which
describes the flux must vanish at some finite distance out
(5.57)

side the actual geometric boundary of the system
Boundary condition (1) is essentially as before. Boundary condition

(2') may be illustrated by deriving the appropriate relations for the inter
face condition in a system in which the flux is a function of only one space
coordinate.
Consider, therefore, two semi-infinite media with different
diffusion properties and with a common boundary at x = 0.
Denote
the medium to the left by the index m = 1 and that to the right by
Further to generalize the problem, assume that at the interface
m = 2.
there is a plane of isotropic sources with strength go neutrons per unit
Assume also that the presence of this source does not
area per unit time.
If j{+'(x) denotes the partial
alter the diffusion properties of the media.
neutron current in medium m at point x in the positive x direction, then
by condition (2')
J'f(0) =i?.+tf>(0)
Jll)(0)
= *<7o+jL2,(0)
('XM>
If we add these equations and introduce the expressions (5.49) for the
partial currents,
we obtain
J*(0)
- Ji(0)
= go
(5.59)
Thus if the source is removed (g0 = 0), the net current through the inter
face must be a continuous function, and it is discontinuous only in the
presence of a neutron source go ^ 0 (see Fig. 5.9). The subtraction of
Eqs. (5.58) yields
*(0) = 0,(0)
(5.60)
independent of the source strength g0, condition (2') requires

Since 0(x)
that the neutron flux be continuous through the interface.
Therefore,
DIFFUSION THEORY
SEC. 5.1]
177
satisfies an ordinary second-order differential equation, the two condi

An alternate
tions (5.59) and (5.60) can be specified independently.
statement of (2') is
(2) The neutron flux and net current must be continuous at an
interface (in the absence of a source)
It
is important to recognize that condition (2') or (2) is arbitrary and

only as an approximation to the original condition (5.53).
Although the requirement that the net current be continuous at an inter
serves
face is satisfactory both from a physical standpoint and analytically, the

The
requirement that the flux be continuous is not entirely meaningful.

x
0
Flo.
5.9
Flux and
net
current around
Vacuum interface,
Flo.
5.10
<(r)
as defined by the diffusion
interface.
difficulty arises from the fact that the flux
equation (5.51a) is not a good representation of the neutron population in

the vicinity of a material discontinuity.
In fact, as will be shown later,
this is precisely one of the situations under which diffusion theory breaks
To place a condition on a function in a region in which it most

A more satis
represents
poorly
the physics is certainly questionable.
factory interface condition for the flux, called the Serber condition, has
down.
but this condition is not useful for dealing with arbitrary

Thus condition (5.60) is generally used for want of a more
been suggested,1
geometries.
accurate and easily applied relation.
Boundary condition (3) is the approximation to the physical require
ment (5.54) that no neutrons return to the system from the vacuum along
This requirement cannot be rigorously satisfied in the
any direction.
present case in which we wish to apply the diffusion approximations
Therefore we impose the less stringent condition that the partial
(5.51).
current from the vacuum be identically zero.
Hence
;-(0) =
(5.62)
where the origin of the coordinate system is placed at the physical bound
This condition may be written in
ary of the medium (see Fig. 5.10).
1 The
Sec. 8.3).
Serber condition
is frequently
used in spherically
symmetric systems
(see
178
REACTOR ANALYSIS
terms of the flux and its gradient at the boundary.
m~~2D
[CHAP. 5
We have by (5.49)
(5-63)*
Equation (5.62) gives an acceptable boundary condition at a mediuminterface for use with the diffusion equations.
However, it
should be realized that this relation places a condition on a function which
was developed on the basis of the diffusion approximation. It will be
recalled that, in the derivation of Eq. (5.37) for the partial current, it was
assumed that the flux was a slowly varying function at the point in ques
tion and therefore could be represented by two terms of a Taylor series.
This requirement is hardly satisfied at any interface, least likely at a
In such regions higher-order terms in the series would
vacuum interface.
be required, and the use of only the first two terms, i.e., the diffusion
approximation, is at best a first estimate.
While Eq. (5.63) is quite easily applied in most problems, it has been
shown that there is an even simpler condition which affords a very good
This approximating condi
approximation in a great many useful cases.
be
on
the
basis
of
the
developed
following
tion may
observation.
Within
the interior of a medium supporting a neutron gas, for example, a reactor,
the neutron density (flux) is relatively high, since in general the neutron
sources (fissionable material) are located in this region.
Neutrons pro
duced by these sources are eventually lost by capture collisions or escapes
Thus a close inspection of the neutron population in
from the system.
the medium would reveal that there is a continuous net movement of
Moreover, this
neutrons toward the outer boundaries of the system.
net outward flow becomes more pronounced as the physical boundary
In the limit, at the outer surface, all
(vacuum interface) is approached.
the neutron velocity vectors will be directed out into the surrounding
Thus at each station, progressing outward from the interior,
vacuum.

vacuum
Whereas the neu

there is a net flow of neutrons away from the center.
tron population in the central regions is augmented by neutrons returning
from scattering collisions in regions beyond, the outermost portions of
Thus
the system are supplied only by the neutron sources in the center.
the neutron density falls off rapidly as the surface of the system is
approached; however, it does not vanish entirely inasmuch as there is
always a finite number of neutrons passing through
the outer layers of
the system.
It would appear then that one could establish a reasonable approxima
tion to the vacuum interface condition by requiring that the function
representing the neutron density vanish at some appropriate point near
Note that our requirement is that the flux

the physical boundary.
The
vanish in the vicinity of the boundary and not at the boundary.
problem, of course, is to determine the distance of this point from the
THEORY
DIFFUSION'
SEC. 5.1]
179
surface such that the resulting statement of the boundary condition best
Let us call this distance e, and
approximates the exact requirement.
suppose that in a typical problem the general solution to the diffusion
is known and that one of the arbitrary constants has been
eliminated by an appropriate boundary condition at some interior point.
equation
The vacuum interface condition may then

remaining
constant.
If
the solution
to eliminate the
above is 0(x), then we
be applied
mentioned
will require that

0(e) = 0
(5.64)
where for convenience we have placed the origin at the surface (see Fig.
If the approximating form of the flux (x), which is now deter
5.10).
mined everywhere in the medium, is to be a good estimate of the actual
flux, then
at x = 0.
it
is necessary to satisfy
Now consider 0(0); if
condition (5.63)
0(t) = 0(0) + 0'(O) +

But since
as closely as possible
small, we can write
is sufficiently
(5.65)
0(e) has been set equal to zero, (5.65) gives
-7
0'(O) =
0(0)
(5-66)
By comparison with (5.63) it is seen that the best value of

condition that the expansion (5.65) be valid, is
t =
in which the last step follows
2D
$\u
e,
under the
(5.67)
from (5.12).
We may summarize
as
follows: to the extent that the equation

0(xo + |Xtr) = 0(xo) + 0'(xo)|Xtr
holds (here we take the coordinate
accurate boundary condition
j-{x0)
may
be replaced
of the boundary to be xo), the more

= 0
(5.68)
by the condition
0(zo + |Xtr)
=0
or
0(xo
= 0
(5.69)
It
is important to note that (5.69) and (5.68) are in fact different condi
tions and, if applied separately to the same system, will yield different
solutions for the flux 0. Although these solutions will differ only slightly
in those situations wherein the diffusion theory is applicable, they will by

no means be identical. 1
The quantity is often called the extrapolation distance, or extrapolation
length, because the analytical solution to the flux is extrapolated beyond
the physical boundary of the medium into a region in which the diffusion
1
An illustration
is given in Sec. 5.3a.
REACTOR ANALYSIS
180
[chap. 5
Since (5.65) includes terms only up to the first

equation does not apply.
order, this process is sometimes referred to as "linear extrapolation."
Note that the use of (5.69) implies that the medium extends out to point
(xo
out to that point; how

device for obtaining a good approximation to the
^Xlr), since the diffusion equation is applied
ever, this is merely a

flux inside the actual medium.
For many
particularly large thermal reactors, the use of the

extrapolation distance, which is small in these cases in comparison with
the dimensions of the reactor, has given very good results; the approxi
mating condition (5.69) will be consistently used throughout this work
to replace the more accurate condition (5.68).
For convenience we shall
also use the notation
cases,
xo
(5.70)
The estimate obtained above for e (5.67) is based on the diffusion theory.
Other estimates can be obtained with the application of more accurate
A somewhat better number than
models, such as transport theory.1
(5.67) is

= 0.7104X,,
(5.71)*
Further, Eqs. (5.67) and (5.71) are each specifically calculated for planar
boundaries; but they are approximately correct for boundaries of small
For boundaries of increasingly greater curvature, e monocurvature.
tonically approaches a limiting value of ^X(r.
We conclude the present section with a summary of the diffusion-theory
boundary conditions which we will generally apply throughout the
remainder of this work. These conditions are the statements (5.55),
(5.61),
and
(5.69):
(a) The neutron flux must be finite and nonnegative in all

regions in which the diffusion equation applies
(b) The flux and net current must be continuous at an inter
face between two different media
(c) The flux must vanish at the extrapolated outer boundary
of the system
In using these conditions in connection with the diffusion theory, it must

that they are only approximations within the limita
tions of the theory to the precise statements (5.52), (5.53), and (5.54).
be borne in mind
5.2
Elementary Source in Infinite Media
Plane Source.
We introduce the application of the diffusion equa

tions to the determination of the spatial distribution of neutrons by cona.
1 S.
Frankel and E. Nelson, "Methods of Treatment
Equations," AECD-3497, September, 1953.
of Displacement
Integral
DIFFUSION THEORY
8EC. 5.2]
181
The
sidering infinite media which contain various elementary sources.
results obtained in these calculations are applied later to neutron-diffu
sion problems which involve more complex source distributions and to
Initially, for the sake of simplicity, we con
systems of finite extent.
sider only steady-state problems and systems in which the spatial dis
tribution of the flux can be expressed in terms of a single space coordinate.
Thus, we begin with the diffusion of neutrons from isotropic plane, point,
and line sources.
In the
case of the plane source, we have a simple
rectilinear system described by a single coordinate x; in the case of the

point source, we have a spherically symmetric system; and in the case of
These analyses are
the line source, the problem has axial symmetry.
followed by an elementary development of the method of Green's func
tion.
The section is concluded with some applications of the method to
the simple singular sources and, finally, to the general problem with arbi
trary source distribution.
The appropriate differential equations for the neutron-diffusion prob
lem in rectilinear coordinates is obtained from (5.51) and takes the form
(a)
(b)
We specialize this system to the
case of a plane source in an infinite

Consider then a plane of isotropic neutron sources of strength
qo neutrons per unit area per unit time (i.e., the sources emit neutrons
with equal probability in all directions from all points on the source
medium.
is,
We assume that, once a neutron is released from the source, the

plane).
source plane is no longer a special region in the medium.
The problem
is to determine the neutron-flux distribution throughout the infinite
medium in the steady state, that
long after the source has begun
emitting neutrons.
If
of
and z.
L2 ^
The general solution to (5.74) may
(x)
be
= Ae*'L
(5.74)
by
where we define the diffusion length
(5.75)*
written in the exponential form

+
4>(x)
4>"(x)
0,
placed at x = 0), then clearly the flux must be independent

Thus 4>{x,y,z) 4>(x), and for x ^
(5.73a) reduces to
is x
The concept of
plane distribution of isotropic
sources reduces the problem to finding the spatial dependence of the flux
in terms of a single coordinate, in the time-independent
case.
denotes the coordinate normal to the infinite-source
plane (which
>

= S(x,y,z)
= -D^(x,y,z)
+ ?a{x,y,z)
J(x,y,z)
-DV*4>(x,y,z)
Be-'11-
(5.76)
REACTOR ANALYSIS
182
[CHAP. 5
We note however that

and applies only to regions away from the origin.
the system is symmetric about x = 0; thus we may consider the solution
for x > 0 alone. Boundary condition (5.72a) requires that (x) remain
bounded for all x. In order that this condition be satisfied for arbi
trarily large x, we must take A s 0.
for x > 0. A more general form which
We have, then,
(x)
= Be~zlL
to x ^ 0 is
is applicable
The constant B is evaluated by applying the special boundary condition

In the present situation, the source is isotropic ; therefore, exactly
(5.59).
one-half the neutrons must be released to each side. Thus,
lim
1*1-0
J(x)
(5.77)
%q0
yields,
in taking the limit
\x\
as
(5.74)
(5.516)
J(x) from (5.10) and note that this is the specialization of

to the present problem.
Substitution of this result into (5.77)
is, 0,
We compute
4>(xA
*(*)
to
(5.78)*
2D
This expression gives the spatial dis

tribution of neutrons in an infinite me
dium with
plane isotropic source (see
Fig. 5.11).
the neutron sources
b. Point Source.
are concentrated at a point, and the emis
If
Source plane-
Fig.
Flux distribution
5.11
to plane source.
due
24>
independent of the angles
sin2
case,
is
For the present
4>
(5.79)
and ^;
fir2
[r*(r)]
In the usual way, we define
(5.80)
F"(r)
*(r)
=0
1
and (5.79) reduces to
(r),
. 4>\
ae{sm6de)
sin
7^
so
r2
>
0.
for
dr2
r2
is
is
sion
is
isotropic, then the problem

convenient in this case to describe the
It
spherically symmetric.
of
polar coordinates with the origin at the
in
terms
spatial distribution
source; therefore, we take >(r) = 4>(r,d,\p), and (5.51a) may be written

The solution
Lq0/2D.
therefore,
r(r)
(5.80)
and obtain from
(5.81)
DIFFUSION THEORY
183
The general solution to (5.80)
which is in the form of (5.74).
is,
sec. 5.2]
there
(Ae"L
<p(r) =
fore,
Be-"L)
(5.82)
so as to satisfy
required that
lim 4irrV(r)
In the limit
-D*'(r)
J(r)
(1
obtained from (5.516),
is
J(r)
(5.83)
q0
r-*0
where
>
is
If
is
is
it
boundary condition
determined by the condition on the source
The constant B
(5.72a).
strength.
go denotes the rate of neutron emission from the origin
(neutrons per unit time), then the total net outflow of neutrons from the
source
given by the limit of the product of the net current in the radial
direction and the area of sphere as the radius 0. Thus
As before,
r/L)e-"'
qa.
Hr)
(5.84).
q-~
is
an alternate boundary condition to (5.83) which gives the

same results.
Since we are dealing with the steady-state condition, the
neutrons supplied by the source during unit time must balance the total
There
throughout the medium during unit time. The only

mechanism of neutron loss in the infinite-medium one-velocity system
absorption; therefore,
2<f.(r)
qo=
For the chosen flux
(r)
dr =
Jq"
is
loss of neutrons
= Be~TlL/r, we
20< dr (5.85)
obtain
g0
dr
(5.85)
= 4irSoL2B = <brDB,
in agreement with the previous result.

The analogous condition
could have been used in the planar problem discussed in Sec. 5.2a or in
the cylindrically symmetric problem to follow.
which
is
c. Line Source.
third form of the one-dimensional problem of
neutron-flux distribution may be derived by considering
line distribu
For
of
convenience
we
tion
take the diffusion equation
isotropic sources.
(5.51a) in cylindrical coordinates and define <(r) = (p,<p,z).
pdp\pd;)

is,
0,
we obtain from (5.83) the relation 4xDB =

The neutron flux due to an isotropic point source
therefore,
as
?W+M-L*-0
(5-86)
For an infinite line of uniform sources and axial symmetry, (5.86) reduces
to
REACTOR ANALYSIS
184
[chap. 5
where p denotes the radial coordinate measured from the line of sources
If we define a new variable x = p/L, this equation takes
at the origin.
the standard form of the modified Bessel equation1
4>"(x)
The general solution
- <p'(x)
- (x)
= 0
(5.88)
is
4>(x)
AI0(x) + BK0(x)
(5.89)
where Iq(x) and K0(x) are the modified Bessel functions of order zero of
Complete tabulations of these
the first and second kind, respectively.
For convenience we
functions are available in the standard sources.2
Aw

Fig.
12
5.12
K0(x)
T~
Q\
|
3
Zero-order modified Bessel functions.
give below the relation between these functions and the regular functions
and Y of the first and second kind.
In general, for the nth-order
functions,
7,(0
iJ.(U)
Kn(t)
\ i+Vn(ti)
iYnffl]
- \ *+lH<M
where H{u{it) is the Hankel function of the first kind and t is real.
The application of the boundary condition (5.72a) shows that the arbi
trary constant A must be taken identically zero, since /o(<) becomes
arbitrarily large as / * oo (see Fig. 5.12). The value of B is obtained by
imposing the source condition [cf. Eq. (5.83)].
lim 2icpJ{p) =
p-0
g0
(5.91)
'See, for example, F. B. Hildebrand, "Advanced Calculus for Engineers," pp.

Prentice-Hall, Englewood Cliffs, N.J., 1950; H. Bateman, "Partial Differ
ential Equations," pp. ;W8 103, Dover Publications, New York, 1944.
* See, for example, E. Jahnke and F.
Emde, "Tables of Functions," Dover Publica
tions, New York, 1945.
160-161,
DIFFUSION THEORY
SEC. 5.2]
185
where the net current J{p) is given by
It
= -D4>'{P)
J(P)
can be shown that
and it follows that B = g0/2irD. The flux distribution due to an iso

tropic line source in an infinite medium is therefore
q0K0{p/L)
*(p) =
(5.92)*
2*D
Method. In the foregoing discussion, the steady-state

neutron flux in an infinite medium was obtained for three different source
distributions, namely, plane source,
point source, and line source. With
d.
Kernel
it is possible to de
termine the steady-state flux from
any distribution of sources in an in
As an introduction
finite medium.
to the technique for calculating the
flux for more general source distri
butions,
consider the case of two
point sources located at the space

points Ti and r2 and of strengths gi
and q2 (see Fig. 5.13).
The neutron
flux at the field point r due to the
point source at ri may be deter
from
Note that
(5.84).
is
in
terms
of the dis
given
(5.84)
tance between the field point and
mined
if the source
medium, the steady-state
(t = 1, 2) were present in the

flux <k(r) would be given by
*.(r)
If the
point
9, alone
-\t-tl\/L
4x/)|r
(5.93)
ti\
the source; thus,
Thus the total flux at
is
are neglected and
if
interactions
it
assumed that
the presence of a point source does not affect the diffusion properties of
the medium in its immediate neighborhood, the behavior of the neutron
flux from either source will be unaffected by the presence of the other.
neutron-neutron
is

these results
given by the sum
(r)
,(r)
4>2(r)
(5.94)
REACTOR ANALYSIS
186
[chap. 5
, i,
Formally, this result follows from the fact that the diffusion equation, of
which
and #2 are solutions, is linear; therefore, the solutions may
be superimposed.
Since the above line of reasoning for two point sources would not be
altered for three, four, or any number of point sources, it is concluded
that, if N point sources located at points r,(i = 1, 2, 3,
N) and of
...
...
strengths qt(i = 1, 2, 3,
N) are present in the infinite medium, the
total flux at a field point r due to all N would be given by
if
4*D\r
is
if
isotropically per unit time in volume dr'
then we define
r'|
r',
track length per unit volume per unit time at
to the S(i') dr' neutrons released per unit time

from the volume dr' about
due
(5.96)
r'
By the arguments given above for the discrete distribution of sources, we

r
can compute the total track length per unit volume per unit time at
due to all sources in the medium by integrating (5.96) over all space.
is
is
The method introduced here

known as the method of Green's func
in fact the Green's
tions or the kernel method.1
The function (5.96)
function, or elementary solution, to the diffusion equation.
The general
solution to the diffusion equation for
system with an arbitrary source
distribution can
be constructed
from these elementary solutions by apply
ing (5.97).
In order to demonstrate the application of this general method to neu
tron diffusion in infinite media, the plane-source and line-source dis
tribution problems (Sees. 5.2a and 5.2c) will be treated using (5.97).
a
Plane-source Distribution.
The plane source of neutrons may be
uniform distribution of point sources over an infinite plane.
regarded as
Consider then the case in which isotropic point sources are distributed
Part
chap.
7.
See, for example, Morse and Feshbach, op. cit.,
I,
0,
uniformly over a plane at x =

such that q<>neutrons are released per
unit time per unit area from all points of this plane. The flux at any
1

*(r'f dl
ii\
discrete distribution of sources may be gen

the total
distribution. Thus
S(r') dr'
eralized to
continuous
number of neutrons produced
about space point
4*>|r
This general result due to
gr^-lr- r.l/L
sec.
DIFFUSION THEORY
5.2]
point x may
on the plane
be obtained
(see
187
by summing the contributions from every point

The contribution to the flux at x from an
Fig. 5.14).
Fig.
5.14
Plane-source distribution.
= track length per unit volume per unit time at x due

to sources in dA at distance
from
s
(x;s) dA
dA
e~,IL
q<>
The total flux >(x)
the integral of this quantity over the entire plane
= 0.
dA
J>(x;) dA
pdpdd and note that
n-
e-'ILp dp
dO
we write
reduces to
qo_
+ xJ, then this integral
f"
*(*)
sJ
(5.98)
AwDs
is
_
=
If
ID
2D jM
JM
e~'IL ds
Lqoe~ \z\IL
2D
in agreement with (5.78).

which
The integral relation (5.97) may be used more directly in achieving this
result for the planar distribution
we specify the source strength in
terms of the delta function (cf. Sec. 4.6d).
We choose for the source
function S(i), which appears in (5.97), the form
if
is
S(r) = A5(x)
(5.99)
the Dirac delta function; the constant

we compute by
requiring that the integral of the source distribution (5.99) over
unit
area of source plane and over all x be go.
Thus,
a
S(x) dx
where 8(x)
is

is,
area dA whose polar coordinates are (p,0) in the x = 0 plane and which
is a distance s from the field point
from (5.84),
dy
dz =
q0
REACTOR ANALYSIS
188
= g0.
...
4^5
&(x)
g0
into (5.97) yields
(*') e-b-fV1- dx' dy' dz'"
\r=T\
r',
If we write the distance

we have denned dr' = dx' dy' dz'.
in terms of the coordinates of the two points
and
then
[z2
^DJJ
[x*
(tf
7/')2
y'Y +
{z
- xT +(y-
[(x
exp
-z')>]i
dx'dy'dz'
-z'Y
- z')*]l\
- 2'p]T
(f
r'|
|r
f
]
go
*(r)
where
The substitution of S(t)
(Z
and
[CHAP. 5
dy' dz'
S(r)
5(x) 8(y)
(5.100)
if
is a
As before, we evaluate the constant

by requiring that the integral of
Thus
unit dimension of source be equal to the specified strength.
over
the number of neutrons released per unit time from unit length of
go
go
exp
A
p'
sin 8'Y~+Jz
sin 0')2
l-*.[p>+(z-z')*)^dz'
zyv
[p2
+(z- z')2]*
p'
sin
sin
8
(p
("
8'Y +
(p
p'
cos
cos 0')2
(2
= g0, yields
*(p') dp'
[(p cos
dz'
p'
Up cos
6
exp
{
fu
f"
*(pA*)
jlm
The substitution of (5.100) into (5.97), using
Jo
lirp
Jq
J*dz
S(y) dy
Jo
6(x) dx
S(t) dr
J'
go
line, then
Jo

is
By transforming the double integral in this relation to polar coordinates

in the y'z' plane (see Fig. 5.14), the result (5.78)
obtained directly.
In the case of the line source in the infinite
Line-source Distribution.
medium, the distribution of point sources may be chosen as
*')2]}
d8'
DIFFUSION THEORY
This form may
-* (p)
If
be recognized as the
exp
cosh
Hankel function of the first kind of
"
exp (iz cosh
t)
zero:1
H\,"(z)
we
z')/ p, it follows that
eft
identify p/L = iz, then
However,
by (5.90),
ffi(ix)
(p)
?-
order
*)
(p,e,z)
= (z
189
(_
we define sinh
If
/_
SEC. 5.3]
K0(x)
is
which
a.
q0K0(p/L)
2*D
in agreement with the solution (5.92).

Diffusion in Finite Media
5.3
The Vacuum-interface Condition.
The purpose of the present
sec
5.
is
to display some of the features of the approximate boundary condi

tion
tions developed in Sec.
and to show how these approximations may be
utilized to specify the situation at the outer surface of a system and at
interfaces within.
We begin with a simple calculation which involves
the application of the vacuum-interface condition.
In the analysis which
follows we examine the steady-state flux distribution in an infinite-slab
is
This calculation
geometry due to a plane source placed on one face.
performed for two different outside boundary conditions on the sourcefree face of the slab, namely, condition (5.57a), which requires that the
partial current of neutrons from the vacuum be zero, and condition

(5.576), which requires that the flux vanish at the extrapolated boundary
The results from these two calculations are compared
[see also (5.69)].
so as to show some of the implications of the diffusion theory as applied
then,
at an outside face of
E. T. Copson, "Theory of Functions of

Oxford University Press, London, 1935.
system.
homogeneous slab of width a in the x direction
a
Consider,
conditions
a
to the boundary

Therefore
Complex Variable,"
and
p. 344, example 22,
REACTOR ANALYSIS
190
infinite in the
[chap. 5
y and z directions, as shown in Fig. 5.15.

x = 0, there is specified a
At the plane
uniform distri
bution of isotropic sources of strength q0

Be
neutrons per unit area per unit time.
cause of the finite extent of the x dimen
sion and the nonsymmetry of the problem,
there will be a nonuniform distribution of
the neutron population along the x direc
tion; the dependence on the y and z direc
Fio.
5.15
width a.
Infinite
slab
ential equation is [see
tions for given x, however, will be uni

Thus, the flux is given by the
form.
function 4>{x), and the appropriate differ(5.75) for L2]
of
*"(*)
The general solution may

4>{x)
= AieZIL
be
- 1 *(x)
= 0
written
Ble~xlL
or
4>(x)
A j sinh
y + Bi
cosh j-
We seek two specific solutions to this problem, one based on boundary

condition (5.57a) and the other on (5.576). In each case we require also
a condition on the source plane which, when used with either (5.57a) or
(5.576), yields the arbitrary constants A\ and Bx or A% and B2.
The boundary conditions to be used in the first case, namely, that the
return current from the vacuum be zero at the source-free face of the slab,
are
(1) >_(o) = 0
(2)
lim j+(x)
x->0
*?
(5.101)
We use for the partial currents the expressions (5.49). It is easily shown
that the application of these conditions to the second of the general solu
tions given above yields the flux i(x):
a. r~\
*<*)
~
x)i
- ^ rshm *( ~ j) + 2Dx csh x(q
[~(2^ + l)V- {2Dx- \ye- \
where we have introduced
c=
u\o\
(5-102)
the symbol
x =
(5.103)*
The extrapolated boundary-condition approximation (5.576) to the

condition (5.57a) yields the set of conditions
(1)
0(a) = 0
(2) lim
T 0
(5.104)
sec.
DIFFUSION THEORY
5.3]
191
is,
where a = a + 0.7104X,r according to (5.70).

The only change here
of course, in condition (1), where the current condition has been replaced
by the requirement that the flux vanish at the extrapolated boundary
x = a. Let the solution of the diffusion equation given above which cor
responds to this set of conditions be n(x) then, one obtains, upon evalu
ating the constants A2 and B2, the result
(5105)
sinh*(a-z)
sinh xa + 2Dx cosh xa
the results
t>i
and
comparing
is
if
We can obtain some insight into the physical requirements which must
the diffusion theory
to be valid in
given situation by
be satisfied
obtained
above.
Specifically,
let us
is
1.
compare the expressions (5.102) and (5.105) in the case when 2Dx
and 2D /a <.
It readily shown that, under this restriction, both of
*,(*)
sinh x(d
...
(5-106)
1.
Thus the solutions of the present problem based on the accurate and
approximate statements of the vacuum-interface condition are in good
Let us examine now
agreement provided that 2Dx <3C and 2D/a
the consequences of these requirements.
We have from (5.12) and (5.75)
(5.107)
3Sa
2D
(2)
(5.108)
1
(1) 2Dx
Therefore
if
is
is
is
The first requirement, then,

that the absorption cross section be small
relative to the transport cross section.
The second requirement
that
the transport mean free path be small in comparison to a, the character
istic dimension (size) of the system. We find therefore that, in order for
the results based on the extrapolated boundary condition to match well
the results based on the no-return current condition, the absorption cross
section must be small as compared with the scattering cross section [note
that for most nuclei 2 a* 2, by (5.46)]. Moreover, the scattering mean
free path (again using X,r ~ X.) must be appreciably smaller than the
diffusion
size of the system.
This condition
in fact
prerequisite
theory
to be valid in
given situation. We conclude then that the
is

2q0
x)
these solutions reduce to the same form, namely,
REACTOR ANALYSIS
192
approximate
are equivalent
[CHAP. 0
and accurate statements of the vacuum-interface condition

only if the medium is slightly absorbing and if the system
is many mean free paths wide (i.e., diffusion
theory must be valid).

result (5.106) may be used to illustrate the effect of
Suppose that,
the diffusion of neutrons on neutron-migration problems.
as in Fig. 2.6, n0 neutrons per unit area per unit time (of the one-velocity
of diffusion) are incident upon one face of an infinite slab of thickness a.
The approximate
1.0
0.8
.-J
\V
.
V
0.6
E
(/>0.4
D=l
i,
cm
20 = 0.01 cm"1
1-
ADx
0.2
si nh >a
20
10
30

a, cm
Fio.
Transmittance
5.16
of slab.
of this slab on the basis of the follow

ing two physical models: (1) Neutrons are not scattered in the slab but
move straight through, unless they experience a capture collision, in
which case they are removed from the beam.
(2) Neutrons enter a
diffusion process the instant they strike the slab.
Let
us compute the transmittance1
For
case 2, the results of the above analysis apply with a source strength
to replace the factor q0/2 which appears in (5.106) [see (2) in (5.101)].
If the number of neutrons transmitted is counted as those received by a
noncollimated counter on the far side of the slab, then for case 2 this is
n0
precisely the net current from the back face. By assuming that the
extrapolation distance is small in comparison with the slab width [which
is implied in the inequality (2) of (5.108)], the net current calculated at
the extrapolated back face will very closely approximate the actual
Then, in the two cases, the transmittance T is computed as
current.
follows
1 :
r,
Case 2:
Tt
=.
Case
Woe
-s.
no
e~z'
-D'(n) _
ADx
sinh xa
ADx
sinh xa
A comparison of these results is given in Fig. 5.10 for the case D = 1 cm

and 2 = 0.01 cm-1 as a function of slab width. The rapid decrease in
1 The
fraction of incident neutrons transmitted through the slab.
sec.
DIFFUSION THEORY
5.3]
193
the transmittance for the diffusion case may be understood when it is

recalled that a diffusing neutron travels a long tortuous path in passing
through the slab and is therefore exposed to capture collisions over a dis
tance much greater than the actual slab width. The nonscattering
medium of case 1, on the other hand, has the distance a in which to cap
ture the neutron.
Note that the graph of the transmittance for the
diffusion medium is broken at the lower limit for small slab widths in as
much as diffusion theory does not apply for a/L
1 [see (5.108)].
b. Elementary
Sources in Finite Geometry.
The treatment above
illustrates the general method for finding the spatial variation of the
steady-state neutron flux in finite media into which neutrons are intro
duced from localized sources.
The mathematical problem is the solution
of the following boundary- value problem:
Differential equation:
DV2<(r)
+ 20*(r)
= 0
(to apply everywhere

except at source points)
(5.109)
Boundary conditions:

(1)
= 0 everywhere
on the extrapolated surface of the\

(i.e., the medium-vacuum interface) except
at]
I
points where there are localized sources
4>
medium
(2)
Net number of neutrons moving away from each source)

point per unit time (in terms of the neutron flux in thatl
region)
must be equal to the source strength
at that]
point
Table
(5.110)
5.2 gives the results of the
application of these conditions to
a few
Each of these single-medium
systems is taken to be composed

The geometries of Table 5.2
length is L.
involve only one space coordinate ; the boundary- value problem described
by the equations (5.109) and (5.110) may, of course, be considered for
more complex geometries (e.g., a finite block of material with a point
source at the geometric center); these involve the technique of the
systems.
of a material whose diffusion
separation
of variables.
One such problem is treated in Sec. 5.5, which
deals with the measurement of the diffusion length.

c. Albedo Concept.
It was shown in Sec. o.le that the interface
boundary condition between two dissimilar media could be expressed in
This approach is generally

terms of the net current and flux [see (5.61)].
useful only if the regions on either side of the interface are only slightly
absorbing and several scattering mean free paths in width; i.e., each
It is to be
region must satisfy the requirements of diffusion theory.
recognized, however, that even in this case the diffusion theory does not
yield an accurate representation of the neutron flux in the vicinity of the
In the case that
interface (within a distance of a few mean free paths).
REACTOR ANALYSIS
194
[chap. 5
one of the media is highly absorbing or narrow in width, the application

This model is
of the diffusion equations is even more questionable.
certainly invalid in the narrow (or highly absorbing) medium; moreover,
its
use
in the "diffusion" medium near the common boundary
Table
5.2
is also
for the Elementary Sources in Finite Geometry
Solutions
Geometry
Slab
Description of
system
Source
strength
Cylinder
Slab with planar sourceon the midplane Infinitely long cylinder with an
axially placed source line
Sphere with point source at the

center
90neutrons per unit area

per unit time
0oneutrons per unit time
plane 90 neutrons per unit length of

source line per unit time
Radius of cylinder is po

Half -width of slab is do
Radius of sphere is Ao
Source point
Explicit
(1) *(3o)
*(z)
statement of j
boundary
= 0
- *(-x)
conditions
(2)
General solu
lim
[-2V(*)1
Or) A sinb
tion to
lim (-2rpD0'(p)]
p0
- + B cosh -
*(p)
"
- go
lim [-iwr*D4t'(r)]
-.0
Ah(P/L) + BK9(fi/L),
*<0
rL
- 90
4sinh--f
Boouh-
L]
differential
The complete solutions are:
Slab:
Cylinder:
*()
Sphere:
*(r)
/So 1
2D cosh oo/L
2<rfl/o(j/i)
[/,(/) K,(p/L)
'1
rirD
,mh{-r)
tinh He/1.
x >0
- K,Q,/L)
Mp/L)\
sinh I
(*?)
It would appear then that the present machinery of the dif

fusion theory is insufficient for coping with interface situations of this
limited.1
type.
We find, however, that this is not exactly the
case.
Granted that the
1 Under these circumstances the flux near the boundary is

highly "directional,"
and the two-term approximation of diffusion theory is inadequate [see (5.30)).
DIFFU8ION THEORY
SEC. 5.3]
195
is,
theory is inapplicable to the narrow or highly absorbing region, it may

still be used as a first approximation for describing the flux in the dif
The approximation is not so good as in the case of an
fusion medium.
interface between two diffusion media, but it is still useful for our present
The remaining problem, then, is the specification of the effect
purpose.
" nondiffusion " region on the neutron population in the diffusion
of the
That
we should like to find a way to specify the boundary
region.
condition on the diffusion medium in terms of the nuclear properties of
This can be done very conveniently by means
the nondiffusion region.
We define the albedo of a given material
of the so-called albedo concept.
as
...
region
return
current from
.
. ,
incident currentL into the region
,_
(5.111)
Albedo =
is
it
information to the adjoining diffusion region by a simple statement on the

In practice, one can either measure
current condition at the interface.
by
the albedo of various materials (and geometric shapes) or compute
In
means of more sophisticated models such as the transport theory.
is
number can be obtained1 and applied directly to the

any event such
remainder of the system as a boundary condition. This approach
especially useful for treating very thin regions, such as plates, foils, or
Thus, for example, one might treat containers and pressure
shells.
vessels in
reactor.
knowledge of the albedo of these components
theories for other major portions of
concept
extremely effective for
permits one to use more elementary
is
Further, the albedo

the system.
describing the properties of thin, highly absorbing plates or rods of the
type used as nuclear-fuel elements.
If the adjoining regions are both diffusion materials, an estimate can
be obtained for the albedo through the use of the partial-current rela
as an introductory example, the case of a plane source
neutrons
per unit area per unit time placed on the face of
q0
semi-infinite slab.
the origin of the coordinate system
at the
source plane, the general solution of the diffusion equation has the form
of Eq. (5.78), namely,
tions.
Consider,
is
An elementary method for
(5.112)
by means of the partial-current condition
[see (5.49)]
j+(0) =
= Cc~ltL
*(0)
"thin"
*'(0)
9,
We evaluate the constant
(x)
If
of strength

if
it
is
is
a
measure of the reflectivity of a medium for neutrons

Thus the albedo
neutron back to the space
of a given speed, i.e., its ability to return
It clear that,
was incident on the medium in question.
from which
known, one could apply this
nondiffusion medium
the albedo of
regions will be discussed in Sec. 5.8a.
(5.113)
ANALYSIS
REACTOR
196
[chap.
This condition states, physically, that the number of neutrons striking a

unit area of the surface of the semi-infinite medium per unit time is half
the source strength (half the neutrons are lost to the void directly from
the source).
Note that this is not the net current, since some of the
neutrons which enter will, after some diffusion, be lost by leakage back
across the vacuum interface; these we have not yet accounted for.
If (5.112) is used in (5.113), we obtain the constant C; the resulting
expression for the flux is
*>
The partial neutron current
iw
out of the medium at the
The current mj+
is,
;_(0)=^(0) + ^'(0)
Albedo
(5.115)
on
(5.116)
This form for the albedo applies only to the semi-infinite medium; for
other configurations, the albedo may be computed within the limits of
denotes the position
diffusion theory from the general relation (where
of the boundary)
It
be emphasized that the diffusion-theory expression for the

even
when the adjoining media are both lightly absorbing,
in
albedo,
error since the diffusion theory does not give an accurate description of
the angular distribution of the neutrons in the vicinity of the interface.
is
should
The essential
involved in computing
parameters
the
albedo
[see
(5.117)] are the diffusion coefficient in the medium and the ratio of the
This
slope of the neutron flux to the value of the flux at the boundary.
ratio will depend upon the geometry, the diffusion coefficient, and the
To illustrate this dependence in the

introduce the relation
of the material.
the semi-infinite
medium we
2D
2A
3
absorption
case of
fh\
\D)
_~
01
into (5.116) and obtain for the albedo a
It
- 2Xlr(Sa/>)V3
(5-U8)
+ 2Xtr(V>)V3
is
0)
the albedo
unity. It
evident that for total reflection (2 =
should be noted, however, that the result for heavy absorption (Sa large)
cannot be obtained from the above relation because this result was
is

||^|g
of course, go/2; thus by (5.111) we have for the albedo

medium
of the semi-infinite
boundary is
sec.
DIFFUSION THEORY
5.3]
197
derived from the diffusion-theory estimate of the partial currents which

requires that 2 be small compared to 2<. Figure (5.17) shows a plot of
the albedo for a semi-infinite medium as computed by (5.118). The
diffusion-theory result is broken off at the higher values of 2a in accord
with the preceding remarks. The upper broken-line curve indicates how
one might expect the actual albedo of the material to vary.
S.,cm_l
Fig.
5.17
Albedo
for
Interface
semi-infinite
Fig.
Neutron crossings of interface,
5.18
is
is
/S.
is
/3
it
is,
/3.
An interesting application of the albedo concept is in the calculation of

the average number of crossings of an imaginary plane by a neutron as the
neutron diffuses through the medium.
The principal ideas involved can
be demonstrated by considering a one-dimensional system and starting
with the assumption that the neutron has made at least one crossing
(otherwise the spatial variation of the neutron flux must be taken into
In order to retain some degree of generality, it will be assumed
account).
that the boundary in question is the common face between two media of
different diffusion properties (see Fig. 5.18). Denote these media A and
B and their respective albedos a and
Thus
will be returned to
neutron which has crossed from
to
[see (5.111)].
will not return to

The probability that
of course,
In the present case we wish to compute the probability p that the
neutron makes precisely n crossings and from this determine the average
number of crossings.
Now the probability that
capture will occur
independent of the number of previous cross
during any one crossing
the joint
ings, and the probability for capture after exactly n crossings
n
crossings and of being captured before the
probability of surviving
independent, the joint
+ l)st crossing. Since each of these events
given by the product of the probabilities for the individual
probability
By this line of reasoning we find that,
the neutron starts in
events.
medium A,
if
is
is
(n
p2
p3
0)
a(l aj8(l pa*(l -
0) a)
(1
po =
p,

medium.
REACTOR ANALYSIS
198
In
- a)(o0)
o(l - 0)(a/3)
general,
p(* = 2n) = (1
p,(v = 2n
1) =
[CHAP. 5
n =
0, 1, 2,
(5.119)
These relations may be checked by computing the sum of the joint proba
bilities p for all v. The sum should be unity; thus,
00
n-0
P. = (1
00
(a/3)"
<0
+ a(l - 0)
n=0
00
(a/3)" = (1
n-0
DO
a/9)
(a/3)"
n=0
(5.120)
We compute the average number of crossings n from the relation
= 2(1
n-0
a/3)
n(a/8)" +
a(l
- 0)
n=0
n =0
oo
(c0)"
n=0
(5.121)
r-^
a
(5.122)
/3,
In the event that a =
n =
5.4
Application
as
?iL+4)
ap
1
n =
of One-velocity Model to Multiplying Media
With the remarks in

Fundamental Mode: Infinite -slab Reactor.
the preceding section as an introduction, the more general problem of
multiplying medium will now be examined.
steady-state diffusion in
As indicated in Sec. 5.1a, the application of the one-velocity approxima
the second in
series of increasingly more
tion to multiplying media
detailed models which we will utilize for determining the nuclear char
This problem will be analyzed for three elemen
acteristics of
reactor.
tary geometric configurations: the slab, the sphere, and the cylinder.
The calculation for the multiplying slab, however, will be treated in
greater detail so as better to display the principal concepts involved.
The general conclusions which can be drawn from the analysis of the slab
problem will later be used in obtaining the solutions for the sphere and
a
is
a.
cylinder.
directions and of width 2a in the x direction
(d
is
and
is
Consider, therefore, planar distributions of isotropic neutron sources in

of infinite extent in both
The medium
an infinite homogeneous slab.

y

The first series is easily identified as a|8(l a/S)-J and the second
The expression for n thereby reduces to

a/3)-1.
(1
the extrapolated
DIFFUSION THEORY
sec. 5.4]
199
the origin of the x axis is placed at the center

reduce the spatial dependence of the
is
in
x
alone.
It
assumed that all available neutrons
flux to variations
(per unit volume per unit time) arise from fission reactions described by
the term v2,, where v denotes the average number of neutrons produced
In the
per reaction and 2/ denotes the macroscopic fission cross section.
half-thickness of the slab)

of the slab.
These specifications
present calculation we will retain the time dependence of the neutron

flux in order to obtain a more general result (cf. Sec. 3.2b) ; thus we define
= (x,t).
The reduction of this result to the steady-state case will
display the motivation for selecting the particular form of the timeindependent solution in the cylinder and sphere problems.
This model in which the neutron source is taken to be vJ,f4> (which
implies that neutrons from fission appear at the one velocity of diffusion)
is not expected to apply directly to operating reactors; however, the
techniques to be used later are well illustrated by this formulation of the
problem, and the results are useful, if properly adapted, to more general
situations. We take then as our neutron-balance relation the timedependent diffusion equation (5.21) along with the assumed source term

S(r,0 = v2/#(r,0
The general one-velocity
diffusion
equation for the multiplying medium
is then
-DW+(i,t) +
(5.123)
- \ J (i,t)
Z.*(r,<) =
is the "one velocity" of diffusion (see Sec. 5.1a).

priate form for the infinite-slab problem is
where
4>(x,t)
+ BH(x,t)
=^jt *(*,)
2a
o(x)
(5.125)
(5.126)*
The solution to (5.125) must satisfy the conditions
in which
The appro
the definition
B* m
(1)
(5.124)
4>(a,t)
= 0
is some specified
(2)
(x,0)
[see (5.72)]
0o(i)
(5.127)
initial distribution.
Equation (5.125) may be solved by the separation of variables;1 there

fore, define functions F(x) and G(t) such that
0(i,O = F(x)G(t)
(5.128)
H. Margenau and G. M. Murphy, "The Mathematics of Physics and Chemistry,"

chap. 7, D. Van Nostrand Company, Inc., Princeton, N.J., 1943.
1
REACTOR ANALYSIS
200
[CHAP. 5
The substitution of this expression into (5.125) yields
*^M +
+"B>
F(x)
_c.
0
DvG(t)
(5 129)
ÂZ[)>
The arbitrary constant cs (here assumed positive) is determined by con

ditions on the time behavior of the flux. Note that a relation of this
type must be satisfied by F and G inasmuch as x and t are independent
variables, and (5.129) can hold only if c is a constant.
Thus (5.129) may
be separated into two ordinary differential equations
F"(x) +
G'(t)
(B2
=
c2)
= 0
F(x)
(5.130)
-Dvc*G(t)
(5.131)
The general solution to (5.131) is

G(t) = G0e-fl""
and if we call
02
= B* +
Go ^ G(0)
(5.132)
c1
(5.133)
we can write the general solution to (5.130) in the form

=

F(x)
cos fix
+ C sin fix
(5.134)
We evaluate the constants A and C by the application of boundary con

dition (1) from (5.127). The result is
= 0 =
= 0 =
F(a)
F(-a)
A
A
cos fid
cos fia
+ C sin fia
C sin fia
(5.135)
The successive addition and subtraction of these equations

requirements that
2,4 cos fia = 0
yield the
2C sin fia = 0
(5.136)
The only nontrivial solutions to these relations are, respectively,

fia =
| (2n
fid =
t(ti +
+
1)
1)
= ana
7a
where a =
^fl
(2n +
where 7 = x (n +
1)
(5.137)
1)
and n = 0, 1, 2, . . . . Thus solutions of the form (5.134) exist for

F(x) provided that fi has any of the values a in the cosine term and the
values 7 in the sine term.
The an and yn are the eigenvalues1 of (5.134),
and for each value of n there exists a solution F(x) given by
Fn(x)
= An cos anx
+ C sin
ynx
which are the eigenfunctions of the differential equation.
'A. Sommcrfeld, "Partial Differential

Press, Inc., New York, 1949.
Equations
in Physics,"
(5.138)
Moreover,
we
chap. V, Academic
DIFFUSION THEORY
SEC. 5.4]
201
X2
when
f)
B2
a2
B2 = w'
when
/3
7s,
c2
follows from (5.137) that

c2
then
it
/3s
note from the definition of /3 [see (5.133)] that for each value of n there
must be a corresponding value for c. Thus if we write c2 =

B2,
takes on values an
takes on values yn
0,
cos ax
Go(Ane-""
(?(<)
is
Fn{x)
C*-H
1,
for various n which must be used in the solution

solution for the diffusion
for G(t) [see (5.132)]. Hence, we construct
. . . .
equation (5.125) in the form (5.128) for each value of n =
These are the values of
sin ynx)
(5.140)
given by the sum1
4>{x,t)
Fn(x) Gn(t)
(5.141)
n=0
0(x)
= Go
if
is
is
4>(x,0)
{A cos anx + C sin yx)
(5.142)
71=0
f*
J-s
The solution of the time-dependent
of the trigonometric
cos anx dx
(5.143)
*o(x) sin yx dx
problem
Given
and any initial dis
now complete.
^rr
alio
4>o(x)
is
j0^3
/
Cn =
An =
The application of the orthogonality properties

functions yields the coefficients
system of certain nuclear properties (implied in B2)

tribution of neutrons (according to 4>0), one can compute the flux dis1

/3
if
is
is
The analysis leading up to the result (5.140) has been concerned only
with boundary conditions (1) of Eq. (5.127). The requirement
that,
this condition
to be satisfied by the spatially dependent portion of
(x,t), then may take on only the values indicated in (5.137). This
in turn places a constraint on the constant
because of the relationship
limited to the values (5.139). Given, then,
(5.133); consequently,
Note, how
value of n, o, yn and in turn X and oi may be computed.
ever, that the evaluation of X and un involves the parameter B2 which
determined by the nuclear properties of the system.
(We will consider
this matter in detail shortly.) The remaining undefined quantities in
(5.140) are the coefficients A and C. These are determined from the
application of the initial condition (2) of (5.127). Thus
we compute
(x,0) from (5.141) using (5.140), we obtain
Margenau and Murphy, op. cit., pp. 240-248.
REACTOR ANALYSIS
202
[CHAP.
| 1
n=0
An cos anx
(5.144)
oo
(x,l)
4>(x,t)~Go
>
follows, then, that as
>
(4n
3)
jY
[î)]-_[I(?i+i)]'
(5.145)
takes the asymptotic form
Ae-fl'x""
cos anx
(5.146)
n-0
is
is
is
Thus with increasing time, the antisymmetric components of the initial

distribution o(x) (given by the sine terms) decay more rapidly than the
We may conclude from this
symmetric components (the cosine terms).
result that, regardless of the initial shape of the flux distribution, the
later distributions will become progressively
more symmetric
with
increasing time, provided of course that no further changes are made in
This behavior
the nuclear characteristics of the system.
a consequence
uniformly distributed
of the assumption that the fissionable material
independent of x.
throughout the multiplying medium, i.e., 2/
one other important conclusion which may be drawn from
There
always one term in the series (5.146) which will
(5.146), namely, there
this single term will yield
predominate, and with increasing values of
We demonstrate this
increasingly better approximations to the series.
of
following
the
argument: Consider the quantity X,
trend by means
which may be written in the form
t,
is
is
B> m
X* m a*
Suppose that we chose the width of the slab
(2/

It
e-B<"
y"X
S1"
Now by (5.139) and (5.137),
according to (5.140) and (5.141).
...x..7,_.
cos anx
Y Ane-D^"
= Go
*(z,Q
C.%
tribution for all subsequent time from the general solution (5.141). The
coefficients involved in this expression are to be computed by (5.143),
and the various eigenvalues, from the relations (5.137) and (5.139).
Let us examine now the distribution of the neutron population in the
present system for large values of t. We inquire, then, into the behavior
of the flux (x,t), given the nuclear properties of the system and a knowl
Therefore, assume that
edge of the flux at some time instant long past.
=
0; then, for all subsequent
the flux is known at some arbitrary time t
time (during which no changes are introduced into the system), the flux
is given by
- S.)
(5.147)
(for fixed composition)
sec.
DIFFUSION THEORY
5.4]
203
such that a particular member of the set X vanishes; i.e., let us take a
such that XJ, = 0, where m ^ 0. Then by (5.147)
P^f-iW-W
All other
in (5.147)]
X are given by [using (5.148)
. p*Li>]'
It
follows that if
(5.H8,
(*)
[(2 +
. _
(2m
(5.149)
m > n
m = n
m < n
X* > 0
<0
= 0
(5.150)
Thus for
is possible for this system1 (unless, by some chance, all An for n = 0,

. . . m 1 are identically zero).
Consider now the special case when m = 0; then,
x
33
(fi)2
("S/
1,
(5-151)
So)
and for all other n > 0
Evidently all the
are positive; therefore, after sufficiently long time

all terms beyond the first (n > 0) will become negligibly small in com
For this case (5.146) reduces to the form
parison to the first term.
X
cos
Ao =
sr.
ccox
= G0Ao cos
(5.153)
The coefficient Aa is given by (5.143)
Mx)
cos
(i)
which is independent of time.
/-a
{x,t) ~(7oi4o
dx
(5-154)
a
if
=
multiplying slab
We conclude, therefore, that
at time
we start
with neutron flux of the form o(x) and choose the composition and the
"Fourier
is
rigorous proof of this statement

given, for example, by R. V. Churchill,
Series and Boundary Value Problems," p. 105, McGraw-Hill Book Com
pany, Inc., New York, 1941.

m 5^ 0, we conclude that (1) all terms in (5.146) with n > m

have negative exponents and therefore decrease with time; and (2) all
terms with n < m have positive exponents and increase with time.
Thus,
the flux is an increasing function of time, and no steady-state condition
REACTOR ANALYSIS
204
[CHAP. 5
dimensions of the slab so that
(0 -
*w
then after some time, the flux distribution in the slab will approach
steady-state condition given by
ast
>
cos
(2a)
)
(x)
<(x,0
00
(5.156)*
>o(x)
cos
^r-1
(5.157)
0
is
negative in some regions of the slab; thus initial distributions of

>
the form (5.157) are unacceptable on the basis of a physical argument.
In the case that o(x) =
there will never be
nonzero value of the
a
if
is
is
neutron flux in the system, no matter what choice

made for the halfwidth, for from (5.143) all An and C = 0. However,
there
even an
is
it
infinitesimally small initial flux,

may be built up to any desired magni
Suppose, for exam
tude by altering the composition or size of the slab.
By
supporting some steady-state flux level o(x).
ple, that the slab
o(x) must have the form1 (5.156)

=
a0cos(^)
At
the composition
is
and the condition (5.155) must be satisfied.

of the slab
changed uniformly so that
(5.158)
<o,
a{x)
the arguments given above
>
(0
<5159
0.
the slab
is
is
Note that this

true only
in composition or form.
if
0 is
0,
and the flux will no longer remain at steady state,

Then by (5.151)
<
will
Because of the assumption that the
but
rise according to (5.146).
simple cosine function, all coefficients An =
initial distribution o(x)
Thus the time dependence will be
for n >
and all C =
for n >
1
Ar
>
0,
|NCLUPCf7i
0,
one could obtain

where m >
steady-state flux of this shape, since
all the coefficients A in (5.146) would be zero except the mth [see
(5.143)], and the slab width o could be chosen so that the exponent Xm in
the with term would be zero. However, such a function as (5.157) with

if
it
flux of the form
the slab with
is
is
= AdG0 and
given by the initial conditions.
Furthermore,
no other choice than (5.155) will lead to steady-state operation [see the
discussion following (5.150)].
It true that
were possible to start
where
homogeneous and contains no irregularities
sec.
DIFFUSION THEORY
5.4]
205
= aoc-fl'x" cos
4>{x,t)
This function
(2d)
given by
(5.160)
t.
is
The initial
shown in Fig. 5.19 for various values of
magnitude of the flux was characterized by the amplitude Cto; thus to
achieve any other higher flux level we
(x,t)>
need wait only until the desired in
achieved by the factor exp
crease
reached at
this level
( Dv\lt).
~^\2
=
suitable readjustment in the
1
is
t',
If
is
o<*i
1%
composition of the slab can then be

made so as to satisfy the condition
The flux will now cease to
(5.155).
rise and can be caused to stabilize at
Fio.
5.19
Flux distribution
in in
finite-slab reactor.
condition such
as (5.159) upon the
multi
is
plying system by external means and thereby exercise control over the
the basis of all reactor control methods (see
neutron population. This
Sec. 1.3b).
is
2,
it
l)wx
2d
0,
(2n
1,
cos
Equation (5.156) gives the flux distribution at steady state in an

infinite-slab reactor (with fission sources only) based on the one-velocity
This result may also be obtained by starting with the steadymodel.
not possible, then, to
state diffusion equation directly; however,
prove rigorously that only the member of the set,
.
is
the one corresponding to n = 0.

which persists in the steady state,
commonly called the lowest mode or fundamental mode.
This term
is
really necessary for the

completion of the discussions given below in which relations analogous
to (5.156) are obtained for the sphere and the infinite cylinder by starting
directly with the steady-state diffusion equation.
b. The Sphere and the Infinite-cylinder Reactors.
The solutions for
in
flux
distribution
multiplying system with spherical
the steady-state
symmetry or in system with axial (cylindrical) symmetry can be carried
out by undertaking series of arguments similar to those used in the case
straightforward and will yield in each
The method
of the slab reactor.
In these applications the timecase a general result of the form (5.141).
The
dependent portion of the solution will have the form (5.132).
part of the solution, however, will involve some
space-dependent
been given are
Thus the arguments which have
is

In practice, one can impose
a a
a a
Conditions of the
the desired level.
and
types (5.155)
(5.159) can be de
rived for other reactor geometries.
REACTOR ANALYSIS
206
[chap. 5
In the case of the sphere (as we will demonstrate shortly)

these functions are the trigonometric functions divided by the radial
coordinate r, and in the case of the cylinder, the ordinary Bessel functions.
It can be shown, as in the slab problem, that long time periods after
new functions.
"starting" the system (i.e.,
>)
will predominate
in the series solutions for both the sphere problem and the cylinder
Moreover, if it is required that either of these systems oper
problem.
ate at steady state, there exists only one term from each of these series
as
one term
(corresponding to the fundamental mode) which is acceptable as a solu

tion.
These terms will correspond to the result (5.156).
We will not use this general approach in the present analysis.
Instead,
we will start with the general steady-state diffusion equation and estab
lish the appropriate solution in each case by a suitable identification of
the fundamental mode, guided in our choice by our experience with the
The steady-state diffusion equation is
slab problem.
with

Eq. (5.124)]
V20(r) + fiJ(/>(r) = 0
(5.161)*
by (5.126). For the spherically

equation reduces to [see (5.80)]
symmetric problem, this
B2 defined
differential
[cf.
-Mr)]"
r
+ B^(r)
= 0
(5.162)
which has the general solution
(r) =-{A
r
sin
Br + C
cos
Br)
(5.163)
This solution must satisfy the boundary conditions (5.72a) and (5.72c).
If the outer boundary of the sphere is given by r = R, these boundary
conditions take the form
(1)
(2)
(r)
must be finite for all r <
JU)=0
._
....
(5-164)
The condition that the flux remain finite for all r, appropriate here since
there are no localized sources, yields the requirement that C = 0. The
values of B needed to meet condition (2) are
~
R
n = 0,
1,
(5.165)
Note that we have retained the symbol B to denote the various eigen
As indicated
values, instead of the notation a used in the slab problem.
in Eqs. (5.137) and (5.165), the eigenvalues form an infinite set; however,
only the first of these, corresponding to the fundamental mode, is required
It is understood, then,
since this mode gives the steady-state solution.
that the B which appears in the argument of
refers to the lowest eigen
value.
Since the quantity B was originally defined by (5.126), this
DIFFUSION THEORY
sec. 5.4]
207
equivalence implies that there must exist a certain relationship between

the nuclear properties of the system and its geometry if the system is to
A detailed
be at steady state [cf. Eqs. (5.151), (5,155), and (5.156)].
study of this relationship is given in a subsequent analysis.
On the basis of the preceding remarks we select the eigenvalue corre

sponding to n = 1 . This choice is entirely consistent with our experience
It
with the time-dependent behavior of the flux in the slab reactor.
n
physically
values
of
are
unacceptable
that
larger
can also be argued
because they predict negative values of the neutron flux within the
However, this argument is not complete, since it does not rule
medium.
out the possibility of a superposition of solutions for the various values of
n; the arguments presented in connection with Eq. (5.149) must be used.
It follows from these remarks that the solution to (5.162) is
=
j j
sin
(5.166)*
Note that as in the case of the slab problem there still remains an arbi
trary constant A which can be determined only from the initial conditions
of the system.
The solution for the cylindrically symmetric system is obtained from
(5.161); the appropriate form for the present application is [see Eq.
(5.86)]
- [p*'(p)]'
p
+ B2*(p)
= 0
(5.167)
In this case we consider a cylinder of radius (R and of infinite height (with

clearly no lengthwise dependence of the flux).
As in the sphere problem,
<
(R and that #(&) = 0.
we require that the flux be finite for all p
Equation (5.167) is Bessel's equation of order zero; its general solution is
<p(p)
AJ0(pB) + CFo(pB)
(5.168)
(ft
The Bessel function of the second kind F0(pS) has a singularity at the
origin; therefore, the constant C must be taken to be zero so as to satisfy
the requirement that
remain finite.
The boundary condition at
requires, in addition, that
=
(5.169)
JoCM)
The Bessel function J0 has an infinite
As in
Let its
*4
is
set of zeros' v\, vt, . .

the two previous cases we select the first root of this equation.
value2 be v\) then the corresponding eigenvalue
(R
See
Jahnke and Emde,
Hildebrand,
where
Vl = 2.405,
etc.
(5.170)
op. ext., p. 126, for a sketch of this function and p. 166 for
zeroes.
1

0(r)
op. cit., p. 575.
[chap.
REACTOR ANALYSIS
208
and the solution to (5.167) is
A7,
(p)
(5.171)*
It

is of interest to compare the shape of the neutron flux for the three
The normalized solutions (5.156),
elementary geometries considered.
(5.166), and (5.171) are shown in Fig. 5.20.
0.4
0.2
Fig.
0.6
a\
0.8
1.0
Normalized solutions for flux.
5.20
In each of the
Criticality Condition and Multiplication Constant.
three geometries for single multiplying media discussed above, it was
found that a certain relationship was required between the composition
of the medium and its dimensions if the system was to operate as a stable
This relation was expressed in the form
chain reactor.
c.
(2,
- 2)
5'
for all three geometries
(5.126)
and
B2 =
\2&)
*'=(!)'
for the slab

for the cylinder
for the sphere
(5.172)*
sec. 5.4]
DIFFUSION THEORY
209
For each geometry, Eq. (5.126) and one of Eqs. (5.172) must hold simul
taneously if steady-state operation is to be achieved.
Note that by
imposing these requirements on each system the composition or size is
specified; thus, if the size of the system is selected, then
the corresponding value of B2, and the substitution of this
yields
(5.172)
number into (5.126) yields an expression for the composition (note that
completely
the cross sections are functions of the fuel concentration).

Conversely,
if the fuel concentration is given, (5.126) yields the B2 value, which
when used in one of the Eqs. (5.172) specifies the size of the system which
will support a stable chain reaction.
A set of equations of the type (5.126) and (5.172) is called the critical
condition or criticality condition.
It will be found that similar criticality
conditions must hold for the steady-state operation of all systems in

which the only source of neutrons is the fission process. An examination
of Eqs. (5.146) and (5.147) shows (in the case of the slab, for example)
that there are three important cases for the quantity
x =
(1)
X?
= 0
(2)
Xg
> 0
toward
(3)
2) ("S/
~
"
]/v^*-A
Sa)
(5-151)
gives, as argued above, steady-state operation

all terms in the time-dependent series have decreasing
zero
at least the first term of the series has an increasing

exponential factor; thus the flux is a monotonically
< 0
increasing function of time
The situations
corresponding to (1), (2), and (3) have been designated

critical, subcrilical, and supercritical, respectively.
Consider now the "multiplication constant" of such a system; as in
Eq. (3.16), the multiplication constant is defined as the ratio of the
number of neutrons produced in a given generation in a reactor to the
number produced in the preceding generation.
If the reactor is to be in
operation, the neutron population must be stationary in
time (on the average), so that neutron gains always just balance neutron
This may be seen
losses, and the multiplication constant is unity.
directly from the steady-state differential equation (using the slab case
steady-state
as an example)
factors so that the flux steadily diminishes
exponential
Xg
Geo
namely,
(0
-D"(x)
+ 2*(x)
v2,(x)
(5.173)
The steady-state solution is given by
(x)
cos
Bx
(5.174)
Sc
REACTOR ANALYSIS
210
[chap. 5
DB*(x) +
20(x)
r2f4>(x)
Integration of this equation over the entire volume of the reactor gives
(in the case of the infinite slab or cylinder the integration is carried out for
a unit height of the system)
DB2* + 2a*
= y2/I>
(5.175)
dx
(5.176)
where we have defined
$ =
4,{x)
Equation (5.175) may
be stated
(Total leakage
+ (total absorption
losses)
verbally
as
losses)
(total fission gains)
The left side of (5.175) is the total number of neutrons lost from the reac
tor (per unit time), and these give rise to y2/<p neutrons on the right side;
thus the multiplication constant for this case is

ZXB2*

If the reactor is not operating
~~
DB1 + 2,
at steady state, the calculation
(5.177)
of the
multiplication constant is more difficult, since, in the differential equation,

A convenient
there is now a time-derivative term to be considered.
method for treating this case involves the use of the quantity vc (intro
duced in Sec. 4. (id), a fictitious number of neutrons per fission, which
may be adjusted so that for any composition and configuration the fission
Thus we choose as the
sources can be made just to balance the losses.
applicable differential equation for the nonstationary system
-Z)V^(r) +
20(r)
= v2,4>(f)
(5.178)
This
where for the steady-state case vc = v, the experimental value.
equation, then, describes a fictitious steady-state system which could be
achieved by using the fuel concentration of the nonstationary system
(implied in the macroscopic cross sections) if the neutron yield per fission
Clearly, if the flux in this system
were vc (in general, different from v).
is to be time independent, then the solution to (5.178) must satisfy a
relation of the type (5.177); but, the solution to (5.178) is of the form
(5.174); we have, therefore, the immediate result that
(5.179)
The value of B2 is given in each case

for all three elementary geometries.
Equation (5.179) states that, if (DB4 + 2)* neutrons
by Eqs. (5.172).
are lost from the reactor, it requires vc2/4> neutrons to replace them;
DIFFUSION THEORY
sec. 5.4]
211
neutrons are actually produced (where v is the true average

Thus the
per fission for the given fuel material).
number
by definition [cf. Eq. (3.16)],
multiplication constant
but, only
v2/4>
Multiplication constant =
is,
of neutrons
(5.180)
is
The usual problem that

where the last equality follows from (5.179).
to determine the multiplication constant of
one encounters in practice
In this situation all the nuclear characteristics of the
specified system.
A;
reactor are known (i.e., all the cross sections, including the fuel, have
been specified), and the geometric configuration and size of the system
have been selected.
The calculation of requires first the evaluation of
B2 from the appropriate relation in the set (5.172).
The substitution of
ve.
The
is
this result, along with the cross sections, into (5.179) yields
multiplication constant
then computed from (5.180).
It
is
*-(f)0+xr
km
is
But, by our previous definitions [see (3.18)], cS//2

the multiplication
constant for the infinite medium in the one-velocity approximation.
For clarity in notation we introduce the symbol kx for the infinite medium.
(5.182)*
The substitution of this expression, along with the definition of L2 from

(5.75), into (5.181) yields
=
+ LJ2)-1
(5.183)*
is
is
Again note that B2 obtained from (5. 172). The multiplication constant
customarily called the "effective multiplication
for the finite medium
the infinite-medium multiplication constant kK
constant," since
multiplied by
reduction factor
the one+ L2B2)~l; this factor
velocity nonleakage probability and will be discussed in detail later. We
summarize the above results by combining the information from (5.151)
(1
is
is
it
A2, X2,
>
and the system

and the system
and the system
is is is
corresponds to
0, 0, 0,
\\
corresponds to
<
<
>
(3)
(2)
corresponds to
(1)
and (5.183):
A;
critical
subcritical
supercritical
We illustrate the procedure described in

Sample Calculation.
means
of a simplified calculation for
the preceding section by
bare
The problem
to compute
homogeneous spherical reactor of radius R.
a
d.
is

in describing the present systems, which take into

account the neutron leakages, to separate (5.180) into the factors
convenient
REACTOR ANALYSIS
212
[chap. 5
the critical fuel concentration and total fuel mass for this reactor as a
For convenience in computation we will assume that
function of its size.
the moderator density Nm in the system is unaffected by changes in the
NF;
as we noted previously (see Sec. 4.8c) this is a

approximation for systems with low fuel concentration.
We
further assume that the diffusion coefficient D is also unaffected by such
changes, being principally determined by the moderator concentration.
fuel concentration

good
Fig.
40
20
80
60
Radius
5.21
100
it cm
Critical fuel concentration and total fuel
120
140
mass for spherical reactor in the
one-velocity model.
Then, in the simple one-velocity

Eqs. (5.12G) and (5.172),
NF\va,
g/>]
approximation,
we must have,
from
- NMgM
D
This relationship is the requirement for criticality.
equation for Np gives the critical fuel concentration
=
Nf _
Figure
5.21 shows a
Nmo*>
/(,
+ DB*
1)
The solution of this

[with
r)
= vo/fof1)
(5-185)
plot of Np/Njn as a function of the radius of the sphere

U236 and beryllium.
The nuclear density of the
for a system of pure
beryllium was taken
as 0.124
cross sections are from Table 2.2.

behavior of the function
nuclei/cms, and the microscopic

is of interest to note the asymptotic
1024
It
N F(R)/N M-
The asymptote corresponds to the

minimum fuel concentration with which a system of U236 and beryllium
can go critical (on the basis of the one-velocity model) ; this is the con
sec.
DIFFUSION THEORY
5.4]
213
centration required for an infinite medium of these materials and can

be obtained from (5.185) by taking the limit as R > . The result is
,(Af)
- i)
MF
The critical mass
of
Um may
be obtained
as a
function of R by the
use of the relation
MF(R)
4* A,
[isaa^(v
1)
fl[2(*>
(5.186)
It is evident from (5.186) that, for

is shown in the figure.
This is the
large R, MF(R)~R, and for small R, MF(R)~R.
Although
model.
the
behavior predicted by the one-velocity
indicated
trend for large systems (R > ) is in good agreement with more accurate
models, the dependence of the fuel mass on R for small systems is given
This function
We will show later, in Chap. 6, that the critical

fuel mass does not decrease to zero with decreasing size but, in fact, first
incorrectly by (5.186).
passes
through

decreases and
minimum
then
as
rises again
to
steadily increasing magnitudes with

further decreases in R. This trend
by (5.186) because
the one-velocity model does not pro
vide an adequate estimate of all the
neutron losses resulting from escapes
through the boundaries of the system.
e. Rectangular
Block and Finite
Cylinder. The methods used in the
analysis of the infinite slab and cyl
is not predicted
inder
and of the sphere are easily

extended to the analysis of the rec
Fig.
Rectangular
5.22
block reactor.
tangular-block and finite-cylinder reactors. The procedure is straight

forward and involves the use of the technique of separation of variables.
Consider first the homogeneous multiplying medium in the form of a
as shown in Fig. 5.22.
rectangular parallelepiped,
The appropriate
form of the diffusion equation which describes the steady-state flux
distribution in this medium of absorption cross section 2a and fission
cross section 2/ is
V*{x,y,z) + B*(x,y,z)
where B2 = (c2/ 2)/Z).
may be written in the form
We look for
4>(x,y,z)
F(x)
= 0
a solution of
G{y)
H(z)
(5.187)
this equation which
(5.188)
REACTOR ANALYSIS
214
[CHAP.

G"(y)
F"(x)
H"(z)
In order for this equation to be satisfied, each terra must be a constant,

Let these constants be k\,
independent of the variables x, y, and z.
k\, and k\; then
F"(x)
F{*)
It
Ki
G{y)
H(z)
C5,19U;
An appropriate
set of
k\ + k\ +
fcf
Bi
(5.191)
boundary conditions may
be stated
in terms of the
In the present calcula

extrapolated dimensions of the parallelepiped.
tion we will ignore the difficulties in the concept of the extrapolated
boundary at the corners which have infinite curvature by arguing that
extrapolated
+ &,y,z)
0,
outer
is
flux
(x,y,z)
example,
4>(
The neutron
(1)
for
to the following pair of boundary conditions:

symmetric about the three midplanes; thus,
4>(x,y,z)
surfaces
where
is
flux
neutron
x,y,z).
of the
(2)
The
flux
vanishes
at
the
block; for example,

half-width of the block in the
rectangular
the extrapolated
x direction (see Fig. 5.22).

It should be noted that boundary condition may be stated in an
alternate manner, namely, that the component of the net current in each
These two statements are equivalent,

of a symmetric function.
of the statement (5.72c).
J,(x,y,0)
= Jv{x,0,z)
and either may be used in the case

the appropriate form
Boundary condition
is
Jx{0,y,z)
Thus,
be zero at the midplanes.
directions
of the x, y, and
sin kxx +
F{x) =
The application of the boundary conditions
cosg
cos kxx
and
F(x)
The solutions of the three ordinary differential equations (5.190) are of

In the case of F(x) we obtain the familiar result
the same form.
A

is
the necessary correction will be small and, furthermore, that the region of
primary importance in the reactor
near the center of the block.
With
this approximation in mind we can confine the requirements on the
yields the solution

(5.192)
where we have selected the fundamental mode by the usual arguments.

The solutions to G{y) and H(z) may be obtained in the same way, and the
sec.
DIFFUSION THEORY
5.4]
215
complete solution to (5.187) is easily shown to be
(x,y,z)
cos
2d
cos
26
cos
(5.193)
2C
We have selected for the eigenvalues fci, k2, and k3 the values correspond
ing to the fundamental modes. It follows that (5.191) may be written
(5.194)

The result (5.193) gives the steady-state flux distribution in the

tangular block reactor with half dimensions a, b,
and c. Note that, as in the case of the more ele
mentary geometries previously considered, the
solution involves an arbitrary constant C which
can be determined only from the initial conditions
on the system.
In order that the system support
the steady-state flux distribution (5.193), we re
quire that
This
rec
condition of criticality. Given the di

mensions of this reactor, (5.195) yields an equa
Given the
tion for the critical fuel concentration.
Fig. 5.23 Cylindrical
fuel concentration, and say the proportions of the
reactor.
The
block, (5.195) yields the size of the system.
multiplication constant for this reactor may be derived by means of the
It is easily shown to be
usual method (see, for example, Sec. 5.4c).
given by
is the
k =
where for criticality, k =
1.
fej
(1
+ L2B2)-'
The value of
(5.196)
B2 to be used in
this relation
is
given by (5.194).
Another geometric configuration which is of practical importance in
Consider therefore a
reactor design is the cylinder of finite length.
in
form
the
of a right circular cylinder
homogeneous multiplying medium
of height 2h and radius (R (see Fig. 5.23).
symmetry, and the appropriate statement
equation is
\h
It
yp
*(p,2)
*(p,2)
This system has cylindrical

of the steady-state diffusion
B**^
has been assumed here that the neutron-fission
= 0
(5-197>
source is given by
216
REACTOR
ANALYSIS
[CHAP. 5
and B1 is defined in the usual way (5.126).

We solve this
differential equation by separation of variables; therefore, take
vZ/<(p,z),
(p,z) = R(P) Z(z)

Equation (5.197) may now
be
(5.198)
written
[PR'(P)Y +' %Q
R(p)1""""1
Z(z)
-B*
(5.199)
As before, let
1
R(P)
[PR'(p)Y
V"/J
m-k*'
Z(z)
-i -*
= k\
k*
(5.200)
The general solutions to these equations are

=
At sin ktz + C, cos k,z

R(p) = A,Jo(k,p) + CpYa{k,p)

Z(z)
(5.201)
The boundary conditions which must be satisfied by these solutions are

Note that the symmetry condition requires that
given in (5.72).
A, = 0. The condition that the flux remain finite at all points within
the medium requires that we take Cp = 0, inasmuch as Y0{p) has a singu
larity at p = 0. The extrapolated boundary condition yields the addi
tional requirement that [see (5.170)]
fc,<ft
where
v\
is the first zero of
Jo.
= x,
k.K =
(5.202)
We combine all these results to obtain the
solution
(p,z)
^(jf)C0SfE
As before (~) denotes the extrapolated

requirement for the cylindrical reactor is
boundaries.
C5-203)
The criticality
and the multiplication constant is given by (5.196).

In this case, how
The general results for the rectangularever, we use (5.204) for B2.
finite-cylinder, and sphere reactors have been summarized in

Table 5.3.
f. Optimum Proportions for the Bare Reactor.
Suppose that a given
homogeneous mixture of materials with a certain ka and L2 is to be made
into a critical reactor.
For the three configurations listed in Table 5.3,
three different total volumes of mixture will be required.
Since for all
block,
DIFFUSION THEORY
sec. 5.4]
three systems k =
1,
217
we must have
Results for the One-velocity
Right cylinder
Criticality condition:
h.
ft
XT
77
(1
A;
1,
is
is
is
is
critical volume and, therefore, minimum fuel mass.

To determine this we require that the volume be a minimum as the ratio
of the dimensions
varied and the buckling
held fixed (i.e., the composi
tion of the mixture
held fixed).
The results of this minimization are
assumed that the extrapolation distance
negligible in comparison
(if
aâ, etc.)
with the dimensions of the reactor; that
(R
V2
Vyh
for rectangular block reactor

of minimum critical volume
a =
is,
is
it
is
is
is
is
give the minimum
(5.206)
for cylinder
1.08/t
tion for
a large reactor),
then
it
(5.206), and neglect the extrapolation distance
Sphere:
B3
ylinder:
_
~
= T23h\
2B3
=
reasonable approxima
may be shown that these volumes are:

t'3
Cube:
(a
is
It of interest to compare the total volume of material needed to produce

a bare critical assembly in each of the three cases considered above.
If
we use for the block and cylinder geometries the optimum proportions
4t<
B*
~~
3B*~
161
150
B*
130
B>
(5.207)
/x\
sin
J008**
/2.405V
&
Radius
xz
s= fc.PNL, and fc, vfptht, pnl =

where
+ L'B')-1,
given for each geometry.
"x,y,z are measured from the center of the block;
measured from the geometrical
center of the cylinder, and
measured normal to the axis;
measured from the
m
center of the sphere;
+ 0.7104X,,.
and B1

<R, height 2h
/2.405P\
*H
cos
2C
fey
2b
cos
TZ
(i)1
X
23
Radius
2c
ry
Buckling,
rX
(s)'
cos
2b
Sphere
Rectangular block
2a
Reactor
Summary op
5.3
Table
R
:
is,
The actual critical volume of the block and cylinder will depend on which
which ratio a:b.c and
of the possible proportions is chosen, that
In each of these two cases, however, the ratio may be specified so as to
REACTOR ANALYSIS
218
[CHAP. 5
It
is seen that the least total volume of material is used in the spherical
Thus the sphere yields the minimum critical fuel mass for
configuration.
This result may ultimately be traced to
a given concentration of fuel.
the greater leakage of neutrons from the cubical and cylindrical forms,
which have greater surface-to-volume ratios.
The question of the most
for
is
efficient shape
a bare assembly
one of making the best use of all the
material involved in the reactor.
For example, the material at the
corners of the cube is very inefficiently used, since neutrons readily escape
from such regions, and in a homogeneous reactor there is as much fuel
invested (per unit volume) in the corners as at the center.
Another measure of the efficiency of use of the material in the reactor is
often taken to be the "peak-to-average" flux ratio.
If the flux profile
is relatively flat over the reactor volume, all regions are contributing
fairly equally to the total reaction rate; however, if the flux is strongly
peaked at the center of the reactor, the fuel in the outer regions is being
poorly used. Some measure of this is the ratio mentioned, that of the
highest value of the flux to its average value over the reactor.
In the
case of the bare reactor configurations
(0)
value of flux at center of reactor

;
?-s
average value of flux over reactor
= V- =
>
Thus
Peak-to-average
Vr denotes
where
ratio =
the volume of the reactor.
fy.
..
ratio
*W df
It may
be easily shown
for the bare assemblies discussed above the peak-to-average

values
(5.208)
that
ratio has the
3.88
3.65
~~
3.29
Block:
Cylinder:
Sphere:
?l*
(5.209)
Note that Ji(n) = 0.519 and vi = 2.405.

It
evident from these
results that the sphere has the more favorable flux distribution.
We conclude the
g. Fission Sources and the Nonleakage Probability.
present section on the application of the one-velocity model to multiplying
media with some remarks about the approximation of neutron-source
strength proportional to flux (see Sec. 5.4a) and the concept of the nonleakage probability [see Eq. (5.183)].
In the problems that have so far been considered, the neutron-source
term S'(r) for fission has been assumed to be vX/4>(i).
We should like
now to examine the implications of this assumption.
Neutrons produced
is

,
.
reak-to-average
considered here,
DIFFUSION THEORY
SEC. 5.4]
219
by a fission reaction occupy the complete energy spectrum; however, the

majority have energies considerably greater than thermal.
These highenergy neutrons are eventually slowed to thermal speeds by experiencing
numerous scattering collisions with the nuclei of the medium.
The
" thermalized " neutrons then enter a one-velocity diffusion process and
wander through the system until they are eventually absorbed or escape
through the boundaries of the reactor. The diffusion equation (5.22) is
useful for describing the spatial distribution of these neutrons, provided
that a suitable expression is developed for the source term.
This problem was previously encountered in the case of the infiniteslowing-down
medium
process for a multiplying
material.
In that
instance we identified the source term for the thermal neutrons as the
In the
slowing-down density at thermal energy q(E,h) [cf. Eq. (4.252)].
present case we require a somewhat more generalized function which takes
Thus
into account the spatial distribution of the slowing-down density.
a suitable description for the source term in the diffusion equation for
thermal neutrons would be

5(r) = q(r,Elh)
(5.210)
where q(T,Etb) denotes the number of neutrons slowing past Elb per unit
Unfortunately,
volume per unit time around space point r in the reactor.
the analytical machinery so far developed for treating multiplying media
does not provide us with a detailed statement of the function q(i,Elb);
For
this approach to the estimation of <S(r) will be deferred until later.
the present we will apply an intuitive formulation which will draw upon
Let us define, then, the following source
the concepts already developed.
function :
S(r) = vZ/(T)plbe
number of thermal neutrons which appear per unit volume per unit
time around space point r
(5.211)
We have used in this definition the quantities:

average number of neutrons produced per fission
probability that a neutron born from fission reaches thermal energy

before it is absorbed [see Eq. (4.283)]
fast-fission effect; the ratio of all fissions to thermal fissions [see
Eq. (4.269)]
Equation (5.211) states that the rate at which neutrons become thermal
at r is given by the fission rate1 at r, 2/<, times the number of neutrons
per fission v corrected by plh, for the losses which occur because of absorp
tion during the slowing-down process. It is important to note that
1
Note that, if
X/ is the rate at which
fissions
occur because of thermal neutrons,
eE/tp is the rate at which fissions occur due to neutrons of all energies.
REACTOR ANALYSIS
220
[CHAP.
(5.211) fails to account for any losses in the neutron population because
Thus in (5.21 1) we have made the tacit
of leakage during slowing down.
assumption that there is no leakage Of high-energy neutrons from the
Moreover (5.211) implies that the migration of neutrons during

system.
the slowing-down process does not alter their spatial distribution (this
is valid in a single-medium system).
Taking the thermal
source function as having the space form of the thermal flux does not
imply, physically, that the fission neutrons slow down to thermal and
remain at the space point at which they originally appeared.
It will be
shown later that in a bare system the neutrons wander throughout the
assumption
reactor during the slowing-down process and always move about in such
way that the space form of the neutron density (flux) at all energies is
the same as at thermal energy.1
In all the preceding calculations in this section the one-velocity

approximation was applied, and the source term was appropriately

This choice is consistent with the form (5.211) in that
selected as v2/<.
it is the special case of that function when epti> = 1 ; i.e., all neutrons born
at r are immediately reduced to thermal (the one-velocity postulate).
It should be recognized that, if the diffusion equation is used to obtain
the thermal-flux distribution in a reactor, it should be written in terms of
the corrected source function (5.211); although this formulation is
incomplete, it is better than the simple form vXf.
The factors which appear in the source relation (5.211) may be
expressed in an alternate set of reactor terminology which displays more
clearly its relation to the multiplication constant of the system. Con
sider then the quantity v2/, which may be written
(5.212)
2^ is the macroscopic absorption cross section of the fuel in the

reactor and 2 is the total macroscopic absorption cross section for the
mixture; i> and/are defined in the usual way. The substitution of (5.212)
where
into (5.211) gives
-S(r)
= i7/ptlle(r)Z0
kxSa(r)
(5.213)
for the infinite medium, kK, is given

Note that the expression used here for kK differs from (5. 182)
by (4.275).
through the additional factors p,h. Clearly, (5.182) is the special onevelocity case of (4.275) used in (5.213). If (5.213) is used as the source
term in the steady-state diffusion equation (5.22), we obtain
where the multiplication constant
-Z)V^(r) +
1 See
2a<Kr)
discussion accompanying Eq. (6.87).
A^.*(r)
(5.214)
sec.
DIFFUSION THEORY
5.4]
This equation may
be
221
written in the form (5.161) if we identify
#2
2)
(*2
- 2)
B1 as
J3 o c
ictlfJ
&
(5-215)
a".
I 1
This statement of BJ is entirely consistent with

The definition of kx in this relation is determined by the
(5.126).
factors selected for the thermal-source term.
Either of the definitions
or
be
the accuracy desired in a
may
depending
upon
used,
(5.182)
(4.275)
particular application.
We showed previously that the multiplication constant k for the finite
medium could be written in the form (5.183) and identified the factor
(1 + L*BS)-' as the nonleakage probability for the neutrons based on the
In the case that the diffusion equation is
one-velocity approximation.
used to describe the distribution of thermal neutrons in a reactor, this
factor gives the nonleakage probability for the thermals.
We demon
strate this relationship by applying two separate lines of reasoning:
According to our usual definition of the multiplication constant of a
system, we can write
with
L2 given
by (5.75).
k =
2i
(3.16)
n0
In our
present situation, nt and n0 represent the total neutron populations

in the first and second generations for a finite system.
Thus if n0 neu
trons are absorbed by a finite medium, these neutrons produce a certain
number of fissions which yield further neutrons.
Not all these second-
neutrons, however, are available to the chain reaction; a

Therefore, if we call
certain fraction leak out through the boundaries.
nf the number of neutrons produced by the reactions because of the n0
generation
neutrons if the system had been infinite and let pNL be the nonleakage
probability for a neutron in the finite system, then it follows that nf pNL
is the number of second-generation neutrons available to the finite system
for maintaining the chain.
Thus, rii
k =
n^
Mo
nf pNL, and
= A.pKL
(5.216)
By comparison of this equation with (5.183), we conclude that

Pnl =
+L2fl2
(5.217)*
By definition,
We can derive this result by an alternate approach.
probability must be given by the fraction of neutrons
Let us compute, therefore, the ratio of the total
absorbed in the system.
number of neutrons absorbed per unit time in the reactor to the total
the nonleakage
J
use
F-g if
ANALYSIS
REACTOR
number produced per unit time.
Pnl
[chap. 5
We write this quantity in the form
2.*(r) dr
jr j-f
222
(5.218)
S{x) dr
If
where the indicated integrations extend over the volume of the reactor.
we use for <S(r) the equation (5.22) and recognize that the cross sections
are independent of the spatial coordinates for homogeneous system, then
,1B,
(5.219)
dr
(r)
*(r) dr + DB*
1 ;
2a
*(r) di
]y?
2a
Jv ?
Pnl
(5.218) reduces to [where we have used (5.161)]
does escape from the reactor, i.e., the leakage probability
neutron
px.,
is
We may further conclude from these results that the probability that a
- Pnl
L2J52
(5.220)
+ L2g2
is
These results will be of frequent use to us in later analyses.

simple
application which may be demonstrated here
the calculation of the
Before computing this
lifetime of a thermal neutron in a finite reactor.
l0,
If
is
it
quantity, consider first the average time spent by thermal neutron in an

infinite medium before
absorbed.
we call this number
then we
mean distance traveled by neutron to absorption

mean speed
\**
may define
0
__
2J,hyth
uth
(5.221)
is
is
is
mean speed
Va,
(2^
mean distance traveled by neutron before

absorbed or escapes from the system
Xh
is
In
vth
and
/X? the absorption mean free path for thermal neutron

the speed of
thermal neutron [see Eq. (4.253) and discussion].
finite medium the mean lifetime
of a thermal neutron
given by
where
it
2i!")t>th
(5.222)
is
We identify the cross section for escape 2'j; (i.e., the number of escapes
Evidently,
from the relation
per unit distance traveled)
(5.175).
=
DB2.
This concept of DB2 as the expected number of escapes per
2
unit path
The
applicable only to finite single-medium systems.
ratio 2^/2^
obtained by computing the total number of escapes and
absorptions in the system by means of the method used for determining
is

Pl
given by
sec. 5.5]
Pnl
DIFFUSION THEORY
[see also (5.220)].

=
With this expression for
(2f + DB*W
(5.222) reduces to
2>lh(l + L'5)
= hpt,h
(5,223)
is,
223
is
is
with the aid of (5.221). Note that I < l0; that

the mean lifetime of
thermal neutron in finite system
less than the lifetime in an infinite
due to the fact that neutrons are lost to the system by
system, and this
escapes, as well as by absorptions.
Diffusion Length
5.5
is
is
is
it
is
is
it
is
is

it
Two physical parameters required in every application of the onevelocity model are the diffusion coefficient
and the diffusion length L.
So far, our interest in the one-velocity model has been confined to descrip
tions of the spatial distribution of thermal neutrons, and these parameters
have therefore been defined in terms of the thermal cross sections.
We
will show later that similar quantities can be defined in the treatment of
multivelocity systems by the energy-group method, or for that matter,
in any system wherein
convenient to describe the "diffusion prop
erties" of a particular group of neutrons in terms of single speed.
The diffusion coefficient and the diffusion length are fundamental
macroscopic properties of a material which are useful in the one-velocity
formulation. Both quantities can be measured directly in the labo
ratory by suitable experiments.
The direct measurement of the diffusion
coefficient, however, entails the use of a pulsed beam of neutrons.
Inas
much as this experiment involves
time-dependent
phenomenon,
discussion of the experiment will be deferred until after
suitable model
has been developed for the analysis of nonstationary problems in neutron
An experiment for the direct determination of the diffusion
diffusion.
based on
length, however,
steady-state phenomenon, and the impor
tant features of this experiment can be displayed by means of the models
and concepts already developed.
Because of the close relationship
between the parameters
and D,
would be desirable to examine the
techniques for their measurement simultaneously.
This
not possible
of
because
the complications mentioned above; thus for the present we
confine our attention to the study of the diffusion length and an experi
ment for its measurement.
It should be noted that separate measurement of both D and
not
Because of the relationship Ll = DfZa,
always required.
clear that
a knowledge of one parameter yields the other, provided of course that
known.
This
precisely the situation encountered in the
experi
2
ment; thus, in order to determine D we require
separate measurement
of 2. The diffusion-coefficient experiment based on the pulsed-neutron
224
REACTOR
ANALYSIS
[CHAP. 5
beam technique, on the other hand, provides a direct measurement not
only of D but also of 2 (refer to Sec. 9.2).

a. The Root-mean-square Radial Distance.
The elementary problem

plane source in an infinite medium gives some
of neutron diffusion from a

indication of the physical significance of the diffusion-length concept.
It was shown in the treatment of Sec. 5.2a that the flux distribution in a
nonmultiplying medium had an exponential behavior given by the factor

In this form one can identify the diffusion length L as an
exp ( x/L).
Thus L is the dis
attenuation, or relaxation, length (see Fig. 5.24a).
tance from the source plane at which the neutron density is reduced to

M
Fig.
5.24
(b)
Neutron attenuation in diffusion lengths.
Each additional distance of multiple L

1/e of its value at the source.
from the source causes an additional reduction by the factor 1/e from the
Figure 5.246 shows moreover that, at any given dis
previous station.
tance from the source plane, the neutron density is smaller in those
materials
which have the smaller value of L.
Thus materials of low
(large 2 or small D) "attenuate" neutrons more rapidly than materials

of large L. One can think of the diffusion length, then, as a measure of
the average depth of penetration into a medium by a source neutron.
In highly absorbing materials, or in materials of small diffusion coefficient,
tends to be small.
It should be noted that a proper
"crow-flight"
measure of the penetration is the
distance from source to
this penetration
point of capture.
The connection between the diffusion length and the

crow flight, or root-mean-square distance, is now demonstrated.
For this calculation we use the problem of neutron diffusion from a
point source in an infinite medium.
Similar calculations can be made
for the plane and line source, but the basic relationship between L and
the crow-flight distance will prove to be the same in all cases.
Consider
then the diffusion of neutrons in an infinite medium from a point source of
For convenience, we place the source

go neutrons per unit time.
at the origin of a suitable coordinate system.
As previously noted, this
strength
situation has spherical symmetry, and the neutron distribution may be

described completely by the radial coordinate r alone.
Let us compute
now the quantity r2 for this system.
We define r2 as the average square
DIFFUSION THEORY
sec. 5.5]
225
If /(r)
is
absorbed in
f(r) di = probability that a neutron

volume di at distance
is
if
is,
radial distance from the origin at which a neutron is absorbed.

the appropriate frequency function, that
(5.224)
dx
(5.225)
f5
then we define
a fr*f(r)
r,
is
is
The frequency function can be computed from our knowledge of the dis
Now
tribution of absorption reactions throughout the entire medium.
number
of
unit
in
per
total
time
this
absorptions
the
system and
qo
2>(r) dr the absorptions per unit time in dt at r. Thus the probability
of a neutron being absorbed in dr at
namely, the function f(r) dr,
simply
/(r)dr
MAr
(5.226)
The diffusion length
is
|f
D
rV-"
dr = 6L>
(5.227)
4a-D
related, therefore, to the mean-square
J0
Jv
is
go
"
radial
(0
(5.228)
based on the quantity r2,

similar
can be shown that for a point source
is
f,
Although the present treatment

calculation can be made for and
it
distance by the equation
if
(5.229)
is
is
a
is
The important fact here that

measure of the average value of the
radial distance squared at which
neutron
absorbed; thus, as pre
viously indicated, one can think of
as the average penetration into a
medium by the source neutrons.
The detailed form of the relations
for a given medium by means of the Monte Carlo technique.
In
is
is
(5.228) and (5.229) (or the cor

responding equations for the other source geometries)
of interest only
when the value of
for particular medium
to be obtained by
purely
analytical method.
For example, the diffusion length
can be computed
calculation of this type one selects suitable neutron source for an infinite
medium of the material in question and traces the detailed trajectories of
a large number of source neutrons from birth to capture.
From these
a

is
it
is
The neutron flux due to a point source in an infinite nonmultiplying

If this
medium was obtained in Sec. 5.2b, and
used in (5.226),
result
easily shown that (5.225) yields
226
REACTOR
ANALYSIS
[CHAP. 5
histories one can compute either r2 or f by an appropriate relation of the

type (5.228) or (5.229), and thence the value of L.
The diffusion length of thermal
b. The Diffusion-length Experiment.
neutrons in a particular material can be determined experimentally by
measuring the axial flux distribution in a long rectangular parallelepiped
The connec
at one end of which is placed a source of thermal neutrons.
tion between this flux distribution and the diffusion length may be demon
by a direct application of the one-velocity model.
Consider
The medium in question
therefore the configuration shown in Fig. 5.25.

strated
Fig.
5.25
Diffusion-length
experiment.
is a parallelepiped of length c and sides 2a and 26. Thermal neutrons are

introduced at z 0 according to some distribution S(x,y).
The only
requirement on this source is that it emit neutrons isotropically and that
it be symmetric about the x and y axes. These requirements are imposed
only to simplify the analysis;
a similar development to that which follows

can be carried out for a more general source distribution.
The neutron sources in the specified system then are all confined to
one end face of the block.
The appropriate differential equation for
describing
the spatial
distribution of the thermal neutrons inside the
medium is therefore
- DV^(x,y,z)
+ 2a(x,y,z)
=0
> 0
(5.230)
The solution to this equation must satisfy the boundary conditions1

=
(1) j+(x,y,0)
(3)
(5)
4>(x,h,z)
(+x,y,z)
We look for
1 Conditions
x and y.
*S(*,y)
= 0
= 4>{-x,y,z)
(2)
(4)
(6)
solution of the form
(a,y,z)
= 0
(x,y,C) =
{x,+y,z)
0
=
4>(x,y,z)
(x,-y,z)
(5.231)
F(x) G(y) H{z).
The
(5) and (6) follow from the specification that S(x,y) is symmetric in
DIFFUSION THEORY
SEC. 5.5]
227
substitution of this relation into (5.230) yields the

~
F(x)
ftl
G(y)
-k\
where
- k\ + k\
set
+/Cj
tf(z)
l>
'
(5.233)
The algebraic signs of these constants have been selected in anticipation

to (5.230).
The general solutions to equations
of the proper solution

(5.232) are
F(x)
Ai
Aj
G(y) =
sin kix + Ci cos hx

sin kty + C2 cos k2y
(5.234)
(5.235)
H{z) = A3e-k- + C3ek"
(5.236)
The application of the symmetry condition yields immediately

A1 = A2 = 0
The end condition (4) requires that

= 0 = Ase-k>*
H(e)
C3 =
It
A3e~u'1.
C3e
(5.237)
follows that H(z) may be written
H{z) = A3e-""[1
- e-tt->]
(5.238)
The extrapolation boundary conditions at the sides require that

F(a)
= C, cos
G(6) =
and in, n = 0,
= 0
with
km =
C2 cos fc25 = 0
with
fc2m
The corresponding
^
^
(2n +
1)
(5.239)
(2m +
1)
(5.240)
eigenf unctions
are, there
fore,
Cicosfci.r
= C2m cos ktmy
Gm(y)
-t
(5.242)
k\n
*L
corresponding
k3mn such that
(5.243)
fcL,
is
follows that, for each set of values of n and m, there

It convenient to define then a number
value of k3.
is
a
It
(5.241)
that
k\
we have
A:,2
Now from (5.233),
k\
Fn(x)
These results may be used to obtain a solution for
(x,y,z).
Thus we
4>(x,y,z) =
write
n,m

which yields
^mn cos fcix cos fc2m?/
e-*--'[l
e~2* <*-')]
(5.244)
REACTOR ANALYSIS
228
[CHAP. 5
the constants A a, Ci, and C2m into a single

coefficient Anm.
The coefficients AHm may be evaluated by the applica
tion of the boundary condition (1) and the orthogonality properties of the
eigenf unctions Fn(x) and Gm(y).
We do not require so detailed a descrip
where we have combined
tion of the flux for this application,
however; for our present purpose
it
In ^(0,0,2)

Fig.
M
5.26
(b)
Axial flux distribution:
(a) linear plot; (b) semilog plot.
will suffice to examine the general form of the flux distribution along the
central (z) axis. This functional dependence is given by
*(0,0,) =
Anme-"-" [1
n,m 0
- e--]
(5.245)
We note however that the quantities k3mn increase rapidly with increasing
values of m and n; thus the higher terms of the infinite series diminish
more rapidly than the first.
Therefore, in first approximation we can
take
0(0,0,2)
~ ôoc-l"[l
e-2*'"<'-J]
(5.246)
The factor in the brackets represents the "end effects" of the parallele
Note that for very long media (large c) this factor approaches
piped.
unity. Thus the function exp ( k30oz), which gives the usual exponen
tial attenuation in a nonmultiplying medium, is combined with a correc
tion factor which accounts for the end effects in the finite medium.
How
in
is
the case of a short medium (c small), the end effect
not
ever, even
significant until the values of z in question approach the magnitude of c,
i.e., when (C z) * 0. Only then is the end effect important. Figure
5.26a shows a typical graph of the approximate function given in (5.246).
It is evident from (5.246) that there exists some intermediate range of z
sufficiently far removed from both ends of the test medium that the
axial flux distribution in this region varies essentially as the exponential
A region of this sort is indicated by the interval AB in
exp ( k3oaz)-
DIFFUSION THEORY
sec. 5.5]
229
Note that we qualify the preceding statement by including

Fig. 5.26.
both ends of the medium.
The immediate vicinity of the face z = 0 is
also a special region of this medium and, from a practical standpoint, can
also introduce certain complications.
Some of these will be discussed in
the next chapter.
Suffice it to say for the present that the flux distribu
tion near the face z = 0 involves the higher harmonics.
Thus both ends
of the medium are poorly described by the simple relation (5.246) and are
to be avoided in the experiment when possible.
The particular impor
tance of a region such as AB to the diffusion-length experiment is due to
the especially simple form of the axial flux distribution in this range.
We have noted that in this range the flux distribution is exponential, and
the exponent includes the constant fc3oo which is a function of the diffusion
length L. Thus a knowledge of fcaoo yields the value of L.
A measurement of the quantity fc30o is obtained from a knowledge of the
"1"
^loo
~t"
activation of these foils due to neutron captures is directly proportional

to the flux level at the foil. Thus a plot of the foil activities as a function
of position along the axis will yield a curve having the spatial form of the
Moreover, if this curve is drawn on semilog paper, it
axial thermal flux.
will display a linear behavior in the intermediate range AB, as shown in
Fig. 5.266.
But, from
The slope of this curve yields the quantity k30owe
have
that
(5.243)
jji
(5.247)
and from (5.239) and (5.240)
it
The quantities
and
are known from the size of the specimen and fc30o
from the slope; thus, (5.248) yields the diffusion length L.
In practical situation may not be possible to obtain a purely expo
The specimen may be too short
nential region such as AB of Fig. 5.26.
for such region to exist. In this case the end effects on the flux distribu
a
tion can be removed by recognizing that these are confined to the quan
Thus flux measurements toward the outer
tity in the square bracket.
If because of
end of the specimen can be analyzed by means of (5.246).
limitation the outer end of the medium
is
not accessible
and the flux measurements must be made near the source plane, then the
end effect might well be negligible; however, the one-term expansion for
0(O,O,z) will no longer be acceptable and some higher terms in the series
must be retained.
In this case
will be necessary to find the best fit to
some practical
it

actual flux distribution in the test specimen.

The usual experimental
procedure is to obtain a measure of the relative flux at various stations
The
along the axis of the parallelepiped by means of metallic foils.
REACTOR ANALYSIS
230
the experimental
one might use
[CHAP. 5
data with a series of exponentials.
4>(0,0,z)
Aooe-1""' + Aioe-k" +
Thus, for example,
.W-*""
The fit to the data is made by adjusting the coefficients

6.6
(5.249)
Anm.
Sample Reactor Calculation with One-velocity Model
A sample reactor calculation is presented in this section to demonstrate

some of the ideas developed in connection with the one- velocity model.
It must be emphasized that this computation, which is based on the onecannot be expected to yield accurate results.
velocity approximation,
This model, in its present form, has severe limitations, and some of these
were discussed in Sec. o.4g. The most serious deficiency of the model is
that it fails to account for neutron losses during the slowing-down process.
Even if the reactor in question is a highly thermal system, the one-velocity
model in its present form is inadequate for describing the neutron physics.
The difficulty is due to the form of the source term used in the diffusion
It was pointed out in Sec. 5.4g that this term should properly
equation.
include certain factors (e and p) to account for some of the processes
which can affect the neutron population at thermal energy through their
But these factors alone do not
influence on the slowing-down process.
suffice. There is still a third factor which must be introduced to describe
the losses in the neutron population due to leakages during slowing down.
This factor is especially important in the case of small reactors wherein
the surface-to-volume ratio of the system is relatively large.
At this stage we are not yet prepared to take all these factors into
Thus the example which follows is not suggested as a proper
account.
Its purpose is to demonstrate the magni

detailed analysis of a reactor.
tude of some of the numbers which may be involved in calculations of this
type and to show a general computational procedure which may be used.
The calculations of the macroscopic cross sections, however, are properly
carried out and should be useful in showing the application of the ideas
developed in Chap. 4. Recognizing these deficiencies, we make no claim
that these calculations are realistic and therefore can conveniently intro
The
duce further approximations which will simplify the computation.
most important of these is the assumption that the source term in the
diffusion equation is of the form vXf4>; thus, in general, we will ignore all
"fast losses."
As an example of the application of the one-velocity model to the
analysis of nuclear reactors, we will determine the critical fuel mass and
maximum flux level in a light-water-moderated reactor which is designed
to produce 20 megawatts of heat.
Let the core of this reactor be a
60- by 32.3- by 55.4-cm parallelepiped, and for the present analysis assume
DIFFUSION THEORY
SEC. 0.6]
231
that no reflector is provided; thus the reactor from a purely nuclear view
point is a simple bare block geometry of the type studied in Sec. 5.4e.
Assume further that there are three basic ingredients in the core, namely,
water for moderation, aluminum for structure, and U2'6 for nuclear fuel.
We specify that the fuel elements in this system consist of flat plates of
Each fuel element is made up of 18 such
aluminum alloyed with Um.
plates, spaced 0.297 cm apart, and the core contains 28 of these elements,
The light-water moderator occupies the gap between
or subassemblies.
these plates and by the process of
Single fuel
natural convection serves in the
element
dual capacity as the coolant for the
Figure 5.27 shows the

system.
core configuration for this system
and the arrangement of the fuel
elements.
Figure
5.28
gives
60 cm
cross-sectional view of a typical fuel

element and shows how the 18 fuel
plates in each subassembly are
mounted into a single unit by means
of side plates.
In this calculation
we confine our
attention to a study of the nuclear

features of the proposed system.
The only engineering specifications
that we require are the total power
output of the system, its geometric
configuration,
and
the
operating
Direction of water flow
Fig. 5.27 Core configuration

fuel-element array.
of
by
The core configura

temperature.
tion shown in Fig. 5.27 is contained in a suitable tank; we will ignore for
the present the effect of this container on the nuclear characteristics of the
The core and its container are designed to allow water to enter
system.
at the bottom of the core and flow upward through the fuel elements
in a direction parallel to the fuel plates.
We specify the entrant temper
ature of the water to be 130F and its exit temperature as 145F.
For
our present needs it will suffice to assume that we can use some average
temperature,
calculations
say, 140F, to describe the complete system.

will be based on this figure.
Our nuclear
The fuel-element configuration
shown in Fig. 5.28 is clearly a heteroge

neous arrangement; however, because of the relatively small scale1 of this
heterogeneity, it is a reasonable approximation to assume, at least as a
first estimate, that the core consists of a homogeneous mixture of water,
aluminum, and Um.
We will
use this
approximation in the calculations
The determination of the scale of heterogeneity is discussed
in Chap.
10
REACTOR
232
ANALYSIS
Our study, then, will
which follow.
multiplying
principal specifications
medium
of the
[CHAP. 5
be based on a bare homogeneous
configuration
Fig.
follows:
in
shown
5.27.
The
and assumptions are as

(1) power,
bare
configuration,
homogene
140F;
temperature,
(3)
(2)
ous parallelepiped 60 by 32.3 by 55.4 cm; (4) one-velocity model.
20 megawatts;
- Aluminum side plates
7.61

Fig.
0.0508 cm fuel
bearing region
I 0.0508-cm
-1
Al clad
M
5.28
(b)
Cross section of fuel element: (a) fuel element; (6) single fuel plate.
We begin the computation by determining the volume fractions of the

principal ingredients of the system. The selected core geometry is uni
form in the vertical direction; therefore, a comparison of the cross-sec
tional areas of its constituents will yield these volume fractions.
From
the dimensions given in Fig. 5.28 we obtain
1.
Cross-sectional
a.
6.
area of metal per element:
Aluminum end plates: area = (8.05) (0.317) (2)

Fuel plates:
area = (18)(0.1525)(6.98)
2. Cross-sectional
area of water per element:

(6.98) (8.05)
cm2
19.2 = 37.0
cm1
vH,0= 0.604
"mo tal
The next step in the calculation is the determination

perature
of the core constituents

jT [see
61.3
cm1
volume fractions are:
t\.t.i = 0.396
properties
5.10 cm1
= 19.2
Total:
The appropriate
Eq. (4.211)].
= 0.656
of the thermal-group
based on the effective neutron tem
An estimate of this temperature requires,
however, a value of the parameter k which involves the absorption cross

section of the mixture.
At this stage of the analysis we can only guess at
this number inasmuch as the value of S for the core mixture is a function
of the fuel concentration, and this has not as yet been determined.
Evi
dently, the proper procedure to be adopted is an iteration. Thus one
should begin with a rough estimate of k and later check this value on the
basis of the resulting critical fuel concentration which is obtained for the
DIFFUSION THEORY
8EC. 5.6]
233
In
the present calculation we will assume that k = 0, arguing

containing large amounts of hydrogen 2, will be rela
mixture
that for a
tively large, and furthermore we expect that the required fuel concentra
Both of these effects yield small
tion will be low so that 2 will be small.
The validity of the assumption k = 0 will
values of k [see Eq. (4.116)].
system.
be examined later.
= 0, the effective neutron temperature Tn m TN, the

We will base our estimates of the nuclear
moderator temperature.
a
neutron
temperature of 140F (333K). Consider
on
then,
properties,
At 20C its properties are <ra = 0.66 barn, Xtr = 0.48 cm,
first the water.
L = 2.85 cm, and D = 0.16 cm (see Tables 2.2 and 5.1). For an effective
neutron temperature Tn = 333K, these yield for the thermal absorption
In the
case
that
cross section1
<ra
[see
=
C'(333)
Eq. (4.245)].
0(TJ
= 0.66
(J^)]1
= 0.549 barn
wherein we have tacitly assumed that the absorption cross section for
For convenience, we will extend this assumption to the
water is l/v.
well.
The macroscopic absorption cross section for pure
aluminum as
water based on the above figure

2iH''(333)
is2
= ^H,0>(333)p(333)a/^[Il0
(0.549) (0.985) (0.6025)/18

= 0.0181
cm"1
Note that the mass density of pure water at 140F (333K) has been used
for the reactor. The transport cross section, which
is required for the calculation of the diffusion coefficient for the mixture,
as the average value
is found to be
- Zg"o(293)p(333)
p(293)
sh,o)(333)
2tr
(333) =
In this calculation
0-985
we have assumed
(0 48)(1)
-=
^-US
cm
that the transport cross section
is
affected by changes in temperature only through the effect on the density

of the material.
The microscopic cross sections for U236 and aluminum are determined
next.
For
U236,
we have from Eq. (4.247)
7J(333) = feYv(333)f.(333)
7/(333)
(0.886) (650) (0.974) = 561 barns
oy(333)/7(333)
(0.886) (540) (0.976) = 467 barns
10 barns
t,
= 2.08
^J =
<r,
= 2.46
1 In order not to complicate the notation, we have used 9 to represent the thermalgroup properties usually given by ath.
'See, for example, U.S. Atomic Energy Commission, "Reactor Handbook," vol. 2,
p. 39, AECD-3646, May, 1955, for density of water as a function of temperature.
REACTOR
234
ANALYSIS
[chap. 5
f function,
which corrects for the non-l/t) part of the cross section, is

Fig. 4.30. The appropriate thermal-group
data for
to
Table
are
aluminum
2.2)
(refer
The
obtained
from
= <ra(293)
^(333)
r~M
(0.23) (0.831) = 0.191 barn
(4) (333)
a, = 1.4 barns
These data may now be used to compute the macroscopic nuclear

properties of the core constituents in terms of the volume fractions with
which
"metai
they
appear
in the core mixture (namely,
In the
= 0.396).
2<,H''
vH,o
= 0.604
and
case of the water
(0.0181) (0.604) = 0.01092 cm"1

= 1.24 cm"1
S(h,o) = (2.05)(0.604)
And for the aluminum (where we assume that in first approximation

vF
= 0)

v,a.,
2-
at
-ai, _
~ ZS
Pa.(Wa1)
(2.7)(0.396)(0.6025)(0.191)
= (2-7)(0.396K0.6025)(1.4)
^^
= 0.00456 cm"1
The transport cross sections for the water and aluminum may be used to
This is an accept
determine the diffusion coefficient of the core mixture.
able approximation in the event that the fuel concentration is relatively
low; this is the case here. Thus we compute for the mixture
=
327r
3[2'> +
=
2[AI)]
(3)(1.24 + 0.0334)
= '262 Cm
The critical fuel mass is computed from the criticality condition for the
The appropriate relation is (5.181) with k = 1. If
one-velocity model.
we solve this equation for 2'aF), the macroscopic absorption cross section
of the Um, we obtain [cf. (5.185)]
+ DB>
~ 1
V
Zj,0)
where 2^' denotes the absorption cross section for the nonfuel components
in the mixture, the aluminum and water.
The quantity B2 is obtained
from (5.194).
If we neglect the extrapolation dimensions for all three
surfaces (i.e., if we use o ~; a, for example), then for a = 27.7, b = 16.15,
and
c = 30
cm,
BH0 (0
+
(0'Om
DIFFUSION THEORY
SEC. 5.6]
If
235
we introduce this result, along with D = 0.262 cm,
and
i?
= 2.08
into (5.185),
=
2J,0) =
0.0155 cm-1,
obtain
we
0.0155 + (0.262)(0.0152)
1 .Uo
Mmmrl
The corresponding value of the critical fuel mass MF is obtained from the
relation (where V = 107,000 cm3, the volume of the core)
M"
P" =
VS/M, _
(107,000) (0.01808) (235)

(0.6025)(561)
~0^>
_-
L346 kg
(5-250)
At this point we interrupt the calculation to check our assumed value

for the effective neutron temperature, using the absorption cross sections
The expression for k required for this calculation is
computed above.
given by
K
-zt
(4116)
that T = 293K, 2 corresponds to the

293K), and 2 refers to the high-energy value
2,200 m/sec value (Tt
It should be recognized that the above
of the scattering cross section.
k
relation for was based on a medium consisting of a single nuclear species
and therefore not properly applicable to the present situation.
How
ever, in the absence of more complete information it will be adequate for
making an estimate of T. We argue, furthermore, that the concept of
an effective neutron temperature is merely a useful tool for estimating
the distributions of neutrons in the low-energy range and should not be
Also, Eq. (4.211) for Tn is only an approximate relation,
taken literally.
a
crude
and therefore
estimate for k should suffice. For this purpose we
use the following data:
where

it will
be
recalled
=
2.(293)
2<H''(293)
2iA1,(293)
+ 2i'>(293)
where
2(H,o,(293)
^H,0)(293)p(333)a
= (0.66) (0.985) (0.604) (0.6025)

18
<4h,o
= 0.01314 cm"1
2-(293)
S(293)
= (0-28)(2.7)(a396)(0.6025)
0i00549
= (001808)(687)
^^(333)
561
m_,
=
thus 2(293) = 0.0408 cm"1.

For the scattering cross section in (4.116)
we will use the value for hydrogen
range,
at the high-energy
REACTOR ANALYSIS
236
= 20.36 barns.
o-<H)
With the volume fraction
= 0.604, this gives
18
Thus
(2) (293)1*
and (using A =
t>H,o
= (2K0.985)(0.604)(0.6025)(20.36)
Nrf*
2.
[CHAP. 5
(3) (333)
/0.0408\
0.811
V.
0386
1)
TH = TN(1
+ IAAk)
347K
Thus our initial assumption was somewhat in error.

We compute, next, the maximum neutron flux at the center of the pro
We begin with an
posed reactor during operation at maximum power.
expression for the total reactor power P in terms of an average thermal
flux
4>.
eZ/^F
(5.251)
X 10~" watt-sec is the energy release per fission.1

average flux is obtained from the relation
= 3.17
dx dy dz
for (x,y,z) the result (5.193), and assume that

easy to show that
(-)
The
(5.252)
use
is
it
If we
o, etc., then
abc
(5.253)
The substitution of this result into the power relation yields
= 20 megawatts,
s'
^ (tt),
and
m ana
= -01808
(650) (0.974)
,M,,:",:' ,m
For
which
close
approximation
= 1513
^(li--S^)(13,420)
a
is
we have
10,5
-utrons/cmVsec
to the flux at the center of the core
Eq. (5.193)].
In the event that this reactor were used as
research instrument,
would be of interest to determine the neutron current at the outer surfaces
of the system.
Presumably, in such an application, the test specimen or
to
sample would be placed as near to the core as possible so as to expose
One can estimate the neutron current at
the highest neutron density.
it
[see
it
4>(0,0,0)

,z)
where
Based on a figure of 197 Mev per fission
(see Table 1.1).
DIFFUSION THEORY
sec. 5.7]
237
the outer surface of the core by means of the diffusion-theory model.

The appropriate relation is (5.48). In the present case, this may be
written
as
(5.255)
wherein we have selected the center of the x = a face of the core as the
Note that the above expression yields the maximum current
test point.
for any point on that face. If (5.193) is used in this relation, it may be
shown that
This yields for the extrapolation

We compute a, using (5.71).
=
=
0.558 and a = 28.26 cm.
If we
0.7104Xtr
(0.7104)31)
6
= 1.513
1015
neutrons/cm2/sec,
j'+(a,0,0) = 2.27

It
distance
use
for
then
X 10u neutrons/cm2/sec
should be noted in closing that the flux level and current obtained in
Better estimates will be

this sample calculation are somewhat large.
obtained in later examples which include the effects of fast-neutron losses.
5.7
The diffusion
Applications of the Kernel Method
of neutrons
from elementary
sources was discussed in
The direct application of these results to thermal neu

acceptable in many situations of practical importance; however,
Sees. 5.2 and 5.3.
trons is
in some
cases wherein the actual source yields fast neutrons, the source
term used in the one-velocity equation for the thermal neutrons must be
altered to account for the behavior of the fast neutrons prior to reaching
thermal energy.
Some aspects of this problem were discussed in Sec. 5.4g
in connection with the selection of a suitable fission-source term for the
diffusion equation [see Eq. (5.211)]. In this latter application it was
observed that fission neutrons appear with large kinetic energies and that
during the subsequent slowing-down process some of these fast neutrons
are lost to the system by resonance capture and leakage.
The discussion
of Sec. 5.4g included a few remarks about estimating some of these fast
losses, and
it was noted that
a more complete
study of this aspect of the
problem would be considered in Chap. 6.

a. First-scattering Source.
Another feature of the fast-neutron source
which must be given careful consideration is the first flight.
When a fast
neutron first appears in a medium it travels outward from its source point
until it makes a collision.
At relatively high energies the absorption
cross section is small (varies as l/v) so that this collision, as well as a
number of those following, is likely to be a scattering event.
In many
HEACTOR
2.38
ANALYSIS
[chap. 5
instances, the largest portion of the total distance traveled by a neutron

Although on the average
during slowing down is due to the first flight.
many additional scatterings are experienced by the neutron, these scatter
As a
ings occur within a short distance of the point of first collision.
reaches thermal energy in the vicinity of the firstcollision point. We can draw, then, the following crude picture of the
result, the neutron
process: a first long flight which accounts for most of the

distance traveled in the fast range, followed by thermalization at the
point of the first-scattering collision.

slowing-down
This rough model of the slowing-down process may be used as a first

approximation in accounting for some of the wanderings experienced by
a fast neutron prior to reaching thermal energy.
It is clear from the
discussion above that in this model the source strength of the onevelocity thermal neutrons at any given point in the medium is simply the
This
first-scattering-collision density of the fast neutrons at that point.
behavior is quite different from that due to the point source studied in
Sec. 5.2b. In the present case a fast source neutron does not join the
thermal population until it has suffered a first-scattering collision, and it
then does so instantly.
The analysis of Sec. 5.2b, on the other hand,
assumes that all neutrons which appear from the source are already at
the one- velocity of diffusion (i.e., are thermal) and enter into the diffusion
It is to be expected therefore that the thermal flux
process immediately.
distribution due to the present model will be more "spread out" (i.e., it
will have larger first and second moments) than the flux given by Eq.
Let us now determine this effect.
(5.84).
Consider, therefore, an isotropic point source of fast neutrons in an
If 2, denotes the macroscopic scattering cross section
infinite medium.
of the fast source neutrons, and the source releases q0 neutrons per unit
time, then the density of first-scattering collisions at radial distance r
from the source is given by
Number of neutrons
available at surface
of sphere of radius r
probability that first'

scattering occurs in
shell of radius r and
thickness unity
(5.256)
We use this quantity as the source term in the diffusion equation; thus,
for a spherically symmetric system,
- DV2(r)
+ 20*(r)
=
4aT2
(5.257)
The solution to this equation may be obtained by the application of the

kernel method.
The gerleral solution is given by (5.97). In the present
case the source distribution <S(r) is given by the right-hand term of
DIFFUSION THEORY
SEC. 5.7]
(5.257)
239
it follows, therefore, that
*(r)
D&rT1
/-
P2|r
(5-258)
e\
The integration is readily carried out by choosing a

with xJ 20/0.
coordinate system, as shown in Fig. 5.29, where the x axis is taken along
the field vector r; then,
0(r)
5oS.
j*
exp
[-*(/-
(r2
p2
2rp cos fl)] sin

2rp cos fl)
dd
[2*
and
*(r)
sinh xp e~2" dp
Q<2'
\c-.r frsl
WDxr [
Jo
+ sinh ht
Jr
(5.259)
It
comparing in the two cases the mean

square distance from the source point
to the point of absorption (as thermal
in the
r22</)(r)4Tr! dr
we
use
(r)
for
(5.260)
S*(r)4xr2 dr
the result
(5.84)
r2
(thermal neutrons appear directly at

the point source), (5.260) yields for
fo"
r2
If
we calculate
cases
f"
two
is,
that
neutrons);
r2 = 6L2
P
where
trons.
Fig.
5.29
Coordinate system.
(5.227)
For the flux given in (5.259)
X,
we compute,
2/xZ\%^-
on the other hand,
2K +
6L2
(5.261)
1/S the scattering mean free path for the fast source neu
Thus the mean square distance to absorption in the case of the

source exceeds that for the point source of Eq. (5.84) by
first-scattering
the term 2X2.
Gaussian Source.
Although the concept of first-scattering source

useful for a number of situations in reactor physics,
it
b.
described above
is

is difficult to compare (5.259) with (5.84) directly. However, a

measure of the relative spread of the two functions may be obtained by
REACTOR ANALYSIS
240
does not provide
a very
satisfactory
[chap. 5
treatment
A superior model is developed in Chap.
of the slowing-down
This model too has

limitations, but for many reactor applications it will give adequate
In order to provide a more complete treatment of the thermalresults.
neutron source problem considered in this section, we will use some of the
results of Chap. 6 in our analysis and leave until later the derivation of
process.
the appropriate
6.
relations.
We recognize, then, that the assumption that a fast neutron is thermalized very near the point at which it suffers its first scattering collision is
As noted previously, a neutron may experience con
extremely crude.
siderable additional travel during later stages of slowing down as it passes
from one scattering collision to the next.
For a medium of the heavier
nuclei, the total distance traversed by a fast neutron can in fact become
quite large in comparison to the distance it covers as a thermal neutron.
The treatment of Chap. 6, which is based on the Fermi age model, shows
that this wandering about during slowing down produces a highly spreadIn the case that the source of fast
out distribution of thermal neutrons.
neutrons is a point source, this dispersion of thermal neutrons is given by
S(r)
qPf_~Zlh
(5-262)
p,h is the resonance-escape probability from the source energy to

thermal energy and Tth is the Fermi age from this source energy to thermal
where
The expression in Eq. (5.262) is developed in Sec. 6.2b.

(see Sec. 6.2a).
we use the relation (5.262) as the source function for the thermal-
If
diffusion equation, then the thermal flux may again be found by using the
Thus if (5.262) is substituted into (5.97), it can be
kernel method.
sinh
sinh xr
rtf-"'1***-"'
dr^j
D*7$k
*M
{*-
shown that
(5.263)
(4r,h)V8x
V*
Vr
source (5.262)
3
slowing-down
is
The mean square distance traveled by the neutrons from the source to
the point of absorption as thermal neutrons may be obtained by using
Eq. (5.260) and the flux given by (5.263); the result for the Gaussian

the relation
Vr
(4rth)i/4x =
(4rth)/4x
As compared with Eq. (5.227), this result shows the greater spreading
out of the neutrons which occurs during the slowing-down process.
c. Isotropic Flux in Slab from First-scattering Collisions.
We demon
strate some of the ideas developed above by computing the isotropic flux
DIFFUSION THEORY
sec. 5.7]
241
in a slab due to first-scattering collisions from a collimated beam of neu

Consider, therefore, an infinite slab of nonmultiplying material
trons.
one face of which is exposed to a collimated beam of thermal neutrons of
strength
q0
neutrons per unit area per unit time.
We specify the isotropic
flux as the quantity of interest because the effect of the slab is to convert
This effect
the parallel beam of neutrons into an isotropic distribution.
materials
most
in
slabs
of
high-nuclear-mass
would be
pronounced
because of the fact that a beam neutron scat
tered by the slab would emerge with a nearly
isotropic distribution (in the laboratory system of
Plane of
sources
Further scatterings within the

coordinates).
slab would increase this tendency; it might be
expected, then, that the steady-state distribution
isotropic

of the neutron population could be well described

by means of an isotropic flux function.
We begin the analysis by developing a suitable

kernel for the specific geometry in question.
2a
Consider, therefore, the system shown in Fig.
Fig.
Slab
with
5.30
5.30. The immediate problem is to determine
plane isotropic source.
the flux distribution in the slab due to an isotropic
This result, which is the
plane source of unit strength placed at x = s.
kernel for the slab problem, may be used directly in the computation of
the isotropic flux due to the collimated beam.
The appropriate differential equation and boundary conditions for this
system are
"(x)
and
+{a)
(2) J+(s)
(1)
(3)
<Ms)
- x*(x)
= 0
= J~{s)
=
= 0
_(-a)
+
>-
where we have used the approximation a
= 0
(5.265)
~ a for
the extrapolated bound

the flux in the slab for x > s
ary conditions and have denoted by 4>+{x)

A similar notation is used for the net
and by -(x) the flux for x < s.
current J. The current relation (2) follows directly from (5.59), and the
flux condition (3) from (5.60).
It can be shown that the solution for this
system is given by the functions
sinh x(a + s) sinh x(a
Dx sinh 2xa
-(x;s)
sinh x(a
s) sinh x(a
Dx sinh 2xa
x)
+ x)
x >
x <
(5.266)
where (x;s) denotes the flux at x due to a plane source of unit strength
at x = s. The subscripts used in (5.266) refer to the notation mentioned
These expressions give a complete description of the kernel for
above.
the slab geometry throughout the region a < x < o.
242
[chap. 5
ANALYSIS
REACTOR
J'a
The application of the result (5.266) to the calculation of the isotropic

flux at any point in the slab due to an arbitrary distribution1 S(x) of iso
tropic sources is straightforward. In the present case the appropriate
integral equation is seen to be [cf. Eq. (5.97)]
=
4>(x)
S(s)
(x;s)
ds
(5.267)
ds
S(s)
S(s) >+(*;*)
JJ
4>(x)
f*a
is
it
s,
different form depending on whether

Because the kernel {x;s) has
convenient to write (5.267) in terms of two integrals; thus,
4>-(x;s) ds
(5.268)
In the present situation we do not count the neutrons in the beam as

part of the isotropic flux distribution in the slab until they have first
These first collisions convert the beam
scattering collision.
experienced
S(x) dx
probability that
number of neutrons
which enter slab per
qQ
slab surface
2,
2i
[Z,6-z<
beam
neutron makes its first

scattering collision in
unit time and area

' of
0,
Thus we define S(x) dx as the

the density of first scatterings in the slab.
If the
number of first-scattering collisions per unit time in dx about x.
beam strikes the slab at face

then
dx about x
= qoS.e-Î+") dx
dx]
(5.269)
e^'
sinh x{a
a)
+
s)
c-z,(*+,) sinh x(s
sinh x{a +
x)
sinh x(a
x)
Dx sinh 2xa
is
<?o2,
(*)
is
is
the probability [cf. Eq. (2.17) and related

2t(x + a)] dx
in dx about x and 2,/2
that
first
collision
occurs
the prob
discussion]
ability that this collision
The substitution of
scattering event.
into
and
the
of
results
use
the
(5.269)
(5.268)
(5.266) yield the equation
where 2t exp
ds
ds
(5.270)
(x;s)
should be given in terms of the extrapolated

By using this approximation we are able to write all the
boundary a.
a.
pertinent functions in terms of the boundaries
Although this
x
properly,
the kernel
o)
is
is
Our purpose in introducing the approximation a ~ a

now evident.
flux
Note that the
(x) given as the integral over the physical volume
of the system (from to
of all sources
within this volume, whereas

is
is is
neutrons into an isotropic distribution, and the resulting flux

given by
the diffusion equation.
The general solution to this equation
given by
identified as
(5.268), and the source distribution S(x) which appears
S(x) neutrons per unit area per unit time.
DIFFUSION THEORY
SEC. 5.7]
243
causes some error, it does simplify the integration and yields more com
pact results.
It
^ sinh *( "
2xa
e-s' sinh x(a +
x)
DW -I*'')
x)
*(X)
sTnh
The completion of the integrations indicated in Eq. (5.270) yields the

solution
erz,x sinh 2xo]
(5.271)
2
,
OA
=1.145 cm"
D =0.357
K =0.771
0.3
0.51
is
of interest to compare this result with an elementary diffusion-theory

calculation based on the assumption that the beam neutrons which strike
cm
"
cm
-2 -1
-3
-5 -4
0.1
x, cm
Fio.
5.31
Isotropic flux due to first scattering collisions in iron slab.
is
the slab enter into

The
diffusion process immediately at the surface.
resulting flux distribution within the slab
given by the solution to the
diffusion equation which satisfies the following boundary conditions:
A(x\
*W ~
3o
(2)
*(o)
(5.272)
j+(-a)
*(a ~
sinh 2xa + 2Dh cosh 2*a
4q sinh
x)
This result
(1)
is
(5
273)
{^'6)
We illustrate the results of this analysis by means of the following

numerical example: Consider
10-cm-thick slab of pure iron for which
the microscopic thermal cross sections are taken to be <ra = 2.53 barns,
= 7.86 g/cm3 and
= 56.
and <t, 11 barns, and
The correspond
=
ing macroscopic properties are: iV
0.0846 X 1024 nuclei/cm3, S, =
0.932 cm-1, 2a = 0.213 cm"1, 2, = 1.145 cm"1, x = 0.771 cm"1, and
p
= 0.357 cm.
Figure 5.31 shows the isotropic flux distribution (j>(x)/q0 in

10-cmthick slab of iron according to the solution (5.271) and using the above
a

0.2
data.
REACTOR ANALYSIS
244
5.8
[CHAP. 5
Problems of Neutron Population in Localized Absorbers
Among the important products of a nuclear reactor are the regions of

high neutron density which may be used for a great variety of develop
ment testing and fundamental research.
The experimental procedure
usually involves placing a sample in the neutron field and determining
the continuous or integrated change in some property. In the evaluation
of the results of such experiments it is usually important to have a quanti
tative knowledge of some features of the spatial (and energetic) distribu

tion of neutrons inside the sample and in its immediate vicinity. Gen
erally speaking the most important effects are produced by two classes of
neutrons: (1) thermal neutrons and (2) very fast neutrons (with kinetic
With respect to the latter
energies of the order of Mev or greater).
group (which produce most of the effects classified under the heading of
"radiation damage"), it may be said that the distribution in energy is
characteristic of the particular reactor as well as of the specific region of
the reactor used. Usually the spectrum of fast neutrons will be little

affected by the test sample itself (although auxiliary devices, such as a

"converter plate," may be used to affect a major alteration) and, except
in the region of absorption resonances, the fast-neutron density may be
expected to be fairly uniform over the sample and its immediate vicinity.
Thermal neutrons, however, suffer significant variations in density in the

region of almost any test sample; this behavior is the subject of discussion
for the present section.
In the general case, the conditions of the experiment are poorly ame
nable to theoretical treatment; the sample will usually be multiregioned
and of complex geometry; the site of the experiment will often be a
channel in the reactor in which the neutron density varies spatially and
in which the streaming of neutrons gives rise to quite different neutron
currents incident on the various surfaces of the sample; the immediate
surroundings of the sample will usually be a complex mixture of various
Consequently, at the
moderator, structural, coolant, and fuel materials.
very outset, any hope of rigorous treatment of all aspects of the problem
and the problem reduces to that of choosing a simplified
model, amenable to calculation, which will yield a satisfactory approxi
mation to some feature of the physical situation of particular interest.
This section is devoted to a detailed consideration of some of the models
is abandoned,
and concepts which have been found to be useful.

a. The Distribution of (One -velocity) Neutrons in Absorbing Regions
in a Uniform Isotropic Neutron Flux Field. Here we consider an absorber
By definition, an isotropic
placed in an "isotropic" neutron flux field.
of
the
surface
of the sample there are
flux is such that on an area dS
dS cos 6 sin 6 dd neutrons per unit time whose velocity vectors
incident l
DIFFUSION THEORY
sec. 5.8]
make angles between
[cf.
'(
and
dS
Eq. (5.36)]
) d
dS
dd
245
with the normal to the surface; thus
number of neutrons incident per unit time
^ a SUIâce (m an isotrpic
whose
of attack have cosines
angles
Aux)
between
and n + dp
(5.274)
n an area
of the isotropic flux and /x = cos 6.

This
4><*is the magnitude
choice for the angular distribution limits the applicability of this model;
in general the angular distribution of neutrons falling on a sample will
not be isotropic; even if the flux prior to the introduction of the sample
where
were isotropic, the very presence of the sample would introduce perturba
However, in many reactor situations
tions in the angular distribution.
the assumption of an isotropic flux will afford a good approximation to
Precisely, the simple isotropic distribution
the actual distribution.
would be realized near the center of a spherical surface of large radius on
which isotropically emitting sources were uniformly and densely packed.

Complete information about this system would involve knowing the

neutron density and angular distribution at every point inside the sample.
In general, such information requires extensive and laborious calculation ;
however, in two extreme approximations the problem may be solved with
The complication in the physical situation is the scatter
relative ease.
ing collisions made by the neutrons before they are absorbed or escape
If, on the average, this number of scatterings is a small

but nonzero number (of the order of 1 to 3), the problem is very difficult,
The two
for it is necessary to follow each neutron history in detail.
extreme cases which are amenable to calculation are: (1) The diffusion
from the sample.
model:
In this
are made, on the average, by the

is absorbed or escapes.
(2) The first-
case many scatterings
neutron in the sample before
it
flight transport model: Here the neutron makes no scattering collisions

in the sample, but continues along its original line of flight until it suffers
an absorption collision or escapes from the sample.
On the basis of these two models, various quantities of interest may be
calculated; those to be discussed here are listed and defined below:
The transmission coefficient:
fraction of all neutrons incident upon surface

a = of a lump of material which pass through lump
(5.275)
without being absorbed

We will demonstrate
mittance,
later that the transmission coefficient (or transfrom a physical standpoint, with our
cf. Sec. 5.3a) is identical,
previous definition of the albedo [cf. Eq. (5.111)].

to use the same symbol.
It is appropriate,
then,
HEACTOR ANALYSIS
246
[CHAP. 5
The self -shielding factor:
average flux throughout volume of a lump

average flux at surface of a lump
(in the presence of the lump)

The over-all self-shielding factor:1
average flux throughout volume of a lump

value of flux at position of lump before it was introduced
for any given physical situa

However, there are
tion, but there is no universal relation among them.
derived quantities directly related:
These quantities
will, of course,
be related
The effective absorption cross section:
"**
that microscopic cross section which, if assigned to

material of absorber lump, would allow a correct
calculation of total absorptions in lump in the
form given in Eq. (5.279)
Total absorptions per unit

..

time in absorber lump

where
Vl
= AL =
,T
N<T,(5.279)
,_
(5.279)
is the number of atoms per unit volume in the absorber lump,

of the lump, and 4>o is the flux in the system in the
is the volume
From (5.277) and (5.278)
absence of the absorber lump.
o-e((
where
<r
is the true microscopic
we have
= <rj0
(5.280)
absorption
cross section of the absorber
material.
The flux correction factor:
that factor by which the quantity Al/'LoVl

=
must be multiplied in order to give flux in
fc
absence of absorber lump
From (5.279) and (5.280)
fc
Cfff
(5.281)
(5.282)
Jo
The general principles for the calculation of the quantities defined in

Eqs. (5.275) through (5.281) in the two extreme approximations may now
be presented for the case of an absorber of general geometric configuration
placed in an isotropic flux.
I. The Diffusion Model.
The diffusion equation
V2(r)
1 See also Sec. 7.4b,
particularly
- x2(r)
Eq. (7.227).
= 0
is applied in the
form
(5.283)
DIFFUSION THEORY
sec. 5.8]
247
where xJ = 2/ D, 2 = N<ra (the macroscopic absorption cross section of

the material of the lump), and D = \tr/3 (the diffusion coefficient of the
material of the lump).
(The method presented here is for a singlematerial lump, but is easily extended to multiregion lumps by application
of the diffusion equation in each region and the continuity of flux and
neutron current conditions at the interfaces.)
The two boundary conditions required on the flux given by Eq. (5.283)
are:
(1) The flux must be a symmetric or nonsingular function \
I
(depending on the particular geometry)
(2) The total number of neutrons incident upon unit area of (
the sample1 per unit time is #0/4
)
\ *(R)
+ ~
[n
V(R)]
\a
(5.285)
where n is the normal vector to the surface and R represents the coordi
is completely specified.
From
we
nates of the surface S; thus,
(i)
(r)

calculate
j(R)
dS
(5.286)
JSi
where
Sl
is the
also
total surface area of the sample, and

j{R)
/.
/o =
\ *(R)
17^
V
L<PSL
-|
[n
V*(R)]
/ (r) di
JVl
J-r I
*(r) di
(5.287)
(5.288)
(5.289)
where 4>sl is the average value of the flux over the surface Sl of the lump.
2. The First-flight Transport
Model.
The calculation of the spatial
variation of the flux in this approximation is not so straightforward as in
In general,
model, and no fixed expression can be given.
the procedure for finding the flux at any point inside the lump is to choose
an infinitesimal volume element dV at that point and calculate the total
the diffusion
track length per unit time in dV due to all neutrons impinging on the
The
lump which succeed in passing through dV before being captured.
1
Note that the condition

0. Near the lump the
directed toward the lump,
means that the presence of
with
sources.
at the surface is not taken to be continuity of the flux

total flux will be reduced from <,, but the part of the flux
by assumption, will not be changed. Physically, this
the lump is assumed not to alter the effective neutron
REACTOR ANALYSIS
248
[CHAP. 5
flux is given by the limit [cf. Eqs. (3.39) and (3.40)]

,.
/total track length/time in dV\ = ,

*
W
(5 290)*
On the other hand, the transmission coefficient is more readily obtained.

we will require the complete directional specification
For this application
of the current and therefore select the form

j(m,<A) dn
^-
4>*n
dn
(5.291)
as given in the first term of (5.32).
Here, ju is the cosine of the angle

of motion of the incident neutron makes with the
normal to the surface of the lump and ^ is the azimuth of the neutron
which the direction
direction
about the normal.
(5.291) is a consistent order of
Expression
with (5.274).
approximation
The integral of (5.291) over all directions
into the lump and over the complete surface area Sl of the lump is ^/Sl4>oSince the probability that a neutron will pass through a distance s of
material without being absorbed is exp ( 2s), we have for the trans
mission coefficient
= cTT
ol<Po
=
Jsl
dS(R)
4r Jsl
/ dS{R)
ir>L
P d
Jo
\
Jo
fi dn
jfof
Jo
e-*-*<*'*M
Jo
rff
d+
(5.292)
where dS(R) is the surface element at surface coordinate R and s(R,n,^)

is the straight-line distance across the lump for a neutron incident at R
For several elementary cases, s does not
traveling in direction (fi,^).
depend upon the position R of the surface; so s(R,0,^) > s(6,\f/), and
(5.292) reduces to
a =
Jo
sin
cos
d8
<r
Jo
(5.293)
Of course, the form of s{8,^) is sufficiently simple so that the integral of

Eq. (5.293) may be performed in closed form in only a few cases.
For the cases to which the first-flight transport model is applicable,
there is a relation between the transmission coefficient a and the over-all
self -shielding factor /0; this is
1
where [cf. Eq. (7.232)]
-a
(^P)/o
2as/
(5.294)
sec. 5.8]
That
DIFFUSION THEORY
249
by (5.295), is indeed the average path length of neutrons

the volume of the lump (with the lump nonabsorbing and an
incident isotropic distribution) may be seen by calculating the flux in the
nonabsorbing lump and equating this to the undisturbed flux.
(All this
s, as given
through
is for lumps with nonreentrant surfaces.)

For the case of very slight absorption, Eq. (5.293) reduces to a simple
If we take, then, Zos(0,^) <SC 1, the exponential may be repre
form.
sented by the first two terms of a Taylor series, and we obtain from
(5.293), with the aid of (5.295),
for small absorption
2as
(5.296)*
The application of the general methods discussed above to several

specific geometries is carried out below and the results are summarized.
The solid sphere:
1.
Diffusion approximation:
(r)
'

1
1
sinh xR
oR
2D
-^Tff.
(*R coth xR
/.
2.
(xR coth xR
2D(xR coth xR
+ 2D(xR coth xR
3
f.
+ 2D{xR coth xR
1)
(5.297)
l)/R
l)/R
(5.298)
1)
(5.299)
- l)/R
(5.300)
First-flight transport approximation:

<Kr)
where Ro
t1
{2e-(l + Ro) sinh r + 2e-r0 cosh
(Rl
- rDiEiiRo -
Sa/2 and r0 = 20r.

s(e,+)
2R
cos
r0)
- E^R*
r0
r0)]|
(5.301)
Also,
(5.302)
(po
Po
+ l)e-"]
- tr* +
Po
pi + 2Poe-' + 2e-'
po
= 2Saff
- 2\
(5.303)
(5.304)
(5.305)
and1
E(x)
I:
e-" dy
(5.306)
1 The
E(x) functions are denned according to K. M. Case, F. de Hoffmann, and
G. Placzek in "Introduction to the Theory of Neutron Diffusion," vol. I, p. 153, Los
Alamos Scientific Laboratory, 1953 (U.S. Government Printing Office).
The func
tions E<>to E, are tabulated on pp. 155-161.
REACTOR ANALYSIS
250
[chap. 5
of the two approximations for the calcula

of
a
10-cm
case
sphere with Xt, = 3 cm, as a function of
The diffusion approximation should be valid for
absorber loading (2).

Figure 5.32 gives

tion of a, for the
a comparison
small absorption, while the transport results should hold at large absorp
tion.
Note that the diffusion result is totally incorrect beyond a certain
value of 2, actually giving negative values of a.
The infinite slab:
1.
a cosh
xx
cosh xa + 2Dx sinh xa

1
+2Dx
2Dx tanh xa
tanh xa
xa coth xa
f.
2.
face)
+ 2Dx tanh xa
First-flight transport approximation (with x measured from left

:
*(*)
8(0,+)
\ 0{E2(Xax)
- *)]}
x > 0
2a
cos
e~z
"
(5.311)
(5.312)
a =
+ Et[X.(2a
XofT* + XqEi{Xo)
Xo
= 22ao
(5.313)
(5.314)
DIFFUSION THEORY
sec. 5.8]
251
The spherical shell:
(xR<,
(5.316)
- 2Dx) coth xT
2Dx) coth xT
xT cosh xT + (x'-RRo
sinh xT
cosh
xT
(xRo
2DxRo/R
2Dx)
+
2D/R + 2Dk*R0) sinh xT
(5.318)
= xA
(5.319)
(1
Rl)
(5.317)
*2(R3
- 2DxR0/R
2DxR0/R
(xRB +
2Dx2Ro
in
fo
2Dx2R<,
-~2Dx^
2Dx ^ + 2Dx
- 2D
sinh xT
R, and
3fl
The average flux inside the shell
(1
|2
a =
First-flight transport approximation:
j,
2.
Po
vT^p)
^Y~rî) exP (-p0
J
4
sv
'
P07)
72)2
for a very thin shell
a =
7
2/4][(l
- |2T +
- 42aT
*o(l
1
4>i
72o
JT4[(1
= 22ft
/VI
72)]
po
873
1)
(1
U
fl
v2Pe-Po(l
6XP
7)
with
3*o
[1
(1
(1
72)
(1
Pi
1
,
- + - T)po]e-'.-^
- - 7') exp (-p Vl - +
ie7)2
Po,, ~
Po(l
7*)*c-"-')
/i
- (~Po Vl
4(1-7)
- 72)2[i(P0 - - #i(po Vl -

2J)x2fio
- 2D/R
+
2D/R
" _
with
(5.315)
Ro)\
R0x
+ sinh x(r
R0)
xT
{ cosh
1)
A = Rto
x(r
= [xRo cosh
0(r)
^R0x
1.
72)
(5.320)
(5.321)
2/2]e-'"'2
dy
(5.322)
(5.323)
(5.324)
REACTOR ANALYSIS
252
[chap. 5
The infinite solid cylinder:
1.
<Kp)
a =
f.
4>oIo(xp)
(5.325)
- 2DxIl(xR)/I0{xR)
Io(xR) + 2Dxh(xR)
1
1
(5.326)
+ 2DxI1(xR)/h(xR)
2I1(xR)
(5.327)
xRI0(xR)
/o
/.
2DnI,(xR)/h(xR)
(5.328)
The integrals involved here

2. First-flight transport approximation.
However, some results have been
cannot be performed in closed form.
1.0i

0.8
0.6
^Sla D
8 0.4
-Cylin der
0.2
-;
Spher e
0.8
0.4
1.6
1.2
2.0
2.4
2.8
3.2
2. s
Fig. 5.33 Comparison of transmission coefficients for an infinite slab, infinite cylinder,
and sphere in the first-flight transport approximation.
2$ 4Z<,Fz,/Sl.
obtained
transmission
coefficient is
a =
with
[z2
AR3
+ 2R\1
f~
Jo
dz
fx
Jo
The transmission
r*(l
cos
J.
The form of the integral for the
calculation.1
by numerical
-a
2ZaR
eos
dfi
(5.329)
(5.330)
coefficient a is given in Fig. 5.33 as a function of
C. Bartels, "Self-absorption of Monoenergetic Neutrons," KAPL-336, May

R. J. Royston, "The Absorption of Neutrons by a Small, Strongly Absorbing
Rod," AERE T/M 117, November, 1954.
1
W.
1, 1950;
sec. 5.8]
DIFFUSION THEORY
253
in the first-flight transport approximation for slab, cylinder,

and sphere.
The values were obtained for the slab from Eq. (5.313), for
the sphere from Eq. (5.303), and for the cylinder from the paper by
dimension
Bartels.
The approximations to a for lightly absorbing
slabs, cylinders,
and
spheres are given below:

Slab: 2a
Cylinder:
Sphere:
2K
2K
- 42aa-' In y2Xaa - f
with
~
a
22a + f
+
- l(22.fl)
a ~
+ i(22afl)2 +
1
(22aa)2
7 = 1.781072 .
22ff2
Infinitely long cylindrical
+
.
(5.331)
(5.332)
(5.333)
This case will not be discussed here,

in the first-flight transport approximation
have been performed by Dwork, Hofmann, Hurwitz, and Clancy.1
b. Flux Distribution in Vicinity of Localized Absorbers in Diffusion
Media. If a localized absorber is placed in a diffusion medium, the
neutron flux in the vicinity of the absorber is reduced from its value in
the absence of the absorber; the effect of the absorber falls off with

but numerical
shell.
calculations
Since there is always a net

increasing distance from the absorber.
current of neutrons into the absorber, the angular distribution of neu
trons incident upon the surface of the absorber is no longer isotropic,
but, in general, has additional terms depending upon the derivatives of
For example, in the diffusion approximation the
the flux at the surface.
angular distribution is [cf. Eq. (5.36)]
j(n)
dfi =
<om
dn
%D4>'nn2
(5.334)
dfi
and 'n are the neutron flux and its derivative in

where
in question, respectively.
The treatment of the preceding
section is rigorously correct where all derivatives of the
while this condition can never be met near an absorber, it
the direction
part of this
flux vanish;
may be true
that the second term of (5.334) is negligible in comparison with the first.
In the following presentation we first neglect the second term in (5.334),
then try to take it into account in simple cases.
In both calculations,
however, we retain the assumption that diffusion theory holds outside
the absorber; for small, highly absorbing lumps this assumption is in
error.
1. Assumption
of an Isotropic Flux Incident on the Absorber from the
Moderator.
As a first model for the calculation of the flux depression
near an absorber we assume that (a) diffusion theory holds everywhere in
the moderator, (6) an isotropic flux of neutrons is incident upon the

surface of the absorber.
Both of these assumptions are closely met if
the flux is nearly isotropic everywhere in the diffusion medium.
The
second assumption
allows the use of the quantities
calculated
for the
1J. Dwork, P. L. Hofmann, H. Hurwitz, Jr., and E. F. Clancy, "Self-shielding

Factors for Infinitely Long, Hollow Cylinders," KAPL-1262, Jan. 10, 1955.
REACTOK ANALYSIS
254
[chap. 5
absorber in Sec. 5.8a; for example, the fraction of the incident neutrons
transmitted through the absorber will, on this basis, be the a calculated
in Sec. 5.8a, which is appropriate for the geometry and neutron proper
Thus the absorber may be replaced, in so far as the

ties of the absorber.
diffusion medium is concerned, by a boundary condition to be applied at
the surface of the absorber which states that the fraction of all neutrons
incident upon the absorber which return to the diffusion medium is the
appropriate a (see also discussion in Sec. 5.3c).
The additional boundary condition needed for the flux in the diffusion
medium will be found from the external conditions, including the neutronsource distribution, at some distance from the absorber.
Such boundary
conditions may be chosen to fit the particular case, but a very useful one,
which may be presented here as an illustration of the procedure, is
tances from the absorber approach its value in the absence of the absorber.
Such an assumption might very well represent the case of an absorber
placed in a large moderator region in which neutrons are being thermalized fairly uniformly over the region.
However, the results which
have a wider applicability than might be suspected from the
For any case in which
restrictive choice for the neutron-source function.
the absorber is placed in a large moderator region not near any strong
localized source of neutrons and in which the introduction of the absorber
follow
does not alter the
distribution or strength of the sources, the flux depres
J'+W
j_(R)
and
lim *(r)
- xV(r)
0(R) ~ 2ZV(R)
PRW(R)
=
(5.335)
W(r)
moderator
is,
sion will not differ appreciably from the forms given.

for the flux in the
The mathematical formulation of this model
(-336)
(5.337)
where So
the uniform source intensity of neutrons, 4>0 the flux in the
region of the absorber prior to its introduction, and a the transmission
coefficient calculated on the basis of the diffusion-theory estimate for the
partial currents [see Eqs. (5.49)]. We find then, as indicated earlier, that
is
the transmission coefficient takes the functional form of the albedo as

defined in Eqs. (5.111) and (5.117) in the case that the current relations
at the surface of the lump are based on the use of the diffusion-theory
Thus, in real physical sense the trans
model in the adjoining regions.
a

obtained by imagining that the neutron sources are uniformly distributed

throughout an infinite medium of the diffusion material and that the
introduction of the absorber does not alter these sources. Then, in the
presence of the absorber, it must be required that the flux at great dis
sec. 5.8]
DIFFUSION THEORY
255
mission
coefficient and the albedo serve to measure the same effect,

namely, the capability of a particular region to return neutrons to the
surrounding medium, be it by "transmission" or "reflection."
It is
important to recognize, however, that the precise analytical form of the
albedo and the transmission coefficient will not in general be the same.
For example, Avhen computing the albedo of a reflector in a reactor
wherein the origin of the coordinate system is located at the "center" of
the reactor, the correct form to use is j-/j+, as shown in Eq. (5.117).
In
the present applications, on the other hand, the coordinate system is con
veniently centered within the transmitting (or absorbing) region, and
this leads to the form j+/j- for the transmission coefficient, as indicated
in Eq. (5.336).
These differences are due entirely to the choice of
systems and should not be attributed to the physical
interpretations.
The use of this model involves the choice of the proper a from Sec. 5.8a,
on the basis of either the diffusion approximation or the first-flight trans
port approximation, depending on the nature of the absorber.
It should
not be expected to give good results in cases where the angular distribu
tion in the moderator, near the lump, differs appreciably from isotropic,
as it would, for example, near a small, highly absorbing lump.
A useful measure of the effect of the lump on the flux in the moderator

coordinate
is the flux depression factor defined below:
Flux depression factor = D =

where
*(R)
(5.338)
a is the original
flux and >(R) is the flux at the surface of the lump.

The results of the application of this model for three elementary geo
metrical forms for the absorber lump, all immersed in an infinite diffusion
In the following, the quantities

medium, are presented below.
x, and
D are, respectively, the neutron flux, the reciprocal diffusion length, and
the diffusion coefficient for the diffusion medium.
Infinite slab:
(x)
0[1
De-"'*-"']
x > 0
(5.339)
=[l+2Dx(j-|;)J"1
(5.340)
where asl (for the uniform slab) is obtained from Eq. (5.308) or (5.313),
depending on the neutron properties of the slab.
Infinitely long cylinder
(p is radial distance from
*>-*>[l-aS$i]
cylinder axis)
=[i+2c*is(;^)r
(5-341)
REACTOR ANALYSI8
256
[CHAP. 5
where aP (for the uniform cylinder) is obtained from Eq. (5.326) or (5.329),
depending on the neutron properties of the cylinder.
Sphere (r is distance from center of sphere)
*(r)
= *o
[l
- ^ -"- j
(5.343)
where aSp (for the uniform sphere) is obtained from (5.298) or (5.303),
depending on the neutron properties of the sphere.
Of course, Eqs. (5.339) through (5.344) are applicable to slabs, cylin

ders, and spheres of any internal structure if the appropriate a can be
determined and if the simple one-dimensional variation of properties is
For example, (5.344) gives the flux depression for a hollow
maintained.

sphere of outer radius R if the aSp is chosen from (5.317) or (5.320).

The model, as presented in Eqs. (5.335) through (5.337), may be
applied to other geometries, the limitation being the difficulty of solution
of the mathematical
Two special cases of great practical
problem.
interest, the finite cylinder and the finite slab, are omitted from the dis
cussion here since no general analytical solution is known (to the authors) ;
each case is treated by approximational methods deemed appropriate.
As an example of the use of these concepts, suppose that the (thermal)
absorption cross section of some material were to be measured by putting
into a moderating region supporting a known

By the neutron activation of the material
neutron flux (thermal) o.
and subsequent determination of the rate of radioactive decay, a value
may be obtained for the total neutron absorptions occurring in the sphere
during the time t of exposure to the flux. The total number of absorp
a solid sphere of the material
Al
in terms of the quantities discussed

If the flux at the surface of the sphere during the exposure is
above.
{R), the number of neutrons incident upon unit area of the surface in
tions
is now to be calculated
unit time (under the isotropic assumption)

these are absorbed, we have
AL
\(R)(l
is
(R)/l;
since
- a)iirRH
(1
a) of
(5 345)
From the definition of Eq. (5.344), the flux at the surface of the sphere is
related to the flux depression factor D and the flux in the absence of the
absorber
o by
4>(R)
= 6,(1
2D)
If it
(5.346)
is appropriate to use the first -flight transport model for the material
of this sphere, then Eq. (5.303) may be used for a, and we have finally
[using Eqs. (5.338) and (5.344) for 2D, and denoting moderator quantities
by the index
and absorber quantities
by the index A]
sec. 5.8]
DIFFUSION THEORY
Al
257
2D.(x.v,ff + 1)(1 +
= irR2(t>ot
1
+ 2DM{xMR +
1)
(i
\1
aA)/R
+ A l
- A/fl
(5.347)
Al is known, the absorption cross section of the material of the

sphere may be determined from (5.347) and (5.348).
2. Application of Diffusion Approximation for Current from Moderator
to Absorber,
(a) Diffusion theory applicable to both moderator and
absorber: Here we drop the restrictive assumption of an isotropic flux
Since
incident upon the absorber and consider the angular distribution given in
Eq. (5.334) which is appropriate to the diffusion approximation. We
consider first the results when diffusion theory is also assumed to hold in
the absorber region and subsequently discuss the first-flight transport
model for the absorber.
In the case that diffusion theory holds in both the moderator and the
absorber, and assuming a uniform distribution of neutron sources as
formulation of the problem is:

before, the mathematical
In
In
the
the
moderator:
- *5,<Mr) Vs^(r) - xAA{r)

^ - ^ - r
4=
V24>m(t)
(5.349)
L>M
= 0
absorber:
where
At the moderator-absorber
(5.350)
(5-351)
interface (coordinate R), the boundary
condi
tions to be applied are

(1)
(2)
- DM[n
<MR)
= 4>a(R)
V*(R)]
- DA[n
V^(R)]
(5.352)
(5.353)
where n is the outwardly directed normal to the absorber surface.
addition
In
we require
(3)
Symmetry or nonsingularity on
(4)
lim
Mr)
= 0 =
A.
<ki(r)
(5.354)
(5.355)
This model gives precisely the same results for the flux depression as
that presented in the preceding discussion [Eqs. (5.335) through (5.337)]
if the a is calculated on the basis of the diffusion model [Eqs. (5.283)
In the following, these quantities are given with the
through (5.285)].
dependence on the absorber properties explicitly shown along with the
spatial dependence of the flux for the three simple geometries previously
used. In this case, of course, the results are applicable only to uniform
absorbers.
REACTOR ANALYSIS
258
Infinite slab:
m(x)
0O[1
Da*a tanh x,ia
Infinite cylinder:
1
4>o
(5.356)
cosh xaZ
cosh x.4a
Dmxm
D.SfXM coth X,|
1)
4>m(p)
x>0
JDe-fr-->]
_l_
[chap. 5
-r
(5.358)
ô(x.ifp)
tfo(x.ft) +
(5.357)
DMxMKx{xMR)h{xAR)
(5.359)
1
4>a{p)
00
PAXAI1(xAR)K0(xl,R)
DMxMh{xAR)K1(xMR)
DaxaIi{xaR)Ko(xuR)

Sphere:
u(r)
= 0o
Mr)
= 0o
-v(r-)
A,/
X^
ft COth
DA
x^
(5.360)
(5.361)
(5.362)
j
X.4
+
+
/O(XXP)
/o(x^fl)
X.tftf
R coth
x^
ft
/?
1
or
sinh xAr
sinh xxi?
(5.363)
(5.364)
(b) Diffusion theory applicable in the moderator and the first-flight

transport model applicable in the absorber: In the case of a small, fairly
heavily absorbing lump placed in a diffusion medium, the most appropri
ate model might be diffusion theory in the moderator and first-flight trans
port in the absorber; however, there is a restriction on the applicability of
this model. The angular distribution of the neutrons emerging from the
absorber will be far from isotropic since those neutrons moving in direc
tions for which the path length in the absorber is short will be preferen
tially transmitted.
Further, if the sample is small in comparison with a
diffusion length in the moderator, the rapid spatial variation of the neu
tron flux in the vicinity of the sample may make the omission of higher
derivatives than the first (the basic approximation of diffusion theory)
invalid. Thus, even though the medium surrounding the absorber is a
"good" diffusion medium in the sense that D/Xa <5C 1, diffusion theory
will be a poor approximation if the absorber is very heavily absorbing.
Thus the model to be discussed here is applied best to an absorber whose
dimensions are comparable with a diffusion length (or greater) and whose
absorption cross section is greater than its scattering cross section but not
so large as greatly to alter the angular distribution of the neutrons in
passage.
DIFFUSION THEORY
sec. 5.8]
259
is,
For the absorber, the general form for the transmission coefficient (for
such geometries that the path length through the absorber is a function
only of the angular coordinate of the direction of motion with the normal
in the
to the surface, and not of the particular position on the surface)
notation of Eq. (5.293),
6.
/,
where we define
=
- \*d+
T
Jo
_/0
Me-s-c.*) dn
d+
JoI
'
^-
fi
At
The boundary condition to
(5.367)
ê-2"'"-*'
dp
(5.368)
on the flux in the diffusion

that given in (5.336) and, using (5.366)
medium at the absorber surface
relationship between the flux
for a, this condition may be expressed as
a
is
be applied
Thus
and its first derivative at the absorber surface.

2D4>'(R)
(j^fy *(R)
(5-369)
For comparison, Eq. (5.336) gives for this relationship

2Z)*'(R)
(yqH;) *(*)
(5-370)
is
it
is
the same in the diffusion medium

Since the model under discussion here
as that presented in Eqs. (5.335) through (5.337), except for the replace
clear
ment of the boundary condition of (5.336) by that of (5.369),
a
comparison of Eqs. (5.369) and (5.370) that the results for the
flux in the moderator and the flux depression are as presented for the
three simple geometries in Eqs. (5.339) through (5.344), except that the
from
(1
is
/2
it is
is
is
/j
/2
is
/i
(1
a)
is
ratio
to be replaced by the ratio
/i)/(l +.A).
a)/(l +
The quantity
just the a previously used and
defined in Eq. (5.367)
However,
defined in Eq. (5.293).
new quantity which must be
Note that
the fraction of
computed for each geometry of interest.
flux of neu
neutrons transmitted by
incident
sample upon which
trons whose angular distribution
an isotropic distribution weighted
with
For the cases of
cosine factor which makes
more "forward."
an infinite slab and a sphere, the quantities
are presented below; no

(5366)
2Z)*S
Jo
a =
= cos
For the distribution j(n) given by (5.334), appropriate
with
for the diffusion approximation, a may be put into the form
REACTOR ANALYSIS
260
[CHAP. 5
simple solutions and no numerical work for other cases are known to the
authors.
Slab, half-width a:
ft
= e~
Xo '--
Sphere, radius
x0)
po
+ xlEi(x9)]
(5.371)
22a
R:
f%
If
- ix0[e-"(l -
-.[2Po
2(1
P)e-"
p?e-']
(5.372)
= 22i?

fraction of the neutrons incident upon an absorber are nonisotropic in distribution, the expressions developed in the first part of
this section for determining the flux inside the absorber, and therefore
also the self-shielding factors, will be invalid. The correct calculation of
these quantities, although straightforward, is laborious.
effort has been directed toward
(c) The foil problem: Considerable
improving the models presented above for the particular case of an
absorbing foil in a diffusion medium, i.e., geometrically, a slab whose
thickness is small in comparison with its extent, but whose extent is
a sizable
comparable to or less than a diffusion length of the diffusion medium.

The corrections involve accounting for the "end effects" and attempting
to take into account higher-order
angular terms (than the two considered
by the diffusion approximation) in the diffusion medium near the foil.

Experiments have been performed in an attempt to check the value of
the proposed corrections.
Some references on this subject are given
below.1
Internal Heating in Localized Absorbers. There are several

mechanisms by which the existence of a neutron flux in the interior of a
body may give rise to heating; of these the most important is the fission
c.
In case the absorber consists of a fissionable material in whole

process.
or in part, and if the material is such that the fission fragments do not
from the sample, the experiment must be designed so that avail
able cooling will be sufficient to keep the maximum temperature inside
escape
the sample below some specified value.

Let us examine the case of the conduction of heat in a uniform body
in which the sources of heat are everywhere proportional to the neutron
If G(r) is the heat
flux (restricting attention to steady-state problems).
generated per unit volume per unit time around point r, then
G(r)
1
W. Bothe, Z. Physik,
- 0*(r)
(5.373)
120, 437 (1943) ; E. Corinaldesi, Nuovo cimento, 3, 131

E. Klema and R. H. Ritchie, Phys. Rev., 87, 167 (1952); C. W. Tittle, Nucleon
ics, 9, 60 (1951); R. H. Ritchie et al., "Thermal Neutron Depression," Health Physics
Division Annual Progress Report for Period Ending July 31, 1959, ORNL-2806,
pp. 133-136, Oct. 29, 1959.
(1946);
DIFFUSION THEORY
sec. 5.8]
261
distribution T(r) is to
is some constant, and the temperature

determined from
where
/3
-fVsr(r)
G(i)
= |80(r)
be
(5.374)
The solution to Eq. (5.374) is to

where f is the thermal conductivity.
The solution of the mathe
satisfy also certain boundary conditions.
matical problem posed by (5.374) can be very laborious and usually
machine
requires
computations; however,
a simple
case
will
be
con
sidered here as an example.

If the absorber is such that diffusion theory is applicable in its interior,
then the flux <(r) satisfies Eq. (5.283), and it may be seen that T(r) must
be of the
form
T(r) = cF^r) +
V2F,(r)
where
C2F2(r)
VJF2(r)
- ^ *(r)
(5.375)
= 0
(5.376)
Fi and F2 are independent solutions of Laplace's equation for the

particular geometry.
Suppose, as a particular example, that a very long cylinder (radius R)
of a dilute uranium alloy is placed in a channel of water which supports a
neutron flux 4>Q. If the water is flowing along the cylinder, and if we
neglect the rise in temperature of the water axially, the boundary condi

i.e.,
tion at the absorber-water
T(R)
If
interface may be taken to be

= To = water temperature
(5.377)
applicability of diffusion theory for the flux inside the

absorber, then from equations (5.360) and (5.375)
we assume the
T(p) = ci +
c2
In
h(xAp)
p
T (
R\ Tl
i*xaK)
41-
Ii(*aR) K0(x\fR)
âHa
umxm
h^R)
Kx{xuR)
(5.378)
where subscript A refers to the absorber and subscript M to the water;

the functions Fi(p) and F2(p) have been chosen appropriately for cylin
drical geometry.
We discard the term In p in (5.378) inasmuch as it is
singular at the axis of the cylinder (i.e., we take c2 = 0). The applica
tion of the boundary condition (5.377) yields, finally,
T(
iW
T
'o
fx,
\(*AR)Kt>{xMR)
\ 1 4I" j^d^di
^
DMxMh{xAR)Kl(xMR)\
("
[L
Io(xAp)
h(xAR)
(5.379)
In this result, we
use
0 = esy1'
where
Sy11
(5.380)
is the macroscopic fission cross section of the material of the
KKACTOR ANALYSIS
262
[CHAP. 5
is the energy release per fission (7.53 X 10~12 calories1).

Of course, the usual experimental situation cannot be represented by any
such simple model; besides the commonly met geometrical and material
absorber and
complications in engineering heat-flow problems, there is the additional

burden of the form of >(r), which may be known only by numerical
calculation [as, for example, in the first-flight transport model for the
A further, extreme, complication may arise

solid cylinder, Eq. (5.329)].
if the neutron properties of the material of the absorber (for example, the
depend on the temperature; then the neutron and heat-flow
problems are interdependent and some sort of iterative technique may
be needed to obtain a solution.
density)

PROBLEMS
S.l Consider a sphere of radius R which has at its center a point source which is
emitting neutrons of one velocity at a rate q0 neutrons per unit time.
a. Find the flux distribution within the sphere, using the boundary condition that
no neutrons return from the surrounding vacuum (evaluate all constants).
6. Show that the flux computed by a vanishes at approximately r = R + 2D.
c. Find the flux distribution within the sphere, using the boundary condition that
vanishes at r = R + 2D.
d. Compare the two expressions obtained from a and c for {r), and show that the
1.
expression of part c is approximately equal to that of part a when D/L and L/R
e. Using D/L = and R/L = 2, compute the net current of neutrons at the surface
of the sphere for cases a and c, and find
(r)
/(e)
Jia)
(r)
f. Compute the absorptions in the extrapolated region of the sphere, using

Show
from part c, and compare this number to the total absorptions in the sphere.
that the ratio of absorptions in the extrapolated region to the total absorption is very
small compared with unity when D/L and L/R <K 1.
If radiation from infinitely removed sources is allowed to pass through a
6.2
sphere of radius R, it may be shown directly that the average chord length of radiation
path in the sphere is given by 8 = %R.
Consider an infinite medium in which there is a spatially constant neutron flux;
introduce a spherical cavity of radius R and use the theorem stated above, along with
the definition of neutron flux (i.e., the flux is the neutron track length per unit volume
per unit time), to show that the first term in the expression for the partial current
in terms of the flux is </4.
6.3
Consider an infinite slab (width 2a) of uniform material described by the onevelocity constants 2o and D.
a. Determine the general expression for the flux distribution throughout the slab
arising from a continuous distribution of plane isotropic sources defined by S(x) m
S0 cos (tz/2<j).
b. The general requirement for criticality of a reactor is that the total number of
Show that the application of
neutrons in the reactor in each generation be constant.
this definition to the slab reactor described above yields the correct statement of the
1 Based
on 197 Mev per fission
(see
Table 1.1).
DIFFUSION THEORY
2G3
criticality equation for this configuration.

It should be noted that this result demon
strates that a flux distribution which is proportional to the source distribution is
equivalent to a system of uniformly distributed fissionable material.
5.4
A point source is emitting thermal neutrons isotropically at a rate qo neutrons
A thermal neutron detector is placed at a distance I from the point
per unit time.
source. With the distance I held fixed, a multiplying spherical shell (inner radius R,,
outer radius Ri and Ri/l
1) is placed around the source so that the source is the
geometric center of the shell. Find the ratio of the counting rates at the detector,
with and without the multiplying shell, under the following assumptions:
1. Diffusion theory holds in the material of the shell; the spherical hole and the
space outside the shell are taken to be void of material.
2. The material of the shell is such that all fissions are caused by thermal neutrons;
fast neutrons from fission become thermal with the same spatial distribution as that
in which they were born as fast neutrons. However, in slowing down to thermal
there are losses by absorption and leakage.
3. With the spherical-sheil geometry, the assembly would not operate in steady
state without the source; however, the composition of the multiplying shell is such that
ffthfc, > 1 (where pu, is the probability that a fast neutron will not escape from the
shell before it reaches thermal).
6.6
A spherical shell of material has inner radius R\ and outer radius Ri.
(There
is a vacuum both inside and outside the spherical shell.)
The material of the shell is
described by neutron cross sections Z0, 2/, and Str (one-velocity model).
o. Find the spatial distribution of the steady-state neutron flux.
6. Find the critical equation.
Assume that the neutron flux satisfies the one-velocity diffusion equation in the
material of the shell and that appropriate boundary conditions are to be applied.
6.6
Consider a sphere of radius R composed of material with the one-velocity
constants Z>, Z, and 2/.
The fissionable material is distributed uniformly throughout
the sphere.
At the center of the sphere is an isotropic source of neutrons of strength q0
neutrons per unit time. Take as the source term S(r) fcx2o*(r)> and U8e the
boundary condition that the flux vanish at the extrapolation radius R.
a. Find the general expression for the flux distribution
in terms of the parameter
B, where

(r)
- (*. -
1)
Note that the flux may have a singularity as r 0 because of the uniqueness of the
center point of the sphere.
b. Determine the expression for
in the following special cases:
(r)
c. Describe the physical situation represented by the values of B3 listed below.

Indicate whether the system would be supercritical, critical, or subcritical on the
basis of the fission sources alone.
Sketch the family of functions given by 4>{r)/(q<s/iirD) in terms of the parameter
d.
B.
In particular, pick out the three special functions described in part
b.
What fraction of all neutrons produced per unit time in a finite, critical, one
velocity, cylindrical, bare reactor escape across the curved surface?
6.7

264
REACTOR
ANALYSIS
[CHAP. 5
A fixed volume V of a homogeneous multiplying material is to be put into the

5.8
form of a right circular cylinder.
o. Find the condition on the height h (in terms of the infinite multiplication con
stant k, the diffusion length L, and the volume V of the material) such that the
cylinder be subcritical.
6. Discuss briefly the physical meaning of results in part a with the aid of a grapli
of k versus the height h.
6.9
Find the fuel distribution and the corresponding flux that give a constant
Assume (1) a bare one-velocity reactor,
power distribution in an infinite-slab reactor.
(2) the extrapolation boundary condition, and (3) that the diffusion coefficient is
unaffected by local variations in fuel concentration.
6.10
The one-velocity neutron parameters 2 and D are constant over an infinite
medium.
Outside a spherical region of radius R, the absorption cross section is made
up solely of radiative-capture cross section; inside the spherical region, the absorption
cross section has the same value as outside but is made up of fission cross section as
well as radiative capture.
For every neutron which is absorbed inside the spherical
region, y2//Z<, neutrons are born from fission; there are no other sources of neutrons.
a. Find the one-velocity neutron flux both for the regions inside and outside the
Give the condition for criticality.
active core.
b. Set up the integral equation satisfied by the neutron flux, using as a kernel the
solution of the diffusion equation for a point source in an infinite medium.
Show that
the solution obtained in part a satisfies this integral equation.
A control rod has been designed for an MTR-type reactor in which the lower
6.11
portion of the rod consists of fuel and the upper portion of high thermal absorber
The two sections of the rod are separated by a gap as shown below.
(nonfuel).
When the control rod is inserted, the gap in the rod will be occupied by water (modera
tor). The problem is to determine whether the presence of this cavity of pure modera
tor in the core will cause excessive peaking of the thermal flux (and therefore increased
Control
rod
Absorber
Note: Reactor core consists of fuel plates immersed in water
Examine this problem on the basis of the follow

deposition) in the vicinity.
ing model: (1) represent the core (moderator plus fuel) by an infinite homogeneous
medium and the gap by a spherical cavity of radius a which contains only moderator;
(2) assume that the one-velocity approximation applies for the thermal flux; and (3)
let the source of thermal neutrons be uniform throughout the core and cavity regions
and take it to be some fixed number S neutrons/time/vol.
a. What are the differential equations and boundary conditions which apply for the
thermal flux?
b. Write down the general solution for the thermal-flux distribution throughout the
entire space.
energy
DIFFUSION THEORY
265
c. Apply the boundary conditions to eliminate one of the arbitrary constants in the
How would you evaluate the others?
general solution for each region.
d. If all the constants are evaluated (not required here), one can show that the flux
has the form (Z moderator < Z core) shown in the sketch.

(a)/
Determine the effect of the cavity by computing the ratio

(allow only one
Note that this ratio is a measure of the increase
arbitrary constant in the final result).
in
at the gap boundaries over the flux computed without the cavity.
6.12
Using the cross sections and assumptions of Prob. 4.8, calculate the critical
radius of a spherical homogeneous reactor for five ratios of moderator to fuel: Nu/Nr
= 100, 500, 1,000, 5,000, and 10,000.
Assume that neutrons are born at E and slow
down, reaching thermal energy at the same space point at which they were born
Calculate also
(i.e., no fast leakage).
a. The nonleakage probability for thermal neutrons pnl6. The critical mass (and plot versus moderator-fuel ratio).
c. Mean lifetime of a neutron in the reactor.
Do not neglect extrapolation distance, work to three-figure accuracy, and take the
moderator density to be 1.65 g/cm3.
6.13
The feasibility of a gaseous reactor can be examined on the basis of a rela
tively simple model. The present problem is to determine the important features
and dimensions of such a system on the basis of some reasonable specifications.
For
this purpose the following simplified model will suffice :
1. The reactor is a bare gaseous sphere of radius Rt>.
2. The reactor is critical and at steady state.
3. Energy sources arise from fission reactions only.
4. The outer boundary of the sphere is at absolute temperature T = T(Ro) m TV
5. The gas mixture is a stagnant system at some fixed pressure p.
6. Energy losses in the gas are due to conduction only; therefore, neglect radiation,
convection, and gravitational forces.
7. One-velocity diffusion equation gives an adequate representation of the neutron
physics.
8. The extrapolation boundary condition is applicable.
9. The diffusion coefficient D is spatially invariant (assume some suitable average
value for the mixture).
10. The coefficient of thermal conduction can be represented by some suitable
average value f.
The problem is to find the flux, temperature, and nuclear-density distributions in
this reactor and its power-generating capacity as a function of the reactor size Re,
the surface temperature To, the pressure p, and the various nuclear and physical
constants of the gas.
REACTOK ANALYSIS
2(56
Part A.
[chap. 5
General Solution
1. Set up the
balance, and the
fuel component
2. Show that
to the form
appropriate equations which describe the neutron-balance, energyequation of state for the gas mixture which consists of a nonfuel and a
List all boundary conditions.
(fissionable material).
the neutron-balance and energy-balance equations can be reduced
Neutron:
VJ0(r) +
= 0
(P5.1)
Energy:
VT(r) +
= 0
(P5.2)
where
'
V'
kD(n +
1)
|w/
- [no
(P5.3)
and {r) is the flux, T(r) the absolute temperature, k Boltzmann's constant, e the
energy release per fission, and n = No/Nr (the ratio of concentration of nonfuel to
fuel). Let

Ko +
jr^r
OX,r
(P5.5)
Here, N(r) denotes the total number of nuclei (of all kinds) in unit volume at r, a. is a
suitable microscopic scattering cross section (may be approximated by nonfuel
components), and a is some constant (say 0.7104).
3. With these definitions,
a. Show that R = fto(l + m)
6. Find the criticality condition
c. Determine the temperature function and show that it may be written (x = r/Ro)
T(x)
T*
+ it)
[(1
+ ,.)'
- *JI
(P5.6)
d. Find also the expressions for the flux <(x) and the nuclear density jV(x), eliminat
ing all arbitrary constants by means of the boundary conditions.
4. Sketch the functions T(x), (x), and N(x).
a. What is the condition on n if the diffusion equation is to yield a reasonable
representation of the neutron physics?
b. Find an approximate expression for the temperature at the center of the sphere
1.
T(0) which is valid when n
5. Critical fuel concentration.
Let D s Xtr/3 l/3ff,.V,,v with A/*.v = p/kTn and
T = 67V Thus we compute a diffusion coefficient in terms of some average tem
If we call
perature of the system 7'v which is defined in terms of TV
2ba.
(P5.7)
"=aW>
then
a. Show that the critical fuel concentration ratio n is
(2
+ M)(,
where v m m//af/\
b. Sketch n as a function of j. for various
$.
- 1)
ft.
267
DIFFUSION THEORY
6. Derive expressions for the total power output of the reactor by two separate
approaches.
Part B.
Computations
7. Assume that a gaseous reactor is a sphere of the type analyzed in Part A, and
Find the critical
let the material be UF6 so that the fluorine acts as the moderator.
enrichment of U*".
a. Derive expressions for a. and <rj,0) on the premise that the US3S enrichment is
negligibly low.
b. Compute these cross sections using the following data:
a.
rjin
F"
U"5
8.2b
3.9
10.0
aa
2.756
0.01
687
"1
0
0
580

8. Compute n, T(0), $, n and the enrichment N13i/N-<3a for the critical system,
Use f = 1,200 ergs/(cm) (sec) (C),
using To = 1000K, ft0 = 10m, and p = lOOatm.
e = 3.2 X 10-4 ergs/fission, a = 0.7, and b = 2.5.
9. Compute the total power output of the reactor.
10. Compute and plot (x), N(x), and T(x).
CHAPTER
FERMI AGE THEORY: THE HOMOGENEOUS

MULTIVELOCITY REACTOR
6.1
The Combined
Slowing-down and Diffusion Equation

a. Assumptions and Limitations.

The two elementary models devel
oped in the preceding chapters allowed us a limited picture of the neutron
population in a nuclear reactor. The first model, which was considered
in Chap. 4, confined attention to the slowing-down process, and all
calculations were based on the infinite-medium concept.

The second
model was developed in Chap. 5, and in this approach the energy depend
ence of the neutron flux was removed by the introduction of the one-
velocity approximation.
The
attention to be
focused on the diffusion process, and the application of the model to the
determination of the spatial distribution of neutrons in a finite system
was demonstrated.
use of this model allowed
In an actual reactor the two phenomena slowing down and diffusion

in fact be separated if a complete description of the neutron
population is to be achieved. As we previously noted, the separation
was made in the preceding work so as to simplify the presentation by
cannot
limiting the number of analytical methods and physical concepts con

sidered at any one time.
It should be recognized, however, that although
each model may give a poor representation of the general reactor problem,
there are numerous special situations in which either the slowing-down
phenomenon or neutron diffusion is the essential feature, and a direct
application with minor adjustments can be made. This was the case
when the one-velocity diffusion model was used to perform a reactor
In this application it was assumed that the system was

highly thermal so that the one-velocity model gave a good description of
the spatial distribution of the thermal flux.
The purpose of the present chapter is to develop a more general model
for describing the neutron population in a reactor, a model which com
bines the principal features of the two already presented.
That this is
indeed a prerequisite for treating the general reactor problem may be
calculation.
268
SEC. 6.1]
FERMI AGE THEORY
269
appreciated from a consideration of the various physical phenomena

In an actual reactor the neutron
which occur in a chain-reacting system.
sources are the fission reactions which take place throughout the fuel
The neutrons produced by fission are distributed
component of the core.
over the entire energy range (see Fig. 4.24) and have an average energy
These fast neutrons from fission are subsequently
in the order of 2 Mev.
slowed down by various scattering collisions with the nuclei of the reactor
This sequence of scattering collisions, which results in a con
medium.
of the neutron energy, is accompanied by a spatial
wandering as the neutron progresses from one scattering center to the
Thus intimately associated with the slowing-down process is a
next.

tinual degradation
spatial movement or "diffusion."

The present diffusion process differs somewhat, however, from the
The previous definition referred to the
previous definition of the word.
spatial wanderings of a particle which suffered little, if any, change in its
It was first used to
speed as it passed from one collision to another.
It will be useful in the work
describe the motion of thermal neutrons.
which follows to extend this concept to include the motion of the fast
By limiting our study to systems in which the slowing
neutrons as well.
down is relatively gradual, we can use the idea of a diffusion process to
describe the motion of fast neutrons in any small interval of energy.
This would be the case in reactor systems which consist primarily of the
For these materials the energy change
heavier nuclear-mass materials.
per collision is relatively small so that a given neutron will suffer many
Note that this extension of
collisions in a specified small-energy band.
the diffusion concept is in accord with the essential feature of the thermal
diffusion process, namely, that the neutron experiences many smallenergy change collisions while in the thermal range.
The combiiHd process of slowing down and diffusion is correctly treated
by means of the Boltzmann equation, which takes into account the angular
distribution of the neutron population as well as its energy and spatial
This more general approach will not be applied in the
dependence.
We will use
present analysis but will be deferred until the next chapter.
instead a less rigorous derivation of the combined slowing-down diffusion
relations which we will later show to be a first approximation to the Boltz
As in the development of the one- velocity model of
mann equation.
Chap. 5, the present model will possess all the pertinent features which
are allowed by the various approximations and constraints imposed on
the physical system, but we will be unable at this time to demonstrate
all the motivations or justifications for establishing these conditions.
Thus we begin our analysis by defining a set of assumptions which will
All the analyses which
form the basis of the neutron-balance relation.
follow, then, as well as each system to which the general equations are
applied, must satisfy the conditions given below:
REACTOR
270
(1)
(2)
ANALYSIS
[CHAP. 6
The medium is isotropic and homogeneous

The macroscopic absorption cross section of the medium is
much less than the macroscopic
scattering
cross section
over large energy-intervals

(3)
The general dimensions of the system are much larger than

characteristic
neutron lengths such as the diffusion length

(4) The regions within the reactor which are to be described
by the combined slowing-down diffusion model are free of
localized neutron sources and sinks
(5) These regions do not contain large voids
(6) Neutron scattering is isotropic in the center-of-mass
tem (C)
sys
(7) The mass number of the moderator is sensibly larger than

unity
Except for statements (6) and (7) we offer no explanation in support of
this particular choice of assumptions.
Statements (1) through (5) will
be justified in Chap. 7 wherein the Boltzmann equation is reduced to the
It is appropriate to remark on the validity and usefulness
present model.
of assumptions (6) and (7) here inasmuch as we will use the continuous
slowing-down model to describe the trajectory of the fast neutrons in
energy space. This model was introduced in Sec. 4.2b in connection with
It will be recalled from
the slowing-down process in an infinite medium.
that application that the actual trajectory in energy space is a discon
tinuous function; however, if the decrements in energy per collision are
small, the slowing-down track may be approximated by means of a con
tinuous function (see Fig. 4.11). It is clear that this would be a good
approximation to the physical system only if the slowing-down medium
Thus- in the present
consists primarily of large nuclear-mass materials.
Furthermore, the require
study we exclude hydrogenous substances.
will also tend to satisfy condition (6) [see Eq. (4.30)].
ment that
It is known from experiment that the scattering in the (C) system is
nearly isotropic for materials with large A, provided that the neutron
Below the ev range, thermal effects
speeds are in the ev to Mev range.
become important; above the Mev range, scatterings tend to be forward.
In the work which follows we will assume that all collisions are isotropic
in (C) and moreover are purely elastic. We will not consider inelastic
Actu
collisions in the present treatment in the interests of simplicity.
diffusion
model
the
combined
slowing-down
ally, those applications of
Such systems
which are most useful involve highly thermal systems.
include relatively small nuclear-mass materials and, except for the struc
tural members within the reactor, do not serve as a significant source of
Al
inelastic scatterings
(see Sec. 4.1a).
FERMI AGE THEORY
271
Neutron Balance in Space and Lethargy.
We begin our study of
SEC. 6.1]
b.
the slowing-down diffusion model by establishing the neutron-balance

For this purpose
condition in terms of the lethargy and space variables.
two
functions:
following
we introduce the
4>(x,u)
track length per unit volume per unit time around

du = r due to neutrons in that volume whose lethargies
lie in du about u
number of neutrons slowing down past lethargy u
per unit time per unit volume around r
Consider a small volume dt of material in a medium supporting a multivelocity neutron flux at steady state, and let us confine attention to those
It is possible to write an
neutrons whose lethargies lie in du about u.
expression for the neutron balance
in the differential
volume dt du in

the four-dimensional space defined

by the coordinates r and u (see Fig.
Sources-
This balance is obtained by

6.1).
for all the processes
accounting
which can add or remove a neutron
from either dx or du. At steady
state, neutron gains into dx du con
leakage
/
1
c
du
+U
q(r, u+du)
Fig. 6.1 Differential volume dt du.

sist of (1) neutrons which slow past
lethargy u, thus entering du from lower lethargies, and (2) neutrons with
lethargies in du about u which appear directly from sources in dx.
are due to (3) neutrons with lethargy in du which are
scattered while in dx and thereby slowed down out of du, (4) neutrons
which are absorbed by nuclei while in dx du, and (5) neutrons of lethargy
in du which escape across the boundaries of dx.
Neutron
losses
By calculating each of the above quantities on the basis of events per

unit time, we may write the balance equation as follows:
q(x,u) dx + S(x,u) dx du = g(r, u.-J-(jfa) dx + 2(w)4>(r,w) dx du
+ ?(r,u) di du
(6.2)
where S(x,u) du = number of neutrons with lethargies in du about

which are born per unit time per unit volume around
from sources
2(x,u) du = number of neutrons with lethargies in du about
which escape per unit time from unit volume around
All the terms in Eq. (6.2) except the last
u
r
u
r
are readily specified by means

of the concepts and functions already developed.
The leakage term 8
causes some difficulty in the present formulation and can be properly
[chap. 6
REACTOR ANALYSIS
272
This compli
identified only through the use of the Boltzmann relation.
cation is similar to that encountered in selecting a suitable leakage term
for the one-velocity model (see Sec. 5.1c). In fact, it is from the onevelocity result that a hint is obtained as to the form to be chosen for the
If the system were such that neutrons

leakage in the multivelocity case.
spent a great deal of time in the lethargy interval du and made many
scattering collisions before being reduced to higher lethargies (lower
energies), there would be some justification for treating all the neutrons
with lethargies in du as a group of diffusing neutrons of a single speed.
Then from results such as Eq. (5.51b) for the one-velocity case, it might
be assumed that the net current of neutrons in du and di around r would
Further, the
proportional to the gradient of the neutron density.
constant of proportionality would be identified as in (5.51c) so that,
finally, we could write for the leakage
be
- D(u) V20(r,u)
8(r,) =
(6.3)
This relation can, in any case, be made as an assumption which is to be

tested by application; however, the approach here indicates that the
(6.3) would be especially applicable to media composed
chiefly of heavy nuclei for which the average gain in lethargy per collision
would be small, and, consequently, the neutrons would make many
collisions in the neighborhood of a given lethargy (i.e., the scattering

assumed form
process in du could be described as a "diffusion").

If we use (6.3) in (6.2), along with the identification
lim
du->0
Then (6.2) may
-D{u)
be
[q{r,u
+ du)
q(i,u)]
q(r,u) du
(6.4)
ou
written
V*>(r,) + Z.() *(r,u) = S(r,u)
- ou g(r,u)
(6.5)*
It is reasonable to assume, under the same conditions, that the relation

ship between the flux and the slowing-down density obtained for the
infinite medium in (4.112) would be a good approximation.
Thus we
= y
write
q(t,u) =
In treating the interval du
{Zt
(u) *(r,u)
rX*
(6.6)*
we are requiring for the

validity of (6.4) that the neutrons make infinitely many collisions in
passing through a finite lethargy interval. In the limit this implies the
"continuous slowing-down" model suggested in Sec. 6.1a; thus, neutrons
take on every value of lethargy in traversing the interval du, since in the
as infinitesimal,
be zero.
In an actual case, of
course, the slowing-down process is not continuous; a given neutron takes
on only a limited number of lethargy values in slowing from fission
For a small (but finite) lethargy interval, it may be
energy to thermal.
limit the average gain per collision must
FERMI AGE THEORY
sec. 6.2]
273
true that most neutrons which slow down through that interval never
take on lethargies in that interval. From this viewpoint it is rather sur
prising that (6.5) gives good results in application. That the result (6.5)
is justifiable to a given approximation by consideration of the energydependent Boltzmann equation indicates that (6.5) is useful because we
are usually dealing with very large numbers of neutrons.
Thus while
the slowing down and diffusion of any one neutron are by jumps, the
behavior of the total neutron population closely approximates that of a
continuous flow (in space and lethargy).
We conclude the present remarks on the combined slowing-down diffu
sion model with the reminder that the set of equations (6.5) and (6.6) are
to be applied
to those situations
which
satisfy
the conditions
(6.1).
boundary conditions,
have been used in situations wherein one or more of these requirements
Although good results can be obtained in some
have been violated.
such instances, usually requiring considerable modification to the original
equations, such applications generally yield erroneous results and must
Equations (6.5) and (6.6), along with appropriate

be used
with caution.
6.2
Fermi Age Theory and Elementary Solutions
a. The Fermi Age Equation.

The purpose of the present section is
to demonstrate the reduction of the combined slowing-down diffusion
equation (6.5) to the Fermi age relation (which is mathematically iden
tical to the standard form of the heat-conduction equation) and to display
the solutions of this equation for several elementary source distributions.
The reduced equation will be given in terms of the spatial coordinates of
the position vector r and the lethargy u.
It will be convenient in carrying
out this derivation to treat simultaneously the analytically equivalent
of the time-dependent one-velocity flux equation which is the

of the steady-state relations (5.51). The time-dependent
equation describes the distribution of neutrons in space and time, whereas
the slowing-down relations give the distribution in space and lethargy.
However, the lethargy and time variables are mathematically
similar
problem
generalization
quantities, each being defined for the real, positive range. Furthermore,
the neutron-balance relations for these two situations are mathematically
Thus the two problems may be easily treated together.

The slowing-down diffusion relations which are to be reduced are given
by (6.5) and (6.6). The corresponding relations for the time-dependent
one-velocity flux distribution are
identical.
-DV^(r,t) +
2a>(r,<) =
4>(i,t)
- |n(r,)
= vn{r,t)
(6.7)
(6.8)
REACTOH ANALYSIS
274
[CHAP. 6
The neutron-absorption cross section 2 and the diffusion coefficient D

The function
are defined for the one-velocity v specified by the problem.
is the corresponding neutron density and is expressed in neutrons
The two differential equations (6.5) and (6.7) may be
per unit volume.
written in terms of a single function by applying the coupling relations
If we omit the source term, the results are
(6.6) and (6.8).
n(r,t)
_,
D(u)
VJg(r,M) +
2(w)
.
.
\Ar q(T,u) =
-DV*n(r,0 + Hjm(r,t)
- q(t,u)
- Jn(r,0
d
(6.9)
(6.10)
We eliminate the linear terms in these equations by introducing integrat

ing factors in the variables u and t, respectively, i.e., by defining two new
functions Q(r,u) and N(i,t) which are related to q(i,u) and n(r,t) by
q{r,u)
Q(i,u) exp

n(i,t) = N(t,t) exp
2al'd'j
(6.12)
density for a system which includes

represented by the cross section 2a and that the exponential
Note that q(r,u) is the slowing-down

absorption
is the resonance-escape probability for the lethargy interval (0,u) accord

ing to the previous definition (4.113); thus Q(t,u) must be the slowingdown density for this system in the special case that 2() = 0. We
observe, furthermore, that n(r,t) is the neutron density for a system with
absorption
and
exp
[- f^HaPdl']
fraction of neutrons which traverse the path vt without being

If 2 = 0, n(r,<) m N(i,t) ; thus we identify the function
N(r,t) as the neutron density for the same system with the absorption
is the
absorbed.
cross section zero.
The substitution of (6.11) and (6.12) into (6.9) and (6.10) yields
H ^(M) - I
=
DvV*N{t,t)
Q(r,)
JV(r,0
These equations reduce to the standard form of the heat-conduction
(6-13)
(6.14)
equa
FERMI AGE THEORY
sec. 6.2]
275
tion by changing from the variable u to t, a function of u, by

(6.15)*
and from
to the function
tj
of
by
(6.16)*
For the transformation u > r we write Q(r,u) > Q(r,r), and for t> ij,
iV(r,<) > iV(r,ij).
(This is a commonly used notational convenience
The application of these
if
misunderstood, can lead to mischief.)
which,
to (6.13) and (6.14) gives the relations
V!Q(r,r)
VW(r,)
transformations
Q(r,r)
(6.17)*
N(t,v)
(6.18)*
which are in the standard form of the heat-conduction equation.

The
these equations (6.17) is called the Fermi age equation; the variable
t is called the Fermi age and will be discussed in greater detail later in this
Each of the above relations is expressed in terms of the spatial

section.
coordinates given by r and a timelike variable t or rj. The solutions to
this class of partial differential equations are well known in the case of
several elementary source distributions.
However, in order to provide a
greater insight into the physical situations involved, we will develop in

detail the solutions to the slowing-down problem for the cases of the plane
and line-source distributions. The corresponding solution to the pointThese methods
source problem is easily obtained by the same methods.
can also be applied to the three elementary solutions to the time-depend
ent problem.
b. Elementary Sources in Infinite Media.
Our purpose here is to
display the solutions to Eqs. (6.17) and (6.18) for the three elementaryIn the
source distributions, namely, the plane, the line, and the point.
of the slowing-down problem, we define the physical system by

specifying the "initial condition" that neutrons of lethargy u = 0 are
released steadily according to some known spatial distribution.
These
high-energy source neutrons subsequently enter a slowing-down process
case
and wander through the system until they are absorbed.

The problem
is to determine the spatial distribution of these neutrons for all u > 0;
that
we should like to know how the neutrons spread out through the
is,
medium as they progress from collision to collision, each time losing

fraction of their energy. The corresponding space-time problem defined
by (6.18) gives the distribution of neutrons which first appear in the

first of
medium from prescribed sources according to some
"initial condition,"
REACTOR ANALYSIS
276
[chap. 6
beginning at say t = 0. This condition gives the number of neutrons

The solution to (6.18) which
released into the medium at that time.
satisfies this initial condition describes how these neutrons are distributed
throughout the medium for all subsequent time.

The calculation to be performed first is the determination of the slowing-down density of neutrons that originate from a plane isotropic source.
We specify that the source emits neutrons of lethargy zero (corresponding
to some high energy) at the rate qo neutrons per unit area per unit time.
This problem is clearly one-dimensional, and, for convenience, we place
the source plane at the origin of the x axis.
Theâppropriate differential
equation is obtained from (6.17).
=
Qfr/T
Û"1
0,
oo
r >
< x <
(6.19)
oo
The conditions which describe the source are
Q(x,t)
slowing-down
t >
Q(x,t)
= 0
(6.20)
follows immediately from the definition of

as
The actual
density in the absorber-free medium.
density
(3)
Note that condition

the slowing-down
Q(-x,r)
dx = g0
obtained from the relations
is
JQ"
q(x,r) = Q{x,t) p(u)
with
r(u)
(6.21)
as given by (6.11) and (6.15).
0(x,s) = {Q(x,r)}
JQ"
The differential equation (6.19) may be reduced to more elementary

to the Fermi age variable t.
form by applying a Laplace transform
define
We
err-Q(x,r) dr
(6.22)
= lim Q(x,t) =
(6.23)
r->0
r>0
6(x,s)
(6.24)
e(jt,8)
<
Thus the Laplace transform of (6.19)
is
Q(x,
0)
is
The initial conditions
real and positive.

where the transform variable
for
x
and
that
all
^
require
(2)
(1)
@(x,s) =
(a) exp
(-x
y/s) + B(s) exp
(x
which has the general solution

(3)
go
neutrons per unit area per unit time appear at x = 0,

= 0
(2) No sources other than those at x
(1)
y/s)
(6.25)
FERMI AGE THEORY
sec. 6.2]
We have left the arbitrary constants A and B
277
as general functions
of the
transform variable s inasmuch as s appears as a parameter in (6.24).

These constants are evaluated by the application of condition (3), the
Laplace transform of which is
/
Jo
e(*,) dx
= 9o
^
*8
(-x,a)
= (x,s)
(6.26)
The substitution of (6.25) into this integral requires that

A(s)
J0
exp ( x
y/~s)
When x is positive, B(s) must
dx + B(s)
be taken
Jo
exp (x
V s) dx
~
4$
identically zero in order that the
Q(x, u)

O<u0
Fia.
6.2
81owing-down density for a plane source.
With
density remain finite (for all positive values of x).
=
y/s,
this choice the above relation yields the result A (a)
and
qo/2
(6.25) may be written
slowing-down
e(
goexp(-xv^)
2
Vs
The Laplace inversion -1{0(.t,s)
yields the solution

) of this transform
to the differential equation (6.19) which satisfies the conditions (6.20).
For both positive and negative values of x the result is
Q(x,t)
=-H@(x,s)\
Recalling (6.21), the slowing-down
9-^-
density with absorption
(6.27)
is obtained
from (6.27) and is

q0p(u) e-x,'iT^
q(x,u)
In Fig.
6.2, q(x,u) is plotted as a
helpful to remember that t

increasing function of u.
is
[4rr(u)]
(6.28)*
function of x for several values of u. It

by (6.15) is a monotonically
as defined
REACTOR ANALYSIS
278
[CHAP. 6
We observe from this figure that q(x,u) has two important features.
First, we note that for any given value of lethargy the slo wing-down
density is greatest at the source plane x = 0; and second, the effect of
the resonance-escape probability p(u) is always to lower the magnitude
of q(x,u) for given x. It is clear from the figure that the high-energy
neutrons which have suffered, on the average, only a few collisions are
concentrated in the vicinity of the source. This effect is in keeping with
the continuous slowing-down model which allows the neutron a degrada
tion of energy that is continuous in time as it wanders through the
medium.1 It is to be expected, then, that the neutrons of lowest energy
Slowing-down
6.3
density at a given space point as a function of lethargy.
will be found farthest from the source (see, for example, the curve u = ut
in Fig. 6.2) and will have distributed themselves more uniformly through
out the medium.
However, regardless of the neutron energy in question,
is
(u
0,
|x|
the greatest density of neutrons of any given energy is at the source

proper.
At some space
Consider also q{x,u) as a function of the lethargy.
point
the slowing-down density will have the form shown in Fig.
>
The slowing-down density, and therefore also the flux, of high6.3.
energy neutrons
near zero)
low because most neutrons which have
is
migrated as far as x will have also suffered several scattering collisions

The density of lowand thereby had their kinetic energies reduced.
also relatively low because of two factors: (1) neu
energy neutrons at x
trons of low energy have diffused far from the source and are spread out
is
low,
over such a large volume that their density in any given locality
and (2) the presence of neutron-absorbing materials causes the continual
removal of neutrons from the slowing-down stream and progressively
fewer neutrons are available in the lower-energy regions.
Let us examine now the analogous problem of the distribution of onevelocity neutrons in the space-time system. Consider, then, an isotropic
v,
is
is
The initial condition

plane source of neutrons in an infinite medium.
released from the source
that a burst of q0 neutrons per unit area
= 0.
These neu
(placed for convenience at the origin) at time, say,
and they retain this speed for all subsequent time;
trons have speed
1

Fig.
Actually, one also expects this to happen in the discontinuous model.
FEKMI AGE THEORY
no energy loss results from collision.

The differential equation
which describes the distribution of the source neutrons in space and time
obtained from (6.18); thus,
oo
<
be satisfied by N(x,rj)
<
are
=
(6.30)
4*
it
Again we find
convenient to specify the source strength in terms of
function denned for an absorber-free medium, N(x,ji). This function
related to the actual neutron density through Eq. (6.12), namely,
n(x,ri) = N(x,ri)e-''"
is a
nm dx
/;
(3)
go
0,
x
=
=
neutrons per unit area at
(2) No sources in the medium except at x =
(1)
(6.29)
oo
The conditions which must
1>>
N(x,n)
rj
dr\
0,
N(x,r,)
dx2
is
that
279
is,
sec. 6.2]
(6.31)
(6-32)*
(4*L>*
as a parameter.
At small values of
n(x,t)
would
show
of time
I,
The x dependence of this function has form similar to q(x,u) shown in

Fig. 6.2. In the present case the curves would be drawn for fixed values
t
mum neutron density occurs at
x
is
0.
sharply peaked form much like the shape of q(x,u0). This curve would
represent the spatial distribution of neutrons soon after their release from
=
the source plane at
As increases, the density function n(x,t)
flattens much like the curves for ui, w2, etc., of Fig. 6.2, indicating that as
time progresses the neutrons wander farther from the source plane and
tend to spread more evenly throughout the medium.
One can also sketch
from Eq. (6.32) a density-time function for given x analogous to the curve
of Fig. 6.3.
Such a curve would show how the density at
specified sta
tion would vary as the initial neutron burst passes by. In this case the
interpretation of the curve would be as follows: At short times after the
burst, the neutrons have not yet had time to reach point x, and therefore
the density would be low; very long after the burst, the neutron density
has fallen everywhere because of spreading out and absorption.
Thus
from the viewpoint of an observer at x, a pulse of neutrons passes some
time after the initial burst at the source.
The time at which the maxi
x

is
= r)(t) = Dvt.
where x2 = 2/Z) and
The solution to (6.29) which
satisfies conditions (6.30) can be obtained by the procedure used for the
The result
lethargy-space problem.
easily shown to be
(6.33)
REACTOR ANALYSIS
280
[CHAP. 6
The calculations of the neutron distributions in both the space-lethargy

and space-time problems for the line- and point-source distributions are
However, a
quite similar and will not be considered here in any detail.
features
of
these
two
of
the
essential
differ
from
problems
markedly
few
the preceding ones, and in order to demonstrate
them more clearly, we
will outline briefly the analysis of the lethargy problem for a line source.
This system is axially symmetric, and the appropriate differential equa
tion is
1
d_
St
P dp
Q(p,r)
> 0, r > 0
p denotes the radial distance from the line source.

Q(p,t) must satisfy the conditions
where
(6.34)
The function
neutrons per unit time per unit length of source

line released at p = 0, r = 0
= 0, r = 0, i.e., for
(2) No sources other than at p
= 0
P > 0, Q(P, +0)
(1)
go
fv
Q(T,r)
dl
S[?^,vUr)U-c)
= qo
The volume integral specified in (3) is over a cylindrical region of unit

height and infinite radial extent.
Equation (6.34) may be reduced to the
modified Bessel equation of order zero by the application of the Laplace
transform
on the age variable r.

of (6.34) is
If
0 is the transform of Q,
as before
then the transform
Hp
p 1
pFpe(p's)
S@M
= 0
and has the solution
S(p,s)
A(s)
7o(p
Vs) + B(s)
Ko(P
Vs)
The coefficients A(s) and B(s) are obtained from condition (3).
of the second integral1
is
these integrals
1/s.
+ B(s)
j*
Z(p V~s)p dp
is
A(s)
The first of
The
relation is found to be
jT"
appropriate
Ko(P
Vs)
therefore A(s) = 0. The value

Thus from (6.36), B(s) = go/2ir, and
divergent;
See, for example, G. N. Watson, "Theory of Bessel Functions,"

Cambridge University Press, London, 1952.
1

(3)
(6.35)
2d ed., p. 388,
FERMI AGE THEORY
sec. 6.2]
281
The Laplace inversion yields the function Q(p,r) from which we obtain
qM,^p
(6.37).
It
is easily shown that in the case of the point source at r = 0, and of

strength q0 neutrons per unit time, the solution is given by
The solution for the corresponding space-time problems may be

obtained in the same way.
These results are, for a line source (go neu
trons per unit length at t = 0),

and for a point source
(q0
neutrons at time
= 0),
c. Kernel Method and Distributed Sources.

It will be convenient to
summarize here the general expressions for the neutron density and the
slowing-down density due to point sources. These relations will be useful
in treating the problems encountered in reactor calculations which have
more general source distributions.

From Eq. (6.38) we obtain, for an
infinite medium, the slowing-down density, past lethargy u at r, that
results from a unit point source of neutrons of lethargy ua at r0, (wo < ),
9(r'Wo)
(6-41)
[L(u;u0W
where the age t(u;u0), in accordance with (6.15), is defined by

r(;o)
f" D(u') du'

/
r(u;0)
- r(.j0)
= r(u)
- r(u0)
(6.42)
In the
same way, we generalize the one-velocity time-dependent problem

by specifying the neutron density at r and t in an infinite medium due to
the release of one neutron at time
n(r,^.A)=-
tQ
from the point
[4rDv{t
In terms of the position vectors r and
r0,
(6.43)
these functions depend only on

r0|, the distance of the field point r from the source
|r
point r0, which replaces the length r in the right-hand members of Eqs.
(6.38) and (6.40), i.e., the distance of the field point r from the source
the length
r0,
282
REACTOR ANALYSIS
[CHAP. 6
r0, u0
.
S(r0)Mo)
p(m;mo) e-lr-riv^(w)
=
yr,
IA
[4ht(u;w0)]1
(6.44)
However, all neutrons which appear from sources at leth

argies less than u can contribute to the slowing-down density past u. The
obtained by integrating the
total contribution of all these sources
expression (6.44) over the lethargy range (0,w) and over all space con
taining sources. For convenience we select an origin for the lethargy
scale such that no neutrons appear in the system with w < 0. If S(t0,u0)
would be so
were, for example, the fission sources in reactor, then u =
the
fission
would
show
neutrons of
negligibly
that
few
spectrum
selected
Wo
< .
is
for
greater than the energy corresponding to lethargy

general notation, the expression for q(r,u) may be written
energy
In
du
S(t0,u0) p(u;u0) exp
Jr
T\
5(r,u^0)Mo)
jf
/[
|fl"rfw0
jy
g(r,u) =
zero.
du
[^(ujwo)]1
a
The application of this result to

specific calculation will be demon
If we
strated for the case of a line source of neutrons of lethargy zero.
assume the source strength to be 30 neutrons per unit time per unit
length of line, then the final expression for q(j,u) should agree, of course,
with the result (6.37). The appropriate statement of the source distribu
tion for this case
is

'
*g(r,w^0,Mo)
neutrons appearing at
is
is,
point 0. This simplification arises, of course, from the assumptions of an

isotropic source and a homogeneous medium.
The general expressions for the slowing-down density (6.41) and the
neutron density (6.43) due to a point source may be utilized to derive
expressions for the corresponding functions in systems with distributed
sources.
Both of these relations represent solutions to linear partial
differential equations; thus the principle of superposition is applicable to
situations which involve either. A procedure which may be adopted for
this purpose will be demonstrated in detail for the lethargy-space problem
The technique presented below is readily applicable to the oneonly.
velocity time-dependent problems.
Equation (6.41) gives the slowing-down density at r past u due to a
point source of unit strength releasing neutrons at ro with lethargy u0.
We now suppose that <S(r0,u0) specifies a neutron source that is dis
tributed both in space and in lethargy, that
the source emits S(to,u)
neutrons per unit time per unit volume per unit lethargy at ro with
lethargy u0. By (6.41), the slowing-down density, at past u, due to the
S(t,u)
g0
(i) S(y) 8(u)
(6.46)
FERMI AGE THEORY
SEC. 6.2]
283
S(i,u) dx dy dz du
where we have selected the z axis as the source line.
is the number of neutrons which appear in the volume element dx dy dz
per unit time and which have lethargies in du about u. The number of
g0
dz
J"^
range
J"^
neutrons emitted by a unit length of source line is easily computed by

integrating the expression (6.46) over the appropriate space and lethargy
dy
dx
i(x)
6{y) 5(u) du
=
neutrons/time/length
q0
0;
is
00
5(u) du =
5(u) du =
/ t
/CO
).
js
formally written for the interval

Note that the lethargy integration
In point of fact, the variable u defined for u >
thus we
( oc oc
have implied that
The proposed calculation for the slowing-down density, in an infinite

medium, due to a uniform line source of zero lethargy neutrons may be
computed from (6.45), using the source (6.46):
t/o)2
(z
x0)s
p(w;wo)e-r,/4r
du0
[4ht(m;m0)]?
/_".
(y
(x
rj
where
(u0) 5(x0) 8(yo)
dx0
dy0
dz0
dy
The lethargy
Zo)2.
integra
x0
and
yo
integrations
dZo
&{Xo)
8ê~",l,rW
dx
reduce this relation to

aZ
The
[$$1
/_".
9(x,J/'z'M) =
/_".
tion yields
[4irr(u)]l
and
Zo
is
The
y1.
it
p1
= x1 +
in the form of the error integral,
integration may be written

easily shown that the final
expression for g(r,u) reduces to the expression (6.37).

The Slowing-down Length.
We conclude this section with some
remarks on the physical significance of the Fermi age.
The age was pre
d.
viously defined by means of the differential equation (6.15).

form this variable
given by
In integral
is
**>
/*
(6.47)*
\U
seen
D(u') du'
~kwv
is
&t
I,
JO
commonly
J0
An alternate expression which
is
t(m;0) = r(u) =
,
,

q(x,y,z,u)
iB .Cx
(6-48)
REACTOR ANALYSIS
284
[chap. 6
instead of 2, in (6.47) results from the particular

choice of the relationship between the slowing-down density and the flux
It should be remembered that this relationship is justi
made in (6.6).
one in which 2< ~ 2,
fiable only in a medium of light absorption, that
The presence of
is,
2<
in which the
t give nearly
the same result; however, for consistency, the definition (6.47) will be
used here and in the remainder of this work.
is
Thus in most cases

[see (4.112) and related discussion].
expected to be useful, the two choices of
present theory
is
is
It possible to draw some conclusions about the relationship between

the age and the slowing-down process by inquiring into the average dis
neutron in slowing down.
The calculation which
tance traveled by
follows
analogous to the one presented in Sec. 5.5a in connection with
thermal neutron from the point at
the average distance traveled by
is
it
it
is
it
enters the thermal range to the point at which

absorbed.
to compute the average distance traveled by
Here the problem
neu
slows through
tron while
given lethargy range.
Consider an infinite
medium in which we have a point source of unit strength which emits
which
" GWe"r"4r<U)
|V(u)]t
<6-49)
Let us compute the average value of the radial distance squared

which
neutrons
from this source acquire lethargy
r2
q{r'u)
at
u by means of the
r2/(r;w) dr
neutron
(6.50)
acquires lethargy
u in dr
where f(r;u) dr = probability that

about
fQ"
r*{u)
relation
This probability
obtained by taking the ratio of the number of neutrons
slowing past u in the spherical shell dr to the total number slowing past u
Thus
throughout the entire medium.
is
^,u)^
dr
(6
JQ
Kr;u) dr
5(r,M)4xr2 dr
51)
The substitution of this relation along with (6.49) into (6.50) yields the
result
=
(6.52)*
r(tt)
is
a
The linear dependence of the average r2(w) on t(m)

general result.
determined by the source distribution.
The constant of proportionality
Thus, in general, the square root of the Fermi age
a measure of the
is
is

0)
neutrons of zero lethargy.

The slowing-down density due to the neu
trons released by this source (arbitrarily placed at =
was found to be
slowing-down length of a neutron.
FERMI AGE THEORY
sec. 6.2]
285
It is seen from Eq. (6.48) that roughly the square of the scattering
mean free path appears in the numerator of the integral for the age and
This leads us to the gross observation
appears in the denominator.
that a large age, and therefore a large slowing-down length, is associated
a large scattering mean free path and a large nuclear mass.
A large
scattering mean free path clearly implies long free flights between scatter
ing collisions; thus for a given energy loss per collision, a neutron in a
medium of large X, need travel a greater distance, on the average, to slow
with
Table
6.1
Age
of Fission Neutrons in Various Moderators"
Material
HsO*
Age to indium
resonance, cm!
Estimated age
from indium
to thermal, cm'
26.7
....
....
....
DsO (0.2% H20)

(4.7% HsO)
(8.2% H.O)

Be (1.85 g/cm)
BeO(3.0 g/cm3)
C(l .60 g/cm)
80.
17.2
Age to
thermal,
cm'.
27.7
109.
93.
78.
97.2
105 10
311 3
53
364
See also U.S. Atomic Energy Commission, "Reactor Handbook: Physics,"

p. 485,
6 This figure is reported by T. V. Blosser and D. K.
Trubey, "An Investigation
of the Slowing Down of Fission Neutrons in Water," Neutron Physics Division Annual
Progress Report for Period Ending Sept. 1, 1959, Oak Ridge National Laboratory,
ORNL-2842, pp. 109-112, Nov. 9, 1959; a summary of the most recent measurements
is given in U.S. Atomic Energy Commission, Power Reactor Technology, 3 (4), 31-33
(1959).
"James W. Wade, "Neutron Age in Mixtures of Light and Heavy Water,"

du Pont de Nemours & Co., DP-163, June, 1956.
.,
through
a given energy interval than it must

Consequently, the neutrons slowing down in
to spread out more than those in the latter,
Likewise, if we fix the scattering
larger age.
E. I.
in a medium with a small X..

the former system will tend
hence, a larger r2(w) and a

cross section, then a neutron
slowing down in a medium with a large nuclear mass (small ) must make
many more collisions (travel farther) to slow to a given lethargy than it
would have to make in a medium of small nuclear mass.
Thus any effect
which
tends to lengthen
Fermi
age.
the slowing
down process will yield a larger
These rather elementary conclusions have considerable significance fir'

the selection of suitable moderators for a given type of reactor.
Presum
ing that a moderator is desired (i.e., the chain reaction is dependent on
the thermal, or low-energy, properties of the fuel), the choice of materials'
For
may in some instances be limited by over-all size considerations.
example, if for some reason a small thermal reactor is desired, it is obvious
BEACTOK ANALYSIS
286
[CHAP. 6
from the results obtained above that the choice of moderators is limited
to those materials which have the smaller values of the Fermi age (from
fission energy to thermal).
The intention here, of course, is to slow down
the fission neutrons in as small a volume as practical ; this can be accom
plished only if the characteristic dimension of the moderator is much
larger than the slowing-down
length (i.e., if \/r <K moderator size).
Table 6.1 lists the Fermi age (to thermal1) of some materials of interest
to reactor technology.
6.3
The Unrefiected
Reactor in Age Theory
Monoenergetic Fission Neutrons. The general results obtained in

the two preceding sections may be extended to systems of practical impor
In particular, we examine here the application of the combined
tance.
slowing-down and diffusion equations to the analysis of systems which
include fission sources.
We begin the analysis by developing the appro
priate relations for a reactor in which the neutrons produced by fission
appear at some fixed energy E0
Elb. As in previous calculations, E0
may be selected to correspond to some average energy computed from
the fission spectrum.
The techniques used in this calculation will then
be applied to the analysis of a reactor in which the fission neutrons appear
It is clear that the first model based
according to some spectrum j().
on the monoenergetic source is a special case of the second if j(u) is
However, both calculations
defined in terms of the delta function 5(m).
will be performed in detail so as better to demonstrate the methods used.
The simple reactor model we propose for these analyses may be defined
in terms of the following assumptions:
a.

(1)
(2)
The reactor is bare, finite, and homogeneous

\
The medium is multiplying; neutrons appear only from I
fission sources
(3) The nuclei of the medium are at rest
(4) The slowing-down process terminates at thermal energy
(5) Neutrons in the thermal range can be treated as a single
group
/
>
(6.53)
As mentioned, our first model is based on the assumption of mono

The use of a monoenergetic spectrum for the
fission neutrons as a first model seems justified in view of the fact that the
Thus a
actual spectrum is a sharply peaked function (see Fig. 4.25).
delta function placed in the vicinity of the peak will yield a reasonable
description of the fission spectrum, especially in regards to its influence
energetic fission sources.
"Age
to
thermal" usually implies
thermal energy.
the age of fission neutrons which have slowed to
FERMI AGE THEORY
sec. 6.3]
287
on the neutron population at the higher end of the lethargy scale, i.e.,
the thermal group.
Practice has shown that this model is entirely ade
quate for many bare-reactor calculations.

The combined slowing-down diffusion equations
for the study of the multiplying medium are
-D(u)
V**(r,u) + S.(u) *(r,u)
S(r,u)
- A q(j,u)
0 < u <
g(r,u) = 2,(m)

- AbVV.h(r)
(t,u)
which will be used
Mth
(a)
0 < u < uth
(6)
+ Z**,(r)
= q(t,uth)
(6.54)*
(c)
Note that the coupling equation which involves the total cross section has
been selected. Experience shows that this relation gives satisfactory
results for the systems of interest here.
The last of these equations
the
of
in the thermal range
describes
distribution
the
neutrons
spatial
(c)
by means of the one-velocity model. It is clear that the source term1 in
this equation is indeed the rate at which neutrons slow down into the
thermal range from above (direct contributions to the thermal group
from the fission process are negligibly few). For the case of the monoenergetic fission neutrons, the source term in the slowing-down
may be written in the following form:
S(i,u) du
=
|outh
2,(u')
4>{r,u')
du' +
^^(r)]
() du
relation
(6.55)
This relation gives the number of neutrons which appear per unit volume
and time around r with lethargies in du about u.
Note that we have
used the delta function S(u) to represent the fission spectrum; the origin
u = 0 may be selected so that it corresponds to the peak of the j(u)
curve.
The factor in the bracket gives the spatial distribution of fission
neutrons; the first term of this expression represents the number of neu
trons produced by fast fissions, the second term, the number produced by
thermal fissions.
The delta function &(u) gives the lethargy dependence
of S(t,u) and states that neutrons which appear from fission enter the
system only at u = 0. Note that the above description assumes that
the space and lethargy distributions of the source neutrons are independ
ent ; thus all fissions in this reactor model have the same energy spectrum,
regardless of where they occur in the system.
The distribution of fissions in the reactor as a function of lethargy, at a
given space point, is given by the function /(r,w), see illustrative sketch,
Fig. 6.4. Thus/(r,w) du gives the number of fissions, per unit volume and
time at r, due to neutrons with lethargy in du about u. Evidently, the
integral of this function over the entire lethargy range gives the total
1 Cf. Sec. 5.4g,
especially Eq. (5.210).
288
REACTOR
ANALYSIS
[chap. 6
number of fissions at r due to neutrons of all lethargies.

Note that the
fissions at r are represented by a "spike" 2'/h,i,(r) 5(u u,h)
thermal
at
ttth.1
the
Clearly,
times the area under the curve in Fig. 6.4, including

utXl, is the bracketed factor in (6.55).
"area" under the "spike" at
The purpose of the present section is to solve the differential equations

For the first case we
(6.54) for two cases of the source function S(r,u).
use the monoenergetic fission source given by (6.55) and look for the solu
tion to (6.54) which satisfies certain boundary conditions.
In a system
of finite extent these conditions

medium.
An appropriate
Fig.
6.4
f(t,u),
are specified on the outer surface of the

was discussed in Sec. 5. le
set of such conditions
the density of fissions
as a function of lethargy.
for the one-velocity model, and suitable approximations for reactor

application were presented in Eqs. (5.72). We will adopt this set for use
in the combined slowing-down diffusion model. In this application the
approximate statement of the outer boundary condition in terms of the
extrapolation length must be satisfied for every value of the lethargy.
Thus we require that
(T,u) =0
(?)
0 < u <
= 0
Mth
(a)
(6)
(6.56)
where f denotes the position vector to the extrapolated boundary of the

reactor.
The extrapolation length, and therefore the value of f, how
ever, is a function of the transport mean free path [see (5.71)] which
Thus properly f is a
function. However, we will assume here that the
lethargy-dependent
extrapolation length is the same for neutrons of all lethargies;2 in so
depends on the neutron
1 The
speed, i.e., lethargy.
delta function t(u uth) is unrelated to the S(u) in Eq. (6.55).

The usual procedure is to select one value which represents well the bulk of the
neutron population; e.g., in a highly thermal reactor one might use the extrapolation
length based on the transport cross section of thermal neutrons.
2
FERMI AGE THEORY
SEC. 6.3]
289
doing, we separate the space and energy dependence of our boundaryIt should be recognized that this is a gross
condition statements.
assumption and can be justified only if the predictions of the resulting
The motivation in this choice is to

model agree well with experiment.
model
to
the
reactor physics which retains the
a
describe
simple
develop
principal features of the physical phenomena and yet is easily applied.
f.
is
is
In the present case of a monoenergetic neutron source,

4>{r,u).
the function S{r,u) in Eq. (6.54a)
properly represented by a delta func
tion in lethargy.
We will not use this formulation here but will instead
apply the requirements of the source as an initial condition in the variable
is
fast flux
u when developing the general solution

statement of (6.54a)
then
for q(r,u).
The appropriate
which applies for all points
o(r,u)
inside the reactor.
< u
<
p(u) q(r,u) =
is
-a() VJ9(r,) +
wth
(6.57)
We have used here the
abbreviations
As indicated above, we look for solution to (6.57) which can be written

as the product of
space-dependent and
lethargy-dependent function.
Further, we will assume that this space-dependent function involved in]
the fast flux describes also the spatial distribution of the thermal flux.
We will determine later the consequences of this assumption and examine
a

is
is
it
is
if
is,
The analytical model we use in this formulation is based on the separa

That
we look for solutions to Eqs. (6.54) which can
tion of variables.
space-dependent function and an energybe expressed as products of
dependent function. This separation of variables can be achieved, how
the boundary conditions are stated in
similar manner;
ever, only
particular
the
choice
made
above.
The
resulting
model, although
hence,
very useful for many reactor applications.
Serious
simple in form,
difficulty
encountered in applying
to actual systems only when the
transport cross section (and therefore the extrapolation length)
highly energy-dependent function. This proves to be the case for hydrog
Under these circumstances, the selection
enous media (see Fig. 4.31).
of
single value of the extrapolation length for the entire lethargy range
of interest may yield large errors in the leakage estimates of neutrons
whose lethargies differ considerably from the value corresponding to the
Even in these cases, however, this approximation
lethargy average
often used to facilitate the calculation.
It convenient in developing the solution to Eqs. (6.54) to rewrite
equation (a) in terms of the slowing-down density ?(r,u) in place of the
REACTOR ANALYSIS
290
it may impose on the system.
any conditions
- F (r) Q(u)
- F(r)*,h
g(r,u)
*th(r)
[CHAP. 6
Thus we select the relations

(6.59)
(6.60)
where the fast flux is given by
- F(r) *()
+(r,u)
with *(u)
(6.61)
The substitution of (6.59) into (6.57) yields
F(r)
()
() Q()
where B! is an arbitrary constant to be determined by other requirements.

(For example, the constant B has the form B = t/R for a sphere.)
From this expression
we obtain the
differential equations
=0
V2F(r) + B2F(r)
r inside the reactor
=0
(6.63)
0 
The solutions to the stationary-wave equation (6.63) which satisfy the

extrapolation boundary condition have been discussed in Chap. 5 in con
In this application, we have from (6.56a)
siderable detail.
The functions
and B2 are determined by the reactor geometry; solutions

of (6.63) subject to (6.65) for the three basic shapes have been summa
rized in Table 5.3.
is
from which we immediately
[-5s
exp
(')
/; Jo"
exp
aV]
r) *.<]
joU
Q() = Q(0) exp

Q'(ti) + [a(u)B2 + 0(tt)] Q(m)
[a(u')B* + fi(u')\ du'}
identify the exponentials
e-fl*
. pM .
Thus (6.66) reduces to
(6.66)
[see (6.47)]
(6.67)
iUty
JXPrgyTobf
(6.68)
tfJJ
Q()
Q(0) p(u)e-*M
(6.69)
The thermal-flux equation (6.54c) may now be written in terms of

We have, with the aid of (6.59) and (6.60),
these results.
[- AhV2F(r)
+ 2?F(t)]
F(r)
Q(,h)
sec. 6.3]
If
FERMI AGE THEORY
291
we rearrange and use also (6.69), then
V'F (r) +
'
l)p*rf%ta'
mPn*:
F(t) = 0
(6.70)
= p(w,h) and r,h = r(wth).

However, we have required that
the thermal flux and the fast flux have the same spatial form; thus F(t)
from either the fast- or thermal-flux equations [(6.63) and (6.70)] must
where
p,h
This requirement can be met only

satisfy the same differential equation.
if the coefficient of the F(r) term in (6.70) is identically B2. If this condi
tion is imposed, it follows that ^th is given by
+*-xf+BW
>
(6-71)
All the functions which appear in Eqs. (6.54) are now completely identi
fied.
These results are summarized
below:

q(T,u) = Q(0) p(M)e-B"(-f (r)
M _
(a)
Q(0)pt<rBW(r)
where F(r) is defined by the relations (6.63) and (6.65).

Note that each
of the results (6.72) involves a constant Q(0) which is necessarily estab
lished by the initial conditions in the reactor prior to the steady state
described by (6.54).
The volume integral over the reactor of the above functions yields the
relations
q(r,u)
di
Q(0)
p(u)e-B*
f F(t) di
(a)
The physical interpretation of these quantities is obvious; for example,

the volume integral of q(i,u) gives the total number of neutrons slowing
This number corresponds
past lethargy u throughout the entire reactor.
to the function q(u) obtained for an infinite system [cf. Eq. (4.151)]. In
fact, for the infinite medium B* > 0, and (6.73a) yields the appropriate
limiting form of the lethargy dependence. It should be noted that in
each of the above relations the integral
dr is some constant which

JR F(r)
must be computed from the known geometry of the system.
REACTOR ANALYSIS
292
[CHAP. 6
The critical equation for the bare homogeneous reactor with monoenergetic fission neutrons may be developed from the solutions (6.72).
By the usual definition of criticality, we require that the number of neu
trons introduced into the system be equal to the number produced by
Thus the source function (6.55)
fissions in the subsequent generation.
The lethargy integral
may be used to establish the criticality condition.
of (6.55) yields
fQUlb
S(t,u) du
[/oUth
2,(u) *(r,u) du + Z?
*tt(r)]
This expression gives the total number of neutrons of all lethargies intro
duced at point r in the reactor per unit volume per unit time; but, this is
precisely the number slowing past lethargy zero because, by assumption,
neutrons from fission appear only at lethargy zero.
g(r,0) =
F(r)
Q(0)
S(r,) du
Thus
2,(u) *() du +
2}V,h] F(r)
This expression may

where we have used the relations (6.59) and (6.61).
be further simplified if we use Q(m) in place of ^(u), and the definition
The function (m) is related to Q(u) by (6.61) and
y(u) = 2/(m)/Z,(w).
= "
[Jo
7(M) P(M)
e_BVCU) dU
The substitution of these rela

the solution for Q(u) is given by (6.69).
tions into (6.74), along with (6.71) for ^lh, yields
(6J5)*
+ DB*\
This relation
where the common factor F(r) Q(0) has been canceled out.
the critical equation for the bare reactor with monoenergetic fission
is
Utb
J-
it
2,() *(r,) du +
dr
given by
jR f
is
= J
to write this relation in an alternate

In the present problem
seen
the fast effect.
is
at times, more convenient
form which involves

that the fast effect
It
is,
neutrons.
2}" *,h
(r) dr
*th (r) dr
Zf
(6.76)
= /0U'h
y{u)
written
du
p(j<)e-"r du + X? Pth<r***
(2? + Z^B2)"1
Pth<r***
+ X?
2>h
pue-""* (2? +
(2?
Dth B2)-1
use the results (6.72), (6.76) may be
is
where the integrations are to be performed over the volume of the reactor.
for
Note that this expression for
finite system and therefore differs
If we
from the definition developed for the infinite medium in Sec. 4.8b.

(6.74)
Z^B2)"1
(6"?7)*
FERMI AGE THEORY
sec. 6.3]
293
Note that this definition of the fast effect differs markedly from the cor
The difference is due
responding result (4.298) for the infinite medium.
to the additional factors in (6.77) which involve the size parameter Bi.
This result, however, is entirely consistent with (4.298) since in the limit
for infinite media B2 0, and the two expressions become identical.
If the relation (6.77) is used in (6.75), we obtain
K
2ih(l+L252)
0)
We recognize the factor (1 + L'B1)-1 as the nonwhere L1 = Dth/2^.

leakage probability for the neutrons diffusing in the thermal range [see
Eq. (5.217)]. The exponential we identify as
c -B>rth
Qth
This identification will
nonleakage probability during slowing

(jown from zero lethargy to Mth
pNL the nonleakage
neutrons, then (6.78) may be abbreviated

1
we use
the
as
interpretation.
into
the
reactor, then
lethargy zero are introduced
This relation may
If
6.3e.
__.
probability of the thermal
= Tje/p,hfflhPNL
be given a physical
(6.80)*
If
n0
neutrons of
neutrons reach u = u,h (i.e., do not escape from the reactor

as fast neutrons or are absorbed during slowing down).
2. Of these rioPtMM thermal neutrons, noPthffthPNL are absorbed by vari
1.
ftopthffth
ous nuclei (i.e., are not lost by leakage as thermal neutrons).

3. noPthffthPni/ neutrons are absorbed in fuel material.
4. WoPth^thPNL/i; neutrons are produced by these absorptions in the fuel;
thus, these neutrons are due to thermal fissions.
5. noPthffthPNL/)je neutrons are produced by both fast and thermal
fissions.
6. For criticality,
the resultant number of neutrons produced by
these fissions must equal the number introduced into the system; so
Eq. (6.80) results.

to note that (6.80) is the condition on the parameters of
If the system is not critical, then we use
the system for criticality.
instead the relation
= no, and
woPthfftbPNL/ije
It is important
k = ije/pti.g'tbPNi.
(6.81)
The
at critical [see discussion accompanying Eq. (5.183)].
k is different from unity
for example, the reactor dimen
sions are too small for a given fuel concentration (i.e., the neutron leakage
Thus, given a fuel, (6.80) defines the reactor dimension for
excessive).
which appears in pNL, 0th, and . If
criticality through the parameter
where
fc
if,
multiplication
is

in Sec.
be demonstrated
notation of (6.79), and call
REACTOR ANALYSIS
294
[chap. 6
both the size and fuel concentration
of a given system are specified, (6.81)

multiplication
the
effective
of
the system.
gives
It is of interest here to identify some limiting cases of (6.81):
1. For the case of an infinite medium, B1 > 0 and (6.81) reduces to
k = ije/Ph = kx
2.
(6.82)
Ignoring the fast leakage is equivalent to saying that fission neutrons

at the point where they are produced, in which case
slow to thermal
j7,h
and
k = rj/pthpNL
(6.83)
Note that, although this expression assumes no losses by escape during

slowing down, fast losses by absorption are still possible.
3. For the one-velocity finite reactor,
(6.84)
In this case fission neutrons are assumed to appear directly at thermal

Thus losses by absorption or leakage during slowing down are
energy.
not allowed, and therefore we take 0th = Pth = 1.
Clearly, in the onevelocity model all fissions are due to thermal neutrons; so t = 1.
4. Large,
highly thermal reactors

k
1)
(6.85)
fc,0thPNL
1)
(t
5.
In this
1].
This relation is a good approximation for such systems because, although

fissions by fast neutrons may be negligibly few, losses by absorption and
leakage during slowing down may be appreciable.
The infinite-medium
multiplication kK is used to represent the factors Tje/pu, since the fast effect
e for the large, finite medium (6.77) is very nearly equal to that for the
infinite medium [cf. Eq. (4.298) for B1
=
No fast fissions
= i?/pthgthPNL
(6.86)
Eq. (6.57) for the slowing-down density yields

source term for the thermal-diffusion equation (6.54c) which
propor
tional to the thermal flux. In general, the source term g(r,M,h)
obtained
from (6.72a).
The factor Q(0) F(r) which appears in this result
given
is
case the solution of
is
is
-AhV,h(r)
=
by (6.74); but, when
(i.e., no fast fissions) this expression reduces
to the form F(r) Q(0) = vZf <lh(r).
If we use this relation in (6.72a),
then (6.54c) may be written
Z?*u,(r)
rtfprtî)
fc.SItothfcbto
(6.87)
It important to recognize that this result satisfies the requirements of

the one-velocity model developed in Sec. 5.4g, namely, that the neutrons
from fission appear in the same space form as the thermal flux. Accord
is

k = 17/pNL
sec. 6.3]
FERMI AGE THEORY
295
ing to (6.87) the system acts as though neutrons slowed down at the same
space point at which the parent fission reaction occurred but with losses
This is by no means what
due to fast leakage and resonance absorption.
actually occurs, since individual neutrons migrate from the point of birth
during the slowing-down process; but, in a single-medium system, wherein
the problem is so defined as to yield a solution by separation of variables,
neutrons are found to diffuse in the same space mode at all lethargies,
including their behavior in the thermal range.
b. The Use of the Fission Spectrum.
With the above model as an
elementary but useful approach to the problem of the homogeneous bare
reactor, we turn now to the more general systems which include the dis
tributed fission spectrum.
All the assumptions established for the
analysis of the monoenergetic fission spectrum model will be extended
to the present

spectrum
itself.
calculation, except, of course, the description of the

We denote this distribution by the symbol j(m); this
function will be obtained, in practice, from experimental measurements.

We choose the origin of the lethargy scale such that j(0) ~ 0; thus, the
spectrum }(w) has the form shown in Fig. 4.25, but with the origin u = 0
shifted far to the left.
The peak in this function occurs in the vicinity of
u ~ 2 (if u = 0 corresponds to E = 10 Mev).
As in the calculations of
Sec. 4.6c, it will be convenient to normalize the function j(w) so that
f"*
i(u) du
(6.88)
This function is related to the fission spectrum in energy space z(E)

through the equation
[** i(u)
JO
Note that u
If
du =
J(Ez(E)
Bth
= 0 corresponds
dE
|j()
or
to E0, and
th
du\
\z(E) dE\
(6.89)
to Eltt.
distribution for fission neutrons, the source func

tion for Eq. (6.54a) takes the form [cf. Eq. (6.55)]
we use j(w) as the
S(r,u) du
[ foU,h
X,(u')
(r,u')
du' +
2^lh(r)]
j(u) du
(6.90)
The immediate objective of this calculation is to solve Eqs. (6.54) with

the boundary conditions (6.56) using the source (6.90).
The procedure
is straightforward and follows closely the analysis of Sec. 6.3a. As
before, we assume that the solutions for the fast and thermal flux can be
written in the forms of (6.59), (6.60), and (6.61); we will determine the
The substitution of these
consequences of these assumptions later.
relations into (6.90) yields
S(r,u) =
i(u)
[ph 2/(m')
Hu') du' + Xf*lh] F(r)
(6.91)
REACTOR ANALYSIS
[CHAP. 6
is,
296
is
we assume
Again it will be convenient to normalize this function; that
In order for the chain
introduced into the reactor.
that one neutron
reaction to perpetuate itself, the yield from the resulting fissions must be
exactly one neutron; therefore,
jR
where
S(t,u) du
fRdr
[f S,()
*{u) du +
2}^.h]
/RF(r)
dr
du
3(w)
(6.92)
Now, the solu
denotes the volume integral over the reactor.
tions F(r) may be taken as a normalized set, i.e.,

=
Moreover, the integral of j(w)
is
(6.93)
fRF{T)di
unity by (6.88); therefore, (6.92) may
be
2,() *() du +
2}^]
(6.94)
S(t,u)
F(t) )()
in
(6.95)
-(u)
Q() V2F(r) + 0(u) Q() F(r) =
-F(r)
Q'(u)
The combined slowing-down diffusion equation (6.54a) for a bare

reactor with the fission spectrum j(u) may now be written in the notation
of (6.57) with the source S(r,u) given by (6.95).
j(u) F(r)
(6.96)
m eSo^'
du>
Q(0)]
(6.98)
0(u).
determined from the requirement that [see (6.95)]:
S(r,0)
- g(r,0)
F(r)
/"
But, by (6.95) and the choice of the lethargy scale,

= 0.
Equation (6.98) may now be written
Q(0)
Q()
(6.99)
Q(0)
<S(r,0)
i(') 9(u;W) p(u;u') du'
0;
where G(u) = B2a(u)

The constant Q(0)
j(u0c/o'-
is
where we have used (6.59), (6.60), and (6.61).

Separation of variables
yields
inside reactor
V2F(r) + B2F(i) =0
(6.63)
= j(tt)
<
<
,
u
j8()]
+
[B2(w)
Q()
(6.97)
Q'()
is

We identify this relation as the criticality condition and will consider

greater detail shortly. If we use this result in (6.91), we obtain
it
[J
reduced to
therefore,
(6.100)
FERMI AGE THEORY
= exp
g(u;u')
p(u;u')
m exp
B2
a(w) dw + B2
/8(w)
du>
297
a(w) dw
= e'BKrW-Hu')]
/}()
where we define
dw]
sec. 6.3]
(6.101)
(6.102)
Note that denotes the fast nonleakage probability and the resonanceescape probability in the usual way [see
A-U=0Eqs. (4.260)]. Clearly, the probability
that
neutron will not be absorbed in
the lethargy interval (0,w)
will
the product of the probability that
not be absorbed in the interval (0,u') and
it
is
traversing
it
will not be absorbed

in (m',u); thus for v! < u,
(6.103)
is
from which follows the relation (6.102).

Equation (6.103) applies since the reso
Fio. 6.5 Slowing-down density
nance-escape probability for any particular
and fission spectrum.
interval
independent of the past history
of the neutron.
The fast nonleakage probabilities are also independent,
follows that
and
it
g(u;0) = g(u) = g(u') g(u;u')

we
identify the functions (6.101) and (6.102)

probability that neutron will not escape
= from the system while slowing through lethargy
a
Thus
(6.104)
g(u;u')
p(u;u')
interval (w',w)
probability that neutron will not be captured
while slowing through lethargy interval (u',u)
(6.105)
(6.106)
Note that with these definitions one can also obtain (6.100) directly by
physical arguments.
Thus to find the slowing-down density past u, we
compute first the increment dq arising from the neutrons born in du'
about u' which slow down successfully as far as u without being captured
or escaping from the system (see Fig. 6.5)
;

p(w;0) = p(u) = p(u') p(u;u')
dq(u;u')
p(u;u') g(u;u') j(u') du'
Then, to obtain the total slowing-down density past u we sum the con
tributions from all neutrons born in the range (w' < u). If we call this
REACTOR ANALYSIS
298
[chap. 6
/"
Q() =
dq(u;u')
j*
number Q(u), integration of dq over the interval (0,w) yields
j(u') p(u;w') g(w;u') du'
is
in agreement with (6.100).

The result (6.100) may be used to solve the thermal-flux equation.
The appropriate relation
[see (6.54c) and (6.59)].
-AkV**(r) + 2?^,(r)
F(r)
|o",h
j(u) 0(uu.;u) p(uu.;u) du
(6.107)
We have used the slowing-down density g(r,M,h) as the source term since
few neutrons are contributed to the thermal group directly by the fission
As in the analysis of Sec. 6.3a, we take the relation (6.60) for
reactions.
the condition on ^th which allows that solution.
and
determine
0th(r)
The substitution of (6.60) into (6.107) yields an equation for F(i), which
when compared to (6.63) shows that
uth
=
2th
+ Ahfl2
?(">;")
Jo
^th
du
(6.108)
(a)
/(')?(;') P(;')d'
(&)
ffr)
2th
(6.109)
uth
+ AhB2
- wik
2,(u)
*(r>u)
j(u') g(u;u') p(u;u') du'
Jo
F(r)
jT"
g(r,u) =
Jo
results
("";) p(u.;u) du
(c)
0.
The space function F(r) satisfies the differential equation (6.63) and the
Note that, except for a constant, the
boundary condition F(T) =
if
we take j(w) = 5(u).

above relations reduce to the set (6.72)
of
the
critical equation which was
We return now to the development
previously recognized as the relation (6.94). This expression may be
y(u) du
is
LiBij
g(-Uth>u')
2lb(l
"th
Jo
2'h
"I
J
+
If
j(m') g{u;u') p(u;u') du'
/"uth
fu
is
expanded by introducing the solution for yp{u) from (6.100), using (6.61),
It easily shown that the critical equation
and ^th from (6.108).
v
du
(6.110)
introduce the definition of the fast effect and the thermal utiliza
tion/, then this equation may be written in the alternate form:
t
we

dis
The solutions to the case of the bare homogeneous reactor with
tributed fission spectrum may be summarized in terms of the above
= 7?e/pNL
J i(u)
g(uth;u)
p(u,h;u) du
(6.111)*
sec. 6.3]
FERMI AGE THEORY
299
The results obtained by use of the Fermi age

Reactor Calculations.
model for the flux distributions and the criticality conditions in a bare
homogeneous reactor may be useful as first estimates in undertaking a
It is not to be expected, however, that these
reactor-design study.
results will give accurate critical-mass figures for an actual design.
Their
primary usefulness rests with the fact that they do yield, with an economy
of effort, reasonably dependable numbers.
These results will suffice, in
many cases, to indicate the approximate proportions and other principal
features of a proposed configuration.
More accurate estimates will, of
course, require more exacting techniques and models, such as the multic.
group treatments
(see Sec. 8.8).
The criticality relations (6.80) and (6.111) will prove to be useful in

making quick estimates of fuel concentrations, given a reactor configura
For calculations of this sort it is entirely within the
tion (or vice versa).
limitations of the theory to assume that the reactor materials are homo
geneously distributed throughout the system, although, in fact, this may
not be the case. It should be noted that the above-mentioned equations
in the parameter B2 through the fast nonleakage prob

ability g and the fast effect e, so that a trial-and-error procedure is required
for solving them. In the situations wherein these relations will be at all
useful, such as thermal reactors, the fast effect will be very nearly unity;
thus, in making a first estimate, it may be convenient to take e = 1 and
then later correct the result when B has been determined from the remain
ing factors in the critical equation.
Better to demonstrate some of the suggestions on the use of the Fermi

are transcendental
age model, let us consider the problem of determining
the critical size and

minimum fuel mass of a uniform bare reactor.
For simplicity, assume
that the basic constituents fuel, moderator, structure, and coolant have
all been selected and an estimate of the relative proportions of these
materials in the reactor is available.
Our problem is to find the dimen
sions of the reactor with these basic characteristics which has the smallest
fuel mass. The procedure is straightforward and involves the calculation
of a number of systems each of different size.
The critical-mass calculation for any one of these reactors entails a
trial-and-error procedure.
One can either assume a size and compute the
critical concentration, or vice versa. We will do the latter.
The next
step then is to evaluate the macroscopic cross sections of the reactor
medium for the entire lethargy range, i.e., 20(m), S/(m), D(u), and , as
well as the thermal-group cross sections at the operating temperature of
the system.
These data may now be applied directly to the computation
of the various parameters which appear in the critical equation, i.e., rth,
In the first trial calculation it is convenient to assume a
pu>, v, f> and L.
value for the fast effect (perhaps = 1) and check it later.
All the neces
sary data having been collected, B2 may be computed from the appropri
REACTOR ANALYSIS
300
[CHAP. 6
This
ate critical equation by trial and error [see Eqs. (6.80) and (6.111)].
first estimate of B2 may be checked by applying it to the calculation of c,
If this value of
which can be obtained from Eq. (6.81), for example.
with the assumed value, the first estimate of B2 will be ade
quate; if not, the computed value of e may be inserted into the critical
As a rule e is very nearly
equation and a second estimate of B2 obtained.
agrees well

unity; consequently, the trial procedure for B2 should converge very

rapidly.
If this procedure is carried out for several fuel concentrations, a family
of reactors of various sizes will be
obtained.
A plot of the critical fuel
mass and reactor size as a function
of fuel concentration will yield a
curve of the shape shown in Fig.
6.6.
The symbol Nf denotes the
fuel concentration required for an
infinite medium of the constituents
Size
used in the reactor study.
The
curve to the left of the asymptote
Fuel concentration
is a mathematical
solution to the
critical-mass
equation
physical significance.
d. Migration Area.
no
having
The critical
equation (6.80) takes an especially

simple form in the case of very large
Fio.
6.6
Critical fuel mass and reactor
size as a function of fuel concentration.
For these systems
reactors.
B2
<C 1 ,
and the fast effect e, which is relatively insensitive to B2, may be computed
as though the reactor medium were infinite [i.e., Eq. (6.77) reduces to
(4.298)].
1
Moreover, if also
written
1
rthB2
1,
7//(p,h0ti,pNi.
then
glh
kKgthpNL
(1
+ BVth)"1, and (6.112) may
^00
=
(1
where
+ B-rtb)(l + L2B2)
M* = rth + L2
and the term in J54 has been neglected.

migration area and M the migration length.
(6.112)
^"00
+ B"-M2
be
(6.113) *
(6.114) *
The quantity M2 is called the

Thus the factor (1 + B2M2)~1
of the combined leakage losses from the reactor due to

escapes which occur during both the slowing-down process and thermal
The number M , then, is a measure of the combined slowing
diffusion.
is an estimate
down and diffusion length so that M* is the mean square distance from
Note,
the point of birth as a fast neutron to death in thermal capture.
FERMI AGE THEORY
sec. 6.3]
finally, that the migration length can serve
301
as a measure of the size of a
reactor.
The identification of the fast none. Fast Nonleakage Probability.

leakage probability g(u) (see Eqs. (6.67) and (6.105)] is demonstrated by
The calculation is carried out
means of a simple probability argument.
by using the technique developed in Sec. 4.3b for computing the reso
As in that calculation, we com
nance-escape probability for hydrogen.
pute first the probability of a particular event associated with a single
The
lethargy interval Ait, and then extend this result to the range (0,m).
probability that the neutron will not escape from the system while slowing
through the interval Am< is denoted as g(u, + Am,;w,); therefore
=
g(iii + Aw,;ttj)
(probability of escape in
The probability that the neutron will

by
Total leakage of neutrons in
Am<
Am;)
Aw( is given
escape while traversing
from system
Am,
jy
Total neutron population
(6.115)
J(r,w.)
g(r,M,)
dS
di
where
Js
and
jv
indicate
over the surface and volume,
integrations
Now
respectively.
J(r,Mi)
g(r,U<)
= Z((m,)
-D(Ui) V*(r,,)
D(m.) *(m.)
<Kr,U = 2.(m,) f2.(m,)

(<) F(r)
f (<) F(r)
VF(r)
We substitute
where we have used Eqs. (6.59) through (6.61).
expressions into (6.116) and apply Gauss's theorem to obtain
- D(ui)
Am,
V0(r,Ui)
Js
- D(iu)
dS
4>{ux) Am<
&(,)
&(,) fy *(r,,) di
*(m>)
fv
fy F(t)
F(r)
these
rfr
dt
The application of (6.63) to the fraction at the right yields B2D{ut) Am*/
Thus the probability of nonleakage while the neutron is in Aw, is
Z,(Mj).
g{Ui
+ AUi;Ui)
2,(u-)
(6.117)
The probability that the neutron will not escape in traversing the interval
(0,w) is obtained by taking the product of the individual probabilities for
- BWjuj)
Am,]
_~ -
YiIn
t\
iIn g(u) =
n [i - 5^*]
1l"i
a
and
OA- ,w [
each Am,; thus

(6.116)
B*D(Ui)
Am,
REACTOR ANALYSIS
302
[chap. 6
(We have assumed here that the neutrons lost by escape from the system
are few in comparison to the total population.)
In the limit as Aut 0,
the summation goes over into an integral, and we obtain
(6.120)*
6.4
of Fermi Age Theory
Diffusion-length Experiment Using Fast Source.
The application
of the general results developed in the preceding sections of this chapter
may be demonstrated by means of two elementary calculations which
have some practical importance.
The first of these is the estimation of
the diffusion length of thermal neutrons in a given material from a
knowledge of the thermal-flux distribution in a block of this material.
A
simplified treatment of this problem has already been given in Sec. 5.5b ;
however, the calculation presented there was based on the assumption
that the neutron source was thermal.
We now discard this restriction
and consider the more general situation wherein the source yields monoenergetic neutrons of some energy E0 > Etb. The present calculation is
of interest, therefore, when "thermal-neutron columns" are not immedi
ately available and fast-neutron
sources must be employed for the
a.
experiment.
is,
The particular problem to be considered is the determination of the

thermal-flux distribution along the axis of a long parallelepiped (com
posed of the material for which the diffusion length is to be measured)
which has a plane source of monoenergetic (E0 > Eth) neutrons placed at
As in the thermal-source experiment of Sec. 5.5b, the thermalone end.
flux distribution is of interest here because the final expression for this
function involves the diffusion length of the material.
For the region in
the block that is far from either end, this expression has a simple approxi
mation of exponential form which
in fact, similar to the result (5.246).
Thus, as in the thermal-source formulation of this problem, the diffusion
length can be obtained experimentally from flux measurements along the
axis of the block.
In general, the application of the Fermi
age model to finite systems
material in question) which contains
is
We
complicated problem.
will avoid the mathematical difficulties involved in solving the more
more simple
general situation described above and consider instead
In place of the block, which may be the actual
geometric configuration.
geometry used in the experiment, we select an infinite medium (of the
which contain localized neutron sources
plane source of neutrons of energy

a
it
allows
con
proposed because
Eo- This particular configuration
siderable simplification in the analysis without distorting the final results,
is

Applications
FEKMI AGE THEORY
sec. 6.4]
303
namely, the expression for the axial distribution of thermal neutrons far
Thus, although this elementary model avoids
the mathematical complications of the actual geometry, it does yield an
from the plane source.
accurate description of the physics of interest here.

Consider, then, the problem denned by the following set of assumptions:
(1) an infinite medium with scattering and absorption cross sections, and
(2) one which contains a plane isotropic source (50 neutrons per area per
time) of neutrons of energy E0 placed at the origin z = 0. The appropri
equations for this system is
ate set of differential
-D(u)
4>(z,u)
+ 2.(u) *(,)
q(z,u) = 2,(u)
-Ah ^
(z,u)
S^Mz)
The initial condition and neutron-balance

(1)
lim Q(z,u)
- ^ ?(*,)
(6.122)
= g(z,.)
(6.123)
requirements
are
= 0
fQ"
(6.124)
Q(z,w) dz = iqo
is
is
where we use, as before, q(z,u) = Q(z,u) p(u). The slowing-down features

of this problem were discussed in Sec. 6.2b. The solution for q(z,u)
This function, evaluated at u = wti>,
the source term
given in (6.28).
in (6.123):
5.
The general solution to this equation was developed in Sec. 2d by use of

In the present case the source S(z)
the kernel method [see Eq. (5.97)].
from (6.125) may be substituted directly into (5.97), and the integration
will yield &h(z).
We note, however, that the system in question involves

is
is
only one space variable z. Thus although the general form (5.97)
by no means necessary, and, in fact, some labor can
applicable here,
kernel relation [see Eq.
be eliminated
by using the one-dimensional
it
(5.78)], namely,
D^/Zf.
takes the form
2AT
For this problem
S(r)
is
where
+*{e)
/_".

0
(2)
(6.121)
<r"l'~rl
df
(6-126)
given by (6.125), and (6.126)
EEACTOR ANALYSIS
304
This result may
be
[CHAP. 6
written in terms of the error function erf (a;), denned by

erf(x) =
(Xe~-'ds
(6.128)

=
qol^'
^-xV^
[2.
Mz)
cosh xz
e~"
- e"
erfÔ
erf(/32)]
(6.129)
VTth
where
Vrih
+ x Vt^
(6.130)
0.
0,
Equation (6.129) gives the thermal-flux distribution, in an infinite

=
medium, that results from
plane source of neutrons at the origin
The resonance-escape probability and the age which appear in this
which corresponds to the energy E0.
expression are measured from u =
is
As previously indicated, the only interest in this result

in its form for
z.
We expect that the asymptotic solution
will give the correct form for the thermal flux in the case of the actual
We obtain the asymptotic
geometry for regions far from the source.
form for (6.129) by introducing the expansion for the error function in the
case of large x, namely,
ertw
1-^(1-27.+
--)
(6-13"
e~" erf(/3i)
e"
erf(/32) =
The difference of the two error functions which appear in (6.129) may
written in terms of this expansion, thus
be
sinh xz
_ 2x Vt,!, exp (-zV4rth

VV (z2/W
xVu)
x2rth)
The substitution of this expression into (6.129) yields [neglecting terms in

exp (-zJ/4T,h)]
This result
is
Ml)(m^y..
(6,32)
to be compared to
^(z) ~ Ae-ôo'
which
(6.133)
the solution to the problem of Sec. 5.5d in the case of an infinitely

For a onelong parallelepiped [cf. Eq. (5.246) for the case ].
space,
is
dimensional
it
>
is

large values of the variable
clear that
[see
fcjoo
so
Eq. (5.248)]
= x
that Eqs. (6.132) and (6.133) are identical in form.
(6.134)
FERMI AGE THEORY
sec. 6.4]
305
Thus, diffusion-length
measurements, in which a block of the material

in question is exposed to a fast-neutron source, can be carried out by
following the procedure used in the case of the thermal source. At large
distances from the source, the thermal flux varies exponentially, and the
diffusion length can be obtained from simple foil activation measurements
along the axis of the specimen.
However, as before, these foils should
not be placed near the ends. At the far end of the specimen (opposite
the irradiated face), the so-called end effects discussed in Sec. 5.5d are
The vicinity of the irradiated face is also to be avoided in

positioning the foils inasmuch as the thermal flux in this region takes a
Gaussian-like distribution such as described in Sec. 6.2a (see Fig. 6.2).
Moreover, if the neutron source is a collimated beam, this distribution
will tend to peak well within the interior of the specimen, and the simple
encountered.
is,
exponential behavior will not appear until much farther inside (see Sec.
5.7c and Fig. 5.31).
Finally, it should be noted that the result (6.132) is generally valid,
that
the thermal flux has the same functional dependence at large dis
is
is
the calculation of the slowing-down

example that will be considered
density, in
slab, which results from a plane source of fast neutrons
This example
included so as to demonstrate
placed at the center.
technique that can be applied to finite systems which contain localized
sources such as planes, points, or lines.
The method of Laplace trans
forms used to obtain the elementary solutions associated with the infinite
is
media (Sec. 6.2)

not useful here because of the conditions which must be
satisfied in the finite geometries, namely, that the flux and slowing-down
0)
density vanish at the extrapolated boundary and that the initial condition
In the case of
(corresponding to t =
be satisfied at the source plane.
the infinite media, the boundary conditions, of course, did not appear,
and the initial condition was easily satisfied by means of simple integral
>
is
s)
(t
it
relation [e.g., (6.20)3]. The basic approach used in these calculations

can be applied to the finite system as well; however, in applying this
method to the present case,
necessary to represent the source by
delta function and to expand the results in infinite series of eigenfunctions
of the transformed
differential equation.
The solution obtained
in this way
not available in closed form and must be expressed by
is
is
means of an infinite series.

No advantage
to be gained in pursuing
this course since similar results can be obtained with much less effort by
the use of other methods.
We treat the present problem by the method of images. Thus we
solution for the slowing-down density due to a
plane source and, by constructing
system of "alternating sources"
outside the slab, reduce the flux at the extrapolated boundaries to zero so
take the infinite-medium

tances from the source regardless of the energy of the source neutrons.
b. Method of Images
Infinite Slab with Plane Source.
The second
REACTOR ANALYSIS
306
[CHAP. 6
Q(x,r) =
(3) Q(a,r)
(2)
dxj
Q(xfT)
q0
(6-135)
Q(-x,t)
=
|"
(1)
Jim
as to satisfy the required conditions.

The differential equation for the
The solution to this equation
slowing-down density is given by (6.19).
must satisfy the boundary conditions:
is
The last relation follows by virtue of the fact that the slowing-down
proportional to the fast flux through the coupling equation
density
2S
Fio.
6.7
_j
Superposition of images.
is,
(6.6), and, in the usual way, we have required that the flux at all energies
satisfy the extrapolated boundary condition. The general solution to
in the case of an infinite medium,
(6.19)
QM
aW"
(6-136)
where Q(x,t) = g(x,7-)/p(r), and gives the slowing-down density in an

infinite medium (with no resonance absorption) due to a plane source of
=
neutrons at x x'.
The conditions required for the slab geometry may be met by distribut
ing an infinite set of source planes of strength +90 to either side of the
slab, as shown in Fig. 6.7. The effect of these alternating "positive"
on the flux
and "negative" sources of neutrons at + 26, + 4a,
Note, for example, how the negative source at
evident from the figure.
inadequate to correct
+25 just cancels the flux at boundary +a but
for the condition at a. The major correction at a
provided by the
in
flux
the
at a due to the
negative source at 2a.
The small error
source at +23
now corrected by
positive source at 45, and so on.
The slowing-down density at due to any particular source of t = neu
is
is
is
...
is

FKRMI AGE THEORY
sec. 6.4]
trons in this alternating system may
QM
'
be
n = 0, 1
is obtained
(6.137)
by superimposing
the
BO
written
gî
The combined effect of this infinite set

contributions of all the sources; thus,
Q(x,r)
307
^j
Q-(x,r)
(_i).e-|^*,.a|vr
(6.138)
B as
In order to demonstrate that this infinite series
is indeed the desired solu
tion to the problem, we must show that it satisfies the boundary condi
tions (6.135).
The first condition is seen to be satisfied if we observe that
rt >
1
= Km
>
g-(*-2no)'/<r 1

n-1
and that
lim
jrr^r /
J -a
Condition
fT**'** dx
,-,0 L(4irr)
is satisfied by virtue of the geometric symmetry of the

We can show that condition (3) can also be satisfied by
examining the boundary x = a. The slowing-down density Q(&,t) is
(2)
image system.
Q(fi,r)=^~ ^ (_l)ne-|3-23|./4,
nm
=
77^[e~5'/4r
e-s,/4'
e-<SS)''4'
Thus it would appear that each term in the series
is canceled
(6.139)
out by
one
other, and we obtain

Q(d,r)
= 0
The result (6.138) satisfies the differential equation (6.19) and the condi
tions (6.135) and is therefore the solution to the present problem.
In carrying out a calculation it is convenient to approximate
A reasonable estimate
infinite series by taking the predominant terms.

of Q(x,t) is obtained from the expression
Q(X,T)
[e-r'Hr
e-U+23)'l4r
the
e~(r-2)<l*r]
(6.140)
REACTOR ANALYSIS
308
[CHAP. 6
for x >
0.
Note that the terms in n > 2 have been neglected since
these terms decrease at least as fast as e~"'liT as n increases.
6.5
Reactor Temperature
Coefficients
a. Temperature
Changes and Reactor Stability. A problem of con
siderable importance to the control of nuclear reactors is the evaluation
of the temperature coefficient of the reactor system.
It is customary to
define the temperature coefficient as the change in the multiplication
(or reactivity) of the system due to a change in the operating

temperature of the reactor.
Thus the temperature coefficient is a meas
ure of the inherent stability of the reactor.
For example, if for some
constant
reason there is an increase in the temperature of the reactor (perhaps due

to a change in the power demands on the system), then if the temperature
coefficient is positive, the multiplication constant will increase, which, in

turn, will cause the rate of energy production to increase and further
raise the temperature of the system.
Such a system would be unstable
and would require a positive and continuous influence of a control system
to maintain itself in the steady-state condition. On the other hand, if
the coefficient is negative, the effect of a temperature rise on the neutron
population (and therefore on the power production) would be to decrease
the number of neutrons (and power), so that the temperature would then
Thus in these systems temperature "disturbances" would

eventually disappear, and the reactor would return to steady state even
in the absence of external control orders.
The evaluation of the temperature coefficient for bare homogeneous
reactors is a relatively easy task and in many instances amenable to
This is not the case, however, for more complex
analytical methods.
decrease.
For
systems such as heterogeneous media and reflected configurations.
these systems, numerical methods are more useful, and changes in multi
plication constant are obtained by making repeated calculations for k for
successive small changes in the temperature-dependent
parameters of the
Here, attention is confined to the bare homogeneous system,
system.
and the emphasis is placed on presenting some of the fundamental con
cepts involved in the computation of temperature coefficients.
The
material presented in this section is meant to serve only as a demonstra
tion of a few of the more important ideas and approaches which can be
utilized in such calculations.
Furthermore, the present methods devel
oped for homogeneous systems may be useful, at times, for obtaining a
preliminary understanding of the temperature response of a particular
reactor system wherein accurate calculations might be unduly involved
and not entirely necessary.
In any case, the temperature-sensitive
The
parameters considered here are of general interest to all reactors.
essential difference in approach taken for the treatment of nonhomogene
sec. 6.5]
FERMI AGE THEORY
ous systems
in
which
is
due primarily
these parameters
309
to the fundamentally
different
forms
may be described for these more complex
configurations.
Temperature changes in a reactor can affect the neutron population

and, therefore, the multiplication constant in two fundamentally different
ways, namely, by changing the nuclear (or microscopic) properties of the
materials and by changing the density and size (macroscopic properties)
of the reactor components.
The various properties affected by tempera
ture are: (1) nuclear: (a) the neutron slowing-down range, (b) thermal
group cross sections as influenced by average neutron energy, and (c)
Doppler effect; (2) physical: (d) density of the reactor materials and
(e) neutron leakage as influenced by reactor size (and migration area).
A change in the neutron slowing-down range is effectively a change in
the thermal base. Thus temperature changes which directly affect the
distribution of the thermal-neutron group do in fact define a new thermal
energy (Eth). Changes in Eth in turn affect such quantities as the age and

the resonance-escape probability which are denned by integrals whose

limits involve this number.
Temperature changes may also directly affect the neutron cross sec
tions, and this dependence appears in both the thermal-group cross sec
tions and in the fast cross sections.
The effect on the thermal group
indicated in (6) above is sensed through the function m [defined by
For most of the quantities appearing

(4.200)] which appears in (4.209).
in the multiplication constant which are functions of the thermal-group
cross sections (e.g.,/), it will frequently suffice to assume that the cross
sections are temperature independent (i.e., are not affected by changes in
neutron speed), as with the c, or to assume a simple l/v dependence in
the thermal range, as with the <ra. The temperature dependence of such
cross sections is easily determined and will be demonstrated shortly.

There is a second temperature effect on cross sections which appears in
the fast range and is especially important in nuclei which possess pro
Even though many of

materials retain a nearly l/v dependence around thermal energy,
the deviation from a simple l/v behavior is not negligible; moreover, in
many instances such materials also possess resonances close to the thermal
nounced resonances, such as the nuclear fuels.
these
These characteristics enter not only into the evaluation of the

range.
temperature coefficient of the thermal group properties, but also into the
temperature coefficient of such quantities as the resonance-escape prob
ability which involve integrals of cross sections over the fast (resonance)
Properly, the fast cross sections which are used in such functions
range.
must be computed from an integral relation of the type (4.210) which
accounts for the thermal motion of the nuclei.
The temperature depend
ence of the fast cross section appears, of course, through
function
2JJ defined
by (4.198) which involves
the distribution
the temperature
of the
REACTOR ANALYSIS
310
TV
material
Thus variations in TN affect
[CHAP. G
2ft,
which in turn affects the
This phenomenon is the

dependent
quantities.
and
is
discussed
in
so-called Doppler effect
connection with the tempera
ture coefficient of the resonance-escape probability.
fast
cross-section
The temperature dependence of the material density and reactor size
A change in
are easily denned in the case of bare homogeneous reactors.
the temperature produces, of course, a change in the various nuclear
densities which are involved in the macroscopic cross sections.
More
over, the thermal expansions which accompany increases in temperature
influence directly the physical size of the system and therefore affect the
neutron leakage.
The Multiplication Constant.

The temperature coefficient of a
reactor is obtained by computing the temperature derivative of the multi
In each of the analytical models so far developed for
plication constant.
homogeneous systems, it was found that the multiplication constant
could be expressed as a product of several factors.
(It will be shown in a
later chapter that this is also the case for heterogeneous and multib.

On the basis of our most advanced model (Fermi age)

regional reactors.)
we obtained the expression
k = j?/plhffti.PNL
The temperature coefficient is computed

derivative of this relation; thus,
1
dk
kdT
1 dri
3T
dplh
plk
dT
dt
dT
(6.81)
by taking the temperature
df
dgth
dT
glh
dT
dpNL
pNL
dT
W-lil>*
This equation gives the fractional change in k due to a unit change in

temperature T. The total change dk due to a change in temperature S T
is clearly
where the derivative is obtained from (6.141) and evaluated at the

original temperature of the system prior to the introduction of the change
In some instances it is more convenient to work with the reactivity a>,
ST.
which we define
u =
It
follows
k =
and
p-1
o>
that in terms of the reactivity the temperature
takes the form

do,
(.
1\
dk
(6.143)*
coefficient
....
which gives the change in reactivity due to a unit change in temperature.
FERMI AGE THEORY
sec. 6.5]
311
Note that, in the event that the reactor was initially at steady state
(* = D,
=
af
af
(6145)
Coefficient of the Thermal Utilization.

The tempera
utilization
for
a
of
the
thermal
homogeneous system is
ture coefficient
obtained from the usual relation (3.23) ; thus we compute
c. Temperature
or
df
2""'
"I
2/ + 2"J
32(F>
2?> ~dT
Z
1
d2'0>l
~d
(6-146)
where the cross sections 2^' and 2^> refer to the thermal group properties
The temperature
of the fuel and nonfuel components, respectively.
derivative of these cross sections is obtained from the relation
32
,,T
(N)
da
dN
._.
(6.147)
N is the nuclear density of the material (nuclei per unit volume)

If we let
and a refers to the microscopic thermal-group cross section.
=
N
SJl/Z3, then,
where
NdT
(6-148)
'~ldT
total number of nuclei of
particular material in the

of the system such that the
1 dl
volume of the reactor is the cube of I. It should be noted that T ^= is the
I d1
usual coefficient of thermal expansion of a material.
We denote this
where
31
is the
reactor and I is a characteristic
dimension
derivative by a, and observe that it is related to the temperature deriva

tive of the density p (mass, or weight, per unit volume) of the material by
the equation
-\t--klr
The derivative of the microscopic cross sections a is obtained from the
definition of the thermal group (4.209) and depends, of course, on a
detailed knowledge of the energy dependence of <r in the vicinity of kT.
Unless otherwise noted, we will assume in this treatment that (1) all
absorption cross sections aa vary as l/v in the thermal range and (2)
scattering and transport cross sections a, and <rlr are energy (temperature)
In the case of a l/v dependence, it was shown previously
independent.
that the thermal-group cross section is also l/v [see (4.234)]; therefore, in
the case of the
aa,
REACTOR ANALYSIS
312
where
[CHAP.
T is the "base" temperature and refers to the initial state of the
If
system.
we use the results
(6.148) through
(6.150) in (6.147), we
lw=-(*a
*r)
obtain
kw=-**
and
The application of the first of these relations to (6.146) yields
j|
-f)(a0
= 3(1
- aF)
(6.152)*
where for convenience we have characterized all the nonfuel components

by the parameters a0 and lo- We find, then, that for a homogeneous
reactor, wherein we have assumed
l/ dependence for the absorption
of the thermal utilization

expansion of the materials
involved. In
system wherein the bulk of the reactor materials are
solids, the temperature coefficient of will be relatively small since the a,
However, in the event that
values for solids are in the order of 10~6/C.
an appreciable fraction of the constituents are fluids (for example, the
moderator may be in liquid form), the/ term in (6.141) may be of com
with the others.

Coefficient
of the Thermal Nonleakage
^-Wf
dU
-pL{vJT
dpNL
is
The temperature derivative of the thermal nonleakage

It easily shown that
obtained from (5.217).
Probability.
probability
is
parable magnitude
d. Temperature
dB*\
W>w)
(6-153)
It
is
the leakage probability of thermal neutrons according to the

definition (5.220). By the usual relation
where pt
is
a

cross sections, the temperature coefficient

depends only on the coefficient of thermal
good approximation to take
D ~ D0, then,
L2~L(1
where L% = D0/2fa0), the diffusion
follows from Eq. (6.155) that
-/)
(6.155)*
length for the nonfuel materials.
It
FERMI AGE THEOKY
sec. 6.5]
But, by our assumption that dau/dT =
313
(6.148), we
0, and the relation
have
We substitute this result, along with (6.151), into (6.157) and obtain, for
the nonfuel components,
dLh =
dT
Jf
~j
L*0(6ao
(6.160)
For the complete system

dL2
dT
L.(6ao
^__L_^)
The temperature derivative of the buckling

Z
we define
(6.162)
B2
Mr
dJ*l
dT
lR
_L
will depend on the geometry of the system, then

(6.163)
dT
(trtt
r=7
- 6a ~
(6-164)
we use also (6.152), then,
dp
pnl
Pl j-Rjf~
- - 3/o, - -L
/)
If
PL
215)
Pnl dT
d%
dpm.
The substitution of (6.163) and (6.161) into (6.153) yields
3ao(2
(6.165)*
~B
Tt"
"aT
(6-166)
ar
B2 term
obtained from the relation (6.163).
The t,
This implies that the Fermi age
term we compute from Eq. (6.47).
model
applicable to the system in question and therefore all the require
ments for the validity of this model are satisfied.
Of immediate concern
is
As before, the
^
^
dT
"
ftn
_L
Probability. We
Coefficient
of Fast Nonleakage
e. Temperature
compute the temperature coefficient of the fast nonleakage probability to
Thus
thermal glb from the expression (6.79).
Tib
'
is
the condition that Xa(u)

to the present calculation
the integral expression for the age
D(u) du
is,(u)
<3C
S,(u).
Thus,
if
is

?1
B2 =
where
If
of the reactor.
will involve some char
(fJ
acteristic dimension
B2
(6.161)
REACTOR ANALYSIS
314
[CHAP. 6
is to yield a good estimate of the slowing-down length for fast neutrons, it

is necessary that the concentration of the highly absorbing materials
In these instances, 2t ~ S and we may
(fuels) in the reactor be low.
This will be the case in many situa
use the form (6.48) for the age rth.
tions of practical
then be written
*i*
dT
a
dT
r,h
Jo
The temperature
importance.
d(m) du =
2.(u)
2.(uth)
fua
dT
J0
derivative of
pw l a
a r|j2,(u)J
aM
rth
may
ic
ifi7^
â0'J
'
jip
We apply again the assumption that <r, and <rtr are temperature independ
ent; thus with the aid of Eqs. (6.151) and (6.159), (6.167) reduces to the
form
(6.168)
6a0Tth
is
if
alh
^-
6a0
j-
= SVth
is
d<7,h
-w
dln\
la
(6.169)*
t;
f.
Temperature Coefficient of i?.

first estimate of the temperature
coefficient of may be obtained from the temperature dependence of the
fission and radiative-capture cross sections of the fuel.
Thus, in the
analysis which follows, we neglect entirely the temperature effects on the
neutron yield per fission
v.
is
7j
When working with the thermal-neutron population, the proper defini

given in terms of the thermal-group cross sections, and not in
terms directly of the measured values as presented in cross-section curves.
It clear then that the present calculation will involve the various
tion of
is
- -rh
is
factors derived previously in Sec. 4.7g.

The quantity
defined by the relation
r\
(3-21)
is
where for this calculation

understood that the indicated cross sections
refer to the thermal-group properties of the fissionable material.
These
are related to the measured cross sections through Eq. (4.247).
Note
it

is
is
is
with ath denned by (6.58). Note that this estimate for the temperature
derivative of rth
based on the assumption that the essential effect of
variations in temperature
to modify the density of the reactor materials.
We observed above that this
true only
the fuel concentration
low.
Otherwise, the relation (6.48) for Tth will not apply, nor for that matter
(6.79) for the fast nonleakage probability.
The temperature derivative of gth obtained by substituting the results
(6.163) and (6.168) into (6.166).
sec. 6.5]
FERMI AGE THEORY
315
that for purposes of clarity and for consistency
with the preceding

analyses we have omitted the superscript "th" in the thermal-group
In order to differentiate these cross sections from
cross-section symbol.
the measured values, we denote the latter by a'. The prime notation
will be extended also to all other quantities referred directly to measured
data.
Thus
and
These quantities
-}
are related by
a(T)
Another useful relation is

1
Note that
- a'(T)f.(T)
'(D
(D
[1
(6.170)
+ a'(T)]Fl+a(T)
1
1
(6.171)
+ oV.
+ '
(6.172)
been computed by J. A. Harvey1 for the standard fuels,

and some values are given in Table 6.2, along with 1 + a', as a function
of neutron temperature T.

This factor has
Table
6.2
The Quantity
+ o' and Its f Factor
for the Standard Fuels
Temperature
Fuel
K
0
293.6
423
575
751
PuJS
J. A. Harvey and
lished work.
1
J.
+'
f l+a
ev
0
0.0253
0.0364
0.0496
0.0648
1.125
1.132
1.115
1.095
1.081
1.000
0.985
0.993
1.005
1.017
1 .207
1 .000
293.6
1.183
0.997
423
1 . 173
1 .001
575
1 . 163
1 .009
751
1.157
1 .018
1.407
1 .000
293.6
1 .416
1 .016
423
1 .039
")7")
1.418
1.423
751
1 .433
1 .093
1 .075
E. Sanders, Progr. in Nuclear Energy, 1,
1 (1956),
and unpub
REACTOR ANALYSIS
316
[CHAP. 6
The substitution of the form (6.171) into (3.21) yields for the temper
ature coefficient of
idT
n:
"il7al;(1
a)-^ W
((U73)*
In this formulation we have tacitly assumed that variations in the tem

perature of the reactor are sensed immediately in the average temperature
Tn of the thermal-neutron population.
(Note also that we have omitted
the subscript n on Tn.)

This is a reasonable assumption in the case of a
In these systems all material components of the
homogeneous system.
reactor core are intimately mixed, and a change in the temperature of
one
implies
temperature
an equal change in the temperature of all.

The neutron
T, being directly related to the moderator temperature
Eq. (4.211)], will likewise experience the same variation. This is

in heterogeneous systems. Here, variations in the
temperature of one component will not in general be immediately sensed
in all the other components.
This looser thermal coupling is especially
Tjv [see
not the
case, however,
significant
in the case of the fuel and moderator
components;
it
is due
as well as to the relatively

Thus a change in fuel
larger mass (i.e., heat capacity) of the moderator.
temperature does not appear instantly as a change in the neutron temper
ature (the neutron temperature in a heterogeneous reactor is also governed
An important consequence, inciden
by the moderator temperature).
tally, of this physical separation of fuel and moderator is the greater

principally to the physical separation of the two
temperature sensitivity of the resonance-escape probability.

The result (6.173) may be generally applied for numerical computation.
However, no attempt will be made here to obtain a numerical estimate of
the temperature coefficient inasmuch as a detailed knowledge of the func
tion fi+a(T) and its derivative
is required.
Probability.
of the Resonance-escape
of some of
only
which
follows
is
intended
to
an
idea
give
The treatment
the factors which enter into the calculation of the temperature coefficient
g. Temperature
Coefficient
There are three such factors:

of the resonance-escape probability.
(1) the usual effect of temperature on the physical density of the materials
involved, (2) the effect of temperature on the l/v portion of the absorption
cross sections, and (3) the effect of temperature on the resonance regions
of the cross sections.
The first two factors are easily accounted for and
The third factor,
can be handled by the methods already developed.
which is the Doppler effect,1 is perhaps the most interesting consideration
Unfortunately,
in any calculation of the resonance-escape probability.
involves
more
nuclear
advanced concepts of
physics than
its evaluation
1 The
resonance
importance of Doppler broadening of resonance lines in the calculation

neutron absorption was first pointed out by J. A. Wheeler.
of
FERMI AGE THEORY
sec. 6.5]
317
have been considered so far; thus, the present analysis is limited to some
brief remarks about the nature of this effect, and the reader is referred to
other sources for more detailed studies.1 However, for several situations
of practical importance to reactor technology, the Doppler effect is a
secondary factor in its influence on the temperature coefficient of a reac
tor.
These situations will be discussed shortly; the formulation which
follows is best applied to such systems.
An estimate of the temperature coefficient of the resonance-escape

probability for a homogeneous system of low fuel concentration may be
For the present calculation
obtained from the general relation (4.113).
we write this expression in the form
p*mexp[If
LiMm\
(6-174)
into the fuel and nonfuel components by
we separate the cross sections
means of the usual notation, then the integral in (6.174), which we denote
/, may be written

[Nrf/KE)
+ Nq^(E)}
dE/E
(6.175)
Let us also separate the scattering cross section for the fuel
a{,F) into its

two parts, the potential scattering <rpF) and the resonance scattering o-'f ' ;
thus2
am = am + am
(6.176)
In this formulation, the
denotes the usual energy-independent portion

of the scattering cross section and the o-'f > the resonance structure.
We
will assume that the scattering cross section a1,0' for the nonfuel materials
has no resonances and can be well represented by an energy-independent
function.
Let
a(pF)
us define, then, the cross section
+ % ,
av m
(6.177)
"
r i+iw
H/>
+ Noa^/Nr]
dE/E
(6.178)
1 See, e.g., H. Feshbach, G. Goertzel, and H.

YamaRuchi, Nuclear Set. and Eng.,
(1), 4-20 (March, 1956); see also Sees. 10.3g and 10.3h, especially Table 10.4 and
Fig.
10.17.
1 See,
pp.
J. M. Blatt and V. F. Weisskopf, "Theoretical Nuclear Physics,"

John Wiley & Sons, Inc., New York, 1952; and D. J. Hughes, "Pile
for example,
398-411,
Neutron Research,"
Mass., 1953.
pp.
258-266, Addison- Wesley Publishing
Company,
Reading,
REACTOR ANALYSIS
318
[CHAP. 6
where we have assigned { some average value by the method outlined in

The above expression is a general form which may be utilized
Sec. 4.5c.
for estimating the resonance integral I, and we will consider such calcula
tions later in conjunction with the analysis of heterogeneous reactors.
For the present our interest in this result is focused on estimates of the
In this calculation it will
temperature coefficient of the integral proper.
suffice to examine an approximate form for the integral based on the
a(/>(E)

Fig.
Absorption cross section of a nuclear fuel.
6.8
that the nonfuel terms involving the

with
the fuel terms.
parison
Thus,
assumption
(E)
Em,
<rlam
are small in com
dE/E
(6.179)
<*p
It
should be recognized that the cross sections which appear in this rela
tion are the so-called effective cross sections which take into account the
thermal motion of the nuclei.
We observed in Sec. 4.7c that only in the
of very-high-energy
neutrons could one assume that the relative
speed (or energy) between neutron and nucleus was well approximated
by the absolute speed of the neutron in the laboratory system of coordi
case
At the lower end of the energy scale, the thermal motion of the
nuclei must be taken into account, and a suitable formula was given in
Eq. (4.210) for computing the appropriate reaction rates. In the present
nates.
case we require the expression

()
/y
for the absorption cross section:
r?(|v
- V|)|v - V|9K(V;TA0 dV
(6.180)
the notation in terms of the

neutron speed v. The cross section a* in the integrand takes various
For convenience
forms, depending on the value of v which is involved.
we separate the integral of (6.179) into two parts, the first extending
where for convenience
we have retained
FERMI AGE THEORY
SEC. 6.5]
319
from Etb to Ei (some energy just below the first resonance) and a second
which begins at Ei and extends over the entire resonance range (see
Fig. 6.8). The energy Ei is so selected that in the interval (Etl,,Ei) the
In the resonance
cross section o\P{v) has approximately a 1/v form.1
range, <r'aF)(v) may he regarded as the sum total of contributions from

We define then
various resonances.
<T(aF)(E)
'E
1
V/KE)

S2
= S
(6.181)
+ c'.P(E)]
(6.182)
Co/v<tv
WE)
^Si +
dE/E
(cq/v)
&i =
where
dE/E
dE/E
"
C'^{E)dE
A'
'
\<x?\E) + a'JKE)}
(6183)
We have assumed that there are no scattering resonances (i.e., <r(,P = 0)

in the interval (Elh,Ei), and in the interval (1,00) the l/<rp term is much
less than unity (i.e., the fuel concentration
is very dilute).
The Co
The cross section which appears in the integral Si
denotes a constant.
is obtained from the reaction-rate relation (6.180) by summing the con
tributions from each resonance, using for the <j the appropriate singlelevel Breit-Wigner formula.2
Our present interest is not in the accurate evaluation of the resonance
integral / but only in obtaining an estimate of its temperature derivative.
If
we use the form
(6.181), we can write the temperature
coefficient of
Pth as
Pth
It
ol
dl Ifrp
Cpdl
iflpoL
follows from our definition of ap [see Eq. (6.177)] and the temperature
of the nuclear density JV [see (6.148) and (6.149)] that
derivatives
=
a ply p
<jp 0 1
The temperature derivative of
$ 1 we
(a0
- aF) In
(6.185)
plE
compute from (6.182).
By changing
variables in (6.182) we have

= 2c
["
>th
.
V{V
,
.
Co/<Tp)
2rJln(l
Pl \
<TpVth/
- ln(l +
\
-)]
(TpViJ]
(6.186)
1 See
1
Eq. (4.234) which demonstrates that, if

op. ext., pp. 391-394.
af ~
1 /v,
then
so also is
(i>)
1 /v.
REACTOR ANALYSIS
320
For convenience
we let Vi >
[i.e.,
[CHAP. 6
we ignore the second term in (6.186)
with the first], then
as small compared
2ap In
Si
(6.187)
wherein we have used the definition c0/vlh
a'/ ''(vit>)
and assumed that
Moderator temperature
7\
T2>T,
Ej
(Neutron

Fig.
6.9
, resonance
energy
cross section in vicinity of resonance)
Doppler broadening of resonance.
Now ^"(iO represents the thermal<C 1 (a dilute system).

group cross section for a l/v dependence; therefore by (4.245),
(6.188)
and the temperature
derivative of Si is
ST
\4T
(6.189)
The temperature derivative of the integral S2 is obtained from (6.183).

has been shown that for many situations of practical interest, in
particular for systems wherein the fuel concentration is very low [o- *
in Eq. (6.177)], S2 is relatively insensitive to temperature variations.1
Thus, although a change in the temperature of the medium causes changes
in the cross section as given in (6.180) through a change in the distribution
of the nuclei represented by 2){(V,7V), these changes are such as to pre
serve the area under the individual resonances (see Fig. 6.9).
This
The over-all effect on the
phenomenon is called Doppler broadening.
resonance integral is small and will be neglected in the present calculation
of the temperature coefficient.
The principal contribution to the tem
perature coefficient of plh is derived, then, through the l/v portion of
al/\ Thus we set dS2/3T = 0, and by substituting (6.189) and (6.185)
It
for example, L. Dresner, CF56-12-147, Oak Ridge National

Feb. 2, 1956; and Nuclear Sri. and Eng., 1 (1), 68-79 (March, 1956).
1 See,
Laboratory.
sec. 6.6]
FEUMI AGE THEOKY
321
into (6.184) obtain
which reveals that the temperature coefficient of the resonance-escape

probability for a dilute homogeneous system is positive in so far as its
dependence on the fuel cross section is concerned.
6.6
Application
to a
Heavy-water-moderated Reactor
The Argonne Research Reactor: CP-6.

We demonstrate the
application of the Fermi age model by calculating the critical mass and
temperature coefficient for the Argonne National Laboratory research
This reactor has a large heavy-water-moderated core
reactor, the CP-5.
a.
and nearly all the fissions which occur are due to thermal neutrons.
It
is to be expected, then, that the Fermi age model will yield a good descrip
tion of the essential features of its neutron physics.
A detailed report on the CP-5 reactor is given in the Atomic Energy
document1 "Research Reactors"; a few of its principal char
acteristics are summarized here.
The reactor core proper is an upright
cylinder of D20 2 ft high and 2 ft in diameter in which are immersed

Commission
MTR-type aluminum fuel elements
The sides and

(see Fig. 5.28).
bottom of the core are reflected by 2 ft of D20, and this region in turn is
surrounded by 2 ft of graphite.
The top of the core is reflected by 2\ ft
of D20 alone.
The fuel elements, of which there are 16, consist of box
like arrays of fuel plates fabricated of a uranium-aluminum alloy, com
For com
posed of 17.5 per cent aluminum and 82.5 per cent uranium.
putational purposes we will assume that the volume fraction of aluminum

in the core is vM = 0.0688 and that of D20, vt>,0 = 0.914.
The reactor
was designed to produce 1,000 kw of heat, and at this power level the
temperature of the D20 is 49C.
b. Critical-mass Calculation.
Our immediate purpose is to compute
the mass of U236 required to make this reactor critical in the hot-clean
condition and to determine the average thermal flux in the core during
full-power operation.
If the core of this reactor were entirely bare, we could use Eq. (6.80)
directly to determine the critical fuel concentration.
Since the present
reflected, we require a modified relation which
will take into account the effect of the reflector.
We observed in Chap. 1
that the purpose of the reflector is to decrease the neutron escapes from
configuration
is completely
the core and thereby reduce the critical fuel concentration of the system.
Clearly, if we were to ignore the reflector altogether, our estimate of the
1
McGraw-Hill Book Company, Inc., New York,
1955.
REACTOR ANALYSIS
322
[chap. 6
critical fuel concentration could be much too large. We should like to

The scheme by which
attempt a somewhat more realistic calculation.
we propose to make such an estimate involves
the idea of an equivalent
Thus we will look for a bare cylindrical reactor which can

be made critical with the same fuel concentration as the actual reflected
reactor.
It is understood that this equivalent bare system will have the
same internal configuration of fuel elements and distribution of nonfuel
materials as the reflected system.
If there were some way to estimate the
size of the bare system, then Eqs. (5.204) and (6.80) could be used to
A suitable procedure, called the
compute the critical concentration.
reflector-savings method, is developed for this purpose in the general
theory of multiregion reactors (Chap. 8). This method allows one to
estimate the increase in the various dimensions of a bare multiplying
medium, over that of a specified reflected configuration, which would be
required if both systems were to be critical with the same fuel concentra
bare reactor.
tion.
These data, which were computed by H. Garabedian and are based on the
two-group theory, indicate that the reflector savings for a D20 reflector
In the CP-5 reactor the 62 cm of D20 on the
62 cm thick is about 23 cm.
sides and bottom of the core is supported by an additional 62 cm of graph
ite. This thickness of graphite reflector, if used alone, would correspond
We will attempt to account for the com

bined effect of this two-region reflector configuration by selecting an
intermediate value of 30 cm for the reflector savings.
That
we are
that
the
of
the
reflectors
presence
at
the
sides
and
estimating
D20-C
equivalent to 30 cm of core material in each of those
bottom of the core
directions.
The reflector-savings estimate for the top of the core 24 cm
this corresponds to a reflector of infinite thickness, according to Garabedian's calculations.
The equivalent bare cylinder obtained from these
Thus we assert
estimates
therefore 60 cm in radius and 115 cm high.
that a bare cylinder of this size composed of the same core materials
would be critical with the same U236 concentration as that required to
make the CP-5 critical.
is
is
is
is,
to a savings2 of about 50 cm.
On the basis of the selected bare-core configuration,
buckling to
"

In the present calculation we estimate the proportions of the equivalent

bare cylinder by utilizing reflector savings data obtained from more
advanced reactor physics models than yet considered in this work.1
be
"Reactor Handbook: Physics," p.

Ibid., p. 484.
494.
we estimate the
FERMI AGE THEORY
sec. 6.6]
For the proposed calculations we will also

r,
= 2.08
Tth
323
use the
= 130 cm1!
p,h
following data:
= t =
1
The fast nonleakage probability glh based on these data may

from Eq. (6.79) ; it is found to be gth = 0.7364.
be computed
We compute next the macroscopic cross sections of the core con

These calculations are based on the microscopic cross-section
stituents.
In the usual way we estimate first the effective
data of Table 2.2.
It can be shown that this
neutron temperature from Eq. (4.211).
We use this figure to compute the macroscopic

absorption cross sections for the D20 and the aluminum based on the
assumption of a l/v dependence for <ra. We obtain from Eqs. (2.54) and
temperature
is 330K.
(4.245)
2(d,o, = 0.00002311
cm-1
and
SÂ"
= 0.0009531 cm"1

The density of D20 at 49C has been taken as 1.08 g/cm3.

The absorption cross section of U23S required to support this system is
computed from (6.80), which, for the present purpose, is conveniently
written in the form
z,n
DB2 42(0)
-r*
>/epth0th
(6 191)
The
2J,0' refers to the nonfuel components D20 and aluminum and has the
value 0.0009762 cm-1.
Because of this relatively low nonfuel neutron
absorption cross section in the core, it is to be expected that the critical

fuel concentration will be low.
As a first approximation we take the
diffusion coefficient in (6.191) to be that of pure D20 at 49C.
If we
assume that the microscopic transport cross section of D20 is independent
of temperature around thermal energy, then the only effect on Str will
In that
in the density of the material.

(see Table 5.1)
be through changes
use the relation
D(49C)
It
= D(2Q?,loP^C)
= 0-890 cm
case we may
(6.192)
that S^' = 0.005775 cm-1. The thermal-group absorption

cross section for U236 at 330K is of1 = 561.4 barns.
For a core volume
of 179,000 cm", this yields a U236 critical mass of
follows
Mf
(f^
= 7183
Physics," p. 770, and corrected for reduced

X Taken from "Reactor Handbook:
D20 density at 49C (see "Reactor Handbook: Engineering," p. 30, for D20 density
versus temperature).
Taken from "Reactor Handbook: Physics," p. 770.
REACTOR ANALYSIS
324
[CHAP. 6
Provided that our guess for the reflector savings is reasonably close, and
that the U236 is available in the nearly pure form, the above mass estimate
should be a reliable figure for the U2*6 required to make the CP-5 critical
at the full power condition before the reactor becomes contaminated with
fission-product
In order to deter
fragments and various other poisons.
mine the fuel loading for the contaminated condition, we would require a
Suitable
knowledge of the fission-product buildup during a fuel cycle.
methods for computing these concentrations are developed in Sec. 9.6.
In the absence of such information we will estimate the critical loading for
the hot dirty condition from the published value of 1.15 kg for the critical
The corresponding value of the fission cross section is
mass.
s'
2'"fe), (nra)
"
0007681
cm"
This result may be used to compute the average thermal flux 4> in the
CP-5 when fully contaminated and at a full power. If P is the reactor
power and
-

BXfVm
the energy per fission, then

10"
=
(3.17
X 10-") (0.007681) (179,000)

= 2.29 X
1013
neutrons/cm2/sec
The project data in the "Reactor Handbook" gives 2.3 X 1013 neutrons/
cm2/sec for the average flux.
c. Estimate of Temperature Coefficient.
The role of the temperature
coefficient in determining the stability of a reactor was discussed in
Sec. 6.5a, and some formulas for estimating the value of this coefficient
for the various factors which appear in the multiplication constant were
It should be recognized that, in
developed in subsequent sections.
calculating the temperature response of these quantities, the coefficients
per se are temperature-dependent functions and may vary in both magni
tude and algebraic sign over the entire operating range of the reactor.
Thus a complete study of the temperature coefficient for a given system
must necessarily consider all possible operational conditions to which the
reactor may be subjected.
A particular situation which is always of
interest in the design of a new system is the initial startup condition ; this
situation may also
be encountered during later stages in the life of the

reactor, although under somewhat modified circumstances, whenever the
reactor is started up following a prolonged period of shutdown.
In each
of these situations the reactor is brought to criticality from a cold con

When steady state is achieved at some specified power (flux)
dition.
level, the temperature stabilizes at some higher steady value.

During
this entire operation the temperature coefficient, as well as the tempera
ture, is varying in time.
Some indication of the thermal sensitivity of
the reactor system can be obtained from an estimate of its temperature
coefficient.
FERMI AGE THEORY
sec. 6.6]
325
We demonstrate the calculation of a temperature coefficient by con

the initial startup condition of the CP-5 and show how this
coefficient may be used to predict the excess reactivity (or multiplication)
sidering
at room temperature.
we compute the
Specifically,
temperature coefficient of the hot clean reactor from the temperature
derivatives of its thermal utilization, fast nonleakage probability, thermal
of the
reactor
nonleakage probability, and resonance-escape probability.

The change
in k for a given change 5T in temperature is then easily computed from
the relation (6.142).
The data required for this calculation have already been obtained in
fn addition to these, we
main part in the preceding study of the CP-5.
will need the coefficients of thermal expansion for the various constituents
of the core.
These are
X 10-yC
X 10-yC
aM = 2.38
aF = 1.94
The linear coefficient for D20
(for a 30.5%
U-Al alloy)
obtain from the relation1 (6.149).
we
Its

value is
aDl0
-^pJf
= 9 876
For the calculation which follows
we
parameters listed below are independent

indicated values:
Bg,b
= 0.002353 cm-2
= 0.7364
X 10-VC
will assume that the reactor
of temperature
and have the
rlh = 130 cm2

plh = f = 1
For convenience in reference we list also the following results from the
for the CP-5 in Sec. 6.6b:
calculations
2 = 0.006751
pL = 0.2367
cm-1
L2 = 131.8 cm2
= 0.8554
pNL = 0.7633
A; = 1
These data apply to the critical reactor at 49C.

We compute first the temperature coefficient of the thermal utilization.
By (6.152),
j^ -
3(1
- /)(oo - a) ~
- /)(a,0 - aF)
coefficient
1 This
p. 30.
= 3.443
X 10-yC
of the fast nonleakage probability

from (6.169), which requires an estimate of alh. Now
The temperature
obtained
3(1
is
Da
value is based on the data given in the "Reactor Handbook:
Engineering,"
REACTOR ANALYSIS
326
For
[CHAP. 6
D,h we use 0.890 cm, and the denominator
relation
{2,
,2{" +
+ A12
ôS,'1"0'
we compute
U>2
The value for Dl0 we take as 0.527 (see Fig. 4.7) and for
Then, by (6.169),
these yield th = 3.974.
1
dg,h
D2
a'i>
^=BTa\T^~6ao
from the
2,
0.4252 cm"1;
_i_2
+
rKdf)
If we assume
that the reactor does not change size with temperature (i.e.,
the physical dimensions of the core tank do not change with temperature),
then
dT
and for a0 = 9.88
X 10-VC, and T

= 0
322K, we obtain
-H.95 x io-yc
The temperature coefficient of the thermal nonleakage probability is

The appropriate relation is given in (6.165)
computed next.
If
s;^ -*[-'-'> -*--]

again we assume that the reactor-core dimensions are temperature
the first derivative is zero, and the remaining terms yield
independent,
dpuL
-45.94 X 10-7C
The estimate for the temperature coefficient of the resonance-escape

probability is obtained from the relation (6.190) which is based on the
More precisely,
assumption that the fuel concentration is very low.
we require that in the resonance range [see (6.181)]
This condition is well satisfied for the concentrations

in the present calculation.
Here
of
U2" with which
we are concerned
and by (6.190)
= 3(a0
- aF) In
Pth
(6.195)
FERMI AGE THEORY
SEC. 6.6]
From the CP-5 data
we have p,h =
1 ;
327
thus the first term is zero.
For the
second term we compute
and
for
(fff) (0.005775)
<r>(kTJNF
S;1"0' = 0.4252
cm"1,
= 0.007072 cm"1
= 322K,
Tm
= 293K,
and
= 0.527, we obtain
^ = 8.20 x io-yc
These results we now use to estimate the temperature coefficient of the

CP-5 at 49C. Substitution into the general relation (6.141) yields
i~
1
The
excess
Hie
multiplication
8k
-49.25 X 10-VC
for the cold clean reactor (at 20C) is com
puted from (6.142); thus,

8k = k

for 8T
k =
I +
29C.
^ ^\
8T
= 0.01428
The multiplication constant at 20C

reactivity is [see Eq. (6.143)]
is
therefore
8k = 1.0143, and the
PROBLEMS
6.1
Consider an infinite nonmultiplying medium.
Within a spherical region of
radius R in this medium, go neutrons per unit time are being emitted uniformly in
Show that the slowing-down density q(r,r) is given by
space at age zero.
Sinh
where r is the radial distance from the center of the source sphere,
erf (f ) =
exp
Jo
e-(r+B)/4r
t(u), and
6.2
Isotropic sources of zero lethargy neutrons are distributed uniformly over a
spherical surface of radius a.
a. What is the slowing-down density due to this source distribution in an infinite
medium?
b. Sketch the slowing-down density q(r,u) as a function of r for various values of u.
6.3
A point source in an infinite medium is emitting neutrons of zero lethargy
isotropically at a constant rate in time. Assume that the slowing-down process may
be described by means of the Fermi age model once the first collision has occurred.
Thus at points of first collision, the neutrons enter a continuous slowing-down process.
REACTOR ANALYSIS
328
Find
[CHAP. 6
for the mean square distance traveled from the source point
Complete the analysis up to the
final step of integration over the radial distance variable.
6. If one were to consider the last scattering as a "free ride," what would be the
Note that neutrons will,
appropriate expression for the slowing-down density q(r,u)?
in general, reach lethargy u at some point r' which is different from r, and the prob
ability of a successful last flight to r must be taken into account.
(Leave this result
in integral form.)
It is desired to increase the thermal flux in the moderator region of a large
6.4
reactor in order to accommodate a certain experiment which is to be performed in
that region. This increase is to be achieved by introducing a very thin foil (of
thickness <) of fissionable material in the vicinity of the experiment.
Use the follow
ing model to examine this situation: (1) The moderator region is an infinite medium.
(2) There are no fission sources anywhere except in the foil, which may be treated as
an infinite plane source.
(3) In the absence of the foil, there is a uniform source S0
of thermal neutrons in this medium.
(4) The foil is entirely transparent to fast
neutrons, but has thermal cross sections 2^F> and 2/. (5) Neutrons from fissions in
the foil appear with zero lethargy and can be assumed to slow down according to
Fermi age theory.
a. Derive an expression for the thermal flux as a function of the distance from the
For this calculation use the kernel method and recall that if the diffusion equa
foil.
tion for thermal neutrons is written in the form
a.
an expression
Zih^h(*)
= S(x)
(P6.1)
is
-ZWttte)
then the solution
(P6.2)
is,
The cross sections in Eq. (P6.1) refer to the diffusion medium, and the source function
S(x) may be defined so as to include all the remaining neutron sources and sinks in the
medium.
The function *>(x;i') is the plane-source kernel and gives the flux at x due to
a unit source at x'.
The final result for #u,(x) is to include only the nuclear constants
of the system, So, and t.
b. Check the solution by showing that it reduces to the proper value when (1) x, the
distance from the foil, is large and (2) the foil is removed.
6.6
Consider a bare homogeneous spherical reactor with moderator and fuel cross
sections as specified in Prob. 4.8 and Prob. 5.12.
a. Find the critical radius, using Nmi/Ntae\ = 10,000.
Include the fast-leakage
correction only in so far as it affects the thermal fissions; that
neglect fast leakage
in computing the fast fissions.
Suppose that a material which has negligible absorption cross section for neu
bare critical reactor. If
trons, except at thermal,
spread uniformly throughout
the amount of this material used
such as to cause
change J2^h in the total thermalabsorption cross section, find the change Sk/k due to the change J2ih/2i".
c. Suppose, further, that the introduction of the absorber specified in part
to
be compensated for by an increase in the size of the reactor; that
more fuel-modera
tor mixture
added to the outer dimensions of the system.
Find the fractional
change Sk/k in terms of the change SR/R.
Neglect the extrapolation distance and
the effect of SR on the fast effect.
Show that by neglecting products of L*Bl and BVth
the result may be written
b
is,
L) SR
+L'fi
2fl'(rth
1
is
Sk
is
is
is
b.

to the point at which neutrons acquire lethargy u.
FERMI AGE THEORY
329
d. An absorber of the type described in 6 is added to the reactor in a such that

Nmoi/Ntb, 10,000 (where .V.ba = nuclear density of the added material).
The
Find the increase in
thermal-absorption cross section of this material is 1.0 barn.
the size of the reactor necessary to maintain criticality.
(Neglect the additional
absorber which would be spread over the additional volume of the reactor.)
By
what percentage does this increase the mass of the fuel?
6.6
a. Calculate the critical radius of the reactor of Prob. 6.5, using JVmod/jVfuei =
1,000.
This time include the fast-leakage factor in all terms of the critical equation.
It is suggested that the calculation for criticality be carried out by assuming various
values of B2 and computing the corresponding k. The B1 which yields k = 1 will
define the critical size.
b. Calculate (Sk/ST) for a reactor in which fission neutrons appear at one energy.
For this calculation assume the following:
This situation could
(1) The change in temperature affects the fuel density only.
occur in the case of a liquid fuel wherein the coefficient of expansion of the liquid is
much greater than that of the solid elements of the system (Sl/ST = 0 for all nonfuel
components; ( = characteristic length).
(2) All microscopic cross sections are temperature independent.

Z,(u)
scattering cross section of nonfuel

components.
c. Compute and plot the leakage, absorptions, and fissions per unit lethargy as a
function of lethargy.
Graph the thermal contributions by a block at uth.
(4) Si(u)
2j0)(u), where 2<0,(u)
CHAPTER
THEORY
TRANSPORT
7.1
Introduction
A complete description of the neutron population must specify simul

taneously the distribution in space, energy, time and direction of motion.
The primary objective of this chapter is to develop techniques and
computational
procedures for determining the angular distribution of
For this purpose we introduce the following function which

gives the most general description of the neutron density :
neutrons.
number of neutrons at time t in volume di

about r, whose speeds lie in dv about v and

n(r,t>,0,<) dr dv dOt = whose directions of motion
lie
in differential
solid
(velocity vectors)
angle
dO
(7.1)
about
direction Q
In
the usual way we define also a neutron flux:

4>(i,v,a,t)
Thus (r,v,(l,t)
n{x,v,a,t)
(7.2)
total track length traveled per unit time by the

neutrons described in (7.1).
In most of the analysis presented in this
chapter it will be convenient to work with this function rather than with
is the
the neutron density.

The general problems with which we will be concerned here deal with
Although this is not the most gen
homogeneous and isotropic media.
eral situation of interest in reactor analysis, the transport model developed
for these media has sufficient generality to treat a large variety of practical
Thus again we avoid questions of anisotropy and
reactor problems.
inhomogeneity, arguing that these are highly specialized situations which
are not amenable to the general methods of reactor analysis and, therefore
lie outside the scope of the present work.
We will lead up to the calculation of the neutron flux defined by (7.2)
by introducing first a somewhat more limited description of the neutron
The appropriate function
population, namely, the one-velocity model.
By omitting, for a time, the problem of describing the dis
is 0(r,Q,<).
tribution of neutrons in energy
space, we can focus attention on the

Thus our first objective
properties of the neutron motion.
of
the
picture
neutron population which gives its dis
is to construct a
directional
330
sec. 7.2]
TRANSPORT THEORY
331
tribution in both space and direction of motion.

This problem is treated
in Sec. 7.2.
The more general model of the transport theory which
accounts also for the distribution of the neutrons in energy space is
considered in Sec. 7.3.
The analytical method employed in both of these presentations involves
the use of expansions in spherical harmonics.
In each case the original
integrodifferential equation, which states the neutron-balance condition,
is reduced to an infinite set of coupled equations in the various harmonics
of the function <. By truncating the series expansion for the flux, this
further reduced to finite size (depending on the accuracy

and
it is shown that in first approximation the resulting equa
desired),
tions yield the diffusion theory and Fermi age models.
set can be
A second general technique for treating the angular distribution of the

flux is presented in Sec. 7.4.
This is the method of integral
Solutions for the directed flux <(r,Q) are derived on the
equations.
basis of the one-velocity model for various media of infinite extent.
The
application of these solutions for the infinite medium to systems of
finite size is demonstrated in the case of the homogeneous slab and
neutron

sphere.
Section 7.5 deals with the application
of the spherical-harmonics
method developed for the one-velocity model of Sec. 7.2 to two situations
of interest to reactor analysis.
The emphasis is placed on various
aspects of the computational procedure involved in such calculations and
on the selection of suitable boundary conditions in terms of the directed
flux.
The last section is devoted to the application of the transport model
of Sec. 7.3, which includes the energy dependence of the neutron flux, to
the calculation of the age in hydrogenous mixtures.
This is the so-called
Fermi-Marshak
metal-water
method.
Computations
are
performed
for various
mixtures of practical importance
results are compared
to reactor design, and the

with several other techniques for estimating the
neutron age.
7.2
The One-velocity Transport Equation
a. Neutron Balance.
We begin our study of the transport model by
introducing the one-velocity approximation.
For the analysis of this
section we use in place of the more general description

following definition of the neutron flux:
m
*(r,O,0
(T,v,il,l) the
n(r,a,t)
= total track length per unit time at time I in unit volume

about point r of y-speed neutrons whose directions of motion
lie in the unit solid angle about Q
The direction
(7.3)
(unit vector) Q is specified by means of the co-latitude
332
ANALYSIS
REACTOR
[CHAP. 7
is
\f>
angle 6 and the aximuth angle

The velocity vector
(see Fig. 7.1).
which describes the neutron motion
given in terms of the neutron speed
through the relation
(7.4)
important to recognize that the function denned by (7.3)

to our previous description
is
It
is
v = vQ
one-velocity
flux
(5.2)
4>(T,a,t)
related
of the
through
the
*(r,0
fa
equation
da
Thus our present definition simply

selects from the entire population of
of speed
those which are
particular direction.
moving in
With the foregoing definitions, we
may proceed to construct, by con
a
neutrons
Fig.
sidering the neutron balance, an equa

tion governing the neutron density.
it
Consider, then, the set of neutrons of

at time which have velocity
volume
dr
about
in
the
element
speed
From conservation requirements
must
vectors that lie in da about Q.
be true that
number of neutrons of
Number of neutrons
moving in direction a
which appear per unit
time from sources in dr
II
direction a gained per

unit time in dx from
scattering collisions
which scatter the
neutron
from all direc
tions a' to a
- Ill
net number of neutrons

moving in direction Q
direction Q which are
that are lost per unit
IV
time by leakage
through the bound
aries of dt
removed per unit time

from dt by being scat
tered into a new
direction
removed per unit time

from dr by absorption
collisions
direction a which are

Direction vector in labora

tory system of coordinates.
7.1
VI
change per unit time in

net number of neutrons
of direction a in dr
(7.5)
TRANSPORT THEORY
sec. 7.2]
333
The first term

Consider now the individual terms in this equation.
describes the rate at which the one-velocity neutrons with direction in
dQ appear from sources in dr. Call this function S; then, the first term
of (7.5) is given by
S(r,Q,t) dr dQ
(7.6)
The second term of (7.5) gives the rate at which neutrons arrive in the
differential element dr dQ by being scattered into the direction about Q
To determine this quantity we compute first
from all other directions.
the number of neutrons of direction Q' that are scattered into the direc
tion Q by collisions which occur in dr at time t. The integral of this
quantity over all directions Q' yields the required term. Now,
S,0(r,O',O dr dQ'
(7.7)
gives the total number of scattering collisions in dr per unit time experi
Of this number a
enced by neutrons with directions in dSi' about Q'.
moving in
from
collisions
/(Q;Q')
dQ
will
these
emerge
fraction
certain
Formally, we define the function
directions that lie in dQ about Q.
/(Q;Q') as the frequency function for the change of direction from Q' to
Q.
Note that
-Q') dQ
[cf.
Eq. (2.11)]
probability that a neutron with an initial velocity

directed along Q', when scattered, emerges from
the collision with a velocity vector which lies in the
_ g.
solid angle dQ about Q
Thus the product of (7.7) and (7.8) gives the number of neutrons initially
directed along Q' which are scattered into the direction about Q while
occupying the volume dr. The integral of this product over all Q' gives
For
the total number of such contributions to dr dQ from all directions.
the second term of (7.5) we have, then,
2>(r,Q',<)
/(Q ;Q') dQ' do dr
The third term of (7.5), which represents the net
(7.9)
losses from dr dQ due
to the leakage of neutrons through the boundaries of dr, may be derived

from a detailed count of the neutron passages through the surfaces of the
This element is
volume element dr = dx dy dz shown in Fig. 7.2.
selected for computational convenience such that its surfaces are oriented
parallel to a Cartesian coordinate system erected in the space in question.
Let us consider first the face dx dy at station z. Now the number of
neutrons with velocity vectors directed along Q (in dQ) which will pass
through this face in unit time is the number of such neutrons included in
the volume which would be swept out by translating the face dx dy a
distance equal to v in the negative Q direction.
(This swept region of
BEACTOR ANALYSIS
334
[CHAP. 7
space is shown in Fig. 7.2.)

Thus the number of neutrons entering
dx dy dz through the face dx dy at z per unit time with directions in dU
about 1 is
n(x,y,z,a,t) da(v k) dx dy
where k is the unit vector in the z direction.
(7.10)
Similarly, the number

x(V
Fig.
7.2
Volume element dx dy
of neutrons with direction Q leaving the volume element through

dx dy face at z + dz per unit time is given by
n(x,y,z + dz,Q,t) di(v
If dz
is taken sufficiently
the
k) dx dy
small, a good approximation
to n(x,y,z + dz,Q,t)
is
n(x,y,z + dz,a,t)
the net loss from
through the dx dy faces is
Therefore
n(x,y,z,a,t) dx dy
~ n(x,y,z,a,t)
+ n(x,y,z,Cl,t)
dz
di per unit time of neutrons of direction Q
dz dSi (v
k) =
0(r,Q,/)(Q
k) dx da
(7.11)
Analogous
where we have used (7.4) and the usual definition for the flux.
considerations for the other two faces lead to the following result: The
sec. 7.2]
TRANSPORT THEORY
335
4>(r,a,t)(a
J k)
*(r,>0(o
Kr,o,0(o
j)
j-x
net loss per unit time from dr of neutrons of direction Q in dQ by leakage

(transport) across all faces is equal to
dr
da
(7.12)
time by absorption; this
is
is
where and are the unit vectors in the x and

directions, respectively,
and the definition for the gradient operator
has been introduced.
Term IV in (7.5)
the number of neutrons lost from dr dQ per unit
given by
2a(r,a,t) dQdr
(7.13)
is
Term
gives the rate at which neutrons are scattered out of dr dQ into
new directions.
This term may be represented by the form (7.7) with
Q' replaced by Q.
The last term VI
the time-rate of change of the neutron density in
This may
jt
be
written
n(i,Q,t) dQdr
=-vjt (r,a,t)
(7.14)
(7.14) into the verbal
V<Kr,O,0 + 2<*(r,0,/)
= -S(r,O,0
*(r,0,<) + Q
The substitution of the results (7.6) through

statement (7.5) yields the equation
dQ dr
f(Q;Q') dQ'
2.<Kr,G', f(Q;Q') dQ' (7.15)*
(7.15)*
a
where we have combined the losses from scattering and absorption into
single term, using the total cross section 2(, and have canceled out the
This integrodifferential equation for the neutron
common factor di dQ.
flux
In the general
usually called the one-velocity transport equation.
is
is
is
very difficult to work with Eq. (7.15); however, in many cases

sufficiently weak
the angular dependence of the quantities involved
that certain approximations may be made which reduce (7.15) to more
In the work which follows we demonstrate
tractable set of equations.
one procedure for doing this which involves the use of expansions in
case
it
spherical harmonics.
Function for Scattering Collision. Before proceeding

will be convenient to exhibit some of the properties
of the scattering function defined in (7.8).
useful geometric device for
as the probability that the
describing this function consists of defining
velocity vector of scattered neutron pierce
particular area on the sur
a
it A
it
b. Frequency
with the analysis

di dQ.
REACTOR ANALYSIS
336
[CHAP. 7
Thus if
face of a sphere drawn about the collision point (see Fig. 7.3).
/(Q;Q')
dl
is the
the
of
the
then
the collision occurs at
center
sphere,
probability that the vector v passes through the area dCX on its surface.
x
Direction of
incident neutron
Unit sphere
Fig.
7.3
Scattering angle in laboratory system of coordinates.

From the definition of
a frequency
function, it follows that
/o/(Q;Q')dQ
is homogeneous
of this assumption is that

the directions LI and
/ can
be a function only of the angle between

We call this angle 0O and define
cos
If
(7.16)
assumptions in this analysis is that the medium

An immediate consequence
and isotropic.
One of the fundamental
in question
we introduce
0o
no
il
Q'
also the function

that
a neutron
moving
in
direction^
Q', when scattered, emerges with a new direction I (7.17)
(probability
whose cosine lies in duo about /i0
then it follows that
/(Q;Q') -gjUGi.jQ')
(7.18)*
Thus the general function

may be replaced in this treatment by the
function t) which applies specifically to the isotropic medium.
We use
this formulation throughout the remainder of this analysis.
c. Expansion in Spherical Harmonics.
The transport equation (7.15)
may be reduced to a set of differential equations by introducing a suitable
series representation for each of the functions which are involved.
In
this formulation we make the expansions in terms of an orthonormal
set of spherical harmonics.
We begin with the frequency function t)
TRANSPORT THEORY
SEC. 7.2]
337
defined in (7.17).
Since this function is defined by means of the single
variable /io, a suitable series representation can be given in terms of the
Legendre polynomials1 P,(no). Thus
PG.0
;0')
>
(7. 19)
^Pi|(mo)
,-0
J*t
The coefficients l), are obtained in the usual way by applying the orthogo
nality properties of the Legendre polynomials.
It follows that
U, =
PnM
(mo;Q')
(7.20)
duo
6.
^.
\i
is
0'
P\.
i
functions
\f/'
= cos
denote the arguments of the direction Q', and let
completely defined by the quantities
and
The
n, \f/',
Legendre polynomial P(mo) may be written in terms of these variables
by the application of the addition theorem2 which involves the associated
Let
and
Clearly, ^0
JL
used in
(7-21)
this treatment follows
P,(M)
(7.22)
<
/3
and
?7
The associated
are related to those with positive index
0
>
/3
rj
This definition applies for

functions with negative index
- sin"
f*()
P?(m'} e*(*~*
P?(/i)
j^TWi
The definition of the associated functions

that given by Sommerfeld:3
< j;.
P$()
(01
ps
intervals given above.

= P,(m)
=0
23)
Note that
(7.24)
for
>
(7.25)
P(M)
1)0
/3
and
-q
which applies for the
/8
p-e =
Some of the associated functions of lower index are given below:
3(i
PJ(m) = 3M(1
P(M) = *(3m2
PrH**)
m2)*
Pr'GO
=
=
PJ(m) =
-Ml
-
i(l
m2)
M2)
-2Pr'(M)
(1
1)
PJGO
(7.26)
m2)1
Refer to Eq. (4.21) and the discussion for definition of the lower-degree poly'
nomials.
A. Sommerfeld, "Partial Differential Equations in Physics," p. 134, Academic
Press, Inc., New York, 1949.
Ibid., pp. 126-129.
'

P,M
REACTOR ANALYSIS
338
[CHAP. 7
The substitution of Eq. (7.21) into (7.19), followed by the application

of Eq. (7.18), yields a series representation for the scattering frequency
function /.
/(0;Q')
JL
|^Jj{ P5W W) e*<*-*"
(7.27)
We follow this general procedure to obtain suitable expansions for the

For these functions, however, it
neutron flux and the source function S.
will be necessary to develop the expansions in the complete set of spheri
In this work we define these to
cal harmonics.
be the
orthonormal set:
y*(Q) ^ F-(M,*) ^ ff-P?Gx)e**

with n = 0,
function is
1,
and
rr(O)
The
n<mThe complex conjugate of this

complex
- H;P;(j*)er**
applies for both positive

are defined by the relation
which

(7.28)
(-l)-F--(H)
and negative
Ran +
[
lKn-
47r(n
conjugate of this
m.
(7.29)
The coefficients
//"
m)!!*
+ m)!
Note that
=
(n+jn)!
(n
to)!
The orthogonality property of the spherical harmonics
fQ
Yl'(Ck)
Y:(Q) da =
rf/i
f* Yl'in,*)
is stated
YZ(nM df
5i5"n
(7.32)
The symbol 8 denotes the Kronecker delta and is zero if n ^ a and unity
if n = ex.
The flux and the source may be expanded in terms of the functions
Thus
(7.28).
K(T,t)YnW
(7.33)
^2
S?(r,0F:(Q)
(7.34)
*(r,Q,0 =
n =0 m = n
S(r,O,0
n 0 m n
TRANSPORT THEORT
SEC. 7.2]
with the coefficients to
339
be computed from the integrals
(r,0
fa
*(r,Q,0
- /Q S(r,O,0
ST(r,0
F:(Q) da
(7.35)
F*(a) da
(7.36)
An inspection of the series for /(Q;l') given in (7.27) reveals that by

regrouping the various factors behind the summation signs one can express
this series in terms of the spherical harmonics (7.28). It is easy to show
that (7.27) may then be written in the form
/(OjC)
,_o 0- -n
W)
^F?(Q)
(7-37)
The transport equation (7.15) may now be expressed in terms of these

The substitution of (7.33), (7.34), and (7.37) into (7.15)
expansions.
yields
f"(q)
n,m
r?(o) y',(Q,) do'
be reduced to a simpler
(r,0
2,
2.j,n)
+Q
C(r,0
- ST(r,0
^(O)
The substitu
2.
where we have applied the orthogonality property (7.32).

tion of this result into (7.38) yields, upon rearranging,
form; thus
rS(Q)
(7-38)
tj
(J
F"(o,)
n,m
c(r'
ij
s"(T't}
n,m
The integral in this equation may
n,m
2>
c(r' r"(Q)
v +
2')
(7. 39)
*2(r,<)
This infinite set may be separated into coupled equations in the

harmonics 0" and 6' by multiplying through by F* and integrating over
all 1 with the aid of the orthogonality property (7.32). For the first,
third, fourth, and fifth terms we obtain immediately

G jt
2,^(r,<)
2.t,a<tf
(r,0
- Si(r,l)
(7.40)
340
ANALYSIS
REACTOR
The leakage term may
[CHAP. 7
by expanding the operator Q
be evaluated
dx
cos
dy
sin
sin
j
e'*(m_U sin
("
gOîOO
*r(r'
(7l42)
- w)PT+i(f)
43)
is
multiplied by Y*'(Q) and integrated over O,

the first term of (7.42)
can be shown that the result
is
The remaining terms in (7.42) may
g)"|*
^_
+1)
to
be reduced in the same way.
^
%+1
For
- - l)(m
(n
(2n+l)sine[(w
^-
3)(2a
the recursion relations
these calculations
gK
(2a
!"(+!L
/.*<-*(=<)
n,m
it
If
("
mP(m)
)P^z{(m)]
(7.45)
By this procedure the contributions of all the terms in

these results are combined with (7.40),
(7.42) may be evaluated, and
can be shown that the set of equations (7.39) reduces to the form
needed.
it
if
are

where we have used the recursion relation1
ew
(7
Tz)
- ^îM
n(fl)
YjH? \[(n+
n,m
m^-^
n,m
functions of
v(r,<)
fl)
and by using the exponential forms for the trigonometric

Then, the second term in (7.39) yields
r:(o) a
(7.41)
dz
4>.
\[<
d
V = ii + sin
\f>
into the form
Ibid., p.
132.
TRANSPORT THEORY
sec. 7.2]
341
'
*fcl(r,0
(7.46)*
The index in this set takes on all integer values a = 0, 1 , 2, . . . , and

a to +a.
For cases of practical importance
/3 takes all values from
Eqs. (7.46) may be reduced to a finite set of simultaneous equations by
taking only the first few values of a, say, a = 1, 2,
5.
Several
applications of Eqs. (7.46) to the calculation of the angular distribution of
...
neutrons
in situations
of interest
to reactor analysis
are presented in
Sec. 7.5.
A first-order approximation of Eqs.

removes
the
effectively
angular
dependence of the neutron flux.
(7.46)
However, this approximation is extremely useful and, in fact, forms the
basis of the diffusion-theory model developed in Chap. 5. For this
d.
Diffusion Approximation.
reason it will be instructive to examine the details of the approximation

and to demonstrate the rigorous derivation of the diffusion equation
(5.21).
It will
that the diffusion approximation is based upon the

that the neutron-flux distribution is nearly isotropic (see
This assumption is equivalent, analytically, to the
Sees. 5.1b and 5. Id).
requirement that the series expansion for the flux (7.33) be well repre
be recalled
assumption
sented by the first two terms.

Consider, therefore, the set of Eqs. (7.46)
=
for values of a
This means that in the series expansion (7.33)
0,1.
all coefficients of
we retain only the terms involving <{j, 4>r\ 1, and
In order that our analysis
higher indices being taken as identically zero.
\,
will also require that the source function S be isotropic

Note that to assume otherwise
to imply
except 5.
that the sources have strong directional properties, and
would follow
that the angular distribution of the neutron population in their immediate
vicinity would also be nonisotropic.
With these conditions imposed on Eqs. (7.46),
clear that this infinite
is
it
it
is
all
0,
be consistent, we
so that
<S

2i(r,<)
REACTOR ANALYSIS
342
[CHAP. 7
(?-48)
!)
(7-47)
<
(I
s<)
<5)
*
*?
-^(S-'5)* + ?i(
G
S<)
-(^ +
**
coupled set reduces to the four equations
*s
s)
'
(7-50)
are obtained from
<*mo
=
=
^(";Q') P'(")
>0o;o0
^mo
(Mo;0')/odMo
(7.51)
|\
j(mo;0') PoCmo)
7o
/_\ /_\
These are
(7.20).
/-i
(7.52)
no represents the average value of the cosine of the scattering

We demonstrate first the reduction of Eqs. (7.47)
angle [cf. Eq. (4.30)].
through (7.50) to the familiar diffusion equation for the case of system
at steady state. Thus the derivatives d/dt
and
follows immediately
i, and
yield
ft1,
for
the
functions
that Eqs. (7.48), (7.49), and (7.50)
it
0,
where
the solutions
(7'54)
<7'55)
notational convenience we introduce

As
D
the symbol
to define the ratio
'--vÛa^-'i)*8
=
3(2,
(7.56)*
Xnn)
is
it
the
We call this quantity the diffusion coefficient and recognize that
function previously given by Eq. (5.44). The substitution of the results
single second-order partial dif
(7.53) through (7.55) into (7.47) yields
we apply also the definition (7.56), then for
ferential equation for
If

tj
The first two coefficients of the scattering function

the relation
(7-49)
"
(I
s^> =
**
2l)
"
(i
-(?i + *) + **-:?ss*
TRANSPORT THEORY
sec. 7.2]
the steady-state condition the appropriate
343
equation is found to be
-V-^
-lla-TyDTy*S-izD+=
<Sjj
(7.57)
may be replaced by the "total" quantities <(r)
and
inasmuch as each represents the isotropic component of their
We compute, therefore, the total functions <p(i)
respective functions.
For this computation we
and <S(r) from relations of the form (7.5).
require appropriate expansions for the steady-state functions; thus, the
relations (7.33) and (7.34) are replaced by the set
The functions
*(r,Q)
5(r,Q)
n ^
<S(r),
and
W
(7.58)
SSW y?(0)
speed
regardless of
S" (r)
S(t,Q) dQ
Y,
and the total number &(r) of neutrons of speed
S(r) =
<ftJS|
(7. 59)
produced by sources
(1)
is
+(r,a) da
+(r)
(7-60)
where we have used the definition
of the spherical harmonics (7.28) and

the orthogonality property (7.32).
The substitution of these results into (7.57) yields
-V DV(r)
+ 20(r)
S(t)
(7.61)*
which in the case of a uniform medium reduces to the usual form (5.22).
e. Net Current.
The net current relation (5.19) introduced in the
development of the diffusion theory may be derived from the first-order
solution to the transport equation.
following definition:
For this purpose
we introduce
is
net number of neutrons which pass per unit time through

unit area at which
so oriented that the normal to its
in the direction
surface
This number can
the
(7.62)
is
J.(i)
be
computed
from
the neutron
density
n(r,il)
di,
which gives the number of neutrons per unit volume around

with
velocities which lie in dQ about Q. The number of these neutrons which

is
Then, the total track length 4>(r) of all neutrons of

given by
their direction of motion,
v,
n,m
344
ANALYSIS
REACTOR
[CHAP. 7
pass through the area dA shown in Fig. 7.4 per unit time is given by
n(r,Q) da](a
[v
dA = <Kr,0)(ti
s)
s)
dA da
(7.63)
where s is the direction normal to dA . This relation is the contribution

to the net current through dA (with normal s) of neutrons with velocities
in dQ about a.
The net current through dA
of neutrons moving in all directions J.(t) dA
is obtained by integrating (7.63) over all a.
dA
dA
f
JO
{s
0(r,O)(O
s)
da
*(r,0)0
(7.64)
da]]
The expression at the extreme right
is
[fa
J,(r) dA
of the
J(r)
Neutron veloc
7.4
ity and the surface dA.
(7.65)
Thus the net current through dA of normal
Fig.
regarded as the component in the

of
vector J(r) which we shall call
Eqs.
define
as
and
the vector net current
[cf.
(7.64) and (7.65)]
may
>(r,Q)a da
(7.66)*
sin
(7.67)
sin
+
k
cos
sin
cos
expansion
and using the spherical-harmonics
for
4>(r,Q)
j(r)
fa
7.3)
given in (7.58)
a
(see Fig.
evaluated by noting that Q may be written in the form

9
is
The integral
be
direction
J(r)
YZ(a)a da
(7.68)
FT.T71
f;(q)
[Y\W
If
sin
sin ecos^ =
sin
(7.69)
tzJjilYlM
YT\a)]
+
cos
We note that
(7.70)
Yl\a)]
(7.71)
and the integrations performed with

of the F,
properties
may be shown that
the aid of the orthogonality
are used in (7.68)
J(r)
it
these relations

J. {!)
form
first-order
27/1
MM
estimate
*rx(r)]j
Mtt
of this function may
be
+ *rl(r)]k
obtained
(7.72)
by using
sec. 7.2]
TRANSPORT THEORY
345
expressions of the form1 (7.53) through (7.55) for

result is [with the application of (7.30) for //"]
J(r)
- D V*
[i
yx *8(r) +
As in the case of the diffusion equation

written in terms of the total flux (r).
that HI = (4r)-*, it follows that
J(r)
yy
4>\,
*8(r) + k
fc1, and
*g(r)
The
<}.
(7.61), this expression may be

we use, then, (7.59) and note
If
-DV*(r)
which is the relation (5.19).

f. One-dimensional Problems.
The general results obtained in Sec.

7. 2d reduce to much simpler form in the event that the flux and the source
In these situations the system
are functions of a single space variable x.
must have azimuthal symmetry, and the angular distribution of the flux
and the source may be completely specified by a single angle, namely, the
angle 6 which the neutron direction of motion makes with the x axis.
If cos 6 = then it is convenient to define the flux in these systems in the
.
M
total track length per unit volume per unit time

around x of all neutrons whose directions of motion
reSpect 0 tne x axjs are such that the cosines
of these directions lie in the conical shell dn about
In our selection of a suitable series expansion for the scattering function

since both t) and
X) we found it convenient to use the Legendre functions,
the Pn are defined by a single variable mo over the same interval (1,1).
It is clear that the flux defined by (7.73) has these same properties in so
Thus we select as a suitable series
far as the variable n is concerned.
for
form
representation
[cf. Eq. (7.19)]
(7.73) the
+(*lM,o
-y
n-0
(7-?4)
(2-!L^)
The coefficients #B(x,0 are obtained by the application of the orthogo

In the usual way we
nality properties of the Legendre polynomials.
*.(x,t)
jlx
compute
4>{x,n,t)
Pn{n) dn
(7.75)
These coefficients are related to the quantities %(x,t) used in the general
expansion of the function (i,Q,t) [see Eq. (7.35)].
spherical-harmonics
is
identical to the time-depend

Note that the analysis of the steady-state problem
ent case considered in Sec. 7.2d, except that all coefficients used in the expansions are
functions only of r.
1

following manner:
REACTOR ANALYSIS
346
This
[CHAP. 7
First, we observe that

(x,fi,t) is obtained from (r,l,t) by integrating the latter function over
all azimuthal directions ^, since (x,n,t) gives the track length of all
is
relationship
demonstrated.
easily
(x,a,t) #
4>(x,n,t)
(*,) HnP? 00
n,m
e**
2t
fQ2'
neutrons whose directions of motion are at an angle 6 to the x axis regard

less of their azimuthal orientation.
We have, then,
*!(*.0 W-GO
(7.76)
n=0
(7.33) for 4>(x,Q,t), the definition

F, and recognized that P(m) = PG0, and that the
zero except when m = 0.
If we compare the result
where we have used the expansion
it
is
if'
(7.28) for the

integral over
(7.76) with the form (7.74),
follows that
(7-77)
<Pn=i
set of coupled differential
equations analogous to the
relations (7.46) may be developed for the one-dimensional formulation of

the transport equation by performing a series of operations on the general
result (7.39).
First we integrate this equation over all so as to collect
the contributions to the neutron population of all processes which yield
whose direction cosines are
irrespective of their azimuthal
This step
orientation.
analogous to the integration performed in
If
we
use
the
fact
that
for one-dimensional system
(7.76).
a
is
neutrons
n-0
\p
2,
integration
("8)
of (7.39) yields
- 2.A tf(*,0 -
HIPM
may be shown that the
0|
then
it
n-v-n
(7.79)
S(x,a,t)
2t
S(x,m,0 ^
is

which
con
description of the source
sistent with the requirements of the present system, namely [see Eq.
(7.34)],
j2J
sn(x,t)
ffJP.GO
n-0
and the coefficients
and
.Sn
30
(2^j)
n-0
P"^}

An appropriate
are related exactly as
and
<.
(7-80)
TRANSPORT THEORY
sec. 7.2]
347
n-0
(sir)
"Fn(M) P(M)
(7'81)
/"I
*2
(am)
I)
2<
*-)
(lit
The second step in the development is to multiply (7.79) through by

1 <
We obtain:
n < 1.
Pa(n) and integrate over the interval
0,
is
where we have used again the orthogonality properties of the Pn functions

and have suppressed, for the moment, the argument of <#J. The integral
evaluated by using the recursion relation (7.43); for the present case
m =
and this equation reduces to
=
mP.(m)
nPn-M
1)Pn+lM
(7.82)
(2a
l)(2a +
(stt)
2g_,
ax
0a-1 ~
3)(2a+
ax
(asrr)
-^
1)
(2 +
HM
2(
1)
21
2-')
This result may be written in terms of the coefficients

and <S by
applying the relation (7.77), and the corresponding one for S. Thus
- -37
of
<fc,(x,0
~ 2,0o(x,<) + (*,<) +
2t)a</)a(x,0
....
is
This set of coupled equations corresponds to the

where o = 0,l,
relations (7.46), and the complete description of the flux and source
to be obtained from the expansions (7.74) and (7.80). As in the more
general formulation, estimates of the spatial and angular distribution of
neutron population can be obtained to various degrees of

trun
accuracy depending upon the value of a at which the set (7.84)
method
cated.
The sample applications of the spherical-harmonics
one-velocity
is
mentioned previously will be performed, using the above set of equations.

We conclude the present discussion of the one-dimensional problems in
1
;
is
transport theory with some remarks about the diffusion approximation

and the relation of the partial currents introduced in Chap.
to the
For this purpose we require the steady-state
directed, or angular, flux.
first-order approximations to the Eqs. (7.84). These are obtained by
based on
retaining only terms for a =
thus, the diffusion theory
>
in
are negligibly small
comparison
the requirement that terms in a
0,

The substitution of (7.82) into (7.81) yields, upon integration,
ANALYSIS
REACTOR
348
to those for a < 2.
The appropriate
[CHAP. 7
set of equations are [using
<S
= 0
for a > 0 and Eqs. (7.51) and (7.52)]
*,(*) + 2o(z) = S9(x)
I L *o(x)
+ (Z' ~
(7.85)
= 0
(7-86)
which when combined yield
-D ^
>(*)
+ 2a*o(x)
S0(x)
(7.87)
*,(*)
-D^*o(i)
(7.88)
where we have used (7.56) and assumed that the medium is homogeneous.
Now, from (7.75),
=
i(x)
f1,,
- (*) - oftotal
*.
track length
al[ neutronsg
Mixdn
_ .
(7.89)
(7.90)
is
it
i)
We can identify 0i from the results of our neutron-leakage calculation

for Eq. (7.5). If we note that 4>{x,n) = n(x,n)(v
and bear in mind
the definition of (7.73),
clear that
is
number of neutrons crossing unit area per unit ime

in a; direc (witn area or'cnted so that its normal
has
in
a
with
x
direction
direction whose angle
tion)
cosine in dn about
is
the net neutron current in the x direction at x.
(diffusion)
*(*,/) =
J(x)
[Ho(x)
from (7.74) in (7.92),
(x,n)n
dn +
(x,ri(i
(7.92)
approximation
iMx)
and use the first-order
Jt
J{x) = *,(*)
j*
if
Thus the integral (7.90)

Moreover,
we write
fi
<b(x
'
|m*i(x)
(7.93)*
we obtain
i*i(x)]
- [io(x) - i*i(x)] - j+(x) - j-{x)
(7.94)

j+(x)
i_(x) =
J*

/_\
*.(*)
4>{x,n)n dn =
j~l
*(x,p)MdM
\,{x)
=
\4*(x)
Ifaix)
- Mx)
(7.95)
(7.96)
sec. 7.2]
If
TRANSPORT THEORY
we note that
349
D',
then it is seen that these results

for the partial currents are identical to the relations (5.49).
A general set of boundary conditions were
g. Boundary Conditions.
described in Sec. 5.1e, and it was pointed out that all these conditions
4>i
D'0
could not be satisfied in detail when used in conjunction with the dif
fusion theory.
In addition, it was mentioned that these conditions could
only when applied with the more precise transport-theory model.

As a matter of fact, not even this is possible, from a practical viewpoint.
To illustrate this point, consider the vacuum-interface condition stated
in (5.54). An analytical statement of this condition for the one-dimen
be met
sional model of Sec. 7.2f would be

4>(0>)
=0
ix
< 0
(7.97)*
sion for the directed flux must be obtained from a truncated set of
equations based on (7.84), and this solution will necessarily involve a
finite number of arbitrary constants.
Thus it is not possible to satisfy
In
difficulty is circumvented by
actual
this
rigorously.
practice
(7.97)
introducing an alternate
interesting to note that the integral for m =

<K0,m)/x d =
5,
(7.98)
yields the relation

(7.99)
in fact the partial

from (7.96) that this integral
current in the direction of negative x, i.e., the return current from the
vacuum.
zero net cur
Thus, condition (7.99) requires that there be
from
should
rent
the vacuum; however,
be recognized that inward
it
is
We note immediately
3,
m =
1,
dM
(0,riPm(n)
It
is
j_x
set of conditions,1 which although not rigorous

are at least consistent with the statement (7.97).
One such set was
suggested by Marshak, and these require that the following integral of
the flux vanish at a vacuum interface:
is
currents from certain directions are still possible, provided of course that
the gains from these be exactly balanced by flows outward from other
This condition on the inward current
precisely the state
directions.
is
ment of (5.57a) used in the diffusion theory.

The vacuum-interface
that
the
condition based on the requirement
integral (7.98) for odd m
vanish at the surface
convenient extension, then, of the diffusionA. M. Weinberg and L. C. Noderer, "Theory of Neutron Chain Reaction," vol.
pp. 11-18, Oak Ridge National Laboratory, CF 51-5-98, May 15, 1951.
I,
theory relation.
1

where for convenience we have specified the vacuum boundary to be at

x = 0. According to (7.97) no neutrons may return from the directions
1 <
Now the variable li takes on all
n < 0, i.e., from the vacuum.
values in the indicated interval; but, in actual computation, the expres
REACTOR ANALYSIS
350
[CHAP. 7
The rigorous statement for the interface condition between two dis
similar media is that the directed flux for each specified angle be contin
If
uous through the interface boundary.
II,
we denote these media
and
then we require that
0,(R,a) = <MR,G)
for all Q
(7.100)*
R refers to the coordinates of
a specific point on the

interface boundary, and Q is any general direction.
Thus (7.100) states
that there be the same number of neutrons per unit volume around
point R in medium I that are moving in direction Q as there are per
unit volume around R in medium II. In the event that the flux is a
where the vector
function of
a single space
variable x, (7.100) takes the form
<fr(x,/0 =
-1
u(x0!li)
<
<
(7.101)
A numerical example of
x0 denotes the position of the interface.
the application of this boundary condition for the P3 approximation1 is
given in Sec. 7.5.
h. Time-dependent Equation in First-order Approximation.
We
conclude the present section with a discussion of the approximations

where
involved in deriving
a first-order solution to the transport equation which

As in the case of the steady-state results
includes the time dependence.
obtained in Sec. 7. 2d, time-dependent relations (7.47) through (7.50) may
also be reduced to a single second-order
partial differential
equation
involving the total track length (t,t).

However, the essential features
of this equation may be determined from a study of the one-dimensional
system developed in Sec. 7.2f.
Consider, therefore, a system wherein
the spatial distribution of both the flux and the source can be specified
by a single variable x and the angular distribution by the variable n.
The appropriate relations for the first-order approximation are obtained
Again we use Sa = 0
from (7.84) by taking only the terms in a = 0, 1.
for a > 0 and for convenience assume that <S is not a function of time;
then,
- -vjt fa(x,t) - 2ao(z,t)
+ So(x)
*!(*,<)
(7.102)
This pair of equations may be reduced to a single second-order equation

in
by differentiating (7.102) with respect to t and (7.103) with respect
to x and eliminating the mixed (t and x) derivative term in fa. The
It has become customary to refer to the order of an approximation in the sphericalharmonics technique by indicating the highest-order harmonic retained from the
infinite set of coupled equations.
1
TRANSPORT THEORY
sec. 7.2]
351
result is easily shown to be
+ 3ZD\
(\
3D
6"
;**^ + ir^*(^
Za*(x,<)
Z>
0(1,0
(7.104)
= 4>{x,t)
where we have used (7.56) for D, along with the fact that <o(x, = 4>{x,t)
In treating specific problems
and 5o(x) = 5(x) [see, for example, (7.75)].
with this equation, it may be necessary to use Eqs. (7.102) and (7.103)
to evaluate the initial and/or boundary conditions.
It is of interest to compare Eq. (7.104) with the general form (5.21)
assumed in the development of the diffusion-theory

Two
distinct
differences are revealed: (1) an additional term in
model.
the second derivative of t has appeared, and (2) the coefficient of the first
Inasmuch as the above result was
derivative term in t has been altered.
obtained directly from the transport equation, one might expect Eq.
(7.104) to be a more accurate statement of the nonstationary problem
Evidently, the diffusionwithin the limitations of first-order theory.
theory model (5.21) implies certain additional assumptions or limitations
which
had been
which have resulted in the omission of the two features noted above.
Some of these are considered in the discussion which follows.
It should
be recognized, however, that the adequacy of the first-order formulation
(7.104) can be determined in the final analysis only by comparison with
experiment since the error involved in the omission of the various higherorder terms is not evident, a priori.
Before discussing the discrepancies between Eqs. (7.104) and (5.21)
be in order to comment further on the nature of Eq. (7.104).
This equation in its general form is the so-called telegraphist's equation.1
it would
The unique feature of this equation is that it describes physical phe

nomena which exhibit both wavelike characteristics and residual dis
turbance effects. The wavelike behavior is given by the second-order
term in
and the residual disturbance effect, by the first-order term in t.

Wave effects propagate with a finite velocity, and time-dependent
neutron phenomena should properly include such a term since time-wise
t,
variations in neutron population cannot be sensed at a given distance

from the source in less time than it takes a neutron of speed v to reach
However, after the passage of this "wave," the disturbance
that point.
persists and this effect is described by the first-order term.
The essential feature of diffusion phenomena, on the other hand, is that
all disturbances propagate with "infinite" velocity, i.e., any disturbance
at any point in the medium is sensed everywhere instantly. It is of
interest, then, to determine the limiting form of Eq. (7.104) for very large
'See, for example, P. M. Morse and H. Feshbach, "Methods of Theoretical
Physics," Part I, p. 865, McGraw-Hill Book Company, Inc., New York, 1953.
[chap.
REACTOR ANALYSIS
352
For this calculation let us impose the condition that

neutron speed.
both vD and 2av remain finite as v . Thus in the limit we require
that the diffusion medium have vanishingly small absorption cross section
In taking the limit, it is con
and a very large transport cross section.
venient to write (7.104) in the form
[l
as
-v
32a(vD)
|2 +
> oo, this equation
(vD)
-(Mn
(vD)
(7.105)
reduces to'
Tr
dt
- Zwi + Dv
d.r2
which is in the form (5.21).

Thus the differential equation previously
selected to describe time variations in the neutron population is a good
approximation only if the absorption cross section is very small and the
transport cross section very large. For those situations in which these
conditions
are not well satisfied,
Eq. (7.104) may yield
a better one-
velocity first-order relation.

7.3
The Boltzmann Equation with Energy Dependence
a. Assumptions and Limitations.

The details of the neutron slowingdown process and the role which it plays in determining the principal
nuclear properties of a reactor have been discussed at some length in the
introductory sections of Chaps. 4 and 6. In order to retain a greater
degree of generality in this development, we select as the basis for the
present model only the most elementary physical features of the processes
In this way we derive
a general statement of the Boltzmann

which
includes
the distribution of the neutrons in
(transport) equation
The only limitations which are imposed on the general
energy space.
involved.
results of this analysis are the two simplifying assumptions listed below:
(1)
(2)
The medium is isotropic and homogeneous

Neutron slowing down is due only to elastic-scattering
collisions with nuclei
(7.106)
The first assumption is customarily made in dealing with problems of

reactor physics.
Problems that involve inhomogeneity and anisotropy
on a microscopic scale are highly specialized situations and fall outside
Gross inhomogeneities, such as multiregion
the scope of this work.
The trans
media, on the other hand, are of considerable interest here.
port model is extremely useful for treating heterogeneous reactor con
figurations and boundary conditions; some remarks have already
made in Sec. 7.2g on the latter application.
been
TRANSPORT THEORY
sec. 7.3]
The purpose of assumption
(2)
is to
simplify the calculation.
353
As
process is due to the joint effects

of both elastic- and inelastic-scattering collisions, and the relative impor
tance of the two processes is determined by the neutron energy and com
By ignoring the inelastic process in the present
position of the media.
previously
noted, the slowing-down
analysis we imply that the system in question is relatively free of heavy

Another
nuclei.
These systems will therefore tend to be thermal.
is
that
a
useful
and
omitting
the inelastic phenomenon
accurate
reason for
model is not yet available for describing the energy spectrum of neutrons
In the absence of such information

emerging from an inelastic collision.
one can always utilize experimental data, but this approach is most
effective when working with the multigroup formulation.
The assumptions (7.106) are the only limitations to be imposed in the
derivation of the Boltzmann equation, and it is important to note that
no specific relation has been selected to describe the neutron-scattering
function.
Thus the results of this analysis will be generally applicable to
In order to demonstrate the
many systems of practical importance.
a particular scattering law is later selected,
namely, isotropic scattering in the center-of-mass system of coordinates.
However, this specific calculation will not be introduced until the general
application of these results
relations have been developed.

The purpose of the analysis
neutron-balance
which follows is to establish a suitable

relation which takes into account the dependence of the
functions upon the neutron energy, direction of motion, time

coordinates.
The balance condition, which is an integrodifferential equation, is then reduced to an infinite, but simpler, set of
coupled integro-differential equations by expanding the flux, the neutron
The deriva
sources, and the scattering function in spherical harmonics.
tion of the Fermi age model from these results is then demonstrated.
b. Neutron Balance.
A neutron-balance condition may be written in
terms of the flux (i,v,Q,t) dr da dv [see Eq. (7.2)], and this relation will
give the net change in the number of neutrons which occupy the volume
dr, whose speeds lie between v and v + dv, and whose velocity vectors lie
in the solid angle dQ about O.
A statement analogous to (7.5) which
accounts for all the physical processes that can affect this population
In this formulation, however, we
may be written for the present case.
must include in addition to the specifications of (7.5) the requirement
that the neutrons involved in the counting process be limited to those
which occupy the speed interval dv about v. Thus Eq. (7.5) may be
extended to the present application if we replace the
immediately
quantity 0(r,O,<) dr dti by (r,v,tl,l) dx dv dQ. There is one important
modification which must be made, however; this occurs in the term which
involves the scattering integral.
Before discussing this, let us first
relevant
and spatial
consider the remaining terms.
- |v
ANALYSIS
REACTOR
354
at
(T,v,a,t) + a
For
V0(r,i,Q,O +
[CHAP. 7
these we can
3S,(i>)
write
- S{i,v,a,t) j dr
*(rf,O,0
dv
da
(7.107)
The first term gives the time rate of change of the neutrons which occupy
the space dr dv dCk, the second, the losses due to the net transport of
neutrons through the boundaries of dr, the third, the losses from this
due to neutrons heing absorbed or scattered ojrjôr_^slowed out"
(of the dv interval), and the fourth, the contributions from sources.
Now for the scattering-integral term. The total number of neutrons
scattered into dr dv da per unit time from all possible speeds v' and
space
a'
is obtained by taking the number of neutrons in

dr dv'dQ! which experience scattering collisions, namely,
directions of motion
2.(y') 4>(r,v',a',t)
and multiplying
di
dv'
da'
it by the probability that
scattering
collision
in
da' throws the neutron into speed in dv about v with direction

in da about a. We denote this probability f(v,a;v',a') dv da and
define [cf Eq. (7.8)].
probability that
with an initial
of motion a', when
= scattered, emerges from the collision with
a speed in dv about v and direction of
speed
f(v,a;v',a')
dv
da
a neutron
v' and direction
motion in
(7.108)
da about a
The product of the total scatterings and this probability integrated over
all accessible v' and a' gives the total gain in the neutron population of
element dr dv da due to the neutrons scattered in from
and
different directions of motion :
higher speeds
/,<
the differential
Q>
*(r/>V)
2-(y')
The total neutron balance
f(v,a;v',a') di
dv'
da'
dv
da
(7.109)
is
is
is
obtained by adding (7.109) to (7.107)

to
the
of (7.5).
If the differential dr dv da
convention
according
sign
found to be
divided out, the result
(T,v,a,t) + x,(v) *(r,t>,o,0 + a V*(r),Q,0

+
4>(r,v',a',t)
S(i,v,a,i)
2,(t/) f(v,a;v',a')
dv'
da'
(7.110)*
where v' = v'a' and denotes the velocity vector of the incident neutron.
Equation (7.110)
the Boltzmann (transport) relation which describes
This relation
the distribution of neutrons in space, velocity, and time.
is

dr dv'
TRANSPORT THEORY
sec. 7.3]
355
general within the limitations of the conditions (7.106).

Specific situations can be described by defining the source function iS and
the scattering function f(v,Q;v',0').
In line
c. Azimuthally Symmetric Scattering from Stationary Nuclei.
is
further
calculation
not
possible
with the concluding remarks above,
is completely
until additional information on the properties of the system in question

This brings us to the third general assumption which will
be imposed in the derivation, namely,
is supplied.
All nuclei of the medium are stationary
(3)
(7.111)
As we well recognize, this assumption implies that the neutron speeds

involved in (7.110) are much greater than the speeds due to the thermal
motion of the nuclei. Thus, all results obtained throughout the remainder
of this analysis apply only to the neutron population well above the
thermal range.
The application of this assumption, along with (1) of (7.106), to the

Boltzmann equation reduces the scattering function f(v,Q;v',Q') to a
the probability that a neutron be scattered into any particular direction

must depend only on its initial and final energies and on the angle between
its original direction Q' and its direction after collision Q. As before,
If we define the directions
we denote this angle 0O (see Fig. 7.3).
=
a' = Q'(6',\t>') and ft
1(6,$), then the scattering angle 0O in the
laboratory system of coordinates may
be
written in terms of the com
where
-O
mo
= cos
mm'
VI
and dtl = sin
fl0
cos
M2
Vl -
m" cos
d8 d$ = \dn d\f/\.
(<A
ponents of these vectors by means of the relation
(7.112)
^')
The frequency func
collision, the neutron speed after collision

and its direction, in dn d\f/ about Q(m,^)
(7.113)
before
in dv about
= scattering
is
f(v,a;v',Q') dv dn d+ = f(vlfi,t;v') dv dp d+
probability that, given a neutron with velocity vector
v'
tion for azimuthally symmetric scattering can be obtained from the

If we include a statement of the condition
general function f(v,Q;v',Q').
may be written [cf. Eq. (2.9) as
in the definition of this function, then
another example of
conditional frequency]
The velocity vector v' used here to represent the condition on the proba
and
bility, namely, that the initial coordinates of the neutron be
For the present calculation, scattering collisions are assumed to be
azimuthally symmetric about Si'. That
equally probable that
the final vector O will lie in any position on the conical surface which has
is
it
is,
p!
v',
is

If the medium is isotropic and homogeneous, then

much simpler form.
the scattering process must be azimuthally symmetric about the initial
Or, more specifically,
direction of motion of the neutron (see Sec. 7.2b).
REACTOR ANALYSIS
356
[CHAP. 7
Q' as its axis and a vertex angle 60 (see Fig. 7.5). Thus the final direc
tion of motion will depend only on the scattering angle between the initial
We define, therefore, the function [cf. Eq. (7.17)]
and final directions.
q(y,/i0;v')
probability that a neutron with velocity v'

scattered into the differential
element
dv duo = is
dv dfio about
and
to describe azimuthally symmetric collisions.
/by
[cf.
This function
Fig. 7.5 Scattering angle 80

in laboratory
system of
coordinates.
is related to
Eq. (7.18)]
f(v,n,f;v')
We will
(7.114)
fio
use the
function
q(p,Mo;v')
(7.115)
in the
henceforth
integral of (7.110).
d. Expansion in Spherical Harmonics.
The
neutron-balance relation based on the assump

tions (7.106) and (7.111) may be written in the form
= S(r,v,Q,t)
2,00
(T,v,a,t) + a v(r,v,a,t)
(i,v',a',l) Z.(i/)
qO,Mo;v')
dv' <V
#'
(7.116)
with the aid of (7.115). As this equation now stands, further calculation
is not possible until some additional information is supplied on the
functions
For convenience, we assume that all these func
S, and q.
tions can be expanded in infinite series of spherical harmonics.
On this
basis we may write for the frequency function q
q(i;,Mo;v')
- (- ^)
P'M
(7-117)
<ko
(7.118)
q,0>;v') =
/_1
where in the usual way we compute

q(f,M<>;v') p,Gx0)
6',
8,
is
it
Expansion (7.117) may now be substituted into the scattering integral

necessary first to express the functions P,(mo)
of (7.116); however,
in terms of the variables
and 41'- The addition theorem for
Legendre polynomials (7.21) may be used for this purpose, and the pro
identical to that used in Sec. 7.2c for the series repre

cedure involved
In carrying out this calculation
sentation of the function t)(fi0;Q')convenient to introduce also the spherical harmonics F"(Q) defined by
is
it
is

~jt(r,v,Q,t) +
TRANSPORT THEORY
SEC. 7.3]
Eqs. (7.28) through (7.30).
After some algebra, it can
q(y,Mo;V)
357
2r
be shown
y$(fl)
q,(t>;v')
Ff(O')
Note that this result is entirely analogous to (7.37).

The flux and source terms are expanded in the complete
0(r,t-,Q,O =
that
X ^(^'O F-(Q)
(7.119)
set
F"(ii)
take
*rOV.O
(7-120)
n 0 m n
oo
n
S(r,A0
(7-121)
n
n = 0 m
with
*r(r,,o =
*(r,A0 do
/0 rr(o)
(7.122)
S(r,v,a,t) da
(7.123)
- /Q Y*(a)
<S|T(r,,0
,s':(r'M)
"
n,m
-"0
f?(q)
Kr(O') Ff(') dQ'
X
/
ij
q,(!,;v,) dv'
tj
*r(r'p,'
n,m
f:(q)
n,m
/S
c(r,u'0 y"(n)
2,(0
2,w]
(7-124)
The application of the orthogonality property (7.32) to the integral over

O' reduces the right-hand side of Eq. (7.124) to
n,m
Y?(a) fv2.(v') :(r,v',t)
qn(v;v')
dv'
and (7.124) may then be written
"
and
<S.
(r/,0
(r,,0
- S5(r,ffl
Z.(i>') q(o;v')
(7.125)
the separation of this equation into

equations in the harmonics for the flux and the
The procedure
is
set of coupled differential
Z,(iO
The next step in the analysis

source
v
+
-t
is
m n
Y*0h4>)
dt>j
straightforward and requires the

/i
multiplication of (7.125) by some harmonic F*(0) and integration over

and
The details of this calculation are identical to those encountered
\f/.

[iwt
The substitution of expansions (7.119) through (7.121) into (7.116) yields
358
ANALYSIS
REACTOR
[CHAP. 7
in the reduction of the one-velocity transport equation to a set of coupled

relations in the various harmonics [see Eqs. (7.39) through (7.46)]. Thus
it is easy to show that [cf. Eqs. (7.39) and (7.125)]
+ S(f)l *g(r,,0 + Si(r,v,t) +
(2a +
+ l+j8)
(2a +
1)
1)
|
.
^+^r'V^|
1)1*
1)
/3
^l(r'M)
+
2)(a +
(2a +
0
1)
3)
/3
(7.126)*
1)
1)
(a
<fc\(T,v,t)
*)(<
1^(^,0
+ l)1*
3)(2a+
j3)(a T(a
(2 + 1)(2,
(2a
D(2
(2a
.
/3
1,
0,
< a.
and a <
e. Isotropic Scattering in Center-of-mass System.
Again, in order to
make further progress in this development, we require additional infor
where a =
q.
it
tion
Before introducing this information, however,
to transform the relevant functions from the speed variable
is
is
is
mation about the system which

to be described by Eqs. (7.126).
The
to specify further details about the scattering func
next step we take
convenient
dv duo = q(w,/i0;v') dv duo
(7.127)
denotes the Jacobian
./
%(u,n0y)J
where the symbol
to the leth
We make this transformation in the case of the func
argy variable u.
tion
new function
such that
by denning
du
(7.128)
dv
w corresponds to the speed v.

Thus (7.127) states that,
the probability that scattered neutron acquires coordinates
and no, then this must be identically the probability that
have also
the coordinates u and fioNote that the Jacobian relates the differential
volume in
no space to the corresponding volume in u, no space.
it
dv duo
is
if
and the lethargy
use the new scattering
(,Mo,V)
function
in the general results

Take, then
series expansion of the form (7.119).
(m,mo;m',Q')
(7.126) we require
In order to
v,

2)(-0
(2a +
I"
d\
_ -(-0
/3l)(2a +
l)(2a
/})(a
(a
(2a
3*)
{dy
(2a +
/3)
(a-/3)(a
+ +
2.(/) qa(v;v') dv'
1)
(a
(r/,0
Jp'
(a
at
|S)
3)
n-0
^--^(u;<Q')
PM
(7.129)
TRANSPORT THEORY
sec. 7.3]
359
/_\
where we have replaced the statement of the initial velocity v' by the
initial direction of motion Q' and the initial lethargy u'. The coefficients
8 are computed from the relation
=
(;',')
(m,mo;m',Q')
Pn(no) duo
(7.130)
is
not yet in its most useful form.

Further
The expansion (7.129)
simplification may be achieved by observing that the frequency function
$(u,no',u' ,Q') may be written in terms of the conditional frequencies
t)(n0;u,u',a') and e(w;u',Q') [see Eq. (2.11)]; thus
- u(mo;u,u',0')
(u,mo;u',o')
fl(w;u',a')
(7.131)
where
given a neutron with
coordinates u' and Q' before collision and
lethargy u after collision, the neutron has
been scattered through angle cos-1 no in (L)
(7.132)
neutron with
Q', the neutron
scattered into lethargy interval du about u
(7.133)
= probability
that,
It
that,
u'
given
and
evident from the definition of the frequency
is
X)
and the energy relation

delta function. This follows
necessarily
(4.18) that \)(no;u,u' ,0')
from the fact that, given the initial and final lethargies of
scattered
is
/n0
which the scattering

which we denote by
can
is
there exists only one angle through
That angle (or the corresponding
f)
neutron,
occur.
E'
(4.18)
1)
+ 2A +
Ay
the neutron energy before collision and
is
where
tj
=
(1
'
obtained from the energy relation
is
the cosine of the
we transform
from the variable
2A,
The angle cos-1
fio
*[(A +
l)e<'-"
is
(4-27)
A*)*
to u and eliminate
means of (4.18), we obtain

Mo(,w')
(1
(A
77
If
'M
"
tj
The cosine
scattering angle in the center-of-mass (C) system.
related
cosine no for the (L) system through the equation
to the corresponding
is
from (4.27) by
- l)e-*]
(7.134)
related, then, to the difference of the initial and final
t)
g.
The integral (7.130) for computing the coefficients 6 may

in terms of the functions
and
If we recognize that
be
written
is
lethargies of the neutron through the above expression.

ij

B(m;u',Q') du = probability
coordinates
is
dm,
,U')
Xf(p0;u,u'
a delta
REACTOR ANALYSIS
360
function, and write

D0*o;u,tt,,0') = tfjio
[CHAP. 7
- f)
(7.135)
then the substitution of this result, along with the application of (7.131),
into (7.130) yields
.(;u',0')
fl(;u',0')
8(Mo
- f) P-(mo)
dô
= 9(;m',0')
(7.136)
This result
is a general relation from which we can compute the coefficients

to
describe
any arbitrary scattering process which satisfies the four
6
fundamental assumptions already imposed.
All we require in any case
is the appropriate form of the function g which relates the initial and final
Thus at this point we must introduce

additional details about the physics.
As we have previously observed there is one scattering law which,
although simple in form, describes reasonably well many of the physical
situations of interest to reactor physics.
We introduce this feature as a
fourth assumption.
lethargies of the scattered neutron.
(4)
Elastic scattering is isotropic in the center-of-mass

system of coordinates
The coefficients 8 which correspond to this particular scattering process

are easily computed if we recall that the frequency function for the change
of lethargy in the case of isotropic scattering in (C) was given by [cf.
'
Eq. (4.50)]
fl(u;u',Q') ^
fT-
u><u + \n (? 13g)
= 0
+ \n
(? 13g)
elsewhere

.(ttju'.O')
P CfVu'-u
1
(7-139)
Thus these functions relate the initial and final lethargies of the scattered
neutron with the cosine of the scattering angle in the (L) system.
We will
use these results in the analysis which follows.
Boltzmann Equation in First-order Harmonics and Isotropic Scatter

The first-order approximation to the set (7.126) is obtained by
truncating the series expansion for the flux at a = 1. This reduces the
infinite set of equations to four simultaneous equations in the harmonics
f.
ing.
\-
All higher-order terms are assumed to be small as

an(l
^Si^rS
In setting up these
compared with these and are therefore neglected.
in
will
convenient
to
use
the
lethargy
it
be
equations
place of the speed and
TRANSPORT THEORY
SEC. 7.3]
361
redefine the flux by means of the relation

4>(r,u,t) \du\ =
(T,v,t)
(7.140)
\dv\
*J(r,u,0
Jui
2,()]
il)
^*?(r'U'
(7-143)
*i(r,',0 Z.(u') i(,V,0') du'
(I
(7-142)
*!(r,',<) 2.(u') i(;u',0') du'

=
- ~ + 2,()1 ^(r.M.i)
(?-141)
Z.(u') i(w;m',Q') du'
4>Tl(T,u',t)
5)
*!(r-"'<)
li)
^(r^t)
--^( +
*(r'M'
(7-144)
where for simplicity we have taken the source to be isotropic.

Note
that this set of equations corresponds to Eqs. (7.47) through (7.50).
For the special case of isotropic scattering in (C), we can compute the
coefficients of the scattering function from (7.139).
We require here
only the first two:
i =
r
a
(7-145)
ra
o =
is is
is
The substitution of these expressions into

given by (7.134).
the Eqs. (7.141) through (7.144) yields the first-order approximation to
the Boltzmann equation for the case of isotropic scattering.
It
important to recognize that even in this form the flux distribution
where
given by set of integrodifferential equations; thus, although the problem

has been considerably simplified, the final analysis
by no means trivial.
An elementary solu
g. Combined Slowing-down Diffusion Equation.
is
theory,
can be obtained
we impose an
the basis for the Fermi age

additional limitation on the
is
tion to Eqs. (7.141) through (7.144), which

if

+ S,(u)
\jt
^3^0;(r'M'-^(^-^)*rI(r'M'
(fy
l{r,u't) 2.(w') 0(u;',O') du'
0(T,u,t) + S0(T,u,t) +
S,(m)
\jt
We have already transformed the scattering function q to the correspond

ing function 8 in lethargy, and the transformation of the remaining quan
tities in (7.126) is straightforward; therefore we write immediately
ANALYSIS
362
REACTOR
properties of the system.
Let us assume, therefore, that
[CHAP. 7
(5) Collision densities 2,(u')%(r,u',t) are slowly varying functions over the lethargy range of the scattering integral
In this
2,(')2(r,M',i) in
case we may expand
From these expansions
point u.
S.(w') *8(r,tt',0
=a
S.(m') tf(t,u',t)
~ Z.(u)
...
Taylor series about the
we select the approximations
2.(u) *S(r,u,f) + ('
- ) ^ [S.(u)
*g(r,,0]
*?(r,u,t)
(7.147)
(7.148)
If these approximations
are used in the scattering integrals of Eqs. (7.141)

(7.144), the resulting expressions may be evaluated in terms of
through
elementary
functions and the integral equations thereby reduced to a

partial
set of
differential equations.
The details of the calculation
follow. We observe first that the range of integration in the scattering
integrals extends from u + In a to u, where a is defined by (4.20).
Then, if we substitute the
(These limits were discussed in Sec. 4. 2d.)
expression for 80 from (7.145) and the approximation (7.147) into the
=-i /
*8(r,u',f)2.(w')e'-du'
_ S.(M) dj(r,,Q
I
+ r^
|-
eu,_du,
Ju+\na
['
[S.(u) 8(r,,0]
- )e-'-" du'
('
(7.149)
Also, the substitution of (7.148) into the scattering integrals of Eqs.

(7.142) through (7.144), along with the expression for Si from (7.145),
^T'U''l) S'(M')feU'"" du'

1
^Mu)tf(T,U,t)
/^/u
lna
V=~a
f"+i
yields the results
The three integrals which appear in these expressions

and
may be shown that
are easily computed,
r^~
Ju+In
e"'-du'
(u'
- )e' du'
a >+lna
lna
dA
afu
it

integral of (7.141), we obtain
-{
(7.151)
(7.152)
TRANSPORT THEORY
sec. 7.3]
363
where we have used the definition (4.42) for and (4.30) for no- The
f which appears in (7.153) is the cosine of the scattering angle
in (C) which corresponds to the lethargy change from u' to u.
function
We substitute the results (7.151) through (7.153) into the expressions

(7.149) and (7.150), and these in turn are used to approximate the set of
Thus,
equations (7.141) through (7.144).
I)
(7-157)
{ry
(7'154)
(7-155)
**
S')
it
^ =~^
four partial differential equations, then, are the first-order

to the Boltzmann equation which satisfy the require
approximations
set of
*!
ments (7.106), (7.111), (7.137), and (7.146).

The steady-state solutions to these equations are easily obtained by
new function 4>(r,u,t) 4>{t,u).
and defining
The
setting d/dt
solutions for the higher harmonics are
0
-4nk-
(7160)
')*'
- 8[*fr0 -
(7-161)
be
-V
DV
written in terms of the total flux
4>{t,u),
(7.162)
which
is
This result may
SJ
-2
1-
The substitution of Eqs. (7.158) through

for the diffusion coefficient.
of (7.154) yields a single second-order
into
the
form
steady-state
(7.160)
equation in
[cf. Eq. (7.57)].
+

This
0 it
-^(-<5K + 7i( + <5)
REACTOR ANALYSIS
364
by the equation
related to
*(r,)
The substitution of
2 is independent
[cf.
Eqs. (7.59) and (7.60)]

S(t,u)
and also
these
[CHAP. 7
functions into
(7. 162) yields,
^fi-
(7.163)
for systems whereiD
of r,
-D()V*(r,) + Z.()
4>{t,u)
S(t,u)
I^
[2.(u) *(r,u)]
(7.164)*
shortly, this is essentially the form assumed for

the combined slowing-down diffusion equation of Chap. 6.
h. Coupling Equation.
It was shown in Chap. 6 that the Fermi age
model when combined with a suitable coupling equation and a diffusion
equation for the thermal flux gave a reasonably accurate picture of the
space and energy distributions of the neutron population in a reactor.
A suitable expression for the thermal-flux distribution has already been
developed, and it was shown that the source term in this relation was in

As will
be demonstrated
In order to combine
fact the slowing-down density at thermal energy.
the description of the fast-flux distribution given by (7.164) with an
equation for the thermal flux, we require, therefore, an expression which
couples the fast flux with the slowing-down density.
We derive here an expression for the slowing-down density which is
based on the assumption that the lethargy values in question are far
Let
removed from the source value.
g(r,i<) =
number of neutrons slowing past lethargy

u per unit time in unit volume around
space
(7.165)
point r
contributions of all the scatterings

which occur at lower lethargies and which can throw neutrons into the
The appropriate expression is
lethargy range > u.
This number
is given by the combined
q(T,u)
f"
S.(u')
4>(r,u')
- -)
du'
(7. 166)
As in the case of the scattering integrals which appear in the Boltzmann

0(r,u') is a slowly varying function over
equation, we assume that
and Therefore may beapproximated by the
the^ interval of integration
first twcTterms oTa TayTor~Series,
^
If
S.(u')
(T,u') ~ 2.(u)
this expansion is substituted
4>(t,u)
{u'
-u)^-
[2.() *(r,u)]
into the integral of (7.166) and the inte
TRANSPORT THEORY
sec. 7.3]
3f)5
gration carried out, the resulting expression for the slowing-down

is
j(r,tt)
?2.(u) *(r,) + a
where
A [2.()
- a) -
0(r,)]
density
(7.167)*
ln* "
2(1
(7.168)
Equation (7.167) is the accurate expression for the coupling relation

between the thermal-flux equation and the combined slowing-down dif
fusion equation.
We summarize the above results:
- D{u) VV(r)U)
+ 2.(u) *(r,)
S(i,u)
g(r,M) = {2.(u) 0(r,M) + a

- AhV2tb(r)
A [2.(w)
^ [2.(u)
+ 2^lh(r)
4>(t,u)]
(r,w)]
(7.164)
(7.167)
= ?(r,Mth)
(7.169)
i. Fermi Age Model.

We develop the relations used in the Fermi age
model by considering successive approximations to the three equations
We begin by solving Eqs. (7.164) and
(7.164), (7.167), and (7.169).
(7.167) for q(r,u); thus,
q(T,u) = 2.(w)
{r,u) +
7[20(w)
{t,u)
y =
where
- D(u)^(t,u) - S(r,u)]
(7.170)
(7.171)*
The first term on the right-hand side of (7.170) is the predominant one
For very small reactors, the second term
for reactors of reasonable size.
can be of comparable magnitude because of the large leakage (represented
here by the Laplacian term).
The number y is of the order of and
ranges from zero to unity (see Fig. 4.18).
In the case of large reactors, wherein the absorption is known to be

small, a good approximation to (7.170) can be obtained by neglecting
the 7 terms entirely.
This approximation is acceptable, however, only
if the source term is also small in the lethargy range of interest.
In a
multiplying medium such as a reactor, the neutron sources are the fissions.
In the derivation of the slowing-down density, it was assumed that the
lethargies in question were far removed from the representative value for
the fission spectrum.
But, at these lethargies the contribution of the
fission spectrum j(w) to the neutron population is very small.
Thus the
omission of the source term does not introduce large errors.
approximation to (7.170) may be taken, then, as
q(r,u)
2,(m) 4>(r,u)
A first-order
(7.172)
REACTOR ANALYSIS
366
and the corresponding
[chap.
expression for (7.164) is
-D{u)V*4>{i,u)
+ 2(w) *(r,)
S(t,u)
qit'u)
which is Eq. (6.5). The source is included in this relation since it may
very well be of comparable magnitude to the other terms. It could be
argued that its omission from (7.170) was permissible only because it
represents a correction on a correction (the y term).
A second-order approximation to (7.170) for large reactors is obtained
by neglecting the leakage term only:
q(r,u)
=a
This expression may
JS.(m) 4>(i,u) + -y20(u) >(r,u)

be simplified
if
q(r,u) =*teZ,(tt)
which is to
we use the
- yS(T,u)
approximation y
(7.173)
*(r,tt)V- &S(r,ti)
then
(7.174)
with (7f164).
A set of equations which has proved to be very useful in reactor cal
culations, wherein the Fermi age inodel is applicable, is
be used in conjunction^

q(T,u)
=\J(w)
-Z)()W(r,M) + 2.() *(r,)
(r,u)J
=
5(r,u)
- ^ g(r,)
(6.6)
(6.5)
which are the equations used in Chap. 6. These relations hold if 2(?i) <3C
2,(u), and u is much larger than the values corresponding to the bulk of
Ke"fission neutrons.
We conclude the present section by summarizing the basic assumptions

that have been introduced in arriving at the Fermi age model as repre
sented by (6.5) and (6.6).
(1) The medium is isotropic and homo
Neutron
down is due to elastic-scattering collisions
geneous.
slowing
(2)
All
of
alone.
nuclei
the medium are stationary.
(3)
(4) Elastic scatter
ing is isotropic in the center-of-mass system.
(5) The angular distribu
tion of the flux is nearly isotropic.
(6) The collision density is a slowly
varying function of lethargy over the maximum allowable jump per
collision.
(7) The neutron constants (2,D) are not functions of the
spatial coordinates.
(8) The reactor is large.
(9) ^ y and 2 <$C 2,.
7.4
Methods of Integral Equations
One -velocity Model
A second general approach to the problem of determining the angular

distribution of the neutron flux may
be developed by
utilizing the methods
of integral equations.
These methods are generally quite powerful, and
formal solutions are readily derived even for nonhomogeneous systems
with complex source distributions. Unfortunately, the mathematical
techniques required for these applications are of a somewhat sophisticated
TRANSPORT THEORY
sec. 7.4]
nature;
moreover,
simple geometry,
367
actual computations, even for systems of relatively

However,
can become quite involved and laborious.
there are specific reactor-physics situations in which the integral-equation

methods are indeed easily applied and can yield accurate results with
A good example of this
is the Serber-Wilson method

The methods and general results developed in
this section serve as a basis for these applications.
In the presentation which follows, we derive the formal solution to the
one-velocity transport equation (7.15) for several media of practical
The general treatment follows that of Case, de Hoffmann,
importance.
The reader is referred to the work of these authors for a
and Placzek.1
reasonable effort.
presented in Chap.
8.
of the general methods and results.
much more complete development
As an
Neutron Distribution in Media with Zero Cross Section.
introductory calculation we derive the neutron-flux distribution in a
system which has zero cross section; media with more complex neutron
a.
If <(r,Q) is the directed (angular)

properties are examined subsequently.
flux in the usual definition, then for the present system to be at steady
state, we must require that the neutron losses from unit volume around

r and from unit solid angle about Q, which can

by the neutrons
supplied from
that
then,
follows,
[cf. Eqs. (7.5) and (7.15)]
balanced
be
It
V*(r,a)
be due to leakage alone,

the sources S(r,Q).
It
S(t,Q)
(7.175)
is convenient for the purpose of integration to write the leakage term

as a partial derivative along the direction Q.
Thus, if s is a vari
able which increases along the direction 1, then
(7.176)
= 8(xja)
(7.177)
V =
and (7.175) takes the form
*(r,Q)
Integration over the variable
*(r,Q)
The function
4>0(r,Q)
ft
it will
be
S(r,Q)
ds
denotes the solution
Q
and
yields
v>(r,Q)
arbitrarily taken
as zero
(r,Q)
(7.178)
to the homogeneous equation
= 0
(7.179)
for all subsequent analyses.
K. M. Case, F. de Hoffmann, and G. Placzek, "Introduction to the Theory of

Neutron Diffusion," Los Alamos Scientific Laboratory, 1953, available through the
Government Printing Office, Washington, D.C.
1
REACTOR ANALYSIS
368
more convenient
[CHAP. 7
way of writing the integral of (7.178) is (see Fig. 7.6)
Jj S(t -
4>(r,Q) =
RQ,Q) dR
(7.180)*
The solution (7.180) states that the total directed flux 4>(r,a) is the sum
total of contributions from all sources which lie on
the segment of a line through r parallel to Q for which
0 < R < <* .
Clearly, sources off this line cannot
contribute to the flux at r (in direction a), nor can
< 0.
The integral equation (7.180) is the solution to the

differential equation (7.175), and this general result
will be very useful in the work which follows.
As
will be shown shortly, the neutron-balance equations
for other, more complex systems may be reduced to
O
Fig. 7.6 Integrathe form (7.175); consequently, their solutions can
tion variable R and
be written down immediately from (7.180).
Before
e
point r.
proceeding with these calculations, however, we will
develop an alternate expression for the solution (7.180) which is obtained
This form will
by transforming the line integral to a volume integral.
be especially convenient for describing the more complex systems.
The transformation is carried out by introducing the angular delta
function 52(ti a'), which is defined
52(a
It
a')
a'
= o
(7.181)
g(a)
(7.182)
has the property that
ja ff(O')
Note that the function
5{a
a'
') da'
**(Q
differs from the usual delta function for direction
Eq. (7.135)]
1), where [cf.
&{a-a'
1)
S(no
(7.183)
1)
a'
1)
Thus the functions 52(0
dQ =
1^
8(mo
a') and 5(a Q'
1) dn0
i(q a') =
^-
equation
Zir
s(a a'
= 2ir
(7.184)
are related through the
l)
8(0
1)
fa
|q2'
We have used here our earlier notation Q a' = cos 0O mo- If (7.183)
is integrated over all directions (i.e., over all possible orientations of a'
to a, see Fig. 7.1), we obtain
(7.185)
The angular delta function

may be used to define the source S(r,a)
which appears in the integral of (7.180).
For this purpose we introduce
52

sources on the line at stations
TRANSPORT THEORY
SEC. 7.4]
the identity
S(r,Q) =
jj
dR
S(t
=
I"
RQ',Q')
la
The above expression may
Ra'
be
variable r' which is related to
a') da'
dp =
R*dRda'
(7.188)
further simplified by using the integration

p
R =
|r
is
extends over
This result may be used to compute the Mai track length (r), which
defined by Eq. (7.59).
The substitution of (7.190) into (7.59) yields
da
Jv~
|r
- r'|
(7.191)
We demonstrate the use of these results by evaluating the integrals

point source. Let
(7.190) and (7.191) for the case of
S(r',a) = S0A{a) 5(r'
r0)
(7.192)
is
a
which states that the source distribution

point source of strength
at r0 and emitting neutrons with an angular distribution
{a).
The substitution of this expression for into (7.190) gives for the directed
<S'0located
V|r-r,y
r0|-
(?
*(r,)
flux
|r

-JTs(< -'i)><(ai)
(7.189)
We use this variable in (7.187) and note that the integration

all space V.
iww
(7187)
by the relations
and
52(Q
s(r-e>i)8ia-i)dRda'
where we have defined the vector
r' = r
(7.186)
written
be
a') da'
|p|
(r,Q)
2(Q
r'|
Then, (7.180) may
S(r,a')
369
m)
REACTOR ANALYSIS
370
[CHAP. 7
which states that there is a flux at r arising from neutrons traveling in

direction Q only if Q, lies along the vector r r0; its magnitude is given
r0|s.
In
the same way we compute for the total flux

So .4
|r
+(r)
rl|)
(|r
by S0A/\r
(7.194)
r,|
at
is
This result tells us that the total flux of neutrons at

r0
precisely
due to a point source

the number of neu
r,
trons which leave r0 that are headed

directly toward reduced, of course, by
the usual 1/r2 factor.
b. Neutron Distribution in Purely
Absorbing Media. The neutron dis
tribution in a purely absorbing medium
may be derived from the general results
for the medium of zero cross section
We begin with
presented in Sec. 7.4a.
statement
+w
of
the
neutron-balance
The appropriate differen

Fia. 7.7 The stations and r'.
tial equation
obtained from (7.175)
by adding an absorption term to the left-hand side; thus,
is
a V*(r,Q) +
S(r) *(r.O)
S(r.O)
(7.195)
is
which states that the only loss mechanism for neutrons from unit volume
and unit solid angle about Q
by transport (Q V<) and
around
is
is
it
Gains are due only to the source S. Note that this

absorption (2<).
formulation does not yet include the neutron-scattering process.
to the form (7.175), for
Equation (7.195) may be solved by reducing
This
which the general solution (7.180) has already been obtained.
integration
the
factor
introducing
transformation
by
accomplished
defined by the equation
the optical thickness,
e^, where
/3(r;r0)
"
/3,
2o(r0
- wO) dw
(7.196)
stations on the line parallel to X which

Thus we define |S(r;r0) as
passes through the field point
(see Fig. 7.7).
the "optical thickness" of the straight-line path joining the field point
with some reference point r0. Call the length of this path
i.e., let
w denotes various
0,
This number, then, corresponds to the variable

derivative Q
may therefore be written
=
(7.197)
s
r|
|r.
The variable
in Eq. (7.176).
The

condition.
30
=
dd 3/3
(7.198)
TRANSPORT THEORY
But by (7.196) and (7.197)
fe
Z.(r
|r
SEC. 7.4]
371
- r|Q) - 2a(r)
(7.199)
e?
The integration factor
op
+ 2a*
|t
We use this expression, along with (7.198), in (7.195) and obtain

(7.200)
If
may now be introduced.
we note
that
then (7.200) may be written
/J""^'
'
y0
S(r0
=
f"
In the notation of (7.190),
with
r'
S(r
R.
If
gives
- wQ)
from (7.196) and obtain
- wQ) dw
S,,(ro
2(r0
u>a) du>
we use
= /3(r
- wO)
JqR
/"'r-r|
y|ro
0'
=
r|
= /3(r';r0)
definition for
use the
0
We
0'
- R;r)
2(r
rfw
(7.202)
(see Fig. 7.7)

diu
- wO) dw
= /3(r';r)
(7.203)
(7.204)
this result in (7.201) and note that
dfi' = 2a de' = S dR
we obtain,
finally
[cf.
Eq. (7.180)]
(r,Q)
fQ"
S(t
- R,Q)e-*fr-R^ dfl
(7.205)*
(7.180).
The expression for the total flux <(r)
the result (7.205).
We first transform
form corresponding to (7.190).
This
is
Thus the expression for the directed flux in a purely absorbing medium
may be obtained from the result for the medium with zero cross section
by merely introducing the attenuation factor e~& into the integrand of
is

where
Ta
Integration over
in the form (7.177).
is
which
p"
{+eB)
bp
computed from (7.59), using
(7.205) to the volume integral
372
ANALYSIS
HEACTOB
[CHAP. 7
*>
f^T-T"
<7-207)
It must be remembered that the Q' which appears in (7.206) and (7.207)
is in fact the direction given by R/R.
We demonstrate the use of these results by computing 4>(t,Q) and
#(r) for the case of a point source. As an example let us take an isotropic
source of strength So at the origin; then,
S(r,a) = - 505(r)
(7.208)
<Kr,0) =
~-t
52
(fl
e-WV
(7.209)
(7.208) into (7.207) gives
*(r)
to be expected.
The special case of
- ^b*-
is
(7-210)
is
homogeneous mgdkras
of great practical impor
In such media the optical thickness reduces to the simple form

0(r';r) =
Za|r'
(7.211)*
r|
tance.
is
as
S(,
- R,Q)e-*. <r -
s-î hîsgp^
/;
and the results (7.206) and (7.207) may be written
For the
(r)
and
'
(7.212)
r>|a
point source (7.208),
case of the isotropic
(7.213)
these solutions yield
r/r)rv
^T-
(7-215)
An interesting and useful application of the concepts developed for the

in the calculation of escape probabilities
purely absorbing medium
from arbitrarily shaped volumes.
If P(r)
the probability that a
is
is

is
which states that the flux of neutrons with direction Q at

zero except
when Q coincides with the line connecting
with the position of the source.
And, in this case, the flux
the total number of neutrons emitted per
unit solid angle (<S0/4irr2) reduced by the attenuation losses according to
e~B.
The total flux may be computed in the same way. Substitution of
sec. 7.4]
TRANSPORT THEORY
neutron born at r within the volume
373
V escape, then the
probability P.*. for a distribution S(r) of isotropic

is defined by
average escape
neutron sources in V
(7.216)*
In general for convex shapes, P(r)
is the total neutron current emerging

V arising from a unit isotropic source
located at r inside V. To compute this quantity we begin with the
probability for escape through the element of area dA on the surface A
of volume V (see Fig. 7.8).
If dQ denotes the element of solid angle
from the surface
of volume

Fio.
7.8
Volume V and orientation of neutron path.
subtended by the area dA from the point r, then rftl/4ir is the fraction of
neutrons emitted by the source which is headed in the direction of dA.
Now suppose V contains a homogeneous material of absorption cross
section 2, and let the distance of dA from r be s, then exp ( Zs) is the
will travel from r to dA without experiencing

an absorption collision.
Thus the fraction of neutrons actually passing
through dA is exp ( Ss) dl/4ir, and P(r) is simply the integral of this
quantity over all directions; therefore,
probability that
a neutron
(7.217)
In
the case of a uniform distribution of isotropic

the expression
we obtain for
sources throughout V,
(7.218)
REACTOR ANALYSIS
374
This result may
,iA:
be
further developed by interchanging
the order of inte

the volume ele
To compute
gration.
ment dr we picture the volume V as
consisting of a series of tubes of cross
section (Q n<) dA and length s0, as shown
[CHAP. 7
in Fig. 7.9. The symbol n( denotes the

inward directed normal to the surface d A
,
iHQ. 7.9
Volume element..
(and n the outward directed normal so
=
n;
that
n). Then the volume element may be written
,,
dr =
(a-
n.) dA ds
(7.219)
or
P'"
47F fA
ja
4irVza
fo
L (1
e~*"
"
'
n<) dS
'
e~z'â
The general problem of computing the
da dA
*)
da dA
(?-22)
probability from
a volume
the average flux in the
volume were it to be immersed in an infinite medium supporting a uniform
This relationship has considerable practical importance;
isotropic flux.

escape
is closely related to the problem of computing
To demon
moreover, the two problems are analytically equivalent.
strate this equivalence, consider a body V of the shape and material
assumed in the calculation of the escape probability to be located in an
isotropic bath of neutrons such that at great distances from V the total
track length is 4>x. Then the number of neutrons incident upon the
body per unit time per unit solid angle from direction Qo through the
surface element dA is
ndA
(7.221)
Now if >(r,Q) is the directed flux at r in V, then the total track length at
r is given by (7.59). Further let >(r,Q;r0,Qo) be the directed flux at r
due to a unit source at r0 radiating neutrons in direction Q0, and let
Then
(r0,Qo) be the source strength.
*(r)
If we
j[n4>(x,a)da
a
f
jIda
a
J u
0(r,Q;r,ao)
S(r0,Oo)
dQ0
(7.222)
for the relation (7.221) and define to be the neutrons released

per unit area per unit time at the surface, then the above relation becomes
u.se
*(r)
This result may
=
be
t5
/ dQ Ja,
/ Qo-n 0(r,Qô,Qo) dQtdA
"wr
Ja
(7.223)
written in an alternate form by the application of the
TRANSPORT THEORY
sec. 7.4]
375
reciprocity theorem,1 which states: The directed flux at r in direction 2 due

to unit source at r0 radiating in direction Q0 is the same as the directed
flux at r0 in direction Q0 due to unit source at r radiating in direction
Q.
Therefore
=
0(r,a;ro,Qo)
*(r0,-O0^,-Q)
(7.224)

(r)
QB'n (T0,-a0;r,
da
t2
-a)
da0d A
(7.225)
which gives the total track length at r inside V produced by sources on

the surface A arising from the presence of the isotropic bath in the
surrounding
Q and
Go,
space.
then
If
we reverse the sign of the variables of integration

is immediately
(r)/4>M
neutrons born at r which
escape through A
p(r)
identified
; thus,
as the
fraction of
(7.226)
and by (7.216), we have

(7.227)*
the average escape probability from a volume V con

uniform distribution of isotropic sources may be obtained by
computing the average flux distribution in the same volume if it were
placed in a uniform field <bx.
^
The calculation of escape probabilities is often most conveniently
carried out by means of the chord method developed by Dirac.2
For this
analysis we refer to Fig. 7. 10, which shows
an arbitrary convex volume V.
As be
fore, n{ is the inward directed normal at
dA and s(Q) is the chord length from
dA in the direction 1. Let us assume
now that the number of chords of length
,
Fia. t AO Chords through vol
j
, ,
,
between s and 8 + as in the solid angle
ume y
taining
is proportional to Q n the cosine

of the angle between the inward directed normal at dA and Q.
Note
that the direction Q in O n, corresponds to the direction in s(tl). On
this basis the total number of chords in V with lengths between s and
dX
ds is
proportional to the integral
f
1
dA
n,
da
Ibid., p. 16, for proof of this theorem.

P. A.M. Dirac, "Approximate
Rate of Neutron Multiplication
for a Solid of
Arbitrary Shape and Uniform Density," British Report M.S.D. 5, Part I, General
Theory, 1943.
*
ANALYSIS
REACTOR
376
[CHAP. 7
where the directions Q over which the integration is performed are selected
If this relation is divided by the total
such that s < s' < s + ds.
number of chords in V, using the cosine weighting mentioned above for
each direction
and s + ds in
a, the fraction h(s)

V is given by
h(s) ds =
all chords with lengths between
ds of
dA f
. Ja=ow
/ dA
da
n,
la- n, da
(7.228)
where the only restriction on the region of integration in the denominator

is Q n, > 0.
The denominator
JQa-
dA
f*
dn
h(s) ds
JA
(7.229)
(7.230)
Ja-ait)
'
dA
J,I
a
itA
rA
may then be defined by
of (7.228) and (7.229) yields

=
Thus, for
n-
da
(7.231)
ttA
Jq
da
sa-n,dA
JQ
da
wA
JI
Interchanging the order of integration

8
(7.232)*
since integration over all values of

extends the a integration to all
This result was previously cited in (5.295).
possible directions from dA.
The expression for
3"W
e_Z"*)
("
"
dA f'm"
(1
irPs; fAdA
PK =
fa(1
computing
may be used to reduce the formula (7.220) for

the escape probability, which we now write in the form
n>)
- e-z-*)(Q-n,)dO
(7.233)
it
s,
where we have separated the integration over CI into two parts.

First
we integrate over all directions leading to chord length
then over all
chord lengths from the smallest to the largest.
follows from
Now,
(7.228) and (7.231) that
4tV
h(s) ds =
dA
a nida

The application
fA
through
s
The average chord length
da
n,
dA
j^s
fA
f**
of this expression is easily evaluated.

in the usual notation, we have
d\p
da = dn
sec.
TRANSPORT THEORY
7.4]
377

1
P.-
(1
and
(1
Q-QW
Q-TLidQ
. <r*.')A() ds
(7.234)*
An important application of this general result is given in Chap. 10 in

connection with the problem of computing the resonance-escape proba
bility for fuel lumps in a heterogeneous reactor.
The evaluation of the chord-distribution function h is easily demon
As an example, consider
strated when the volume V has a simple shape.
Let chords of
a slab of width a (see Fig. 7.11).
length s make angle 6 with the x axis, and let
Then for any chord s' with inclination
n = cos 0.
'
B1 , we have Q n, = cos 0' = ft', and with dO = dn' d\p',
(7.228) reduces to
1
*w*.-^
f
]AdA
fn+d*
swjQ
his) ds =
or
,dty'
~ 2fi
ds
8*
dii
(7.235)
Fig. 7.11 Chord

keeping in mind that n implies chords of length s.
through slab.
The result (7.229) has been used in evaluating the
For this case the average chord length is
general relation for (7.228).
shown from (7.230) to be s = 2a.
c. Multiplying Media with External Source.
We examine here media
which exhibit both elastic-scattering and fission interactions between
neutrons and nuclei in addition to the absorption property considered
in Sec. 7.4b. In the present development we introduce an alternate
formulation of the one-velocity transport equation (7.15) which combines
into a single integral term both the scattering and fission processes. This
is accomplished by defining a new quantity c which is given by
c(r)
r2,(r) + S.(r)
2,(r)
(7.236)*
where the various cross sections and v have the usual definitions.
Accord
c is the average number of neutrons produced per collision.
The neutron-balance condition may be written in terms of this number.
ing to (7.236),
2,(r) c(r)
Eq. (7.15)]
*(r,Q') /*(0;Q') da'
jQi
a V0(r,O) + 2,(r) >(r,a)
[cf.
For the steady-state system it takes the form
S(r,Q)
/*(tl;tl')
dCk
= probability that, when

direction
12'
suffers
secondary (product)
tion in dQ about Q
neutron with
collision,
yields a
neutron with direc
it
where
(7.237)*

C2t
(7.238)
REACTOR ANALYSIS
378
[CHAP. 7
Note that this function describes the general collision process consisting
The average yield per collision corresponding
of scatterings and fissions.
It follows that the integral of (7.237)
to f* is the quantity c of (7.236).
multiplied by c2, gives the number of neutrons which emerge from all
In the case of a
collisions with direction in unit solid angle about Q.
nonmultiplying medium c = 2/2( and f*(Q;a') = /(0;Q'), as given by
(7.8) . Finally, it should be mentioned that the source function S which
appears in (7.237) does not include the contributions from fission and
therefore differs from the function S given in Eq. (7.15).
In the present calculation we assume that all collisions are isotropic in
(L)
therefore,
/*(Q;Q')
(7.239)
The integral which appears in (7.237) now reduces to the expression for
the total flux <(r) [cf. Eq. (7.59)], and (7.237) may be written
a V0(r,Q) +
2(r)
*(r,fl)
Z,(r) c(r) *(r) + S(r.O)
(7.240)

If
this equation is compared to (7.195), we observe immediately that the

two relations are of the same form, the only difference being due to the
2( which appears at the left of (7.240) (in place of the 2), and the addition
of a source term at the right which augments the external sources S.
Thus the solution to (7.240) may be obtained by applying the result
In place of the source function S which appears in (7.205) we
(7.205).
use the two-term source of (7.240) and modify the definition of the
optical thickness j? so that it is given in terms of the total cross section 2,
in place of 20. In this way we obtain the solution
<Kr,fl) =
4ir
Jo
2,(r
- R) c(r - R) *(r - R)e-r-R;r) dR

+ jj S(t - R,Q)e-0 dR
(7.241)*
The expression for the total flux is obtained from the general result (7.207).
Again noting that the source consists of the two functions on the righthand side of (7.240) and that /? is defined in terms of 2,, we have
^
'P
4>(t)
\s(t',a') + ~ c(f) 2,(r')
(?-242)
Both
where Q! = R/R [as used previously in Eqs. (7.206) and (7.207)].
of these relations are integral equations and may be solved by means of a
The
Neumann series1 which is developed by successive approximations.
'See, for example, H. Margenau and G. M. Murphy, "The Mathematics and
Physics of Chemistry," pp. 504-505, D. Van Nostrand Company, Inc., Princeton,
N.J.,
1943.
sec. 7.4]
TRANSPORT THEORY
379
method, although straightforward, is laborious, especially if the conver

The method
gence of the series is slow so that several terms are required.
consists of making successive estimates of the function which appears in
is
<S,
For example, a first estimate of

is obtained by ignoring the
(7.242).
The resulting expression for involves
4> integral at the right of (7.242).
which
This solution may
only the integral of the function
specified.
The solution to the general time-dependent
a V*(W)
is
transport equation
+
i
^
(r,O,0
easily developed
In the notation of the present method, the non-
from the above results^
stationary one-velocity
system
is
it
then be used in the

second estimate for
integral of (7.242) and
Higher orders of accuracy are obtained by continuing this
obtained.
process. When 4>{r) has been computed to the desired accuracy,
may
then be used in (7.241) to obtain the directed flux <(r,a).
2(r) 4>(r,Q,<)
2,(r) c(r)
4>(r,t)
(7.243)
we have omitted the external source term
to avoid undue
The solution to (7.243)
obtained by seeking
number
complication.
a such that (T,Q,t) may be given in the form
(T,a,t)
(7.244)
4>{T,a)e<
is
is
The technique by which this

discussed under the section
accomplished
If we substitute (7.244) into (7.243), we
on the Serber-Wilson method.
2(r) *(r,Q)
V(r,tl)
Q,
obtain immediately
2((r) c(r)
(7.245)*
<f,{r)
This quantity may
= 2, +
(7.246)*
"effective-removal" cross section and

(total) cross section by the addition to
be called an
related to the actual-removal
is
2<
system with total-removal cross section

to that of the actual system.
d.
"time-absorption"
in
term a/v. If solutions of the form (7.244) exist,

then we have succeeded in reducing the time-dependent problem to an
equivalent steady-state problem which describes the neutron economy
of
and a source term identical
Source-free Infinite Multiplying Media. Of particular interest in

technology are multiplying media which contain no external
reactor
V0(r,O)
2,(r) (r,Q)
= 2(r) c(r)
jQi
neutron sources (i.e., sources other than the fission reactions).

these problems, the transport equation takes the form

is
wherein
Kr,a')
/*(";')
da'
For
(7.247)
REACTOR ANALYSIS
380
[CHAP. 7
If
the distribution of neutrons emerging from collisions is isotropic in (L),

then the solution to this equation is
*(r,Q)
if
4tt
Jo
"
2(r
- R) c(r - R) >(r - R)e~^'-^ dR
(7.248)*
The total flux is given by
If
the medium in question is homogeneous,
*(r,U)
*(r)
*(r
jf
= c2(
4?
- R)e"*.
- rT
0(r>-^-'"l
JK
|r
its magnitude
>(r) = e,Br
(7.251)*
these
by substituting
Thus
into (7.251).
cX,
If
we define
be
written
= r
g-S'lr-r'l+i'B
|r-rt
4ir
Jv
dp
(7.253)
To evaluate the
P2
(see
r' Wr'
then this equation may
where we have canceled out the common factor e,Br.

integral we define
= p(p,0,^)
dp dn di
dp
= cos
functions which
Eigenvector
and coordinate system.
7.12
with
(7.250)
is determined
~-%}r
Fig. 7.12).
Then (7.253) yields
e*8"'
is
a
Thus
"4r
4t
solution to (7.251) for an infinite medium, provided that
!H(H
(7.254)*
General Solution for <p(r) for Infinite Media. The general solution
obtained by superimposing the elementary solutions
to Eq. (7.251)
is
2.

pip.i,^)
Fio.
dff
Solution for <p(i) in\lnfinite \M edia.

The elementary
solution to Eq. (7.251) for the infinite medium
is e'3*.
The quantity B is an eigenvector, and
r',
Elementary
reduces to the form
rfr'
We now demonstrate some of the properties of

will be useful for later applications.
1
1.
/S
written
(7.211), and the above solutions~may~be
"
then
TRANSPORT THEORY
sec. 7.4]
eiBT
if
381
we define the eigenvector B by the relation
B = Bu
where u is a unit vector (a direction), then the general solution may be
obtained by integration over all directions u; thus
*(r)
|u
a(u)e'Bu
du
(7.255)*
The eigenvalue B is computed from (7.254), and the arbitrary function

a(u) is analogous to the arbitrary constant which always appears in the
steady-state solution to the diffusion equation for a multiplying medium.
Its value can be determined in these cases only by specifying the initial
condition of the system.
3.
(r)
Is Solution
Laplacian of
(r)
V*4,(r)
Stationary-wave Equation.
using (7.251) and (7.255), we obtain
=
to the
WBu
Ja a(u)
du =
-B2
If
o(u)e'Bu
we compute the
du
(7.256)
Thus the solution (7.255) for the infinite medium satisfies the differential
equation
V2<Kr)
#2<Kr) = 0
We recognize this relation as the appropriate neutron-balance equation

for a multiplying medium at steady state in the one-velocity approxi
mation [cf. Eq. (5.161)]. Solutions of the transport equation are there
fore also solutions of the "diffusion equation" (properly called the
stationary-wave or Helmholtz equation) . Conversely, solutions of the dif
fusion equation will also satisfy the transport equation exactly in the
Solutions of the diffusion equation for finite
case of infinite media.
geometries, on the other hand, do not; however, if these solutions are
applied in regions far from boundaries, the transport equation can be
approximately satisfied. In these regions the angular distribution of the
flux is nearly isotropic (which is a limitation of the diffusion theory), and
the diffusion theory results should give a good approximation to the
Thus solutions to the diffusion
spatial distribution of the neutrons.
solutions
to
the transport equation.
equation are the asymptotic
4. Directed Flux >(r,a) in Infinite Media.
The general solution for
>(r,12) in an infinite homogeneous medium is obtained by substituting
the general expression (7.255) into (7.250) ; it is easy to show that inte
gration over R yields
(r,Q)
'
4ir
['
Jo
dR
[ a(n)e--**
Jn
dn =
ff
4ir
Ja
W
1
f*
+ iBu
O/S,
(7.257)*
The application of this result to the infinite-slab problem is demonstrated

in the analysis which follows.
REACTOR ANALYSIS
382
[CHAP. 7
An approximate solution for the angular

Infinite-slab Reactor.
distribution of the neutron flux in an infinite slab of multiplying material
may be derived from the general solutions for infinite media developed
in Sec. 7.4d. The present formulation of the problem follows the work
of A. H. Wilson.1
Consider a uniform infinite slab of half-width a oriented with respect
The problem is
to the x axis, as shown in Fig. 7. 13.
to determine 4>(r,Q) throughout this medium.
The
general solution for tf>(r,Q) in an infinite medium is
given by (7.257); however, this solution does not
apply directly to the present problem since one di
mension of the medium is finite.
Nevertheless, it
is possible to develop an approximate solution to
Fio. 7.13 Infinitethe slab problem from the results for the infinite
e.
slab reactor.
medium.
The directed flux in an infinite medium wherein the spatial distribution

of neutrons is given by a single variable x, and therefore by a single angle
In this special case (x,8) is symmetric in
8, has been given by Wilson.
solution for
x and 6, and the elementary
(x,6)
may be written in the

form
cos
etB*
cos
1
1
iB
-
2^
Bx + cos
sin Bx
2<i
cos
.)
l +
Y2
cos2
(7.258)
As usual, the eigenvalue B is to be computed from (7.254) ; this may be

demonstrated by substituting (7.258) into (7.250).
The above expres
sion, then, is an exact solution to the transport equation.
However, it
has no direct physical application for two reasons: (1) it applies only to
the infinite medium; and (2) not all the solutions (7.258), for the various
In spite of these facts, (7.258) can
values of B from (7.254), are positive.
be used to construct physically useful solutions.
The technique is to
replace the actual finite body by an infinite medium for which (7.258)
gives mathematically exact solutions and so construct the solution that
the regions of the infinite space, wherein <p < 0, are outside the boundaries
This approach is analogous, then, to the method of
of the real system.
images which was used for the Fermi age solution to the slab problem
In general, this method cannot yield an exact solution
(see Sec. 6.4b).
to the medium with a finite dimension since it is not possible to satisfy
1 A. H. Wilson, The General Properties of the Transport Equation and Its Use in
in Bodies Having Spherical Symmetry, A.E.R.E.
Determining the Multiplication
Harwell, M. S. 10.5A, Dec. 7, 1944.
TRANSPORT THEORY
sec. 7.4]
383
However, an approximate
the boundary condition completely.
can be obtained which can be useful for actual applications.
The boundary condition
we should like to satisfy is [cf.
(a,e) =0
<
<
solution
Eq. (7.97)]
(7.259)
Inasmuch as this condition can be satisfied only in part, we select instead

the less restrictive requirement that there be nqjieutrons returning to the
slab along directions normal to the faces; thus, we ask that
(o,ir) = 0
If
this condition is applied to the solution
(7.260)*
(7.258), the following equation
is obtained:
2(o
This equation

(7.254).
slab.
It
cot Ba =
(7.261)
with the eigenvalue relation

The joint solution yields the criticality requirements for the
is convenient to write this pair of equations in- the form
is to be solved simultaneously
As in the diffusion-theory problems, the specification of the cross sections

For example,
(i.e., fuel concentration) yields the size a, and vice versa.
if the fuel concentration
is specified, then S( and c [from (7.236)] may be
and Eq. (7.2626) yields B. The substitution of this result

into (7.262a) gives the half-width of the slab a.
Since boundary condition (7.260) satisfies the correct relation (7.259)
computed,
to determine the directed flux (a,0) for

Wilson has performed this calculation for the case
The resulting values of (a,6) are shown by the
c = 4/ir and Ba = 7r/4.
broken curve in Fig. 7.14 to a relative scale. The representation of the
physical situation by the approximate condition (7.260) is evidently not
too poor.
As is to be expected, 4>{a,6) 5^ 0 for any value of 6 different
from ir; however, over a sizable angle (a,0) <K <(a,ir/2).
There is an alternate condition to (7.260) which yields somewhat
better values of B. This condition corresponds to the diffusion-theory
requirement that the partial current from the vacuum be zero [see (5.57)].
The appropriate statement in terms of the directed flux (r,Q) is [cf . (7.64)]
only at
other
a-,
it
is of interest
directions.
dA
Jin
0(a,a)(O
s) da = 0
(7.263)*
REACTOR ANALYSIS
384
[chap.
where dA is an element of area on the boundary surface a, s is the out

ward pointing normal to dA, and the integration extends over all Q
directed into the medium from the surrounding space.
In the slab
problem,
= 4>{x,6), and (7.263) reduces to
/a
(t>(a,0) cos 0
sin
= 0
0 dO
dQ = sin
where we have used the differential
(7.264)
dd d\j/.
This condition
1.2 1
-V-
1.0
0.8
0.6
ai
u
TO
No return current
c = 4/ir

% 0.4
M
X
"S
(a,v)=0
=
" c 4/t
fl/2,
=1
Ba=1.017
B/2,
=1
Ba=jr/4
V
s v.
0.2
-0.2
-0.4
20
40
60
80
100
120
140
160
180
Inclination to x axis S, degrees
Fig.
7.14
Directed flux at surface of slab.
will yield 4>(a,6) < 0 for some values of 6; but the effect of this on the
estimate of the inward current will be compensated for by the contribu
tions from the remaining directions.
The substitution of the infinite-medium
yields, after integration,
cot Ba
B
2,
(1
(
(Vg)
In
[1
solution
tan-1
(7.258) into (7.264)
(B/Z,)]
+ (B/S,)*]
(7.265)*
A comparison of this result with (7.261) reveals that the right-hand side
of (7.265) is always less than B/Xt; thus these two conditions give different
This discrepancy is well
values of a for a given set of cross sections.
demonstrated by the sample computation for c = A/ir and B/S( = 1. In
the
Ba
of the boundary condition

4>(a,ir) = 0, Eq.
0.785, whereas the no-return current condition
case
(7.261)
yields
(7.265) yields
TRANSPORT THEORY
SEC. 7.4]
385
= 1.017.
Thus the two conditions presented above would give

different estimates of the slab thickness for a critical system.
This discrepancy is due to the fact that the neutrons in each case are
Figure 7.14
emitted from the slab in a nearly isotropic distribution.
shows this tendency for both cases (the solid-line curve refers to the
Ba
noticeably
no-return current condition). Only if the neutron emission is mainly

along the outward normal to the surface will the two methods agree well.
This is the case for spherical geometries, an example of which follows.
f. Bare Spherical Reactor.
We consider here a second demonstration
of the technique of constructing approximate solutions for the directed
In this
flux in a finite system from the results for the infinite medium.
It
example we analyze a homogeneous multiplying sphere of radius a.
was shown previously that the directed flux in an infinite homogeneous
medium could be written in the form

0(r,O)
*(r
- R)e-*' dR
(7.250)
The elementary solution to this equation is e'Br, where the eigenvalue B

is computed from (7.254).
It was also shown that this solution satisfied
the stationary-wave (diffusion) equation (see Sec. 7.4d).
Now, the
only solution of the diffusion equation (5.161) which is spherically
symmetric and finite everywhere is [cf. Eq. (5.166)]
*r)
Ajel*:
(7.266)
This function will also satisfy the integral equation (7.250) ; thus, we may
use this relation for the total flux 4>(i) under the integral sign.
The result
is
*(r O)
-cS'A f"
4*
(B|r-
dR
-R\)e~^
R|
sin [B(r* + R* f"
(rf + R* - 2rR)l
Jo
4r Jo
cL,A
sin
2rRM)*]e~1-*
dR
(7.267)
where n is the cosine of the angle subtended by the two vectors r and R
Note that in the present system 4>(r,Q) 4>{r,0).
(i.e., i R = rRy).
All the statements which have so far been made apply to the infinite
medium.
We now assume that the solution (7.267) is
valid in the interior of the specified sphere. If (7.267) were an exact
solution for this finite system, then we could satisfy the boundary con
dition 4>(a,0) = 0 for tt/2 < 0 < t. Since this is not the case we attempt
to satisfy this requirement in part.
As in the slab problem, if we ask
that
> 0 everywhere within the sphere, then the appropriate condition
homogeneous
to be satisfied is
*(<*,) = 0
ANALYSIS
REACTOR
386
[CHAP. 7
which is identical to (7.260).

The application of this approximate bound
ary condition to (7.267) yields the equation
cX,A
ir
f"
e~z' sin
a
Jo
B(a + R) dR
+ R
cZtAez*
4*
t"
e~z- sin
Bs
ds
(7.268)
This equation gives the eigenvalue B which satisfies condition (7.260).

The integral which appears in this relation is more conveniently written
in the form
f"
e-*"sin Bsds
^.
Zl
[EiCSfi
J.
2i
[-
,_.a<_iB)
- iBa) - Ei(X,a
e_.(2l+iB)]
+ iBa)]
ds
- Im #i(2,a + iBa)
(7.269)

where E\{z) is the exponential integral [see Eq. (5.306)] and Im 2?i(z)
denotes the imaginary part of 2?i(z). The exponential integrals of com
plex argument have been tabulated and will be discussed in detail in
connection with the Serber-Wilson method.
From the above results we find, then, that the condition
yields the requirement
-Im
E^Xfi + iBa)
= 0
4>{a,ir) =
(7.270)*
This condition is to be used in conjunction with the eigenvalue relation

As in the case of the set (7.262), (7.270) together with (7.254)
(7.254).
yield either the size or the fuel concentration, given the other.
The boundary condition which requires that there be no return current
from the surrounding vacuum [see (7.263)] may also be applied to the
Wilson has carried out this calculation, and the
present problem.
detailed analytical statement of this condition is given in the previously
mentioned report. 1 He finds, however, that the result agrees with (7.270)
to within a few per cent.
Since the resulting expression for the no-return
current condition is quite complicated, it is suggested that the more
simple form (7.270) be used.
We conclude the present section with a few remarks about the calcula
tion of the directed flux 4>{r,d) within the finite sphere. These calcula
tions have been performed by Wilson, and the reference report presents
several series representations
of this function which are suitable for com
These include special expansions for both B/2, J5> 1

and B/Xt <SC 1, as well as a general solution which is given in terms of
Bessel functions and Legendre functions of the second kind.
putational purposes.
lIbid.
TRANSPORT THEORY
sec. 7.5]
7.5
387
of Spherical-harmonics Method
Applications
In this section we illustrate the application of the spherical-harmonics

Both
method by considering two problems of some practical interest.
deal
with
one-dimensional
and
the cal
steady-state
systems,
problems
culation is carried out on the basis of the one-velocity model developed
in Sec. 7.2f. In the present applications the general time-dependent
relations given in (7.84) reduce to the following set of differential
equations
-2,4h>(x) + S,l)0(x) + Sn(x) =

(7.271)*
where n = 0, 1, . . . . Our immediate objective in each case will be to
obtain solutions of this set of equations which satisfy the particular
Since the method
boundary conditions that define the two problems.
we
will

value of
use is independent
re), we
of the order of the approximation
(i.e., the
will truncate the
set (7.271) at n = 3.
This yields the
which is the next more accurate model after
so-called P3 approximation
It is of interest to note in this connection that only
the diffusion theory.
yield higher degrees of accuracy.
the odd-order approximations
For
=
=
(Pi), re
1 is the first-order approximation
3 is the second
example, if re
The approximation corresponding to re = 2 yields the same order
order.
equation (second) as does the re = 1 approximation.
Thus
calculation yields no new information about the angular
the re
distribution of the flux, and so we must go to re = 3 for a more accurate
In the same way n = 4 is of the same order as re = 3, and so the
model.
differential
= 2
after P3 is 7\ and so on.

of
the one-velocity transport model is obtained
The P3 approximation
from Eqs. (7.271) by retaining all relations which involve the harmonics
for re < 3.
These are
next better approximation
(a)
- - Tx 3 +
-2(*2 + 2.th*2 +
S2
S3 =
2(03
+ 2,t)33 +
where we have neglected the
'=
<4 term
-J-
<2
in (d).
,
^
(7.272)
(<0
REACTOR ANALYSIS
388
In the first calculation which follows
[CHAP. 7
we compute in detail the angular
distribution of the neutron flux in the vicinity of an interface between

This example demonstrates the application of the
two dissimilar media.
interface-boundary condition (7.100). The second example deals with
a slab of multiplying material, and in this calculation the emphasis is
placed on the specification of the vacuum-interface condition and on the
est imation of the critical width of the slab by means of the various models
which have been developed.
a. Angular Distribution of Neutrons at an Interface.
Consider
two
media of different nuclear properties which have the plane

There exists throughout this entire
boundary.
The problem is
space a uniform one-velocity isotropic neutron source.
to compute and compare the distribution of the one-velocity directed
semi-infinite
x = 0 as a common
flux in the two media according to the Pi approximation.

For this application the general steady-state solution to the transport
equation in the P3 approximation (7.272) reduces to the somewhat simpler

form.
dfa
^ + Zo =
So
(a)
dfa
, ,
dfa
dx
= 0
(d)
in which we have defined
2,
2,
2,4>3
2.D
(7.273)
(7.274)
and have recognized that for an isotropic source Sn = 0 for all n > 0.
Since the above equations describe the directed flux in a homogeneous
medium which contains a constant source, the application of these
equations to a two-medium system requires that a separate solution be
obtained for each region and the two results matched at the common
surface x = 0.
We solve the four simultaneous equations in (7.273) by eliminating the
three harmonics fa,fa, and fa and thereby obtaining a single fourth-order
differential equation in fa.
The solution of the fourth-order equation
may then be substituted into (7.273) and corresponding expressions
The procedure is straightforward
obtained for the higher harmonics.
Suffice it to say that the calcula
and will not be discussed in detail here.
tion may be carried out by first eliminating fa from (7.273c) by means of
(7.273d), then fa and finally fa by performing several differentiations of
the Eqs. (7.273a), (7.2736), and (7.273c).
The resulting expression for
TRANSPORT THEORY
sec. 7.5]
o may
389
written in the form
be
d*o
d2a
dx
dx2
7<o
= SiZjiSo
(7.275)
where
Z
P ~
3
35S3
92qSi
42
Equation (7.275) is an ordinary differential
Its solution
coefficients.
o(x)
y = SoSiSj
i" 352, i"

152i
equation
(7.276)
with constant
is easily shown to be
= ae
+ be-"" +
ce"z
+ aV
So
(7.277)
and d are constants to be established from the interface

condition, and the roots of the characteristic equation, m and n, are
wherein a,
6, c,
given by
m
2a
(7.278)

The solutions for the various harmonics are obtained by substituting

Eq. (7.277) into the set, Eqs. (7.273). Thus Eq. (7.273a) yields i{x),
Eq. (7.2736), 4>t(x), and Eq. (7.273d), 3(z). The results are
=
4n(x)
2{x)
e~ - - e + - e~ - - c
(7.279)
(n
n
m
\m
J +-/
iMae-* + ^Mbe-"" + ^Nce"* + iNde-"1
f/2ix - |2i0
-So
*^
(7.280)
.(*)
3
(
142;
amMemx
6mMe_mT
cnNeni + dnNe~
+ 32!/)
(7.281)
where
/ and
are constants of integration
and
(7.282)
If
the solutions (7.279) through (7.281) are used in Eq. (7.273c), then a
comparison of the coefficients of like functions of x reveals that both
must be identically zero.

The above relations are the general solutions which apply to each of
the two media in question.
Specific solutions are obtained for each
and
medium by requiring that the flux remain finite for all x. This immedi
ately identifies several of the integration constants, and it may be seen
that for the medium which occupies the space x < 0 (which we denote
REACTOR ANALYSIS
390
[CHAP. 7
by the index M), Eqs. (7.277) through (7.281) reduce to the

= aemuz
4$*\x)
4>{m(x)
<W\x)
+ ce>" + =^
(7.283)
e-* +
f
nM
\mM
iMâe"-* + ^Nttce""*
/"(ac) =
set
-2^'
e"* J
{flmuMMe*
(7.284)'
v
(7.285)
+ cnMNMe'*')
(7.286)
If
we denote by the index F the medium which occupies the space x > 0,
then the appropriate solutions are
=
=
4>[n(x)
<&"(*) =

*<3"(x)
be" +
de-"
2i,">
"
+
+
(7.287)
(7.288)
iMr6e-"** + iNwde-"'*
(7.289)
i4cF
(6^f* "
+ dnFNre-'*)
(7.290)
The parameters
m,i, n Af and 2V< are computed from (7.278) and (7.282),

using the appropriate cross sections for the particular medium i.
The constants a, b7 c, and d are obtained by the application of the inter

face condition (7.101), which, in the present notation, may be written
M0,m)
-1
= 4>r(fl,v)
<
The Pz approximation for the directed flux (x,/x)

ing the series [cf. Eq. (7.74)]
Hx,M) =
J) (^r^)
<
(7.291)
is obtained by
*-(:r) p(,l)
terminat
(7-292)
i-O
at
s =
<Hx,M)
If
we
of
3; thus, introducing the definition of the functions P.(m), we have
**o(x) +
*2(*)0V
1)
+ i*,(x)(5/i*
3M)
(7.293)*
apply this approximation to each side of (7.291), then for all values
we must have that
#"'(0)
#"(0)
(7.294)
The application of the condition

since n is an independent variable.
(7.294) to Eqs. (7.283) through (7.290) yields the following four simul
TRANSPORT THEORY
SEC. 7.5]
taneous algebraic equations in a,
a-b
[
mM
mj?
and d:
b, c,
c- d =
S0
1^
[
- A/F6
Um
A/ji,a
391
n'
J
+ NMc - NFd
L
= 0
(7.295)
The solution of these equations for the four unknowns when substituted
into the harmonic relations (7.283) through (7.290) gives a complete
description of the directed flux through the two regions M and F.
We illustrate the above method by means of a sample computation.
Let

If
we assume
2<,M)
2'*"
that
X)n
0.0022^>
S^>
0.99821*"
= 0
for n
20 = 2( 2,l)o = 2a
2JP>
= 0.5
> 1, then
2i
W>
jtf0 = 0
(7.296)
it follows that
= 2, 2,1)! = 2,
/i0
(7.297)
^1
22
23 2|
2<
These data may be used to compute the parameters m, n,
M, and N.
The results are:

nM = 0.054772,' *>
JV* = 0.00003333
mM = 1.9742jM>
A/.w = -0.9921
=
2.4412
A//, =
-0.7481
= 0.98762<f,)
JVF = 0.5382
The substitution of these values into Eqs. (7.295) yields for the coefficients
a =
c =
-0.09324
-8.634
The solutions for the harmonics

(0M)(x)
[M>(x)
= 10
are then
- 0.09324e1"42''"'i -
S O.SloSe0
^(x)
= 0.2517
d = 0.5212
b
8.634e0
06477S'1J"*
064772'""1
(7.298)
= 0.04625eI974J;<(,"
0^>(x)
0.039 13e1974S'(J"*
and
4>\P(x)
\">(x)
^)(x)
<tiF)(x)
= 0.5 + 0.2517e-2-4412'<r)l
+ 0.5212e--987s'<"*
= O.OSlSOe-2-4411'"'' + 0.2635e-0S,87<s'(')l
= -0.09414e-2-441-"'^
= -0.09849e"2 "I*."1*
+ 0.1402e--987,z'c')*
+ 0.05936e- M7eS'",J!
(7.299)
REACTOR ANALYSIS
392
[CHAP. 7
The corresponding solutions for the Pi approximation (diffusion theory)

are of interest for purposes of comparison.
The appropriate differential
equations are obtained from (7.271); these equations were developed in
Sec. 7.2f.
The general solutions for
and
i(x)
-4xDe"
(7.300)
+ CxDe~"
Ac-
(7.301)
(7.302)
-^Dê"""
(7.303)
tiPW
Ce- +
(7.304)
#"(*)
CDrxFe-"'
(7.305)
.4
F,
4>[M,(x)
and
by the application of the boundary
This yields
- (i/^)]
and
The angular distribution for the flux {x,p)

(*,/,)
io(x) +
(7>306)
given by
(7.93)
%4>x(x)n
The appropriate solutions to these equations for the system defined by

found to be
tf(x)
4>\M){x)
and
Table
oFKx)
\F){x)
7.1
= 10
8.728e0-""z'<J"*
= 0.3153e "87S'""*
= 0.5 + 0.7722e-1-2262'""*
= 0.3153e-1-2252'1"1
for various x and

by (7.298) and (7.299).
lists (x,p)/tf>0(x)
approximations
given
(7.307)
(7.308)
to the P3
The numbers in the
the constants (7.296)
is
according
parenthesis are the corresponding values obtained from the diffusiontheory relations (7.307) and (7.308).
These results have been used to compute the total flux 4>o(x) throughout
the vicinity of the interface.
Figure 7.15 shows this distribution at
various stations measured from the interface x =
in units of mean free
0

(7.88)
are
Ae + Ce~" +
We compute the constants
-D'0(x)
*o(x) =
4T>(x) =
condition (7.294).
(7.87)
In the region A/ these reduce to
2//).
and in the region
*,(x)
is
x2
= S0
=
Za4>o(*)
where
f-
-Dtfix)
paths
for each region.
The particular choice of constants given
TRANSPORT THEORY
sec. 7.5]
393
in (7.296) represents reasonably well the relative properties of
(M) type material and an absorber
(scattering)
a moderator
It
{F) type material.
is of interest to note the relative suppression of the flux in the absorber

5|
1
P3 approximation
Pj
approximation
****
*"
****
Interf ice
0>
>
Scat erer
ol
Absorber
65432101

Distance
from interface in mean free paths
Fig. 7.15 Total neutron flux in the vicinity of the interface between
scattering materials.
23
absorbing and
region as compared to the scattering region.

Note also that the relatively
higher neutron density in the scatterer results in a net flow of neutrons
into the absorber region.

Table
Distance
from
7.1
Angular Distribution of Flux at Interface According to

Pi and P3 Approximations
-1
-1
interface
\,
0.3708
0.2724
0.2(100
0.3352
(0. 1B54)
0.3268
JV"
0.3379
x,c>
0.3725
0.4956
(0.3716)
(0.2431)
(0.3477)
(0.3141)
0.4022
(0.4044)
(0.6523)
0.6154
(0.8718)
0.9859
(0.6395)
(0.5000)
0.5593
(0.5000)
(0.8046)
0.8548
(0.6859)
(0.5000)
0.4231
(0.7568)
0.7890
0 7215
0.4190
(0.3605)
(0.2209)
(0.3088)
(0.5000)
0.4546
0.3441
0.7453
(0.6284)
0.6579
0.4862
0.2558
(0.1282)
0.6269
(0.5000)
(0.7791)
0.9348
(0.5956)
(0.6911)
Figure 7.16 shows the relative distributions of the directed flux at

various angles to the x axis. Sketches a through e in the figure reveal
that, at all stations in the neighborhood of the interface, there is a
aoi.ovaa sisaivnv
jvho]
/,
jaquinu jo suoj'jnau Suiaoui ui aq1} suoiôaaip o > ii > X i

aqi
Aûofoui jo aq^ suoainau aara papâq oûi aqi joijaûi suoiSaa jo
'-a-i
aq^ jaqjosq'B
Xjuq %v jCiaAi^tqaj aSaiq saoutj^sip uiojj aqi aoBjaaiui ui
qôq mpaui jBaaAas) uraaui aaaj (sq^d saop aq^ uoi^nqu^sip
auiooaq
ApAi^Tqaa jaSa^j

"Oidojôsi
SIMX s! UMOqs u! ^L'l
'fyl'L
asaqj, s^nsaj a^tujsuouiap
(I)
()
uoipajiQ auisoD jo ucuinau 'uo|)Oui n
1
aaueisiQ
oij
0
wojj oDcpajui u; ueaoi osjj sqied
IJ9AV
()
jujnSjuy uopnquisip jo
xny puncuw am aaujjajui uaa.uiaq ub

jaqjosqu pun JaJa;}eas 'dJK) auoin aajj :(q}d (d) 'JMWOS I
(q) 'jaja^Bos
I
^ ijjj^ (a) :aauija}uj (p) 'jaqjosqB
-f
() 'jaqiosqra
(/) aajai}B3s jo
jo paôajip xng joj JBJ8A8S sau\v\ jo li
'jaq.iosq
.' (6) ajisoduioD qaâ^s
917,
jjjv
uoa-jnau
fjjyj
ijJIV
fjjjv
sec. 7.5]
TRANSPORT THEORY
395
the assertion made previously on the validity of the diffusion theory,

namely, that the flux distribution according to this model gives an
accurate representation of the flux only in regions far removed from
boundaries.
Figure 7.16gr is a composite sketch of the directed flux for several values
It allows a comparison of the relative density of neutrons moving
of ji.
in a particular direction to the total neutron population at each station.
b. Infinite-slab Reactor.
Consider an infinite slab of multiplying
material with width 2a operating at steady state.
The neutron sources
in this system are the fission reactions, and on the basis of the one-
vZ,
jlx
velocity model we find that the number of neutrons produced per unit
volume per unit time is given by the expression
4>(x,n)
dn =
= S(x)
vXf<f,o(x)
(7.309)
yzT^
2i
dx
(7310)
(7-310)
are given by (7.274), and the symbol
S.po
vZt = 2
- vZ,
is
22
IA
_
Z'Q
where the Z for all n >

,
defined by
dx
d0
djn
dfa
^+^
(7.311)
is
The general solution to Eqs. (7.310)

given by the relations (7.277)
through (7.281). Note that in the present notation, however, the
analogous
surface of the slab.
5,
3,
n =
1,
0(O,M) dn =
0}:
P.GO
(7.312)*
to the requirement (7.98) which

imposed at the
The substitution of the expansion (7.292) into (7.312)
is
This
is
condition at the center of the slab
(x
is
symbol 20
Because of the symmetry of the present
replaced by Z.
some
of
the
arbitrary constants which appear in the general
problem,
solution can be eliminated.
For this purpose we introduce the following
Pn(M) d" =
*-(0)
(^r) *,(0)
/-i
yields

The appropriate set of equations which describe this system in the P3

approximation are obtained from (7.273), using the definition (7.309)
for the neutron sources. Thus,
(7313)
ANALYSIS
REACTOR
396
[chap. 7
Thus it follows that
<^
0,(0)
If
we
= 0
= 0
0,(0)
(7.314)
these requirements are imposed on the relations

obtain the pair of equations
Mm(b
which
a)
+ Nn(d
constitute
c) = 0
(6
the determinant of these equations

acceptable solution to (7.315) is
b
o{x) may
a = 0
it follows immediately
written in the form
be
c) = 0
(d
condition on the quantities
Since
From this result
i -
a)
(7.279) and (7.281),
(6
and
a)
(7.315)
(d
c).
the only
does not vanish,

c = 0
(7.316)
that the general solution
for
[see (7.277)]
o(x) = A cosh mx + C cosh nx
(7.317)
0i(x)
fa(x)
i(x)
sinh mx
sinh nx
(7.318)
$ AM cosh mx + CN cosh nx
1423
(7.319)
nNC sinh nx)
(mA/ A sinh mx +
The arbitrary constants A and C

boundary condition (7.98). For the
(7.320)
are obtained by applying the outer

case n = 1, this yields at x = a
*o(o)
"
2l{a)
*'2(a) =
/Mm) p,(,l) d"
**(a)
X C^^)
j-0
/-i
- 803(a)
-o
The application of
(7.322)
conditions
l+-:N+
_ tanh ma
/23
12niV
iZii
.r
ma
(52V/8) cosh na
+ +
(5A/ /8) cosh
tanh na
22o
on
.
,r
-N
tanh ma
m
s
-z-M
cosh ma + (22o/m) sinh ma

cosh na
(22o/n) sinh na
-\
\
+
A C
C,
to the expressions (7.317)

and
and these
through (7.320) yields two equations in the unknowns
may be reduced to the form
these
two
3,
(7.321)
and for n

and the Eqs. (7.310) yield
tanh na
(7.323)
(7.324)
sec. 7.5]
TRANSPORT THEORY
397
where (7.324) may be identified as the criticality equation for the infinite"
"
slab reactor in the P% approximation.
Consider as a numerical example of the slab problem the system defined
by the following
set of parameters:
2 = 0.22,
2J
-0.0462,
Zi
= 0.5
JTo
= 0.62,
22 = 23 =
M, and
These data yield for the quantities m, n,

m = 1.9372,
M = -1.022
2/ = 0.12,
= 0.82,
2.
= 2.46
AT
(7.325)
2,
the values
n = 0.2965t'2,
-0.05820
The substitution of these results into the criticality equation (7.324)

gives the half-width a of the infinite-slab reactor which will be critical
with the composition
This solution
(7.325).
o2, = 4.196
is
found to
be
P3 approximation
(7.326)

The value of A/C is computed from (7.323), and the result yielded
A/C = -0.00002898. Thus the total flux according to (7.317) is
o(x)
- 0.00002898 cosh (1.9372,x)]
= C[cos (0.29652,*)
is
(7.327)
This system may also be analyzed on the basis of the Pi approximation.

The total flux in this case is given by o{x) = A cos Bx. In order to
obtain a consistent comparison of the Pi and P3 calculations we discard
the usual extrapolated boundary condition o(a) = 0 in favor of the
no-return current condition j_ (a) = 0. From the diffusion-theory expres
sion for j_ given in Eq. (5.49) we obtain the condition
tan Ba =
(7.328)
And, from the expression for the buckling (5.126),
using D =
we have
(D'-K1-5!) (-!-!)
1/3(2, The substitution
of the data
jl^2.).
(7.325) into
these relations gives
^
for the
Pi
= 0.2878
approximation.
0o(x)
Ba
= 1.261
o2, = 4.382
The expression for the flux

=
is
cos (0.28782,x)
It is of interest also to compute the half-width a on the basis of the

integral-equation technique developed in Sec. 7.4e. For the boundary
REACTOR
398
condition (7.260)
ANALYSIS
[CHAP. 7
= 0, we have the requirements
>(<!,*)
(7.262).
In the
present problem
e
= 1.046
and from (7.2626) we obtain B/2t = 0.3781. The substitution of this

result into (7.262a) yields the critical width Ba = 1.209, or in mean free
paths,
S,o = Ba
If
we apply instead
= 3.198
(7.330)
the no-return current
condition (7.264), the

criticality equation is given by (7.265). For fi/2, = 0.3781, we obtain
Ba = 1.343. The corresponding solution in mean free paths is
2a = 3.552
(7.331)

We summarize these results below:
Pi
Pi
4.382
4.196
*(o,x)
= 0
3.198
No return
current
3.552
It
is of special interest to note that the diffusion-theory methods, even

in the Ps approximation, yield significantly different solutions from those
of the integral (transport) theory model.
This is to be expected in the
present application since the system in question is relatively narrow
Mean free paths from center or slab ( 2, x)
Fig.
7.17
mations.
Total flux distribution in
an infinite bare-slab reactor by
Pi
and
Pt approxi
TRANSPORT THEORY
SEC. 7.6]
399
Note, however, that the Pa calculation

(roughly seven mean free paths).
results than does the Pi
integral-equation
with
the
more
closely
agrees
Finally, it should be recognized that, if a lower fuel con
calculation.
centration had been selected, the slab width a would have been larger
and all four solutions would be expected to agree more closely.
The total flux distribution in the slab reactor defined by (7.325) is
shown in Fig. 7.17 for both the Pi and the P3 approximations.
7.6
Computational
Methods for the Neutron Age
The purpose of this section is to display three computational methods

for the neutron age that are especially useful for hand calculation.
These methods are based on various approximate models which involve
It is convenient
of the physical system.
to discuss these at this point since they serve well to demonstrate the
application of the Fermi age model and the transport-theory methods
some rather gross idealizations
developed in the preceding sections.

The particular utility of the first model, which was developed by Fermi
The second model
and others, is in application to water-metal mixtures.
for extending these results by similarity techniques, and the

third is useful for materials of large nuclear mass. It should be pointed
out that each of these methods was designed primarily as a fast hand tool.
With the advent of the fast computing machine there is perhaps less
interest in these methods ; nevertheless, the development of the analytical
models is helpful for discussing some of the physical features of each

is a rationale
system.
a. Fermi-Marshak
Method for Hydrogenous Mixtures. A very
instructive and useful application of the general Boltzmann equation
(7.110) is the calculation
of the neutron
age in hydrogenous
systems.
This problem was first considered by Fermi,1 and later by Ornstein and
Uhlenbeck2 and by Marshak.3 The present treatment, which draws
principally from the work of Marshak, is based on several simplifying
about the physical situation that render the method
assumptions
especially attractive for computing the neutron age in water-metal
systems.
Other more accurate models for describing the neutron popu
lation in general hydrogenous systems are developed and discussed in
considerable detail in Chap. 12.
Special attention is given there to
provide a complete mathematical treatment along with a comparison of
The analysis which
the analytical and physical features of each model.
follows, on the other hand, is intended primarily as a demonstration of
1 E.
'L.
Fermi, Ricerca sci.,
7 (2),
S. Ornstein and G.
E. Uhlenbeck, Physica,
13 (1936).
MT-17 Series T.I.B. (1943).

R. E. Marshak, Revs. Mod. Phys.,
4, 478 (1937); and
19, 3 (1947), esp. pp. 200-201
R. E. Marshak,
and 209-211.
400
ANALYSIS
REACTOR
[CHAP. 7
the transport-theory methods
and to derive a suitable formula for

For these reasons some of the purely mathematical
computing the age.
steps are omitted from the analysis and left for detailed discussion in
Chap. 12.
We know from our earlier study of the Fermi age model that the age
is proportional to the quantity r2(w), the mean radial distance
squared at which fast neutrons from a specified source reach lethargy u;
the constant of proportionality in this relation is dependent upon the
nature of the source [see, for example, Eq. (6.52)].
Our ultimate objec
For this purpose we introduce the
tive, then, is to compute r2(w).
following model:
t(m)
(1)
The slowing-down medium
is infinite, homogeneous, and
isotropic
\
1
0 appear in this medium at a I

(2) Neutrons of lethargy u
constant rate from a single plane source of unit strength /
located at x = 0
I
The nuclei of all nonhydrogenous materials in this

>
medium are of infinite mass
(
(4) Elastic scattering is isotropic in the center-of-mass sys-V
(7.332)
terns of coordinates
1
(5) The absorption cross section of the medium is small com- I
pared to the scattering cross section
/
The system defined by these assumptions is evidently one-dimensional

and at steady state, and in this case the general Boltzmann relation
reduces to a somewhat simpler form.
The appropriate equation is
obtained from (7.116); in terms of the lethargy variable this may be
written
a V*(r,ttfQ) +
+
*(r,u,Q) = S(r,ufi)
hfu-
S,(M,) du'
L *(r,u''a,)
(M-";M'.Q')
da'
(7-333)
(7.114) and the transformation (7.127).
In the one-dimensional system the transport term reduces to the deriva

A more convenient
(x,u,Q)
and <S(r,M,0) * S(x,u,Q).

pair of functions is defined by the integrals
*(*,w,fi) =
S(x,u,n) =
%[**
as given in (7.78), and 4>(t,u,Q)
{x,u,Q)
jo
tive
S(x,u,Q)
d+
For an isotropic plane source producing

(7.76).
neutrons at lethargy u =
the function S{x,u,Q) =
Therefore,
after
S(x,u,n)
0,

(3)
5(x) S(u)
(7.334)
monoenergetic
(l/4x)
S(x) S(u).
(7.335)
TRANSPORT THEORY
SEC. 7.6]
If
401
Eq. (7.333) is integrated over the interval of
^,
and the expressions
(7.78), (7.334), and (7.335) are applied, the result is
fa
(x,utli) + S,(u) (x,u,n)
hL
$ S(x) S(u)
s,(w'} du'
fa-
*(x'M'-'1') ("mo;u',o')
da'
(7. 336)
Let us define
between r2(u) and x2(u).

function f(u;x,y,z) such that
the relationship
is
6.
Equation (7.336) gives a complete description of the neutron distribu

tion in the system specified by the model (7.332). If the solution of this
equation were obtained, the resulting expression for (x,u,n) could be
used to compute the quantity x2(u), the mean-square perpendicular
distance from the source plane at which neutrons reach lethargy u; this
calculation could be carried out by using a technique similar to the one
2d.
The immediate problem, then,
to determine
employed in Sec.
f(u;x,y,z) dx dy
dz
probability that neutron of lethargy

= zero released at origin reaches
(7.337)
variables, this function may be written
are independent
Since x, y, and
in the form
S(u;x,y,z)
Mup) f,(u;y) ft(uys)
(7.338)
fi
are the frequency functions for the individual variables.

We use this form to compute r2(u) thus,
where the
fr*f(uyt,y/)dxdydz
x2
y2
r2
J2m
wherein we have used
x*(u) =
/i(u;x)(s2 +
y*
oo
(7.339)
z2)
dx
Since
z2.
x2fi(u;x) dx
(7.340)
obvious that
Thus the
F(u) = x*(u)
age may be computed

=
x2(w) in
turn
x\u)
(7.341)
from the relation
iffy)
ix*(u)
(7.342)*
obtained from the definition
The quantity
is
t(u)
T\u) = Zx\u)
y*(u) +
x2dx
J-=^
J_m
dx
Ja ff
is
f"
Mu?)dz[m oo h(u;y)dy
+
ac
?()
/""
72(u) m
it

lethargy u in dx dy dz at station (x,y,z)
which corresponds to the relation (6.50)
(x,u,n) da
4>(x,u,n)
da
[see also (6.51)].
(7.343)*
402
ANALYSIS
REACTOR
[CHAP. 7
The calculation of the age requires therefore the solution of Eq. (7.336)
for the flux 4>(x,u,n). In carrying out the calculation (7.343), it is actually
more convenient to work with the Fourier transform of the collision
This technique is generally useful for
density than with the flux proper.
computations of this type, and we will have occasion to apply it to other
problems later, specifically to the calculation of the age using the ChristyWheeler kernel and in the analysis of hydrogenous systems (Chap. 12).
We develop the appropriate relations required for this technique in the
Since our primary interest here is to demonstrate
analysis which follows.
an application of the Boltzmann equation, we will not examine all the
details of the mathematical analysis but will instead concentrate atten
Thus the present
tion on the reactor-physics aspects of the problem.
analysis is to serve as an outline, and the reader is referred to the report
by Marshak and to Chap. 12 for further analytical details.
In accord with the remarks above
we replace the flux
in Eq. (7.336) by
the collision density F, where

F{x,u,n) = -Zt(u)*{x,u,y)
use also the mean free path
Uu)
-\-
the approximation being based on assumption
(7.345)
(5), Eq. (7.332).
The
substitution of these definitions into (7.336) yields
\ Hu)
h jo
du'
fa
S(x)
F{x'u',tl') (u>"o;',a')
We compute next the Fourier transform
F(z,H,M) + F(x,M,M)
!T{
Uu)n ^
da'
(7.346)
of this equation.
If
we
lX(()fM]<pU,W,M)
F(x,u,n)e't* dx
(7.347)
-^L
v 2jt
fa-
jo
du'
*(*'W''M,) (".w.-u'.O')
da>
(7-348)
to obtain an approximate solution to this equation which

may be used to compute x2(u). This may be accomplished by reducing
the above form to
set of coupled equations through the use of suitable
and 8.
The procedure
is
series expansions for the functions
is
Our purpose
2^
of (7.346)
is
then, the transform
<p(?,m,m)
!'
5{F(x,u,n)\ =
^=
call
[1

We
(7.344)
similar to
TRANSPORT THEORY
SEC. 7.G]
We define, therefore, the following
that followed in Sees. 7.2 and 7.3.

expansions in spherical harmonics:
*(,u,M)
(W,Mo;u',a')
403
- (^T^) **(*,) PW
m-0
(7-349)
(7.350)
-(u;u',0')
-a
a-0
y(fl)
rr(O')
where (7.350) corresponds to the expansion for q(y,^0;v') given in (7.119).

The coefficients <pm and 8 are obtained in the usual way by applying the
orthogonality properties of the spherical harmonics.
The substitution of these expressions into (7.348) yields
27
(^)
X
m-0
p.W
*.(,)
*-a'M,)
*a(Uy'Q,)
(^^j
m-0
/o"
- *,()*,]
=0
K(Q)
3- -a
rr(a')
Pm{n')
da'
(7.351)
easily evaluated with the aid of the orthogonality

we note that P(m) = y(tl)///.
properties of the spherical harmonics,
With this reduction, (7.351) takes the form
if
is
The integral over Q'
m-0
*-
set of coupled equations
multiply through by various P(m) and integrate over all

we obtain
the case n =
n.
in $m
we
Thus, for
if
This relation reduces to an infinite
/;
(^) "- w
(7.352)
e*i(f,M)
*o(,u') o(u;u',Q')
^"
#0({,u)
0,

[i
d'
(7.353)
REACTOR ANALYSIS
404
[CHAP.
and for n > 0,
.(,)
[(n +
-5
*-(*,') (;'
"
dM'
(7-354)
The systems of interest to the present application are hydrogenous

It is convenient from the standpoint of
mixtures of various substances.
the neutron physics to separate such mixtures into hydrogen nuclei and
all other nuclei.
If we denote by the index H the quantities that refer
to the characteristics of the hydrogen nuclei and by the index 0 the
properties of all other materials, then the scattering function 8 may be
written
(7.355)
v(//)
then
2 Z
A m =a-
Let
- A = ^-
= A(w>
and
But,
with S. =
+ ^^r
we know from our previous
for isotropic
and (7.138)]
scattering
In the
case of hydrogen,
(7.357)
A)<0)
system
[see Eqs.
(7.131),
= *(mo~ f)c"'~"
1 a
this yields
(1
study of the Boltzmann equation that
in the (C)
S(,Mo;u',Q')
(7.356)
= 5(Mo
- f//)e'-
(7.135),
(7.358)
(7.359)
where from (7.134)
in
By assumption
= e("'-u)
(7.360)
Eq. (7.332) all other materials have infinite mass;

therefore the scattering from these nuclei is isotropic in (L), and the
change in lethargy is taken to be zero. It follows that the appropriate
scattering function must have the form
(3) of
(0)
$ g(M'
u)
(7.361)
Note that the integral of this function over all u and no is unity. If this
function and (7.359) arc used in 3, the scattering function for a hydrog
enous mixture may be written in the form
(u)MoV,Q')
A(u')
(mo
<;)"-"
HI
- A(u')] S(u' - u)
(7.362)
TRANSPORT THEORY
SEC. 7.G]
The coefficients
For example,
from the integral relation
8 may be computed
o =
A(w')e'- +
, =
M(')e|
405
- A(u')l
[1
('
(7.130).
- u)
These functions may now be substituted into Eqs. (7.353) and (7.354)
and the various harmonics
obtained from a simultaneous solution of
a finite set of these relations which yields an acceptable approximation
function
to the transform
<.
The relationship of the function $ to the mean-square distance i2(m)

is now demonstrated.
To begin with we note that the formula (7.343)
may be written in terms of the collision density F; thus
f
^
x2dx
}-
F(x,u,n) da
(7-364)

The function F, however, is the inverse Fourier transform of
F(x,u,ri
ff-M*,u,M)}
-=
('
HZ,u,n)
i.e.,
d$
(7.365)
Moreover, 4> is given by the expansion (7.349) ; it follows, then, that the
integral in the denominator of (7.364) reduces to
^F(W)
dU
-~
30
jja J" (^) *-(,) P.(*0*-*
d{
m=0
-^mrsf'.MM'-",dl
(7-366>
we have applied the definition (7.28).

The integral which
appears in the last equation is the first of a set of functions called the
moments of 4>0. The general definition for the with moment is,
wherein
FTKx,u)
_ (-D
V2ir
f"
J-
{-*o(C,w)e-*
d{
(7.367)
Note that
F$>(x,u)
m F0(x,u)
-j= f"
V2ir J
is the Fourier inversion of the coefficient
JL
F(x,iM)
*0({,u)-***
<p0(,w).
ôFû)
(7.368)
Evidently,
(7.369)
REACTOK ANALYSIS
406
In an analogous fashion
we introduce
*iT"(,w) =
[CHAP. 7
also
F^x,u)e^ dx
-=
/
2ir J _
V
(7.370)
and observe that
-L=
V 2ir
Fo(x,M)e'^dx
(7.371)
S>0(,M)
/""
*i,0'(i,w)
is
If these relations are used in

the Fourier transform of Fa(x,u).
the integrals of (7.364), we obtain the results
x*dx
F(x,u,n) da
dx
^M0,u)
(7.372)
~B
-w*mu)
. F"(x'u)
x2
F0(x,u) dx
--m [S*a'tt)L.
The expression (7.364) may then
?()
1K4
be
written
[^f^l
- -
<PoU,)
J{_0
(7-373)
Thus we find that the calculation of the age requires

knowledge of the
zeroth-order harmonic of the Fourier transform of the collision density.
This function o(,M) may be obtained from the solution of Eqs. (7.353)
and (7.354).
Since only *0 and *i are involved, one needs to solve only
the relations
and substitute
l-K
Jo
^-
/""*o(f,M')o(tt;',Q') du'
Jo
- i\t(u) *i(,w)
+
V 2t
- *X.() *o(,u)
d5
*o(f,M)
(7.375)
"*i(,m')i( ;u',a')du'
the resulting expression for
$>o
into (7.374).
is
primarily of a mathematical nature

The remainder of this analysis
The details of this calcu
and of little interest to the present purpose.
lation have been carried out by Marshals and are presented in the refer
ence report.
In this calculation Marshak uses the coefficients for the
scattering function given in (7.363) and obtains an approximate solution
for
3>o
from Eqs. (7.375).
the age of

J~
277
/_
F{x'u',l) da =
f-.dx
J/
which
neutron
This result, when applied to (7.374), yields for
of lethargy
u, which first appears in the system
TRANSPORT THEORY
sec. 7.6]
at lethargy a < u, the following expression
+ h(u;a) + h{a)U{u;a) +
This equation applies when u
defined
a.
407
76(M;a)
I<(u;a)h{a)
(7.376)*
The various functions in (7.376) are
7,(a) = X<(a)e-*
Ii{a)
X,(a)e*

(7.377)
and
(7.378)
Age calculations12 for fission neutrons in various metal-water

of interest
to reactor
design have been
mixtures
out using the Fermiare given in Fig. 7.18 for
carried
Marshak equation (7.376). These results

aluminum, stainless steel, and zirconium as
function of the metal-to-
water volume ratio.

b. Equivalent Mean-free-path Method for Water-Metal
Mixtures.
An essential feature of the model used in the Fermi-Marshak calculation
of the age in hydrogenous systems was the assumption that all nonhydrogenous materials in the mixture have infinite mass.
Thus this model
attributes the entire mechanism of neutron slowing down to the scatter
ing collisions with hydrogen nuclei.
The heavier materials in the mixture
serve primarily to "spread out" the neutron population, and the amount
of this spreading-out effect at any particular energy is determined by the
relative magnitude
of the macroscopic scattering cross section of these

and of the hydrogen at that energy.
Since the macroscopic
scattering cross section of the hydrogen in most metal-water systems of
interest to reactor design is appreciably larger than the cross section of the
materials
1 L.
W. Nordheim, G. Nordheim, and H. Soodak, "Slowing Down of Neutrons in

and in Graphite," Metallurgical Project CP-1251 (A-1827), Jan. 20, 1944.
1 Nancy M. Dismuke and M. Ruth Arnette, "Age to Thermal Energy
(0.025 ev)
of Fission Neutrons in HjO-Al Mixtures," Mon P-219, Dec. 5, 1956.
H,0
REACTOR
408
ANALYSIS
[chap. 7
nonhydrogenous materials, 2^', over the entire slowing-down range, the

details of the energy structure of X',0) affect little the age of the mixture.
And, in first approximation, the spreading-out effect of the nonhydrog

enous materials may be described by an appropriate energy-average
Alumini im
- water
a t68'F
Stainless
steel
at 4 503F
- wat er
^\

DO
0.2
0.4
0.6
Metal
FiO. 7.18
- water
Zir conium
v/ater
at 68 F
0.8
1.0
1.2
1.4
volume ratio
Age of fission neutrons in water mixtures by Fermi-Marshak
method.
Thus the principal factors which determine

scattering cross section.
the age in metal-water mixtures are the volume fraction of water and the
average scattering cross section of the metal.
To illustrate this point further, consider the water mixtures of two
different metals a and b.
Let the average scattering cross section of
metal a be greater than that of metal b. If the volume fraction of
water is the same for each system, then we would expect that the age in
system a would be greater than the age in system b.
However, if the
two metals are not too dissimilar [i.e., if S<' is not much different from
TRANSPORT THEORY
sec. 7.6]
409
for
2*'], it is clear that there exists some volume fraction of water
> w*', such that the ages in the two systems are
system a, where
identical.
If one could somehow determine this value of vl\ then a
knowledge
of the age in system
could be used to estimate the age in
system a.
A clue
as to how two different metal-water
mixtures might be compared
Fermi-Marshak expression for the
We observe
age.
that this result is completely specified by the mean free path in the
mixture.
Thus, if as argued previously, the details of the energy struc
ture of the metal cross section are not a primary factor in the determination
of the age, then any two systems with the same mixture mean free path
would exhibit the same age.
One might argue, then, that the comparison
is obtained from the
could be based on a representative
value of
X(.
for the present the question of how one is to determine an

appropriate value of X( and inquire first into the details of the comparison.
If Aj denotes the mean free path in a water mixture of metal i, then by
Let
us defer
the usual relations

Xi0
where for the present systems
X(
~ X
^y^]"
and
Pi m density f of metal i
p m density t of water in system containing
y, = volume fraction of metal i
v\p m
(7.379)
metal
volume fraction of water in system containing
i
metal
The other symbols
in the equation have the usual definitions.

volume fractions are related by (3.26); thus,
vt
+ v?
i
The
(7.380)
Now consider water mixtures of two different metals a and b. If the

for a given slowing-down range is to be the same in the two systems,
then by our arguments above we require that
age
For the
(7.381)
(7.379), and introducing also the relation

the condition
yields
(7.381)
XJ we use expression
(7.380) for the
>,
-aRP- +
Aa
^
A
ib
v)
-~a'
Jib
Aw
- *)
(7-382)
Suppose that the age has been determined accurately by some method for
system a and that these data are available as a function of the volume
fractions of metal and water as shown in Fig. 7.18.
These results may
t These densities refer
to the pure material at the specified temperature.
ANALYSIS
REACTOR
410
[CHAP. 7
then be used to estimate the age in system b, since given a volume fraction
vi, of metal for system b, we can compute from (7.382) the volume frac
tion va for system a which would have roughly the same age. Using
this value of va we could read off immediately from the age-versusvolume fraction curve for system a the age for system b which contains
a volume fraction vb of metal.
The expression for
va
in terms of
in the form
Va
obtained from (7.382) may be written
vb
- x) + (x -
9a
1)
(7 383)

X
It is sometimes more convenient to work with the metal-to-water volume

fraction Vi, where
Vi

-g
be
and
- r^yi
(7.385)
written in terms of this ratio; thus,
- + (x v(ea
elY+W^ld
1)
Vb(6b
Va ~
1)
(7-386)
This relation, or (7.383), may be used to relate one metal- water system
We observe that the comparison is based on some suitable
to another.
value of the microscopic scattering cross sections of the water and metal
in each system and on the densities of the various constituents.
Evi
dently this technique may be used to extrapolate data from one metalwater system to another as well as from one temperature to another.
The selection of representative scattering cross sections to be used in

If we assume as a rough
(7.384) may be based on flux-averaged values.
approximation that the flux is l/E, then we may compute the a, from
the relations
">
=_
El
J
tb;-
E
(IE
=
Ml
'
Mo
Ju0
(r.(u) du
(7.387)*
Ei
< E0 and (Ea * Ei) is the slowing-down range of interest.

The equivalent mean-free-path technique has been applied to several
systems of engineering interest, and the agreement with more accurate
where
Table 7.2 lists some values for the age to

of fission neutrons in stainless steel-water mixtures at 450F
for various metal-water ratios.
Column 2 gives the age as computed
methods is surprisingly good.
thermal
TRANSPORT THEORY
sec. 7.6]
411
from the Fermi-Marshak relation, column
4 the age as estimated by the

and
column
method
6 the age estimated from
using A1-H20 data,
present
Zr-H20 data. Columns 3 and 5 list the metal-water ratios for the alumi
num and zirconium mixtures which correspond to the stainless-water ratios
Table
Age of Fission Neutrons
7.2
at
Volume ratio of
stainless to water
t, cm*
Fermi-Marshak
0.3
0.6
1.0
were
(2)
(3)
(4)
(5)
(6)
40.0
46.5
53.0
61.5
0.332
0.500
0.652
0.799
41.2
49.0
54.5
60.0
0.472
0.779
1.038
1.346
41.5
48.5
54.5
62.0
computed
The scattering cross sections for these cal

(7.387). These are listed in Table 7.3,
from
densities:
Table
SS-HA
450F
0.834 g/cm'
Pu = 7.75
= 39.3 barns
= 6.8
c. Age
Calculations
From Zr-H20 data

Zr-H30 ratio
along with the appropriate

From Al-HsO data

T
listed in the first column.

culations
Steel-Water Mixtures
A1-H20 ratio
(1)
0
in Stainless
450F
7.3
AI-HjO,
Zr-H20,
68F
68F
pIa,) =
1 0
2 .7
PAl
a\kn
42 .7
1 .5
by Christy-Wheeler
P<?r)
io
6.43
P7.T =
-(Zr) =42.7
= 9.0
Kernel.
The calculation
of
the age on the basis of the continuous slowing-down model has two
serious shortcomings.
The first arises in connection with the scattering
properties of materials at high energies, and the second because of the
use of the slowing-down density in place of the flux in the calculation of r2.
The first of these considerations is important when the scattering cross
energies and/or exhibits forward direction
In both
tendency [see, for example, Eq. (4.31) and related discussion].
instances the neutron will travel large distances from its point of origin
before the scatterings become sufficiently frequent and isotropic to
section
is small
at high
justify the
Usually after the

use of the continuous slowing-down model.
first collision the neutron energy is markedly lower, and these considera
tions become less important. Thus it is possible to improve the con
tinuous slowing-down model by computing from the actual source dis
tribution the distribution of first scattering
collisions
and using these
412
REACTOR
ANALYSIS
[CHAP. 7
starting points for the slowing-down process. This approach

is another extension of the first-collision source model introduced pre
viously to illustrate the kernel method (Sec. 5.7).
The second correction is an attempt to improve on the use of the
slowing-down
density in the calculation of r2 by introducing a more
statement
in terms of the neutron flux. If the usual coupling
accurate
relation [for example, Eq. (4.93)] were exact, then the use of either the
flux or the slowing-down density would yield the same result; since this is
not the
correction is required.
The distinction becomes important
when determining the end point of the neutron trajectory involved in the
If this is to be represented by an absorption reaction
r2(w) calculation.
at the specified lethargy, which is the phenomenon observed in making
age measurements with the use of absorber foils, then the neutron flux
as the

4>(t,u)
is required.
Since in general a neutron reaches lethargy u at some
point r' different from r, a relation is required between >(r,w) and q(r',u)
which accounts for local spatial variations in the neutron distribution.
As in our earlier work, we compute the age from the value of r2(u) for
For this purpose
a point source of fast neutrons in an infinite medium.
we select a point source of unit strength at the origin of a suitable coordi
The calculation of r2(w) is based on the
nate system in the medium.
continuous slowing-down model which has been modified to include the
first and last flight corrections proposed above.
Consider, therefore, a
source for the continuous slowing-down process which consists of the
first-scattering centers of neutrons emitted from a point source at the
If
the scattering cross section of zero lethargy (source) neutrons

is 2,(0), then the number of first-scattering
collisions per unit time in
is
at
given
by
dr0
r0
origin.
S(r0) dr0
These are the sources for the slowing-down

scattered at r0,
G(|r
f0|)
dl
(7-388)
4Tr2
"
process.
%rr()II
For each neutron
(?-389)
slow past lethargy u per unit time in volume dr' around r'.
The first
scattering sources then give rise to a slowing-down density at r':
q(T',u) =
In order to determine the flux
G(|r'
4>(t,u)
- r,|)S(ro) du
we must first convert the expression
for q(i',u) into a source strength.

This is accomplished by multiplying
by
du/l,
where
is
the
average gain in lethargy per collision; this
q
product gives the number of neutrons actually occupying the interval du.
To obtain the flux at r we integrate this source with the transport kernel ;
TRANSPORT THEORY
sec. 7.6]
413
thus,
*(r,u) du
where
T^
- r'|)
7\|r
^j
[q(r',u)
dx'
- r'D - hr=?\>
(7-390)
also follows immediately from Eq.

Note that the above expression for
the
of
case
a homogeneous, nonmultiplying medium with no
(7.242) for
It is of further interest to note that
direct external sources at lethargy u.
this relation for
reduces to the usual coupling equation when q is
spatially independent [cf. Eq. (4.93)]. For the present case

4>(r,u)
T(\x
- r'|) j
G(|r'
- r0|)S(r0) du
(7.391)
This expression has been referred to as the Christy-Wheeler kernel.

It will be more convenient in this calculation to work with the neutron
From the
flux (x,u) from a plane source than with the function (7.391).
function (x,u) we can compute x2(u) and then, with the aid of (7.341),
obtain r2(u). An expression for (x,u), analogous to (7.391), is
=
4>(x,u)
- f"
\
T(\x
- x'\) dx'
Jj
oo
G{\x'
- xu)S(xa)
dx0
(7.392)
It should be recognized that in this relation the function G is the plane

Gaussian kernel [see, for example, the form (6.28)], and T and S are the
The explicit forms of these func
one-velocity plane transport kernels.
tions are not required for this calculation; if desired for other purposes,
these may be obtained from the integral relation (7.191) by introducing a
suitable delta-function representation1 of the source [see Sec. 5. 2d].
The quantity x2(u) may

expression for
4>{x,u)
be obtained,
formally, by substituting the
from (7.392) into the relation
/
*2() = J-=^
/
x2(x,u) dx
(7.393)
${x,u) dx
As in the
case of the Fermi-Marshak model, it will be more convenient

to evaluate the above expression by introducing the Fourier transform
of the neutron flux.
Therefore, let
SiKz,u)|
1
(Z,u)
~j=
4>(x,u)e*t*
dx
These kernels have been computed in the Case-de Hoffmann-Placzek

report.
p. 4 in the reference
(7 394)
paper.
See
414
Now, the nth moment of
It
ANALYSIS
REACTOR
(,)
[CHAP. 7
is given by [cf. Eqs. (7.367) and (7.370)]
f*
i"
({,) =
x'4>(x,u)e*** dx
^=
(7.395)
follows, then, that

*m
/_",
J
Thus (7.393) may
be
(x,u)
dx = -s/2^9(0,m)
- ~ V2^(2,(0,w)
***(*)
written
[cf.
(7.396)
Eq. (7.374)]
a*
e(f,)
(7.397)
x2(w)
The calculation
it
T(\x
- x'\) dx'
T,
S,
j"
J"
T*G+S =
G(\x'
x0\) S(x0) dx0
(7.398)
follows that
= 6(,W) =
The numerator of (7.397) may

this purpose we require
~ 0(,u)
= d"gs
0()
be computed
+ g"s8 + s"g8
5{(x,u)\
and
ff{7VG*,S} = 0(*)ffU)s()
(7.399)
g(l)
(7.400)
s($)
from this expression.
2(s'g'd + sg'6' + s'gd')
For
(7.401)
B(*>")
= fl"<) 0(0) s(0)
T-
it
that
1
seen
\Si
is
0,
S,
T,
G,
{.
where the primes denote differentiation with respect to

The deriva
tives indicated here may be expressed as the various moments of the
If
transforms of the functions
and
as defined in Eq. (7.395).
=
we note that all the odd moments of these transforms vanish at
g"(0) s(0)
6(0)
+ s"(0) g(0)
9(0)
The application of this result and (7.400) to (7.397) yields

X
jHu)W - _
^(0)
p(0)
i"(0)
s(0)
_ r(0)
8(0)
I.
N. Sneddon, "Fourier Transforms,"

See, for example,
1

is
G,
If
of x2(m) requires, therefore, the evaluation of (,u).

This may be obtained from (7.392).
g(), s(), and 0() denote the
Fourier transforms of the functions
and
respectively, then, since
the convolution1 of these functions, i.e.,
(7.392)
pp. 23-27,
{7-WZ)
McGraw-Hill
TRANSPORT THEORY
sec. 7.6]
415
The terms at the right are the x2(u) values for the kernels G, S, and T,
respectively; for example,
g"(0)
Thus
But,
z2(w)
r2(w) = 3x2(m)
_9(0)
+ xj(u)
(7.403)
FJ(u) + F|(u) + FJ(u)
(7.404)
a(ii) + x%(u)
therefore,
=
7*(u)
The first term at the right is immediately recognized as the r2(u) com
puted for a point source on the basis of the Fermi age model [see Eqs.
(6.49) through
(6.52)], since1
*oW = jGdT
= 6t0(m)
(7.405)

The last two terms are easily computed; thus
SSdi
r()
'
[2.(0)]'
pe-z.Wrdr
(7.406)
[Z.()]s
The substitution of these results into (7.404) yields an expression for r2(u).
If we define the age t(u) for this system by the relation t(m) = %r2(u),
then we obtain
t(w) = tg{u) +
This expression for the

Table
7.4
Age
Material
Beryllium
Graphite
3[Z.(0)]2
+'
(7.407)*
3[Z.()]
age offers some improvement
over the expression
of Fission Neutrons in Beryllium
and Graphite
Fermi age
Christy-Wheeler
Experiment0
79 . 7 em'
84.0
97.2
364
364
352
U.S. Atomic Energy Commission, "Reactor Handbook:

Physics,"
p.
485,
tq(u) obtained from the Fermi model and should yield reasonably good
results for nonhydrogenous materials.
Some sample results are given
in Table
7.4.
1 The symbol tq(u) denotes the age based on the Fermi continuous
model.
slowing-down
REACTOR ANALYSIS
416
of Fission
d. Age
neutrons
Neutrons.
Inasmuch
[CHAP.
as
primary source of
of particular interest
the
in a reactor is the fission reactions, it is
to develop a suitable expression for the average age of neutrons which

Various methods have
appear according to a specified spectrum j(u).
been developed for computing the age of neutrons slowing through a
particular lethargy interval, and three of these have been discussed in

this section. The method selected for a specific application is deter
For our
mined, of course, by the material composition of the system.
present purpose we require only that there be an acceptable method for
computing the age t(m;m') for the interval (u,u'). These results may then
be used to compute the average age
Table
to the spectrum
corresponding
Histogram of Fission Spectrum
7.5
=0
(u
corresponds to
Mev)
10
An
it
Am
0 0.5
0.5 1
1 1.5
0.021
0.104
0.207
0.231
0.183
0. 117
0.067
3.5 4
4 4.5
0.035
4.55
0.0093
0.0056
0.0034
0.0014
0.0002
1.52
f (w)
2 2.5
2.5 3
3 3.5
Consider,
therefore,
i*
0 0151
5 5.5
5.5
6.5
6.5 7
6
a normalized
function
$(?*)
which
describes the
If t(m;m') is the
distribution of neutrons appearing from fission sources.

age of neutrons of lethargy u which first appeared from j at lethargy u',
then we define the average age to u for the spectrum from 0 to u by the
relation
f
If
f\(u;u')
j(u') du'
(7.408)
}-^j-u
the integrals are to be performed numerically, this expression may be
approximated
by
N
Y
"-'
^
t(u;u) j
A?<
(7.409)
^ J-AWn
n-1
where the quantities un may be denned by the relation

n-1
A?im
-y
Awn
n >
At/
UlS-y
sec. 7.6]
TRANSPORT THEORY
In many instances it will
be convenient
417
to work with the spectrum g(w)

One that has been used frequently is given
in the form of a histogram.

in Table 7.5 (see also Fig. 7.19). It should be recognized that the
quantity j* listed in this table is the fraction of neutrons which appear
1
II
All 2
Fig.
7.19
^11 .
4
1
>
Histogram of fission spectrum.
from the fission spectrum in the indicated

3* and j are related by the equation
lethargy interval.
Thus the
Si = i Aw

In this notation the average
age
to thermal of fission neutrons
is given
by
f(u)
r(u;un)it
(7.410)
CHAPTER
REFLECTED REACTORS
8.1
Introduction
The bare reactors discussed in the preceding

chapters can easily be improved from the viewpoint of neutron economy
by the addition of a surrounding nonmultiplying medium which serves to
a.
Reflector Savings.
reflect back some of the neutrons escaping from the core and thereby
allows them further opportunity to cause fissions.
The combination of
in
a reflector depends, of course, upon the usual
materials to be used
of materials compatibility, economy, strength, shielding,
On the basis of neutron conservation alone, there are two major
benefits to be obtained from the addition of reflectors:
requirements

etc.
(1)
(2)
Neutron escapes from the core are reduced

Neutrons are moderated in the reflector
In the models which have been used for a bare reactor, the vacuum
has been treated as a perfect absorber for neutrons; thus, the addition
of any material whatsoever, no matter how thin or how highly absorbing,
to the outer surface of the core will be of some advantage, since it will
always return some fraction (albeit small) of all neutrons incident upon
it from the core. In our work with the diffusion-theory model we found
it useful to describe the reflective properties of a medium by means of
This concept is especially appropriate in the consideration
of multiregion reactor systems.
Since the albedo is the fraction of
neutrons incident on the surface of a medium which are reflected back, its
the albedo.
value depends upon the thickness of the medium as well as upon its
It is clear that, although a thin reflector will
absorption cross section.
to the core, it will also allow a certain
and thereby be lost to the system.
As the
thickness is increased, the fraction of neutrons lost by escape through
It is neither profitable
the outer surface of the reflector is decreased.
nor practical, however, to employ reflectors of very large thickness.
always
return
some neutrons
fraction to diffuse through
Although the albedo increases with increasing thickness (Fig. 8.1), little
is gained by extending the thickness beyond a few diffusion lengths.
The second advantage offered by a reflector is derived from its moder
This characteristic is especially important in dealing
ating properties.
418
419
REFLECTED REACTORS
SEC. 8.1]
with thermal (or partially thermal) reactors.
In all reactor cores, fission
neutrons appear at all energies, and if the nuclear properties of the system
are designed so as to exploit the thermal characteristics of the fuel, it is
Depend
necessary to introduce a moderating material into the system.
ing upon the engineering requirements of the reactor, it may not always
directly with the
be possible or desirable to mix these low-mass materials
In these cases the reflector regions may be composed of moderating

Thus
materials, and much of the slowing down allowed to occur there.
reflectors in many reactor systems may serve as major sources of thermal
fuel.
This arrangement has some advantage over systems in which

neutrons.
the fuel and moderator are intimately mixed, since the slowing down
Core
01
4L
3L
ZL
Reflector thickness

Fig.
Albedo as a function of reflector
8.1
thickness.
Flo. 8.2
Hectors.
Slab reactor with infinite re-
the resonance range of the fuel takes place in regions which

contain no fuel.
This generally results in an increase in the resonancethrough
escape probability.
The moderating properties of
a reflector are not required
(or desired),
the principal concern is to
of course, in fast reactor systems.

There,
minimize the amount of low-nuclear-mass materials so as to realize the
"fastest" neutron spectrum possible. Thus the primary function of a
reflector in fast reactors is to reduce leakage losses.
There
third advantage to be gained from the use of an outer reflector

on a reactor, namely, that the reflector may also serve as part of the radiaation shield (or the shield may in part serve as a reflector).
This effect
has not been included in (8.1) since for the purposes of this study the
essential features of multiregion systems are embodied in the first two
is a
considerations.
A convenient measure of merit of a reflector is the so-called reflector

savings. The reflector savings gives the decrease in the critical core size
of a bare reactor which could be realized by the addition of a particular
reflector.
To illustrate this idea consider, for example, an infinite-slab
reactor of half-width a.
equation
In the one-velocity approximation, the criticality
is given by
vZ, = 2 + DB2
where
5r/2o.
(5.126)
This relation defines the critical fuel concentration
420
REACTOR ANALYSIS
[CHAP. 8
is,
for the dimension a. Now assume that this slab is reflected by two semiinfinite media of graphite as shown in Fig. 8.2. The albedo for a semiinfinite medium is given by (5.116), and for the properties of graphite
listed in Table 5.1, this relation gives approximately 0.92; that
92
per cent of all neutrons leaving the faces of the slab are reflected back.
Thus the leakage losses expressed by the DB2 term in (5.126) are con
siderably reduced, and the multiplying slab with the proposed reflector
could be maintained at criticality with
markedly lower
of fuel.
On the other hand,
we wish to retain the original
(bare-slab) fuel concentration, the width of the core in the reflected system
could be reduced.
It this difference in critical size which we call the
reflector savings; thus
the critical size of
bare system and or
a<>
a
configuration
reflected system with the same fuel concentration,
that of
if
is
is
if
concentration
Reflector savings =
This number
the addition of
relative
(8.2)*
is
saved by
a reflector.
The reflector savings depends, of course, on the
nuclear properties of the core and reflector as well as on the
a measure
of the amount
thickness of the reflector itself.
of core material
In dealing with reflected configurations,
called the equivalent bare dimension and, as

of Sec. 6.6,
may be
useful in determining the nuclear properties of complicated
ao
is
the number
commonly
extremely
it
previously demonstrated in the sample calculation

multiregion reactors.
b. Analytical Models.
is
is
The general problem of treating the slowing

down and diffusion of neutrons in the reflected system
analytically
difficult, even in the case of a highly thermal reactor in which the principal
interest
centered in the diffusion of thermal neutrons.
The principal
are involved in making the following calculations:
Losses of neutrons by absorption and leakage as they slow down to
thermal: the presence of the reflector introduces new absorbing materials
and a new spatial dependence for the flux, leading to complicated expres
difficulties
1.
sions for the absorption
and leakage probabilities.
Spatial distribution of the neutrons which enter the thermal group:

because of migration of the neutrons during the slowing-down process,
the spatial dependence of the slowing-down density at thermal will not
be the same as that of the thermal fissions; moreover, there will be an
2.
apparent source of thermal neutrons in the reflector, since some neutrons

are thermalized there.
In all the preceding chapters the focus of attention was on the develop
ment of fundamental concepts and suitable models for describing the
The two basic
neutron population in elementary reactor geometries.
models so far examined,
applied to bare reactors.
how these methods
model and the Fermi age, were

to show
the present chapter our purpose
the diffusion
In
may be extended,
is

a0
then we define
with suitable
modifications,
to
REFLECTED
SEC. 8.1]
REACTORS
421
more practical configurations, which, as we have already observed, invari

These modifications are complicated by
ably involve multiregion media.
The nature of these prob
listed
above.
the two primary considerations
lems is well displayed in the case of the continuous slowing-down
diffusion
model.
It will be recalled that in developing the age-diffusion model some dif

ficulty was encountered in working with boundary conditions involving
This problem was circumvented by eliminating
the energy variable.
the energy variable from the boundary condition statement for the outer
If the continuous slowing-down diffusion model

surface of the reactor.
is applied to practical systems involving several regions of dissimilar
composition, it is clear that a further complication arises in establishing
We observed above
workable relations for the interface conditions.
that because of this dissimilarity in material composition the spatial
Thus it is not possible
forms of the neutron flux at various energies differ.
to describe the spatial distribution of neutrons of all energies by means of
function, and so analytical solutions are not generally possible
for these systems. One way in which this difficulty may be overcome
a single
involves the
down
use of erwr^y grmi-pn
In this method the entire slowing-
range is divided into various energy intervals, and an appropriate

equation of the one-velocity diffusion type is used to describe
differential
the spatial distribution of the neutrons of each group.

But, since the
various groups are coupled through the slowing-down-density term, this
approach frequently yields a formidable set of simultaneous equations
which are best handled by means of fast computing machines.
The
computational techniques involved in this method are discussed at some

length in Sec. 8.8.
It is clear from the preceding discussion that a very accurate description
of the neutron population in an actual reactor requires extensive computa
tional effort. For the present, our attention will be directed to less
exacting
models which
are amenable
to hand calculation.
We begin
with the application of the one-velocity
diffusion model to multiregion

media.
Since this model cannot account for the distribution of neutrons
in slowing down, its use in analyzing reflected configurations serves to
determine only the reduction in neutron escapes that can be achieved by
It is important to note, however, that many of

the methojjs_jind_ results developed in the one-velocity treatment are
directly applicable _to the multigroup model.
The second basically different approach to the multiregion problem is
the two-group model.
This model attempts to describe the distribution
a
particular reflector.
in both space and energy by separating all neutrons into

thermal neutrons and nonthermal (fast) neutrons.
The spatial distribu
tion of neutrons in each group is given by a one-velocity diffusion relation
in much the same manner as in the multigroup treatment; however, the
of neutrons
REACTOR ANALYSIS
422
[CHAP. 8
two-group model developed here differs markedly from the general multimodel in one essential feature.
The two-group treatment is an
exact analytical solution of a particular set of differential equations which
describe an idealized reactor system, whereas the solution from the multigroup
Although the two-group

group method is an iterated (numerical) result.
model would appear to be very crude, at first glance, it has proved to be
useful and effective for preliminary analyses of complex
extremely
systems.
8.2
The One-velocity Model
a. General Equations.
We have noted that in reflected reactors,
which are dependent at least in part on thermal fissions, the reflector
It is clear that, on
serves as an important source of thermal neutrons.
the basis of this fact alone, the spatial distribution of the slowing-down
density at thermal cannot be proportional, everywhere, to the distribu

tion of thermal fissions.

Since this is a fundamental premise of the onevelocity model, it is to be expected that the application of this model to
reflected systems will yield at best a poor approximation to reality in
In fact, the most that can be said with

most cases of practical interest.
this model about the distribution of thermal fissions is~that it_is propor
tional to the thermal flux in multiplying regions and zero everywhere
However, there is some advantage to be gained from this applica
else.
tion since this model will at least describe the reflective characteristics
Furthermore, it often yields a
of the various nonmultiplying regions.
useful first approximation to the neutron distribution, and the analytical
treatment of the one-velocity model will be helpful in developing the
multigroup methods.
Throughout the analysis which follows we assume that the reflected
reactor consists of a central core of multiplying material which is inter
nally homogeneous, and surrounding the core, although not necessarily
covering all its surface, is another internally homogeneous medium which
Thus all the reactors
is nonmultiplying and which we call the reflector.
This does not limit our analysis,
considered here are two-region systems.
however, since all the fundamental ideas and techniques involved are well
The
demonstrated by these systems and are more generally applicable.
analysis of systems having three or more different regions can be carried
These calculations are merely more laborious.
out by the same methods.
The basic relations which give the distribution of neutrons in the core
(C) and reflector (R) of a one-velocity
- DcV*Mr)
system may be written
2<c'= rZ/*c(r)
rZ/*c(r)
(a)
(8.3)
-DmW+m{t)
2<fi(r)
= 0
(b)
REFLECTED REACTORS
SEC. 8.2]
The boundary conditions to
423
functions
be satisfied by the
4>c
and
4>R
are
assumed to be:
(l) Symmetry
in the core, or nonsingularity of the flux at the
core center
Continuity of the neutron flux
reflector interface
The flux must vanish on all extrapolated
(3)
reflector interface
Continuity of the net current at all points of the core-
^)
at all points of the core-
outer surfaces
of the assembly
Throughout the remainder of this section we will be concerned with

the solution of Eqs. (8.3), using the conditions (8.4) for three basic
For convenience in notation we write the differential
reactor geometries.
Wc(r)
V^(r)
BJ*c(r) =0
B2
- . [2/ - Z]
(ft)
=04-7)~
if
is
0)
b. Spherical Geometry.
The case of a spherical core (radius
and
particularly simple
spherical shell reflector (outer radius Ri)
spherical
In this case the general solutions of Eqs. (8.5) are
assumed.
symmetry
is
cosh xRr
n
CR
sinh xnr
K,
r<
sin
/ts
sinh xR(Ri
(8.8)
(2) and (3) in (8.4) yield
MRo)
MR*)
9)
The continuity conditions
<P(r)
c,
Br
r)v
, .
<Pc(r)
and discard the cosine term in

easily shown that the solutions take the form
we
(8.7)
is
apply this condition to
0,
= ft! + 0.71X'*'
with
Also,
(8
If
4>k(Ri)
is
=
Since the function cos Br/r
we take Cc = 0.
singular at
condition
in
for
the
by
(8.4), we have
present case,
(4)
it
6)
AK
cos
Br
Cc
(8
.
sin Br
Ac
= 4>u(r)
. .
(r)
c(r)
,
>c(r) =
,

equations in the form
ANALYSIS
REACTOR
424
When the results (8.8) are used in the first of

between A and C is obtained; thus
=
sinh
[CHAP. 8
these relations, an equation
sin BR0
hr(Ri
(8.10)
Ro)
If Eqs.
(8.8) are used in (8.9), and the flux relation is divided by the cur
rent relation, the constants A and C cancel out and the following equation
results :
BRo cot BRo
- [xrRo
coth x(#i
Ro)
1]
(8,11)*
In order that the system operate at steady state, the above relation
is
is,
This
ship must be satisfied by the parameters of the core and reflector.
in fact, the criticality condition for the one-velocity two-region
relation
spherical reactor.
The computational procedure involved in the application of this relation is straightforward and follows closely the methods
used in bare-reactor calculations.
that
The only difference, of course,
the size and material
it
is
is
is
is
is
and its constants are x and DR. The problem

to determine the critical
concentration of fuel in the core.
first to assume
The usual procedure
This
value for Dcdetermined in large part
easily done, since Dp
is
(presumably known) and only

by^the nonfuelônstitujnôf_îe-ore
slightly affected T)y"the fuel concentration.
Given Dc, then, the righthand side of (8.11)
completely determined, and the left-hand expression
may be solved for the corresponding value of B. The substitution of this
result into the B2 equation of (8.5) yields the fuel concentration.
Having determined the quantity B, the flux relations (8.8) are now
is
is
Note that, as in the bare-reactor results, the expres

completely defined.
sions for the flux include an arbitrary constant
obtained
(or C) which
from the initial conditions.
convenient in some cases to evaluate
(It
this constant in terms of the total power output of the reactor.)
Of
is
the comparison of the results for the reflected sphere

and for the bare sphere. These are given in Table 8.1.
Before discussing the analogous treatment of other geometries, some
of the physical consequences of the addition of the reflector will be con
sidered.
The results which follow are developed for the special case of
some interest here
the sphere, but similar relations can be obtained for the cylinder and the
rectangular-block
reactors.
We begin with
study of the reflectedsphere system in which the extrapolated
thickness (Ri Ro) of the
reflector
small compared to both the core radius Ro and the diffusion

length in the reflector Ls.
(Admittedly the diffusion theory postulated
is

transcendental
properties are here related by
may be used, consider the following situ
ation: A spherically symmetric reactor ol core radius Ro and outer reflec
tor radius /?i
to be analyzed.
The reflector material has been selected,
To illustrate how
equation.
REFLECTED
SEC. 8.2]
REACTORS
425
above does not apply well in this case, but the qualitative behavior is
correctly presented.)
Now, for small values of the argument,
the hyperbolic
is
cotangent
given by the expansion
coth x =
(8.12)
we assume that Dc ^ DR, then the application of the first term in the
above expression to the criticality equation (8.11) yields
If
BR0 cot BR0
But, by assumption, the right-hand side

Table
8.1
Comparison
(8.13)
is a large number; thus
and Reflected-sphere
of Bare-
Flux in core
Definition of B*
- (,S/ -
Geometry
A' sin Br
This
Ro
transcendental
Ro
Ri
0.
Ri
s,
(8.14)
in (8.13) gives
BRo
when
Ro) + 1]
is
use of this relation
BRo
an expression for cot x in the case x = ir
cot
The
- S.]
- 1 - - UkRo coth
and
Ro
1.
therefore,
Bfto cot BRo
e)
[X/
(ir
We require,
S.)
- -r
state
sphere
^4sin Br
Condition for steady
where
Reflected
Reactors
~
R\
Ro
(8.15)
Thus in the limit for very thin reflectors the

(8.11) goes over into the bare-reactor result.
have assumed Dc DR, the extrapolation distance
relation
(Note that, since we

the same for core and reflector.)
The approximations introduced
is
above may be used to estimate

very thin reflector to
reflector savings realized by adding

Bare sphere
BRo^t.
the
bare spherical
REACTOR ANALYSIS
426
[CHAP. 8
Let the critical radius of the bare system
reactor.
definition
be
Rao',
then by
Reflector savings = R0o
Ro
(8.16)
where Ro is the core radius of the reflected reactor having the same core
If we apply the requirement that xr{Ri Ro) <SC 1, then
composition.
the criticality relation takes the approximate form~~
BRo cot BRo
- (-
an
Dc
Ro
\Ri
Ro
l)}
(8.17)
Since, by assumption, the reflector thickness is also small compared to Ro,

the right-hand side of (8.17) is very large, and it follows from the approxi
mation (8.14) that the above relation may be written
-r~-^
Dc\f
BRo
BRo
If we
that T /Ro
use the assumption
l)
<C
1,
with
f - R,-
Ro
(8.18)
then the solution of this equation

for Ro yields
We know from our relations for the bare spherical reactor that Roo =
therefore, the substitution of (8.19) into (8.16) gives
Reflector savings
In many
thin reflector
r/B;
(8.20)*
thus, the addition of a reflector of

extrapolated thickness
Ri
R0 to a bare spherical reactor allows the
critical core radius to be reduced by the amount f. The linear relation
1
ship (8.20) does not hold, of course, if the inequalities xr{R\ Ra)
cases of interest
Dc
DR
In these cases, the critical

and (î Rq)/Rq<K 1 are not satisfied.
for
the
reflected
must
be
computed from (8.11) and the
radius
system
It should be borne
result substituted into (8.16) to obtain the savings.
in mind also that the result (8.20) applies specifically to spherical systems,
although this expression is frequently used as a general relation for other
geometries when making quick first estimates for the effect of a thin
reflector.
The general approach used above may also be applied to a system

In this case we impose the restriction that
having a thick reflector.
Ro) >s> 1 and furthermore that xrRq ^> 1 ; i.e., both the reflector
xh{Ri
and the core size are to be much larger than the diffusion
of the reflector material.
Because of the former requirement,
thickness
length
coth
xr(Ri
Ro)
1,
and (8.11) is approximated
BRo cot BRo at
- ^ (XsRo +
1)
by
(8.21)
RKFLECTED REACTORS
sec. 8.2]
427
By the latter requirement, the right-hand side of this relation
is large and
negative, and therefore the approximation (8.14) applies for the left-hand
side. It follows that
T
DCLR
and by (8.16)
Reflector savings
DCLR
thick, reflector
(8.22)*
An accurate calculation for the reflector savings based on the solution

of the criticality equation for the onevelocity model will yield a curve of the
The approxi
shape shown in Fig. 8.3.
mations (8.20) and (8.22) apply for the
extreme values of reflector thickness.
The problem of the spherical one-veloc
ity reflected reactor may also be studied
Reflector
Fig.
thickness
Reflector savings.
8.3

Thus let
by means of the albedo concept.
us denote the albedo of the spherical shell reflector by a, where
ot
fraction of neutrons incident upon the reflector from

the core which are returned to the core
This number
(8.23)
function of the inner and outer radii of the shell (R\ and
transport and absorption cross sections of
In general, a is also dependent upon the angular
the material of the shell.
distribution of the incident neutrons; in the present analysis we select
the distribution which is consistent with diffusion theory.
The one-velocity flux in the core which satisfies the differential equation
,
(8.5a) must now meet the boundary conditions
is a
Ro) as well as the (one-speed)
(1)
(2)
4>c(r)
j^(Ro)
for r
a??{Ro)
< Ro
(8.24)
The solution of the core equation which satisfies condition (1) in (8.24) is
given in (8.8).
The substitution of this expression into condition (2) in
(8.24) yields the relation
Dc
Ro
(1
- BR0 cot BR
which corresponds to (8.11).

Regarding
has the limiting forms [using (8.14)]
0
1
~*
(8.25)*
>-K^-:)
+ 2DC
BR0 cot BR0 =
R0 and
for
Ra
Dc
as being fixed, (8.25)
428
If
REACTOR
the concentration
ANALYSIS
[CHAP. 8
of materials in the core is held constant, the reactor

the radius Ro as the albedo of the
reflector is increased ; this behavior
may be kept critical by reducing
of the critical radius is illustrated

in Fig. 8.4.
schematically
The
albedo of the reflector may be
Critical core
radius Jf
increased by (1) increasing the re

flector thickness, (2) increasing the
macroscopic scattering cross sec
Albedo
Fio.
radius versus reflector

albedo for fixed core composition.
8.4
Critical
tion of the reflector, and (3) de

creasing the macroscopic absorp
tion cross section of the reflector,
Fio.
8.5
Completely reflected cylindrical
Fio.
8.6
Cylindrical
reactor with side
reflector.
reactor.
Sec. 8.6 by means of an approximate
method which entails an iterative
procedure.
B2<pc(p'z)
Mp'z)
*c(p'z)
tp
1. Cylindrical Core with Reflector on the Curved Surface.

The geometry
The
and dimensions of the system in question are shown in Fig. 8.6.
appropriate forms of the differential equations (8.5) for the case of axial
symmetry are
\tp
(a)
(8.26)
The solutions to these equations are to satisfy the boundary conditions

In each of the equations (8.26) the flux
and
separable into
(8.4).
is

c. Cylindrical Geometry.
The general problem of a finite cylinder of
multiplying material completely surrounded by a reflector cannot be
For the present we will consider
easily solved analytically (see Fig. 8.5).
two less general cases.
The completely reflected reactor is treated in
that
dependent component;
4>c{p,z)
*aM
The substitution of
4-
dp
[pF'c(p)}
-B2
(a)
\\
Gc(z) =
AJ0(\cp) +
(8.30)
A2Y<,{\cp)
CJo(\ap) + C2K0(\ap)
cos p-cz + Ai sin p.cz
Fa(p)
as follows
Fc(p) =
Ga{z)
= 2
(8.29)
G^W ~
^
_ Mc
GcW ~
=
s^Wlwr
pF*(p)
The general solutions to Eqs. (8.29) are
d sin
GR(z) = C3 cos iiRz +
(8.31)
nRz
order to meet the boundary conditions
(1) in (8.4) the coefficients

zero,
F0(Xcp)
must
be
taken
to
be
since
singular at
C<
=
Now consider condi
and sin ncz
not symmetric about
is
0.
that the flux be continuous all along the

The application of this condition to Eqs.
(2) in (8.4) which requires
This equation can
cos^^=
a(po,z)
C3 cos
be satisfied
(u^Ci/o(Xpo)
for all values of

pc =
AiJo(\cpb)A3
Mfi
If
CiK0(\apo)]
(8.33)
AJ o(Xcp)
We now satisfy
condition
cos nz
c(p,z)
3,
we absorb the arbitrary constants Ai,

and C3 into
A, the solutions for the flux may be written in the form
<Mp,z) = KVo(Xsp)
C2K0(\rp)]
(4) in (8.4)
32.
if
4>c(po,z)
interface.
core-reflector
and
(8.31a)
(8.316) yields
tion
is
0,
At, Ai, and

p
, .
is
"c
-xi
In
CrC(z)
=
~
Gc(z)

into (8.26) yields the equations
are independent variables, the above relations can be satis

constant; therefore, we set
each term
-ApF'cW
pFc(p)""cv"J
It
GR{t)
fied only
if
and
,R o7)
Fab)
these relations
Vrr
fit c(p)
Since
= Fc(p) Gc(z)
M|
429
X2,
REACTORS
is,
REFLECTED
SEC. 8.2]
cos hz
single constant
,g
that the flux vanish at every
REACTOR ANALYSIS
430
point of the extrapolated
[CHAP. 8
outer boundary of the system.
Thus, over
the end faces,

-)4>c{p,h)
4,R{pfi)
= 0 = AJo(\cp) cos ixh

= 0 = [C,/o(Xp) + C2Ka(\RP)]
cos y.K
In general, the /i's in these two equations
are not identical since they

involve the extrapolation distance outside the core and reflector ; however,
for cylinders with dimensions large in comparison with the extrapolation
distance in either the core or the reflector, the fi's are approximately equal.
In any
of (8.35) can be meTonTylf
case, the conditions
"
cos nK = 0
(8.36)
that the difference in extrapolation distance for the core and reflector
would be difficult to take into account, since we have already required
Equation (8.36) can be satisfied only if
that the p's be the same.
so

(8.37,
taking, as in previous work, the lowest eigenvalue.

The requirement that the flux vanish over the extrapolated
surface of the reflector gives a relation between Ci and C\; thus,
Cl
If
we call
^
L0(p)
the fluxes may now be written
<pc(p,z)
<P(p,z)
(8-38)
~7o(AW~
/o(XfiPi) Xo(Xsp)
/o(Xp) X0(XRp1)
o(Xcp) cos
= CLo(p) cos
curved
(8.39)
TTZ
(8.40)
where C and A are constants to be determined from the interface condi

tions on the flux and net current, (3) in (8.4); these are
AJo(Xcpo)
DcAJ'0(\cpo)
= CLo(po)
= DnCL'0(po)
CR4n
lS'"'
One of these equations may be used to give a relationship between ^4Nand

C; however, A and C cannot both be determined from (8.41) and, as in
all reactor problems, there will remain an arbitrary constant which is the
flux level at steady-state operation.
The ratio of the two equations is
independent
of
A and C and gives the criticality condition for this system:
DcJ o (XcPo) _
o(Xcpo)
DrL'q
(pq)
Lo{po)
sec. 8.2]
REFLECTED REACTORS
431
The results of the analysis given above might be used as follows:

Suppose that a system is to be built of the type shown in Fig. 8.6 with
p0 and pi and the height 2h specified.
Also, sup
has
been
that
the
reflector
selected
composition
pose
(i.e., x% is deter
In the core, the concentration of all materials is known except
mined).
that of the fuel.
If the fuel concentration is low, as will be the case in
many real systems, the diffusion coefficient Dc for the core may be
the values of the radii
assumed to be known, since it will be only slightly affectecTby small

Then, the criticality conditôn-fS^^)
changes in the fueI_cgjicentration.
may be used to^djvterjmnjTxînasmuch as all the other quantities in this
relation are now known?
Sote that from (8.30) and (8.37)
X2
y()
" 4-
(0
(8.43)
With the value obtained for \c, the buckling may

A
be computed from

(8.44)
The fuel concentration
required to maintain
this system at steady state
is calculated from (8.5)

2
(8.45)
The results given above include, of course, the

case of the very long cylinder for which the
quantity (ir/2h~)2 may be neglected in com
parison to the other terms.
2. Cylindrical Core with End Reflectors Only.
Figure 8.7 shows the geometry and dimen
sions for the cylindrical core with end reflec
tors.
The work of the preceding
case may be
over here through the separation of
variables since Eqs. (8.26) through (8.28) are
valid for the present calculation.
However,
it is convenient in fitting the present bound
taken
Fig. 8.7 Cylindrical

tor with end reflectors.
reac
ary conditions to make a different choice for the algebraic sign of the
constants.
Instead of using relations (8.29), let
pFc(p)
[PF'C
()]' =
goto
Gc(z)
-\%
-4
i
(8.46)
Gr{z)
REACTOR ANALYSIS
432
It

A
-\\ +
= B2
n%
The general solutions for the present

=
Gc(z)
n%
G(z) =
Aa
If the
m 0.
(8.47)
case are then
In order to meet the requirement that the flux

we must have A2 = C2 =
AJofrcp) + AtY0(\cP)
C,Jo(XP) + C2Ko(Xp)
A3 cos ncz + Ai sin mc2
C3 cosh hhZ + Ct sinh p.z
Fc(p) =
F(p)
[CHAP. 8
(a)
(6)
(8.48)
(c)
(d)
be finite and symmetric,
flux is to be continuous across
the core-reflector interface, we also require
\c
= Aft = X
(8.49)

Consider now the reflector located on the positive 2 axis.

(Note
that the flux in the other reflector may always be obtained by sym
The extrapolated boundary condition at a yields n(p,a) = 0
metry.)
or Gr(S) = 0. We obtain immediately from (8.48d) the relation
C3 =
If we use
C4
tanh Ma
(8.50)
the results so far derived in the expressions for the flux, we have
AJ0(\p)
c(p,z)
<Pr(p,z)
= (Vo(Xp)
cos ncz
sinh p.(a
(g
51>
2)
The flux is also to vanish all along the curved surfaces ; thus,
(pi,2)
and it follows that
= <MP1,2)
= 0
(8.52)
and, as before, it is assumed that the extrapolation

for the core and êjêctor.rnaterials.
distance is the same
The continuity conditions at the interface may now be applied to

obtain a relationship between the constants .4 and C and the criticality
condition for the end-reflected cylinder.
These are:
c^
M^v
Catv-K ca^*<JDcmc tan
AcoB
sinh pft(a
Mcft
h)
h)
nch = Dp. coth p(a
'cL
(g 53)
(8.54)*
The calculation of the critical parameters of the reactor proceeds as

before, with the relations obtained above and with
REFLECTED REACTORS
sec. 8.2]
433
We turn now to the consideration of

Geometry.
rectangular systems ; again there is no simple solution to the problem of a
How
rectangular-block core completely reflected as shown in Fig. 8.8.
d. Rectangular

ever, there is one geometry which is readily solved by methods similar

to those used for the reflected cylinder.
Consider a block core with two
Fig. 8.8 Completely reflected

tangular-block reactor.
Fig.
rec-
8.9
Block
reactor with
slab-reflected
ends.
reflecting slabs of the same thickness on two opposing faces, as shown

in Fig. 8.9.
As before, the differential equations for the flux in the core
and reflector are given by (8.5) and the boundary conditions, by (8.4).
In Cartesian coordinates, these
dx2
c(x,y,z) +
^ c{x,y,z)
are
4>c(x,y,z)
+ B2Mx,y,z)
d2
dx2
d2
n(x,y,z) + Q-j n(x,y,z) +
dz2
= 0
(a)
= 0
(b)
x\R{x,y,z)
Here, as in the cylindrical case, the variables may be separated.
Fc(x) Gc(y) Hc{z)

= FR(x) GR(y) HR(z)
4>c(x,y,z)
4>R{x,y,z)
(8.56)
4>R{x,y,z)
Assume
(8.57)
then Eqs. (8.56) take the form
C(y) _,_ H^)

c(x)
+
+
Gc(y)
Fc(x)
Hc(z)
F (*) | G'^(y)
#(z)
'
/..\ 'i
GR(y)
FR(x)
HR(z)
in
=
=
-B2
(8.58)
V2
Choose
F'cix)
Fc(x)
F'R\x)
Fr(x)
-xf
M?
G'tiy)
Gc(y)
GR(y)
-M
"Ml
X2
-Ml
Hc(z)
HR(.z)
(8.59)
REACTOR ANALYSIS
434
[CHAP. 8
then Eqs. (8.58) become
x!
xl
xr
B:
(8.60)
The general solutions for the various functions in (8.59) may now be
written (taking account of the symmetry requirement to eliminate the
sine functions)
Fc{x) = A \ cos \\X
Gc(ij) = A2cos\2y
Hc(z) = A3 cos Xjz
Fr(x)
Ci sinh mx +
GR(y)
cosh mx
(8.61)
HR(z)
be continuous
across the
is,
In order to meet the requirement that the flux

core-reflector jnterfacej-that
Ei
= C2 cos my
=
cos n&
c(a,y,z)
R(a,y,z)
(8.62)
where Co =
\iy
A'i/Ci,
cos Xjz) =
C'(sinh
set_
we
X3 = Ms
X2 = M2
(8.63)
In order to meet the requirement that the flux vanish at all points of the
outer boundary of the system, we must have
X2 =
m =
7T
7T
X3 = 113 =
26
26
"%
= tanh /xid
(8.64)
extrapolated
has been made that the extrapolation distance

the same outside the core and the reflector.)
With these results we
(Again
is
the assumption
cos
cos
(8'65)
^Z
sinh m(3,
^-
cos
^|
B(x,y,z)
cos Xix cos
x)
fa{x,y,z)
obtain for the fluxes the expressions
apply only to the region on the side of the positive x

axis; the flux in the rest of the reactor
obtained by symmetry about
=
x
0.
is
These solutions
through the core-
ix\{3,
a)
sinh
tan
Xi<z
= Drmi coth
m(cl
and
cos Xia
a)
^>DcXi
reflector interface yields a relationship between the constants

and the criticality condition for this system; thus,
be continuous
The requirement that the net current
L j

Co cosh fna) cos p2y cos
cos Xia(cos
or
(b)
(8.66)*
REFLECTED REACTORS
sec. 8.2]
The parameter m is obtained from Eqs.
The substitution of this result into
435
(8.606)
and (8.64), which yield
gives Xi; then B2 may be
(8.666)
obtained from (8.60a) using (8.64):

B2 = xl
Gi)2
^/
+
(868)
()'
Fig.
8.10
Block reactor with slab-reflected sides.
all the boundary conditions with functions in which the variables have
been separated.
An approximate technique for treating these problems
is discussed in Sec. 8.6.
e. Criticality Relation.
is,
The one-velocity model as used so far in this

section is based upon the choice of a thermal-neutron source term which
is proportional to the thermal core flux itself, that
vZ/0c(r) [e.g.,
Eq. (8.3a)]. Since neutrons are born at high energy with the spatial
distribution of the thermal flux in the core, the
use of this source term
it
is
neglects the diffusion of neutrons during the slo wing-down process. In

the case of bare system
true, of course, that this form of the thermal
source term
made for the migration
correct, even when allowance
is
is
during slowing down, since the flux preserves its spatial form at all
reflected reactor the spatial
energies.
However, as we have noted, in
dependence of the flux will in general be different at different lethargies.
In
is
the analyses which follow, an attempt

made to account for the
wanderings (and losses) of fast neutrons in reflected systems.
In order to obtain some idea of what
involved in estimating these
is

The block geometry considered here is the only rectangular system

As indicated earlier,
for which an analytical solution has been obtained.
the completely reflected geometry is difficult to treat analytically.
Unfortunately this is also true for other more simple configurations such
as the block core reflected on four sides, with or without corners (Fig.
The difficulty in all these cases is that it is not possible to meet
8.10).
ANALYSIS
REACTOR
436
[CHAP. 8
effects, let us assume that the following neutron-balance
relation1 is valid
for the core region:
-DcV^dr) + 2r*c(r)
*Zsp&2+c(r)
(8.69)
We define the factor
in this equation as the fraction of neutrons, born

at high energy in the reactor, which is not absorbed in slowing down to
The factor g* is the corresponding fraction which is not lost
thermal.
In the present case of a reflected reactor
by escapes from the reactor.
these quantities could not be correctly calculated from the relations
previously derived for the bare systems. Whatever estimate is made of

these factors for a particular reactor, it must account for the absorbing
and moderating properties of the reflector as well as the core.
Suppose that we accept the above equation as applicable to the reflected
reactor, recognizing that the quantities pg and g& are to be computed
'
This relation may then be used
by some, as yet unspecified, procedure.
to obtain an expression for the multiplication constant.
In the usual
way, we define
neutron production
total neutron
where
losses
di
fc cfr)
2/Pggg
_ Dc
denotes integration
^ v>c(r)
dr
dl + s,c,
^ 0c(r)
over the volume of the core.
The volume
integral of the Laplacian term, which gives the neutron losses from the
We have from
core, may be computed from the differential equation.
(g
0c
71)
we call the term in the brackets B2
with
fc
is
vc
(8.72)
Z*>]
reduces to the form
v;
therefore, we can compute
and the relation for
the system
critical,
in the usual way; i.e.,
[*e2/PS0g
If
B2(f>c,
then VJce =
fc
B* m
If
|^2gg*
2gj
v^c +
(8.69)
obtained from (8.73) by setting
The quantity vc denotes

in steady state.
system
1.
vc
fc
where
is
Vc

the neutron yield per fission
required to maintain
the
sec. 8.2]
REFLECTED REACTORS
437
The quantity Bt is computed from the criticality relation for the

particular reflected configuration in question.
We summarize these
relations below for the three elementary geometries previously considered.
1.
Sphere:
Dc(BRa cot BR0

=
2.
-Dh[hrR0
Cylinder:
1)
coth xh(Ri
1]
DrL'0(pq)
La(po)
o(XcPo)
/o(Xp)i;Co(XPi)
u)
\*c
'
X2
end-reflected:
Dene tan iich = DRnH coth fiR(a
4-4 +
B-
fx = 2.405
-*
(*)' h)
Cylinder:
3.
"

Io(\rPi)K0(\kp)
B2 =
Ro)
side-reflected:
DcJ'o&cpo)
L0(p) =
[CHAP. 8
REACTOR ANALYSIS
438
We have noted that the probabilities p and
#5, in general, are depend

is
possible in some instances,
ent upon the properties of the reflector.
however, to use for these quantities the results of the bare reactor calcu
This is the case, for example, in large thermal systems of
lations.
spherical geometry in which the slowing-down properties of the core and
It
Since systems of this type are of

reflector are very nearly the same.
some practical importance, it is of interest to demonstrate this result.
Let
us consider, therefore, the
following general relation for the diffusion
of thermal neutrons:
-Z)VV(r) +
2a(r)
G(r;r') dt'
fc 2^(r')
(8.74)
where <(r) = thermal flux at r in reactor, including all points

in both core and reflector
=
G(r;r')
probability that a neutron born at r' becomes
thermal at r
(8.75)

By our previous arguments vS,f{r') dr' is the number of neutrons appear

ing per unit time in dt' around r' at high (fission) energies. Thus the
integral term is a generalized expression for the rate at which neutrons
Equa
are entering the thermal group at r for any point in the reactor.
tion (8.74) is rigorously
correct (within the limitations of the thermalif the proper kernel G(r;r') is chosen for the
diffusion model)
In general, G(r;r') depends upon the highsystem under consideration.
energy cross sections of the reactor materials and the geometry of the
group
We will not attempt here to choose an accurate kernel but

will concentrate on finding a kernel which will be a good approximation
medium.
in some useful cases.

Let us limit our analysis, then, to the spherical system mentioned
above.
As a first approximation we will assume that G(t;t') is given by
the
infinite-medium
kernel of Eq. (6.38) :
k(r)
G{t;t')
Reflector
Reflector
It
slowing-down
(8.76)
would be reasonable to suppose
that the source function represented

by the integral would be strongly
peaked in the core region and that
Fig. 8.11 Flux
thermal reactor.
distribution
<fc
large
the thermal flux would be peaked

also, as indicated in Fig. 8.11. This
is consistent with our assumption

properties of the core and of the reflector are nearly
core is the primary source of thermal neutrons, and
decays rapidly from the center of the reactor, little error
that the slowing-down

Thus the
the same.
since the flux
in
sec. 8.2]
REFLECTED REACTORS
439
will
be made in extending the integral of Eq. (8.74) over all space when
describing the flux distribution in the core; therefore, take
-DcV^dt)
= G{r;r') dr' (8.77)

2<= fm vZf4>{f)
vZf4>{f)
G{r;r') dr'
fm
(8.77)
Moreover, because of the relatively small thermal flux in the reflector

and the rapid decay of the kernel G for large values of |r r'|, we
would expect that in the reflector region the integral term will be negli
Thus we assume that the neutron balance in the reflector
gibly small.
is given by
= 0
-Z)W(r) + 2>(r)
which is identical to Eq. (8.36).

The integral of Eq. (8.77) may be evaluated by expanding the flux
4>(r') in a Taylor series about point r; the first three terms are
*(r')
*(r) + (r
- r')
- r')
v>(r) + *[(r
V]V(r) +
(8.78)

The substitution of this expression into (8.77), along with (8.76) for Gx,
yields the following results for the three terms:
vZ,
vZ,
jm
fm
(r)
GK(r;r') dr'
- r') V(r) G(r;r') dr'

- r') V]*(r) Gx(r;r') dr'
[(r
fm
(r
r2/Ptll(r)
= 0
=
(8.79)
^/rthptllW(r)
Since the point r is in the core, the fluxes on the right of these equations
The application of these results to the core equation
are core fluxes.
yields
(Dc + v2/P,hrlh)V^c(r)
If
we define
&
- n+
where
fc
"
[vZ/Plb
-^
then (8.80) reduces to the form (8.5a).
33
2<C)]c(r)
= 0
rfer- |s
Since the differential
(8.80)
(8-81)
(8.82)
equations
for the flux in both the core and reflector are now identical to the set
(8.5), it follows that the multiplication constant is given by the expression
fc
=
2[1
Jfc.ro)]
(8'83)
This equation would be expected to give good results in the case of a

large thermal reactor with slowing-down properties much the same in the
core as in the reflector.
If this result is compared to (8.73), it may be
440
[CHAP. 8
REACTOR ANALYSIS
that the analysis above has amounted to obtaining an estimate for

That this is indeed a reasonable estimate may be established by
reviewing the results obtained for the bare reactor, using the Fermi model
In that case
(6.81).
seen
gfc.
ga
e~B^
If
for B2Tth <C 1 (i.e., large reactors).

in (8.73) for
then
u
K
*2fP*
2<(1
L^J)(1 +
JSVu.)
I +
(8.84)
Tth
this form had
been used
>Z/P*
2ic'[l + B\L%+
rth)]
directly
(g 85)J
^
But this is the relation (8.83) when kx ~
1 and pft is taken as pth for the

core material; thus, the two results (8.83) and (8.85) are comparable.
It is of interest to note also that (8.85) is in the form (6.113) if we identify
L%

8.3
To, as
the migration area M*.
Reactors with Spherical Symmetry
Serber- Wilson Condition
The accuracy of the methods used in the one-velocity

a. Introduction.
model developed in the preceding section is limited by the interface con
The two
ditions selected for the solution of the differential equations.
conditions imposed in these calculations were the continuity of the
The purpose of the present
neutron net current and the neutron flux.
section is to select a new set of boundary conditions which will give a
more accurate representation of the neutron flux in the regions of impor
tance, particularly in regions wherein the selected analytical expressions
for the flux are known to be valid.

In this calculation, the continuity of net-current condition will be
This choice is made because the net current out of any region
retained.
is proportional to the volume integral of the neutron density (or flux)
Thus in computing the neutron leakage from a volume
over that region.
V having surface area A, one might use a relation of the type
Leakage
dA =
-D frV2dr
(8.86)
It
is evident from this expression that so long as the computed flux >
gives a reasonably good picture of the spatial distribution of neutrons
over most of the region, fairly large errors can be tolerated over small
portions of the region, for example, near boundaries (see Fig. 8.12).
The present argument, then, is based on the fact that the net-current
condition is an integral-type relation and requires simply that a certain
volume integral of the estimated flux approximate closely the integral for
The continuity-of-flux condition, on the other
the exact calculation.
hand, although physically acceptable, is not always the best condition
sec. 8.3]
REFLECTED REACTORS
441
to
This is especially true when using the diffusion-theory approxi

use.
mation for the flux distribution, namely, that the flux is well represented
by an isotropic term and
a term in the first derivative.
is not the case near outer boundaries

and discontinuities
in the material
of
the system.
Thus the
composition
Of course, this
Calculated
flux
Interface
that the diffusion-theory

expressions for the flux be continuous
requirement
in
these regions can introduce

errors in estimating the current
interface.
In the present treatment
/-Actual flux
large
at an
we discard
the condition that the flux be contin
Fig.
r
8.12
Flux distribution
around
interface.
uous at an interface, and select instead

a relation of the integral type (as in the case of the net-current condition).
The boundary conditions to be applied here in place of the continuity
conditions
(1) and (2) of (8.4) are as follows:
Continuity of net current at interface

for neutron flux must satisfy the transport
Expression
(2)
equation at the center of the reactor

(1)
(8.87)
The net-current condition (1) of (8.87) is the same as that previously

used.
Condition (2) of (8.87) replaces the continuity-of-flux condition
and requires that the general solutions which are selected satisfy the
accurate equations for the flux distribution in the region of greatest
importance, namely, in the central region of the core. This is known as
the Serber-Wilson condition and is an integral relation which is applied
at the center of the reactor.
Thus, discontinuities in the expression for
the flux may very likely result at the interface; however, it is recognized
that this is of no great consequence since the solutions obtained will not
be valid in these regions, in any event (Fig. 8.12).
The method based on the Serber-Wilson condition is especially suited
to the treatment of reactor systems which are spherically symmetric.
We will develop the basic ideas involved in this method by analyzing,
on the basis of the one-velocity model, the simple case of a spherical core
by an infinite reflector.
However, the method is generally
applicable to finite systems,1 to multiregion configurations, and to multiThese problems will not be discussed here.
group (energy) calculations.2
In order to demonstrate the effect on the criticality requirements of the
surrounded
1 A. H. Wilson, The General

Properties of the Transport Equation and Its Uses in
Determining the Multiplication in Bodies Having Spherical Symmetry, AERE-M/
S-105A, Dec. 7, 1944.
Bengt Carlson, "Multi-velocity Serber-Wilson Neutron Diffusion Calculations,"
Los Alamos Scientific Laboratory, Report LA-1391, Mar. 24, 1952, declassified.
REACTOR ANALYSIS
442
[CHAP. 8
Serber-Wilson condition, the critical radius of a spherical reactor with an

infinite reflector will be computed for various core and reflector composi
tions, and the results will be compared with the computations based on
the method of Sec. 8.2b (using the continuity-of-flux condition).
b. One-velocity Transport Equation and Asymptotic Solutions.
The
development
of the present method is based on the general one-velocity
transport equation for nonstationFor this calculation

ary systems.
we introduce the following assump
tions: (1) the core is a homoge
neous multiplying medium with no
external (nonfission) sources; (2)
r-r'
the reflector is a source-free homoge

neous medium; and (3) scattering
is isotropic in the laboratory system
Coordinate system.
of coordinates.
Sec. 7.4c
a Vtf(r,G) + 2(r)
It
was shown pre

equation for
that the appropriate integrodifferential

system of this type could be written in the form
viously in
where
*(r)
fv
4>(T,a)
c(r') S,(r')
S(r) c(r)
(i)
(7.245)
*(r') G(r;r') di'
and the volume integral is to extend over all space.
(8.88)
The quantity
is
given by
c(r) =
and the effective absorption
cross section by
2(r)
- S,(r) + ?
The kernel G which appears in the expression for

G(r;r')
and
/3
is given by (7.196)
(7-236)
>Z/(r^Z'(r)
with
(7.246)
(t)
is defined
- rT
e-/S(r';r)
47r|r
(8.89)
2 replaced by 2.
In the present analysis our interest
is in the
total flux distribution
throughout a system consisting of a spherical core of radius a and

infinite reflector (Fig. 8.13). For convenience in analysis, we place
4>{r)
an
Since this is a simple

the origin of coordinates at the center of the core.
two-region problem, the integral of (8.88) is conveniently separated into
the form:
*(r)
= ccZ((C)
G(r;r') di' +
fc c(r')
c2
G(r;r')
fR (r')
dv'
(8.90)
REFLECTED REACTORS
sec. 8.3]
443
This relation gives the flux at any general point r in the system in terms
of integrals involving the flux in both the core and the reflector.
The
notation ( )c refers to properties of the core and ( )n to properties of
the reflector;
Jc
denotes integration over the core, and
tion over the reflector volume.
jR denotes integra
Also, from our previous definition
of
c,
we have
Cc '
vXf +
2(C>
'
_
=
CR
2(ft)
(8-91^
Sp
The kernel G(r;r') is to take on the appropriate form, depending on the

relative position of the field point r and the position of the integration
Since we have assumed that the two regions of the system
variable r'.
are homogeneous, the 2 which appears in the expression for /3 will be
/3
reduces to relations of the type
/3(r';r) = 2|r
r'|
constant over each region, and
(8.92)
The Serber-Wilson Condition.

The Serber- Wilson condition
established by requiring that the solutions selected for the flux distribu
tions throughout the core and reflector satisfy rigorously the integral
At this point in the system, (8.90) takes
equation (8.90) at the origin.
on an especially simple form by virtue of the fact that all integration
paths extend along radial lines. Thus, using (8.92),
is
c*2>
w,
>
p(r')e-fc-5*(''
dr'
*R(r
)e
MXr,2
]r f
<br(i')e-fcr' dr'
ccZP
jc f
MO)
(8.93)
We are dealing
(i)
(r),
with
here
a spherically
and dr = 4irr2 dr.
With
symmetric
system,
these substitutions,
so
that
(8.93) reduces
Jo.
J"
(8.94)
over the core into the
Thus
dr
[fQ" c(r)e-~zc
c2^'
4,c(r)C-^ dr
]
= cc2<
<f,c(0)
1=1Jo
Jo
<Mr)e-fc-2K(r-a) dr
to separate the integration
("
form
be convenient
c2JR>
f~
will
It
Mr)e~fcr dr
= cc2ic
c(0)
to
0(r)e--^f<'-) dr
(8.95)
For the functions <c(r) and ^s(r) which appear in (8.95) we use firstorder approximations of the solutions to the transport equation.
The
basis for this choice
derived from the following argument.
It was
shown previously in Sec. 7.4d that the elementary solution to the oneis

c.
velocity transport equation
for the source-free
infinite multiplying
444
REACTOR
medium was the function
e*8-',
[chap. 8
ANALYSIS
where the eigenvalue
B is given by the
it is seen that the
In the present case,

transcendental equation (7.254).
substitution of this form into (8.88), with G given by (8.89) and
terms of 2 in place of 20, gives
We observe, however, that e,B-r is also

for the multiplying medium
Vs0 +
B2
f}
in
a solution of the diffusion equation
B real
= 0
(5.161)
Thus the
in Eq. (7.256) and the subsequent discussion.
general solution of the integral equation (8.88), which can be constructed
from the elementary solution [see, for example, Eq. (7.255)], will also
as demonstrated
Now the only solution of the diffusion

satisfy the diffusion equation.1
equation which is finite everywhere and which is spherically symmetric is

(r)
(8.97)
solution to Eq. (8.88) at points far

removed from the boundaries of the medium because in these regions the
angular distribution of the neutron flux is nearly isotropic, and the dif
fusion-theory results will give an adequate representation of the spatial
Thus sin Br/r is sometimes called the
distribution of the neutrons.
This function should
be a satisfactory
asymptotic solution to the transport equation.

The transcendental equation (8.96) may be written more symmetrically
^(IH
wherein we have introduced
the quantity
(898)
defined by
average number of neutrons produced per

collision in the system described by (7.245)
vXf
+ Z'
(8.99)
Note that
c2 = c2,
(8.100)
It
is seen from Eq. (8.98) that the requirement that B be real implies
that c > 1, which corresponds to a multiplying medium [see (7.236)].
(In the present context, multiplying means capable of supporting a selfsustaining
plying,
1
In the event that the medium is nonmultichain reaction.)

1 and B is imaginary.
Then, it is convenient to define
<
Wilson, op. cit.
REFLECTED REACTORS
SEC. 8.3]
445
B = ix, and Eq. (8.98) takes the form
tenh-K
w.th
k m
2/
The corresponding expression for the flux is obtained from (5.161) and
is seen to be
V2
- x2
In the analysis which follows

(8.102) to represent
c and
= 0
x real
(8.102)
we use the solutions
4>R,
with B
to Eqs. (5.161) and

and x defined by Eqs. (8.98) and
(8.101).
It will
in developing the present analysis to introduce

another new quantity if, which is related to c through the equation
be convenient
= c(r)
<p(i)
Thus, physically,
<p is
(8.103)
the excess of neutrons, over the incident particle,

If we use the usual definition of 2(
2, = 2. + 2, + 27
~
("
DZ/ ~ z?
<p =
then
(8.104)
(8.105)
it
Material
-1
Pure absorber (no fissions)

Pure absorber (with fissions)
<fi
given below:

>
the average number of neutrons produced per absorption (tj)
greater
than unity. These parameters may also be defined in terms of the effec
where
if
It
is
g/
<p
was
tive absorption cross section 2. The appropriate definition for
given previously in (8.99). The corresponding expression for the net
neutron production per collision
then
(8.106)
+2S-
follows from (8.105), (8.99), and (8.106) that
?2
= ^2,
-"
(8.107)
the solution
The asymptotic form to be used in the core region
The
solution
for
the
reflector
Eq.
of
corresponding
region
(5.161).
(8.97)
is

which is released per collision.
446
REACTOR
[chap. 8
ANALYSIS
is given by Eq. (8.102) ; the only solution to this equation which is spheri
Thus for the present
cally symmetric is the negative exponential e~'T/r.
analysis we select the following asymptotic expressions, denoted by the
symbol
*~c
where
- =^
(8-108)
These are
from (8.98) and x from (8.101).

in the forms
is to be computed
conveniently
summarized
=
with tan =
l + fc
tSE=r--1 + -
(8.109)
(8-110)
**'mm
Note that, if the material composition of the core and reflector is specified,
the parameters B and x may be obtained directly from the above tran
scendental relations.
The quantities w and k have been extensively
tabulated1 as functions of <pc and <?r.
The expressions (8.108) are now applied to Eq. (8.95). First we note
that
must satisfy the integral equation (8.88) ; therefore
*?(r)e-Scr dr
j'
The substitution of the asymptotic solutions 4> and

with the application of (8.111), yields the relation
cc2e2<*
=
c*S{S)eJ
0?(r)e-J<^ dr
(8.
% into
1)
cc2Sc>
1
1
(8.95) along
,
*g(0)
jQ"
0J(r)-!' dr
(8.112)
is
is
[Note that the use of (8.111) allows us to cancel the term at the left of
The equation (8.112)
the
(8.95) with the first term at the right.]
Serber-Wilson condition.
It based on the requirement that the solution
be satisfied
the system
the reactor =
is
This condition must
if
Net number of neutrons produced throughout
is
is
it
for the flux satisfy the transport equation at the center of the reactor, and
replaces the condition of continuity of flux at the core-reflector interface.
The second boundary condition to be used
the requirement that the
net current be continuous at the interface.
For the present calculation,
this condition
more conveniently stated in integral form; thus,
(8.113)
to be at steady state.
See, for example, D. Sweeney, M. Goldstein, and B. Carlson, "Tables of Functions

Associated with the Serber- Wilson Method for Neutron Diffusion Calculations in
LA-1364 (May 19, 1952), vol. II, LA-1365 (May
Spherical Geometry," vol.
1952), vol. Ill, LA-136G (Apr. 30, 1952); also, K. M. Case, F. de Hoffmann, and
G. Placzek, "Introduction to the Theory of Neutron Diffusion," pp. 56-57, Los
Alamos Scientific Laboratory, 1953.
5, a
I,

- ^~
REFLECTED REACTORS
SEC. 8.3]
447
However, it applies with equal validity to noncritical systems (nonstationary systems) since these systems can always be described by
means of an equivalent
steady-state
problem
the use of the
through
"effective-absorption" cross section 2 [cf. Eq. (7.245)]. If we use the

parameters appropriate to a system with "time absorption," as implied
in 2, Eq. (8.113) may be written
2(r) *(r) dr
fr p(r)
This integral may
be separated
= 0
(8.114)
into the core and reflector terms; thus in
terms of the solutions (8.108),
fc *cW
dr =
-?R2fi
Jr flj(r)
dr
(8.115)*
This integral equation along with (8.112) provides the boundary con
ditions to be satisfied by the solutions 0.
d. Criticality Equation.
The criticality equation for the spherical
reactor with infinite reflector is obtained by dividing the flux condition
Before performing this opera
(8.112) by the current condition (8.115).
In
tion, we first evaluate the integrals involved in these expressions.
for
and
will
be
used.
every case, the asymptotic solutions (8.108)
<pc
4>r
The first integral to be computed is that from the left-hand side of
r>
sin Br dr
From the right-hand side of (8.115)

0(r) dr = 4*vl
Bl
(sin aB
- aB
cos
aB)
(8.116)
we obtain
re~"r
= 4*
Jo
f (r) dr
Jc
*j[
(8.115):
dr
(ax
+ l)e-M
(8.117)
f"
A\
e-
is
e~lcr sin
Br dr
AEi(alR
ax)
(8.118)
The term from the left-hand side
the exponential integral.1

of Eq. (8.112) gives [cf. (7.268)]
where
p-Za'-*r Ar
T
-Im
Ei(2ca
where Im
+ iBa)
/n
E\
f"
and from the right-hand side of (8.112) we have
(8.119)
Ei(z) denotes the imaginary part of Ei(z).

The substitution of the above results into the relations (8.112) and
(8.115), followed by the division of (8.112) by (8.115), yields the criti1

Pc2c
Case, de Hoffmann, and Placzek, op. at.
ANALYSIS
REACTOR
448
[CHAP. 8
cality condition
ccZSe-c
<Pc^c
Im Ei(Sca + iaB)
"y^, (sin aB
aB
cos
cifS<*(ax)21(a2
<pHlR{ax
aB)
+
+
ax)e<-*+"
1)
(8.120)

This result was obtained by both Wilson and by R. Serber and gives the
critical radius of the core for a specified core and reflector composition
We will discuss this relation in more detail shortly.
(or vice versa).
1.5
Fig.
8.14
2.5
3.5
4.5
The functions Q(y,u) and S(y).
In order to simplify the notation,
and
aB = a2c tan
we introduce
to
= x tan
o>
the quantities
= y
(8.121)
Some of the functions on the left side of (8.120) have been tabulated.1
It will be convenient, therefore, to rewrite the criticality equation in
terms of the tabulated functions (see also Fig. 8.14):
S(V) =
Q(y,o>)
1
Ibid.
yl (sin
-y
cos y)
- - e-"-" Im El (2ty
\tan
(sh + *)
o}
co
(8.122)
REFLECTED REACTORS
sec. 8.3]
If
449
these relations are used in (8.120), we obtain
S(y)
We have used
wn
<Pr2iR
for
here the definition
(8,123)
from (8.110), along with (8.121)
to obtain
ax =
= nf
<c2a
(8.124)
Equation (8.123) is the criticality equation for the equivalent stationary

reactor.
The solution of this equation for a system of given size and
composition yields the multiplication constant of the actual reactor.
Before dealing with time-dependent systems, we first examine the prob
lem of the stationary, or critical, reactor.
e. Critical Systems.
The steady-state system may be described in
In every
only the 2 terms which appear in the preceding analysis.
instant 2 becomes 2(, and the results derived above take the forms sum
marized below.
The criticality equation (8.123) becomes
<pc
S(y)
<pH
The characteristic equations (8.109) and (8.110)
Xpa
Equation (8.125)
also,
ta^
fc
with tan
+ ^
are now
u>
with" =
= 2iR,a
and
and
*~an
J^r.
(8.126)
(8-127)
spr)
= x tan u =
aB
(8.128)
is
it
is
the criticality equation for

reactor operating at
In its present form, this equation relates the properties of
steady state.
the core (left side of the equation) to the properties of the reflector
(right side). Thus, given the properties of the reflector, (8.125) deter
mines the critical fuel concentration (or size) of the core. In making
will be convenient to call the core
computations with this relation,
term Tc(x,<(!c) and the reflector term Tii{$,<pr).
Note that &> deter
is
mined by <pc [according to (8.126)] and that

determined by <pr (8.127),
that the only variables which remain are x and
once the ^>'s (i.e.,
cross sections) have been specified.
that
The criticality condition
is
f,
so
Tc(x,<pc)
= TR([,<pK)
(8.129)
x
is
is
is
f,
Thus, given
to find
value of
known.
The problem, then,
TK
such that Tc = TR. If the core composition 2p
specified, this value
a

terms of our present notation as the special case when the reactor period
This corresponds to a > 0 and the effective absorption
1/a is infinite.
cross section 2 2, [see Eq. (7.246)].
Formally, this change affects
450
REACTOR ANALYSIS
[chap. 8
of x gives the critical radius o from relation (8.128) ; if the magnitude of a

is specified, (8.128) gives the composition 2jC) of the core.
The solution of Eq. (8.125) has been carried out in parametric form,
and the results are presented in graphs suitable for criticality computa

Figure 8.15
tions for a range of reactor compositions of practical interest.
is a plot of Tc(x) for various values of <pc, and Fig. 8.16 is a plot of T(f)
for various values of <p. It should be noted that x and f are related by
Core radius in core mean free paths (x =
Fig.
8.15
2 P a)
Core properties.
the ratio of the total cross section in the core to the total cross section
in the reflector; thus
X
*f
=
2p">
(8'130)
The application of these figures to an actual reactor computation

proceeds in the following manner: (1) Compute <pc, <p, and 6 from the
properties of the reactor.
(2) Select a value of f and from Fig. 8.16
find TR{$,if>R).
(3) By using this value for Tc, select the value of x from
the appropriate curve (for the given <pc) in Fig. 8.15.
(4) See if this
value of x and the assumed value of {" satisfy Eq. (8.130) ; if they do not,
repeat the process until the selected value of f and the corresponding
value of x do so.
REFLECTED REACTORS
sec. 8.3]
451
The critical core radius x (in mean free paths) has been computed for
some combinations of <pc and <pr which may be encountered in reactor
These data are presented as a function of d, the core-reflector
design.
For these cases, the critical size can
cross section ratio, in Fig. 8.17.
be obtained
directly from this figure if the core and reflector compositions

For other combinations of <pc and ipR, the procedure outlined

are given.
above can always be applied.
Core radius in reflector mean free paths
Fig.
8.16
(()
Reflector properties.
The results based on the Serber-Wilson method have been compared

with those obtained from the continuity-of-flux condition.
The critical
radii obtained by using the two methods are plotted in Fig. 8.18 as a
function of 6 for the case <pc = 0.2, <ps 0. The Serber-Wilson calcula
tions were based on Eq. (8.125), and the continuity-of-flux condition was
based on (8.11).
The appropriate form of (8.11) in the present notation
is
Pcx cot fax =
where
P2C
= 3^c
and
- - Prx
0
0%
= -3<pR
(8.131)
(8.132)
The flux distributions for the case <pc = 0.2, ipR = 0, and 6 = 0.8 are
shown in Fig. 8.19.
It is of interest to note the discontinuity in the flux
due to the application of the Serber-Wilson condition.
Also, note that
the critical mass computed by the continuity-of-flux condition exceeds
the Serber-Wilson estimate by as much as 40 per cent for small reactors.
aoiovas sisaivnv

Z9f
SIO
ZO
VP
90
80
OlJetf JO 9JO0 SSOJO U0IJD3S
O'T
<ivho]
Jl:>3l>aJ SS0J:> UO'l3eS
']
OE
UI)L Xjir
ni >-
B
3
a
o
UOS|l/\
91
ZQ=
0="
o
20
so
Ojjey (0 3J03 ssojo uoipos 0) jopoijBj ssojo uoipss (0)
'olA gl"8 uosueduiog jo sijnsoj uiojj uos|i;\Y-joqjog uoiiipuoo pun uiojj -X^mupnoo
xng-jo uonipuoa
sec. 8.3]
f.
453
REFLECTED REACTORS
Noncritical Systems.
We return
now to examine the analysis of
The neutron distribution in these

systems which are not at steady state.
cases is described by the integrodifferential
It was
equation (7.243).
noted in Sec. 7.4c that a solution to this equation could be written in the

0.5
2.0
1.5
1.0
Radial coordinate
Fig.
form (7.244).
by integrating
8.19
2.5
3.5
3.0
in core mean free paths
4.0
[rZ|]
Flux distributions.
The corresponding expression for the total flux is obtained

(7.244) over all Q
tf>0M)
thus
=
(8.133)
4>{i)e'
is yet to be determined.
In dealing with the nonstationary system we found it convenient to define an equivalent steadystate system by introducing the effective cross section 2.
The relation
where the period
l/a
which determines this cross section, and consequently the reactor period,
is Eq. (8.123).
Our purpose in the present calculation is to obtain a
procedure for computing a from (8.123) and, subsequently, the multipli
cation constant k of the reactor.
It will be convenient at this point to work, not with a, but with another
parameter y which is related to a through the equation
y =
;J
(8.134)
The effective absorption cross section can be written in terms of y by

substituting the above expression into (7.246). Thus, for the core
2C =
Sf>(l
+7)
or
I=
+ 7
(8.135)
KEACTOB ANALYSIS
454
For the reflector, using (8.130) for

2*
r =
i +
70)
[CHAP.
0,
= s,(S,(i
+ ye)
(8.136)
We can now write the cross-section ratios which appear in (8.123) in

terms of 7.
Observe first that, from (8.107),
?c2c
?2fi
- 7)
-
2f >(*>c
= 2<'(^
(8.137)
yd)
Thus the cross-section ratios in (8.123) take the form
The parameters u and
1+*?
^'
and
1+^L
(8.138)
may also be written in terms of 7 ; the appropriate

relations are easily found to be
tan

+
= -j
1
1
<pc
and
1+7
With these results (8.123) may
w^; "^r
be
k
r
j
1
tanh
written
<pr
+:
+ yd
1
z
1
- VtT^;
,Q 1QQv
(8.139)
(8-140)*
It
should be noted that, for the steady-state condition, 7 0, and this

equation reduces to the form (8.125).
The problem of computing the multiplication constant A; of a noncritical
reactor entails the solution of (8.140) for the value of 7 which satisfies
the specified geometric and material properties of the system in question.
The multiplication constant can then be computed from the number 7
by a simple procedure, as will be demonstrated shortly.
In the usual
problem of interest the size and composition of the reactor are specified;
thus x = 2,<C)a, f = 2,(R)a, ipc and <?r are known, and from these 8 can
by means of the relation 6 = x/f . The determination of

7 involves a straightforward trial-and-error procedure for the solution
of (8.140).
The procedure is first to select a 7 and compute u and k
from Eqs. (8.139). Next, the values of y and f are found from Eqs.
(8.135) and (8.136); note that
be computed
x =
x(l +
7) =
r1
tan w
(8.141)
These values of y and f are then inserted in (8.140), and, if they satisfy
the selected value of 7 is the solution for the system in
the equation,
SEC.
REFLECTED REACTORS
8.3]
455
A sample calculation based on this scheme is included at the

question.
end of the present section.
An expression for the usual multiplication constant A; may be derived
in terms of the parameter y. We start with the core equation of the
equivalent steady-state system
Q
and integrate
result is
[cf. (7.245)].
V*c(r,a) + 2c*c(r,0)
dr
jaa- Vc(r,a)
Now the multiplication constant

k
dQ + 2e*c = cc2,(C)*c
L c(r)
*c =
where
-L 2pcc*c(r)
(8.142)
Q and over the core volume.
over all directions
Jc
dr
The
(8.143)
(8.144)
is denned by
total production (n + 1st generation)

total population (nth generation)
production
leakage + absorption

(8.145)
For the actual reactor
not at steady state,
A; j
and (8.145) takes the
1,
form
=
If
ccZjC)3>c
leakage +
we write the total absorptions

k
2*c
in terms of 2C by using (7.246), then
ccZf^g
leakage + 2c^c
- yZf^c
The sum of the first two terms in the denominator

identically
[cf. (8.143)]; therefore,
cc2c
k =
cc
of this expression is
-y
(8.146)*
A sample calculation in which this method was used is given below.

The computations are for a spherical reactor with infinite reflector
having the following specifications:
i=
<Pc
2ca
= 3.0
= 0.3
<pR
2o
-0.03
= 3.3
= 0.91
The problem was to determine the multiplication constant of this system.

The values of the various functions appearing in (8.140) have been
computed for three selected values of y, and the results are listed in
REACTOR ANALYSIS
456
[CHAP.
By interpolation it was found that y = 0.1455.

sponding value of A; is found from (8.146); it is k = 1.127.
Table
8.2.
Table
8.2
Computation of the Multiplication
Function
to
t
y
Q(v,*>)
S(y)
f
E.lfd +)]
Tr
Tc(x,fc)
Tc/Tr

8.4
The corre
Constant
y = 0.130
y = 0.142
y = 0.150
0.617
0.599
3.39
2.41
0.0674
0.424
3.69
4.04 X 10~J
1.468
0.748
0.510
0.596
0.615
3.43
2.32
0.1066
0.425
3.73
0.579
0.620
3.45
2.25
0.1375
0.433
3.75
3.29 X 10-'
1.360
1.590
1.168
3.5 X
10-*
1.405
1.230
0.895
The Two -group Model
a. Reactor
The analytical model of a reactor which
Equations.
describes the neutron population as two separate groups of diffusing
neutrons, one fast and one thermal, has a great deal of usefulness in
reactor calculations.
To begin with, it is sufficiently simple from an
analytical and computational standpoint to permit calculations to be

made rapidly.
Moreover, the relative simplicity of the method allows
one to trace rather easily the various factors which influence the physical
features of the system as predicted by the model.
Also, for many
reactors, particularly large thermal reactors, the method gives acceptable
first estimates of the important characteristics.
Finally, there is sufficient
latitude allowed in the choice of certain parameters which appear in the
calculation
that any given "family" of reactor types can
be analyzed
with increasing accuracy as experience is gained in the selection of the

best values for these parameters.
This feature becomes increasingly
important as the engineering practicality of certain reactor types is
established and experimental information becomes available for compari
son with the predictions of the model.
Physically, the two-group model assumes that the behavior of the fast
neutrons in a reflected reactor can be represented by a single diffusion
equation (in each region) in terms of suitably selected fast-neutron cross
sections.
The thermal neutrons are represented by a second group in
Thus in the application of the model to the general
multiregion reactor, two differential equations are introduced for each
the usual way.
region, one to describe the thermal

group.
The solutions
group and one to describe the fast

to these equations in each region are coupled to
REFLECTED REACTORS
sec. 8.4]
457
the solutions in adjacent regions by appropriate boundary conditions

for each group, along with requirements on the solutions at the center of
the reactor and at the outer boundary.
The sources for the thermal
group are in each region proportional to the fast flux, and in those regions
containing fissionable material the sources for the fast group are propor
tional to the thermal flux. The following treatment is focused on the
analytical statement and solution of a particular case, the two-region
reactor with outer region nonmultiplying.
The methods developed for
this case are easily extended (in principle) to more general situations.
The analytical statement of the two-region system is as follows.
Consider first the diffusion equations for the fast and thermal neutrons
in the core:
-D,V2^,(r) +
Fast group:
Thermal group: -Z)2V22(r)
+ 24>2(r)
v2/*s(r)
pc2î(r)
(a)
(8.147)*
(b)
appears in (a) is to be chosen so that 2(/J)<i(r)

is a good estimate of the slo wing-down density of fast neutrons out of
the fast group at r. The source of fast neutrons is taken to be the neutrons
The parame
produced by fissions caused by thermal neutrons j<2/#2(r).
ters
Di and
2^' which
are the thermal-diffusion
coefficient and thermal-absorp
tion cross section in the core, respectively, computed in the usual way.
The source term for the thermal neutron equation (6) is taken to be the
number of neutrons slowing down out of the fast group, reduced by the
number lost in actual absorption during slowing down.
We define
fraction of neutrons leaving fast group which
reach thermal (in core)
arbitrary, since there
is
still the choice of
is,
and assume that pc is given by the resonance-escape probability in the

core.
It should be noted that this selection
to
certain extent,
2^' to be made.
Thus, we
have taken (8.148) as

convenient definition and reserved all necessary
in specifying the thermal-source
term pc2$i for the
adjustments
we assume the
following
Fast group:
Thermal group:
-D,V24(r) + 2*(r)
Z>3V23(r)
+ 2J?*(r) =
set of equations:
quantity 2^'.
For the reflector
(c)
(8.149)*
= pA2g>0,(r)
(d)
Both of these quanti

2^> corresponds to 2'^' for the core.
Note that the form (c)
ties are called the fast removal cross sections.
explicitly assumes nonmultiplying reflector material, since zero source
a
The parameter

The quantity
2(i'<#i(r)
As in the core, O4 and

of fast neutrons has been taken in the reflector.
of
in the usual way;
are
thermal
the
reflector
calculated
properties
2^4)
REACTOR ANALYSIS
458
[CHAP.
the source term pR2$3 is entirely analogous to the corresponding term

discussed above for the core. The quantities Dx and D3 represent the
fast-diffusion coefficients in the core and reflector, respectively; these
must also be chosen in some way to represent the entire fast range.
b. Group Constants.
We can obtain some idea of how one might go
about identifying and computing the fast-group constants from a study

Consider first an infinite multi
of two simpler multiplying systems.
is
no
In
this
case
there
spatial variation in the neutron
plying medium.
flux, either fast or thermal, and the set of Eqs. (8.147) is reduced to the
form
2g>*! =
2>*2 = pc2#>*i
vXffr
If
we take the ratio of these two equations, the constants

cancel out, and the resulting expression may be written

(8-150)
and
(8-151)
A comparison of this relation and the appropriate criticality equation

for this system which was obtained in Chap. 4, Eq. (4.274), wherein
t = 1, reveals that the quantity pc may be correctly identified as the
This was previously indicated.
resonance-escape probability to thermal.
The removal cross sections may be identified from a similar analysis
for the bare multiplying system. It follows from our study of the Fermi
age model that the spatial variation in the neutron flux is of the same
functional form at all energies. Thus we may take
*i(r)
- *iF(r)
'
*,(r)
and
*F(r)
(8.152)
F(r) satisfies the differential equation V*F + B2F = 0 and B1 is

In the usual way, we
given by the various relations listed in Table 5.3.
where
have required that both the fast and thermal fluxes vanish on the extrap
olated boundary surfaces of the reactor; the extrapolation distance Is
assumed to be the same for the fast and thermal groups.
The application of the above relations to Eqs. (8.147) yields the set
Dtftfr
As before
SJ}>0,
we take the
the constants
and
ratio of
D2B* +
be
these
2?>*, = pc2j>>*i
equations
(8.153)
and upon canceling out
tf>2 obtain
vZ,
This may
D2B^ +
v2f*t
_ Drf* +
pcZ
2f
(8.154)
written in the form
2(1
+ L\B%{\ + L\B*)
With L? ~
(8-155)
REFLECTED
sec. 8.4]
459
REACTORS
If
we compare this result with the Fermi age analysis of the bare reactor
(6.78) for the case wherein there are no fast fissions ( = 1), then we can
immediately identify the factor (1 + L\B2)~l as the fast nonleakage
probability; thus,
+ LIB1)-1
(1
For
a large reactor e-fl,rth
(1
(8.156)
B2Tth)~', and we note, then,
and
L?~rth
m e~B^
Zg>
that
(8.157)
This is one way in which one could define tfieTemoval cross sections 2.
As experience
would certainly be an adequate first approximation.
is gained from the comparison of theory and experiment, better estimates
can be made for this parameter.
The fast-diffusion coefficients may be computed from a suitable expres
Thus we take
sion for the neutron leakage.
<0

It
If we assume that, in first approximation, the neutron-energy spectrum

is the same for all r, and in fact varies as 1/E, then it follows from the
above expression that an acceptable definition of Di would be
=
Finally, from (8.157),

Zg>
Tth
D(u)du
Jo
Wth
MthTth
JO
(8.159)
D{u) du
(8.160)*
In a medium in which the scattering cross section varies only slightly

with lethargy and in which the fast-absorption cross section is negligible
in comparison with the fast-scattering cross section, (8.160) may be
approximated
as
2g>
(8.161)
r^-
Thus the "mean free path to removal" from the fast group
*
_1_
2$
X.Mih
{
_ / mean
free
path\ /average number of
\for scattering
^collisions to thermal
is seen to be
\
/
The parameter
is usually calculated by choosing a
[see Eq. (4.47)].
fast-diffusion coefficient for the entire group and dividing by the age to
thermal as in (8.160).
Note that (8.147a) may be written directly in
terms of the slowing-down
density
-rthV2g 4where we have used (8.160).
thus,
=
vXffa
(8.163)
REACTOR
460
ANALYSIS
[CHAP.
General Solution.
Consider now the analysis of the reflected reactor

by the differential equations (8.147) and (8.149); it is
understood that the parameters which appear in these equations are to
The general
be chosen by some means such as that outlined above.
solutions to these equations, for each region of the reactor, will be given
c.
as
described
In selecting these
as a linear combination
of two known functions.
functions we will be guided by our previous experience with the onevelocity model. The proper weighting factors (coefficients) for the two
functions in each combination will be determined by the requirement that
each general solution satisfy the appropriate differential equation from the
set (8.147) and (8.149) as well as the various boundary and symmetry
conditions imposed on the system.
We begin with the differential equations for the core.
Let Z(r) and
W(i) be two functions such that the following linear combinations give
the fast and thermal fluxes in the core:
*t(r) = AiZ(t) + CiW(x)

*,(r) ^ A2Z(i) + C,W(t)

We define the functions Z and W
as the solutions
V(r)
V2Z(r) + B\Z{r) = 0
The parameters B\ and
(^
l0*)
to the equations
+ B\W(t)
= 0
(8.165)
that the expressions (8.164)

Thus we substitute
#2
Eqs. (8.164) into (8.147) and apply (8.165). The result may be written
in the form:
for
and
B2 are to be chosen so
are acceptable solutions of Eqs. (8.147).
Z{t)[AiDxB\ + ZgAx
Z{r)[A2D2B\ + 2?A2
- vSfAi]
-+ W(t)[C1D1B*
pc2g>Ax]
+ W(t)[C,D,B\ +
Ci2<
C22
- C,w2,]
- pcC&tf]
=0
^'i0O;
= 0
Bi 5* B2, Z(r) and W(t) are independent functions. It

that
the above equations can be satisfied only if each coef
follows, then,
ficient is identically zero. This requirement yields the relations
Since in general
S/
Ax = A
\DzB\ +
Zffl ~
_ Cl f pcS<
,2,
J
[d2BI +
\DlBl
Or, if we cancel out the factors C\ and
f [DiB\ +
[[DiBl
+
+ 2g>][DsS| +
2>][D2Z??
2ff>"
(8.167)
Ai,
2]
2]
.pcS/S*"
= vpcZ/Zff
^,1D8;
SEC.
REFLECTED REACTORS
8.4]
Thus B\ and B\
are solutions
(B* +
461
of the quadratic
4)(B* +
k\)
equation
1!ff
(8.169)*
^.-^p
(8.170)
We find, then, that if the expressions (8.164) are to be solutions of the core
equations (8.147), with Z and W given by (8.165), the quantities B\ and
Bi must be taken as the roots of the B1 equation (8.169). These are
seen to be
82
--
^4-^
\ [w
k')2
4K?Ki(fc-
"
f /fV
^
(8-i7i)
Since our primary interest in this model is in the application to reactors,

it is clear that fc > 1. In this case
[{k\
4f"

It
follows, then, that if n* and

solutions to (8.171) are
B\ =
B\=
4kWK
!)]
>
(8-172)
K\
X2
are denned as positive and real then the
K\
small and positive

large and negative
n1
X2
With these choices, Eqs. (8.165) take the form

V2Z(r) + MZ(r)
= 0
V*W(r)
- \*W(t)
= 0
(8.174)
The solutions to these equations have already been discussed in much

detail.
The n equation yields for Z the trigonometric and ordinary
Bessel functions,
and the
equation, the hyperbolic
and modified Bessel
functions.
We summarize the general solutions for the flux distributions in the

core with the aid of the results (8.167).
If we define
. _
_
PC*?
_- PcW 2
>2X2
iM2 +
^
-DiX2 +
*,(r)
=
=
(8.175)*
gjj>
vZf
*(r)
73
2%>
AiZ(r) + C,W(t)
0lAiZ(r) +aiClW(t)
'
>
^1'0J*
Note that, since in all geometries there will be conditions of symmetry

or nonsingularity on
and <2, the functions Z and W must be chosen
^>
REACTOR ANALYSIS
462
to meet those conditions;
^jjj
gjS-
Vc''<7
[CHAP.
thus Z and W do not involve any undetermined
constants.
We consider next the reflector equations (8.149), which may be written
- S*.(r)
- S*(r)
V2*3(r)
V^(D
= 0
(8-177)
V(4)
V<3)
with
K's^
K'sfe
(8'178)
The solutions to the above equations are to vanish at the extrapolated

By the usual arguments [refer to discussion
outer surface of the reflector.
following Eqs. (6.56)] we specify that the extrapolation distance be the
p^same for the fast and thermal groups. Thus, if ft' denotes the extrap-
f^0
vjy pP6
olated outer surface of the reflector,
*,(ft')
*(ft')
= 0
(8.179)
We look for solutions to Eqs. (8.177) which may be given as a linear

combination of the functions U and V, where these functions satisfy the

equations
V2C/(r)
- K|f/(r)
= 0
=
C7(ft')
and
- k\V(i)
V2F(r)
V(ft')
= 0
= 0
(8.180)
(8.181)
Now, the form of the fa equation is seen to be the same as the reflector
relation used in the one-velocity model [cf. Eq. (8.5b)], and it was found
that the solution to that equation could be given in terms of a single
function.
It
follows, then, that
*,(r) = Sitf(r)
(8.182)
where <Si is some constant to be determined.

The function
other hand, must be expressed in the more general form:
*(r)
- TiV(r)
The substitution of
these relations for

the
yields
requirement
(8.177)
Tt
Thus
<0
we have
-o,Si
+ 7W(r)
with a,
and
*,(r)
Sitf(r)
*(r) =
(8.183)
into the
<fn
rnrf2^
for the reflector fluxes
TiV(r)
on the
a3S1C7(r)
equation of
(8-184)
(8.185)*
d. Criticality Condition.
The general solutions (8.176) and (8.185)
give the thermal and fast flux distributions in both core and reflector in
terms of the functions Z(i), W(r), f/(r), and F(r). These functions, in
turn, are defined by Eqs. (8.174) and (8.180) and by the symmetry and
REFLECTED REACTORS
sec. 8.4]
463
The .onlyquantities still unde

extrapolated boundary requirements.
termined are the constants A\, C\, T\, and Si. TheslTmay be computed
from the core-reflector interlace conditions which have yet to be applied.
These conditions are, physically, the continuity of flux and net current
If we denote by R the coregroups.

reflector interface surface, then the appropriate analytical statements are
in both the fast and thermal
(1)
(2)
(3)
(4)
i(R)
*,(R)
- ZWi(R) dA(R) - Z>,V*,(R)

*,(R)
*(R)
- ZW4(R)
- D2V2(R)
<JA(R)
The application
<*A(R)
(8-186)*
dA(R)
(D,E/')Si
{-DiZ')Ax
(aJT)C,
+
{-V)Tx
(a3U)Sl
(oiZ)Ai
(-a,Z)2Z')A, + (-o,7),W')Ci + (-o,Dl/')Si
=
=
(a)
(b)
(c)
(d)
(8.187)
The symbols Z, W, U, V and Z', W, U', V denote the functions and

^
their normal derivatives evaluated at r = R. M. 5. O-^
The above set consists of four simultaneous algebraic equations in the
Since these equations are homogeneous, it is
four unknown constants.
not possible to obtain explicit expressions for all four quantities, and, as
in all steady-state reactor problems, there will remain one undetermined
Thus three of the
constant (the flux or power level in the reactor).
The only nontrivial
unknowns will be determined in terms of the fourth.
solution of this system of equations requires that the determinant A of the
coefficients of A\, Ci, T\, and Si vanish; that
D3U'
atW
aiU
-a3DtU'
-a&iZ'
This expression may
-atDiW
-V
-DxW
OiZ
-U
-D,Z'
is,
JJ
(8.188)*
DAV
the criticaUtij conditionîor the

reflected reactor in the two-group approximation, since
involves both
the group constants and the dimensions of the reactor.
In the usual
regarded
as
it
be
way, the specification of the fuel concentration or the core size allows the
determination of the other through the use of the above relation.
Unlike
the simpler models and applications previously considered, however, the
somewhat more involved.
The details of an
procedure
procedure will be discussed later in connection with a sample
and some computational approximations which have been
computational
appropriate
calculation,
is

+ (D7')r,
=
=
(-W)C,
(z)jt, + (mci + (-itys,
to the general solutions (8.176)

of
equations in the four unknowns
and (8.185) yields the following set
Au Ct, Tu and Si:
of these conditions
VaJ
j>
HEACTOK
464
ANALYSIS
[CHAP.
For the present we confine

found to be useful will be demonstrated.
to
a
outline
of
the
method
and the necessary steps
general
our remarks
involved in a computation.
Suppose, for example, that the shape and dimensions of the reactor
have been specified, and that the composition of the nonfuel components
The problem, then, is to determine the critical fuel concen
is known.
The calculation may be carried out by assuming various fuel

loadings and finally selecting that loading which satisfies all the system
tration.
requirements
determined
above.
It
should be recognized that, in this
procedure, it is necessary to recompute several of the nuclear constants

As a rule,
KU k2, . . . with each assumed value of the fuel concentration.
only a few of these constants are markedly affected by variations in the
so that as a first approximation one need make
When a fairly reliable esti
adjustments only for these quantities.
mate has been obtained for the critical concentration, the entire compu
fuel concentration,
tation may be repeated, using the corrected values of the less sensitive
parameters.
Once a value has been assumed for the fuel concentration,
and all the

nuclear
parameters computed, the corresponding values of X2 and n2

These values, along
may be obtained from theêneraLrelation (8.171).
with the appropriate data for the various functions Z, W, . . . , may
If the value of the determi

then be inserted in the determinant (8.188).
nant is found to be identically zero, the assumed fuel concentration is in
facTlhe critical concentration,
if not, the procedure is repeated until
this condition is satisfied.
In the event that the problem of interest is the determination

yi^f
of the
multiplication constant of a given reactor system, the procedure outlined

For this case, all the dimensions and nuclear
above is slightly modified.
constants are fixed and the immediate objective is_toxJeirjmine_a_'e
such that the determinant (8.188) vanishes.
This is best accomplished
by selecUn^-varirjTO~raulei"of^^^\^cTrTs to say vc), computing the
corresponding values of X2 and fi2, and finally obtaining the value of A.
That value of k, call it kj,, which satisfies (8.188) gives the actual multi
plication constant k. Thus if kx corresponds to the experimental value
of v , then the multiplication of a noncritical system is given by
(8.189)
It is
of interest to note that the general method outlined here is entirely

analogous to the one used for the bare-reactor calculations, using either
For example, in the case of the
the one-velocity or Fermi age models.
Fermi
age model, the problem was to solve the equation
REFLECTED REACTORS
sec. 8.4]
for
critical dimensions from an equation such
B2 and then compute the

=
(for a sphere).
as B
465
In the two-group
t/R
model, the equivalent
relation is obtained from (8.169), which may be written

v'Sfl
Si2'
Pc
L\ =
where
+LV)U +LW)
(1
L\
and
-2
(8.191)
-2
Equation (8.190), along with the determinant (8.188), may be regarded

fllislnodel.
as the critTcuTityreqliirenients^Tor
[Note That the last two
factors orTThe~Tett sideof (8.190) give the fast and thermal nonleakage
As in the bare-reactor calculation, (8.190)
respectively.]
obtained
n2 (and X2) and the critical size (or concentration)
from the determinant.
e. Application
to
Cylindrical, and Rectangular Block
Spherical,
probabilities,
is solved for
The functions appearing in the determinant (8.188) are sum

marized for a few principal geometries in Table 8.3.
Note that in each
geometry, such as that of the block core, the boundary conditions apply
for all y and z along the core-reflector interfaces.
As used in (8.188),

Reactors.
however, the function cos n2y cos nsz appears as a factor in every term,
and since this factor is not identically zero, it may be factored out of the
determinant, which, then, becomes independent of y and z.
f. Comparison with One-group Method. Some idea of the errors in
the criticality requirements which can be caused by the omission of the
effects of the fast neutrons in a reactor can be obtained from the compari
son of the results based on the two-group model with those of the one-
Table
Summary
8.3.
As
model.
group (velocity)
a representative
case, we consider a spheri-
of Functions for Principal Reactor Geometries

Spherical Geometry
Z'(R)
w(*)
W'(R)
cos
= M
iiR
sin /xfl
sinh Xi?
=
=
X cosh
XB
U(R)
W
sinh \R
R*
sinh
=
_
V'(R)
V'CR)
Kt cosh
sinh k,(R'
- R)
ks(R'
R
R)
sinh k}(R'
TP
R)
R)
<i cosh
R~
sinh Kl(R'
R)
- R)
REACTOR ANALYSIS
466
Table
Summary or Functions
8.3.
[CHAP.
for Principal Reactor Geometries
(Continued)
Cylinder
=
Z(R)
Z'(R)
WB) =
W'kX)
J(liR)
(Side Reflector)
cos l,z
IJAUR)
cos Uz
h(l,R) cos hz
/,/,(i,fi) cos l,z
l/(R)
IT'(R)
[/o^K) KoM')
/o(i.S') #<,(/.)] cos /2z
U[li(ltR) KSaR') + h(Uft') Ki(UR)\ cos i2z
[IoihR) K0(hR')
/,(*,#') K0(lR)] cos In
hUiVJi) Kt(hR') + IoihR')
Ki(ltR)] cos 1*
=
=
Cylinder (End Reflectors)

= J<t(m\p) cos m2ft
= mjy0(mip) sin m,h
= J0(m,r>) cosh m,h
Z(R)
Z'(R)
W(R)
W"(R) = miJo(mip) sinh mih

2.405
m|
= Jo(>",p) sinh m4(5 A)

= 4/(mip) cosh m((3
t/(R)
U'(R)
V(R)
V'(R)
Jtsimip) sinh m6(3

= m6J0(iip) cosh
X' +
A)
ml
h)
mi
m}
+ m|
h)
toi(<2
TO?
Block Geometry
cos nia cos nty cos n>z
Z(R)
Z'(R)
;
W(R)
R"(R)
H*
- nl - n\
ni sin n,a cos n2j/ cos
n%z
cosh n,a cos n2j/ cos nz

n< sinh na cos n,y cos njz
ni
l/(R)
i/'(R)
V(R)
K'(R)
+ n + n\
n, =
f.
2C
11
25
= sinh ns(3
a) cos n2t/ cos n2z
= ns cosh nt(<2 o) cos n2i/ cos njz
= sinh n( a) cos n2j/ cos njz
= ne cosh 7i(3 a) cos n%y cos njz
J +
n\ + n\
= k\ + n\ + n\
cal reflected reactor which has the following material properties:

Core:
Di
2}J>
= Z)2 = 1 cm
= 0.01 cm-1
pc = 0.90
2
2/
= 0.05 cm-1
= 0.0229 cm-1
Reflector: D3 = Dt = 1 cm
2J," = 0.01 cm"1
pR = 0.90
2< =
0.0049 cm"1
= 22 cm
REFLECTED REACTOKS
sec. 8.4]
467
denotes the extrapolated reflector thickness.

Table 8.4 lists
the critical radii of the core for this system based on the two models.
For comparative purposes the results for the bare reactor, with the core
This table gives also
properties listed above, have also been included.
used in each calculation
the basic relations

numbers.
Spherical
(^Critical ÂDrl of
Cores )in Various Approximations

in ail cases)
(Fast fissions neglected
pth(1 + L'B'>"'e"<',r", =
&' extrapolated
(^f,) Ml
reflector
1)
DC(BR cot BR
-D{*XR
>V')-'(1 +
76.8
(6.83)
187
(8.155)
1S7
66.5
(8.73)
(8.11)
(8.85)
coth
L.V')-'
.('
=
R) +
186
(8.11)
(8.188)
(8.171)
180
(8.190)
.(S'
- R)
D[*R coth
1)
PcU + BHLC> + ru,")!"'
two-group
BR
"
1]
oore, one-
oore,
L',B*)"'
1]
Pe(1 + Lc'B')~'
(cm)
0. Reflected
L''B'yH1
(^)
DC(BR cot
th
velocity (using
migration area)
pth(1
4. Reflected core,'1
one-velocity
Tw)
Critical
radius
(5.183)
3. Bare core, twogroup
outer boundary.
4,
is 2
is
The strictly one-velocity calculations, methods

and
are seen to
give much smaller critical radii than the others; this indicates that the
fast leakage
Calculations and give
quite important in this system.
5,
identical results because of the large size of the bare core. It interesting
to note the comparative accuracy of the "migration-area" calculation,
which predicts, in this case, very nearly the same result as the two-group
method.
The form of the flux predicted
two-group
calculation,
yields a thermal-flux
is
also of interest the one-velocity calcu

The
yields the general functional shape shown in Fig. 8.20a.
6,
lation,
4,

leakage)
+ L'B')
2^)
P'h(1
2. Bare core, one-
velocity (with fast
"
z!r)
1. Bare core, onevelocity (no fast

leakage)
5. Refleoted
Reference
equations
Critioal equation
Method
8.4
Table
and the reference equation
where
distribution, which,
as
ANALYSIS
REACTOR
468
[CHAP.
shown by Eqs. (8.176) and (8.185) and Table 8.3, is a linear combination
of two functions in both the core and the reflector:
For the thermal flux in a spherical core,
Mr)
For the thermal flux in
Mr) *(r)
c,
t(r)
(sin
r
a spherical-shell
[sinh Kt(R'
y.r
sinh Xr)
(8.192)
reflector,
- r) + c, sinh ks(R' - r)]
(8.193)
and Ci are constants to be determined by the calculation

described above.
The spatial form of these functions, as well as of the

where
4>r, Co,
Fig. 8.20 Comparison of the flux distributions

reactor for the one- and two-group models.
in core and reflector of a spherical
fast flux, are shown in Fig. 8.206.

The dip in the thermal core flux and
the peak in the reflector thermal flux result from the thermalization of
neutrons in the reflector; as indicated by the slope of the curve at the
core-reflector interface, there is a net flow of thermal neutrons from the
reflector to the core. A numerical example of a reflected spherical reactor
is considered in detail in Sec. 8.5.
The evaluation of the

g. Evaluation of Criticality Determinant.
criticality relation (8.188), although straightforward, is generally labori
ous and time-consuming.
This is especially true in the case of cylindrical
geometries wherein the functions Z, W, U, V and their derivatives which
each involves several terms.

It is possible,
to achieve some simplification in the calculation if one takes
advantage of the fact that many of the terms in the determinant are
relatively insensitive to variations in the fuel concentration (or the size
appear in the determinant
however,
of the core).
This feature
determinant and rearranging
may
be
demonstrated
by expanding
some of the resulting terms.

We begin by dividing the first column of the determinant
the
by the func
REFLECTED REACTORS
sec. 8.4]
469
tion Z, the second by W, the third by U, and the fourth by V.

way we obtain the expression
1
Z'
w 33
^"
Each of these functions
a3
= 0
(8.194)
Z)2a2co
If
U'
^
77
(8 .195)
is,
where
-Dtf
a,
-DlU
A =
In this
of course, evaluated at the interface, e.g., in

= R.
the case of the sphere, at
We next expand (8.194) in terms of the minors of the fourth column.
It can be shown, after some algebra, that the resulting expression may be
written in the form
D1Z)2co(a2
ox)
DJhufHfli
Z)!Z)/3(a2
gi) + D3DtW(ai
at)
a,) + ^{a^D^ + asDiZ)*)
+ a2D2P3)
a^(q3Di4
In many
right-hand side of this equation

rela
tively insensitive to the fuel concentration, the influence of which
felt
through the value of the parameter p. This dependence
easily seen
from the quadratic equation (8.171) which defines p. Note, in particular,
that
change in the fuel concentration affects directly the value of 22)
a
is
is
is
cases of interest the
and thereby the values of
<c2and
kK.
The computationaj_procedure
to be followed in using (8.196) consists
in selecting various values of the fuel concentration, computing the cor
X)
from (8.171), and evaluating each side of

The right-hand side, call
the relation (8.196).
(m), and the function
then
be
as
function
of
the
fuel
may
plotted
concentration, and the
(m)
(and
it
responding values of
of the two curves yields the critical concentration.

Since
slowly varying function of n, one
the function (ji) will be, in general,
will suffice to determine the curve of
or two estimates of
versus fuel
The function (n), on the other hand,
rapidly varying
concentration.
of
For example,
particular value of
(8.197)
for
cot pR
reduced to one, or so, estimate of 0(/x) and
(i.e., fuel concentra

a
is
Thus, the evaluation of

tion)
relatively simple form.
reflected spherical reactor,
in the case of
and, moreover, has

a
function of
y.
is
intersection
simple calculation
f.
The procedure suggested here

discussed in some detail in the sample
are presented.
calculation of Sec. 8.5, and appropriate curves of and
is

(8.196)*
REACTOR ANALYSIS
470
Sample Computation
8.5
[CHAP.
Using Two-group Model
Nuclear Package Power Plant. As an illustration

of the use of the two-group model in reactor calculations we consider a
small nuclear power plant for generating steam and electricity. The
essential nuclear and engineering features of this reactor are as follows:
The core consists of a cylindrical array of MTR-type fuel elements
The water,
mounted within a cylindrical pressure shell filled with water.
a.
Description of
which serves both as coolant and moderator, enters the pressure shell
through a set of pipes, as shown in Fig. 8.21a, and first flows downward
Control

rod
Pressure
shell
55.8
(a)
Fig.
8.21
Nuclear package power plant: (a) actual geometry;
(6)
idealized model.
in an annular region surrounding the fuel elements, then upward through

the fuel elements, and finally out of the shell through another set of
The temperature of the entrant water is 432F, and that of the
pipes.
The 450F water, under pressure, then passes through a
outlet, 450F.
heat exchanger and produces lower-pressure steam for a turbogenerator
system.
The inlet water which first passes through the annular region surround
ing the core serves as a side reflector, and the water below and above the
The individual fuel elements, which are assem
core, as an end reflector.
bled vertically, consist of a box structure of parallel plates similar to the
The structural material of these elements is
design shown in Fig. 5.28.
stainless steel, and the fuel-bearing plates are fabricated from an iron
clad on both sides with stainless steel.
The fuel elements are

55.8 cm long, and the cylindrical cluster has a diameter of 55.8 cm.
The
annular region of water which forms the side reflector is 17.8 cm thick,
and the end caps are 91.1 cm thick. The volume of the core cylinder is
alloy of
U2S6
136,400 cm5, and the volume fractions

vw = 0.797
!>B8
of its constituents
= 0.110
vA
= 0.093
are
(8.198)
REFLECTED REACTORS
sec. 8.5]
471
where vw refers to the water, i'SB to the stainless steel, and vFA to the fuelbearing alloy.
Our primary interest in this calculation is the determination of the
critical mass of the hot clean reactor and the radial distribution of the
fast and thermal flux throughout the core and reflector.
An accurate
analysis of this system must necessarily take into account the completely
reflected cylindrical geometry shown in Fig. 8.216.
However, since this
would entail a somewhat involved calculation, we will approximate the
actual configuration by an equivalent reflected sphere of the same compo
sition.
This will reduce our computation appreciably and yet not obscure
A
any of the essential steps in the application of the two-group model.
study of the effect of the corners in the completely reflected cylinder will
until the next section.
For the purposes of the present calculation
be deferred
we assume
that the equiva
lent spherical system1 is one which has the same core volume as the
actual geometry and a spherical-shell reflector 17.8 cm thick.
On this
=
basis the radius of the spherical core is R
31.93 cm, and for con
venience we take the extrapolated outer radius of the reflector to be

R'
= 50 cm.
It
important to note that the simplification proposed

here, and in particular the selection of the equivalent system, is not to be
construed as necessarily representing a good approximation.
Its only
is
function is to clarify the presentation.

b.
In the calculation which follows we will use the

Group Constants.
set of data.
For the densities of the various constituents of the
following
core we take:
Material
Water (450F)
Stainless
Alloy
Fuel
where by the term
"alloy"
Density
0 . 832 g/cm
7.75
7.85
10.9
diluent which serves as the

carrier for the fissionable material.
According to the notation of (8.198),
the volume fraction of the core occupied by alloy plus fuel is vk.
For
convenience in computation we define two additional volume fractions
such that
Vyk = v, + vA
(8.199)
we mean the iron
where vF is the volume fraction of the core occupied by the actual fission
able material and vA the fraction occupied by the diluent.
This notation
will be useful when selecting various fuel concentrations (i.e., values of
vF) for evaluating the criticality determinant.
1 Alternative procedures for
determining equivalent systems is given by F. G. Prohammer, "A Comparison of One-dimensional Critical Mass Computations for
Completely Reflected Reactors," ORNL-2007, Mar. 1, 1956.
REACTOR ANALYSIS
472
The nuclear concentrations

table are easily computed.
= 0.02219
= 0.009169
NBB
Nw
of the components
in the above
listed
These are
NA = 0.08443
N, = 0.02459
nuclei/cm3
1024
[CHAP.
X 10"
X 102*va
X 10"ty
(8.200)
where we have used the relation
(8.199) and the data of (8.198) and

assumed that the average temperature of the water in the core is 450F.
The group constants for this reactor have been computed from the
methods described in Sec. 8.4b, using the data (8.200).
These are listed
in Table 8.5.
Table
Two-group Constants
8.5
Flux group
for Package Power Plant
Core
Reflector
TC = 45.3 cm1
Fast
Dt =
1 .44
= 1.0
PC
y(l)
Di
Thermal
TK = 40O cm1
cm
- D,/tC
= 0.03179 cm"1
Dz
1.85 cm
PR
1.0
D,
= 0.229 cm
L\
- 0.04625 cm"1
- 0.204 cm
- 16.64 cm'

0.01226
The thermal-group absorption cross sections will

following data (for neutron temperature of 450F)
*
The corresponding
concentrations
macroscopic
(8.200) are
2^'
2<a8)
s(1>
A1
TC
2)
#4
cross sections
Zf
based on the nuclear
= 0.152i>A
= 10.92y
(8.201)
k,
= 0.02208 cm-2
'"-=-=
2>
= 1.80 barns
= 444.1 barns
<r(aF)
= 0.01002 cm-1
= 0.01705
and the values of the various
from the
of
= 0.4518 barn
= 1.86 barns
a?8'
be computed
cm-'
0.0250 cm-2
= 0.06010 cm"2
id = 0.1486
cm-'
k3
= 0.1581
cm-1
k4
= 0.2452 cm-1
(8.202)
Evaluation of Criticality Determinant. In the present section we

trace through a sample calculation of the criticality determinant, using
For this calculation we assume one
the procedure outlined in Sec. 8.4g.
vp
value of the volume fraction of fuel
(i.e., the fuel concentration) and
In practice, several
evaluate the functions f and 0 from the Eq. (8.196).
c.
REFLECTED REACTORS
sec. 8.5]
473
values of vP will be required to establish the shape of the functions f

and 0. The value of vF we select for the computation which follows is
vF = 0.006; since this does not correspond to the critical fuel concentra
tion,
be equal to 0 in this case.

= 0.006, we obtain vA = 0.087 from (8.199), and by (8.201)
will not
For v,
Zf
2"'
= 0.01322 cm-1
Z/> = 0.06552 cm-1
2,
thus
= 0.1058
=
k\
= 0.05496 cm-1
2(2)
fc
(8.171).
= 0.4620
^%f^
We compute next the roots of the quadratic
cm"'
cm"!
= 1.278
equation
(8.169), using
These are found to be

p.
= 0.07606
X =
0.6999
(8.204)
The various functions which appear in Eq. (8.196) may now

We require
a>

(8.203)
= ijY =
=
0 =
4f
V
=
=
be computed.
- 4 0.6686
-0.1905
-k3 coth Kf - 4
K
- K -0.2766
-** coth
X
coth
\R
kaT
(8.205)
where we have used the definitions for W, U, V and their derivatives as

listed in Table 8.3 for the case of the reflected-sphere reactor, and
T = R' R. For completeness we list also the constants
=
Dlfi* +
at =
'
-4.983
(8.206)
These data, (8.205) and (8.206), may now be used to compute

values of 0 and f from (8.196).
For 0 we obtain
- oi) -
+u(a3DiDt +
ft
>1D2o,(a2
a2D2D3)
DiD^jai
DiDtPia* +
a*)
+
+
o3)
D3D^<3(ai
iK^D,/), +
-0.1598
The function
the
- oj
a3Z)1)4)
(8.207)
we compute from
= ~ = n cot tiR
-i
-0.1193
(8.208)
REACTOR ANALYSIS
474
[chap.
using the spherical functions given in Table 8.3.

and (8.208) are not identical, the determinant
Since the results (8.207)

(8.188) is not zero, and
therefore v? = 0.006 does not correspond to the critical fuel concentration.

-0.3,
Figure
volume
8.22 gives a plot of the functions
fractions
near the critical value.
and
9 over
The solution
a range of
fuel-
to the present
problem is seen to be
(vF)ianl = 0.006425
It
is of interest to note the shape of the functions
f and
8.
Evidently,
is a very slowly varying function of vF, whereas f is a very rapidly varying

This behavior was previously mentioned in Sec. 8.4g.
function.
The critical mass of U235 for the present system is easily computed
from the volume fraction
M,
p,vFV
vF.
Thus
= (10.9)(0.006425)(136,400)
= 9.55 kg
Flux Distributions. The calculation of the fast- and thermal-flux

distributions requires the evaluation of the various constants which
For this purpose we need also
appear in the Eqs. (8.176) and (8.185).
the criticality values of those nuclear constants of the system which are
functions of the fuel concentration.
These may be obtained by the
d.
SEC.
REFLECTED REACTORS
8.5]
The appropriate values are found to
procedure outlined in Sec. 8.5c.

2*>
2/'
= 0.01316 cm-1
oi
475
= 0.07016 cm"1
kK = 1.316
4 = 0.4820 cm-2
=
= 0.2842
-4.859
a2
o3 =
=
=
0.08057
X
0.7146
/x
The flux distributions
fa
The constants A
i,
cm-'
= 0.1104
(8.209)
-6.461
are given by the relations
fa = AXZ +
=
be
dW
S,U
fa
Ci, Th and S\
= aÂyZ
+
TyV - aÛ
fa
=
a^CyW
(8.210)*
are determined from the set of equations
have already
recognized, these equations are

Since,
(8.187).
homogeneous, it is not possible to obtain explicit relations for all four
The most information we can obtain from these equations
constants.
Thus, the computational procedure
is the ratios between the constants.
If this
is to solve for three of the unknowns in terms of the fourth.
as
we

procedure is followed in the case of Eqs. (8.187a) and (8.1876), the fol
lowing relationship is obtained between C\ and A\\
c.=f(|^:)^
The functions which appear in this expression are given in the notation of
Eqs. (8.195), and in the usual way, these are to be evaluated at the coreThe result (8.211) may be used to determine Si
reflector interface.
from the relation
St =
+ WC1)
jj{ZAl
(8.212)*
[which comes from Eq. (8.187a)] and 7\ by
Ti
+
^ [AxZ{ai
a3)
Cjr(a2 +
a,)]
(8.213)*
These three expressions, then, give Ci, Si, and Ti in terms of A\.
We now apply these general results to the present problem.
For this
This value of Ai
purpose we use the data (8.209) and let A\ = 12.41.
ie.,
is based on the normalization of the function fa;

requiring that fa(0) = 1. Then
Ai
= 12.41
Si = 0.6986
Ci = -1.548 X
Ti
lO"10
= -0.8135
we obtain
A\
by
(8.214)
REACTOR ANALYSIS
476
40
30
20
10
[CHAP.
Radial distance r, cm

Fig.
8.23
Radial distribution
of fast and thermal flux in reflected
sphere
reactor.
The substitution of these results into the flux relations (8.210) yields, for
the radial distributions,
*<r) =
12.41
S-^-r
r
Mr)
= 3.527
4>t(r)
= 0.6986
*(r)
1.548
X 10-
7.552
sinh k3(R'
-0.8135
10(8.215)
r)
sinhK<(^-r)
r
4.513
sinh
*(/*'
r
- r)
It is of interest to note
These functions have been plotted in Fig. 8.23.
the peak in the thermal flux near the core-reflector interface.
8.6
Completely
Reflected Systems
a. One-group Model.
A general analytical solution of the completely
This creates no particular
reflected reactor is presently not available.
In these
hardship when only a rough reactor calculation is required.
which is
cases the effect of the "corners" in reflected configurations,
generally small, can be entirely neglected since only a very small portion
Thus to ignore
of the total neutron population occupies these regions.
the corners introduces small errors into the critical-mass, calculations
even though the estimated local flux distributions may be in considerable
error.
When more accurate critical-mass estimates are required, these
sec.
REFLECTED REACTORS
8.6]
477
effects cannot be ignored, and in the absence of analytical solutions, it

has been necessary to develop various approximate methods.
These
methods are generally of an iterative type and involve the use of the
In the dis
concept mentioned previously.
cussion which follows we describe two such methods which have been
applied with some success to the one-velocity model and to the two-group
model.
As a first example, consider the analysis of a completely reflected
"equivalent-bare-reactor"

cylindrical reactor on the basis of the one-velocity
Fio.
8.24
model.
A typical
Actual reactor and equivalent side-reflected cylinder.
Let the dimen

system of this type is shown in Fig. 8.5 (cf. Fig. 8.21).
sions of this reactor be as follows: core, radius p0 and height 2h; reflector,
Given values for these dimensions and the
composition of the reflector, the problem is to determine the critical fuel
concentration in the core.
To solve this problem we study first a cylindrical reactor with side
reflector only which has the same reflector material, the same fuel concen
tration, and the same core and reflector radii (p0 and pi, respectively)
as the actual system.
Such a system I is shown in Fig. 8.24.
The
only feature of this geometry as yet undetermined is its length (21).
Thus our next objective is to find I such that the two systems shown in
Fig. 8.24 are critical with the same fuel concentration.
One plausible method of obtaining a first guess for I is that of finding
the height of a completely bare reactor,
(radius pi), which is critical
with the same fuel concentration as that in the core of a critical cylinder
reflected on the ends only (reflector material taken the same as that for
radius
pi and height
2a.
III
REACTOR ANALYSIS
478
[CHAP.
the actual system) and having the dimensions shown in Fig. 8.25, II. Thus
the value of I is determined by requiring that the two systems shown in
Fig.
critical with the same fuel concentration.
8.25 be
In this approxi
for the com
the procedure for calculating the fuel concentration

pletely reflected cylinder would be as follows:
Fig.
1.
8.25
Equivalent end-reflected and bare cylinders.
Calculate the buckling B^ for the end-reflected cylinder; thus,

(8.216)
B,1,
critical with the same buckling;
B"
<8-217)
Equation (8.217) yields an expres
^-
Computation of critical fuel concentration

I
value of
(8.218)
of system
I.
sion for
be
I:
where
is
b;"
will
Gi)'
thus, compute
III
()'
with nc given by (8.54).

2. Find I such that system
3.
With the
obtained from (8.218), find

(8.219)
where Xc
is

mation
given by (8.42).
Since the fuel concentration
of the actual
SEC. 8.6]
REFLECTED REACTORS
system is to be the same as that of system
I,
479
we set
(8.220)
i-[2,-
Z<]
Bl
(8.221)
It should be noted in using this procedure that there does indeed exist
an I for which Eqs. (8.219) and (8.220) hold accurately; however, the
value obtained by steps 1 and 2 is only an approximation, and one whose
validity could be ascertained only with great difficulty.
b. Two-group Model.
The general approach used in the case of the
one-velocity model is easily extended for use with the standard two-group
method described in Sec. 8.4.
As in that calculation, the primary objec
is
tive here
the estimation of the reflector savings in the direction of each
axis of symmetry of the reactor.
One scheme which has been devised

for this purpose employs an iterative procedure involving the analysis of

a sequence of elementary reactors each of which incorporates only one
In each of these calculations it
feature of the actual reactor geometry.
is assumed that only one face (or pair of faces) is reflected while the others
For example, reactor 1 in the sequence could be

reflected on the Xi face and bare on the x2 and x3 faces.
Reactor 2 could
be reflected on the x2 face and be bare on the Xi and x3 faces, and so on.
The important point here is that each of these elementary systems has a
This is
sufficiently simple geometry to allow an analytical solution.
are taken to be bare.
carried out with the aid of the two-group model, and each calculation
yields an estimate of the reflector savings for the face(s) which has been
reflected.
By progressing down the sequence of reactors, a complete
cycle is eventually made around the actual reactor and a reflector savings
obtained for every face.
It should be evident from these introductory remarks that the crux of
the entire matter is the selection of the dimensions (or bucklings) of the
bare faces for each of the elementary reactors in the sequence.
this point that we Introduce the iterative procedure mentioned
It
is at
above.
In carrying out the calculations for the first complete cycle around the
reactor, it is necessary to make a first rough estimate of the reflector
savings (i.e., bucklings) for all three faces.
These data may then be
used to obtain an improved estimate for the savings.
For example, in
analyzing reactor 1, the rough estimates for the bucklings in the x-i and
x% directions may be used to compute a better value for the savings in the
The same procedure may be followed in the case of the
Xi direction.
other two axes x2 and x%. This first cycle yields, then, an improved value
By repeating this procedure

for the bucklings in all three directions.
for several cycles, one can easily obtain a convergent value for the
REACTOR
480
ANALYSIS
[CHAP.
It has been
bucklings (reflector savings) for each of the principal axes.
found in practice that this procedure does indeed converge very rapidly
and the results appear to be independent of the initial guess which may
Unfortunately it has not yet been
be used in computing reactor 1.
shown precisely what reactor is analyzed by this rationale since an exact
solution of a completely reflected system is not presently available.
How
ever, criticality estimates obtained by this procedure have been found to
agree surprisingly well with experimental measurements on the ORNL
Swimming Pool Reactor.
The application of the two-group model to each of the elementary
reactors in the iterative sequence described above requires the solution
For this calculation it is convenient
of the criticality equation (8.169).
to write this relation in the form
1
k'
=
(1
^4 p
=
where
(8-222)
+ L|B)(1 + tcB*)
- I ^ - rc
=
L\
(8.223)
multiplication constant required to main

kc is the infinite-medium
tain the prescribed system (viz., geometry) at steady state. Note that
this quantity is entirely equivalent to the number vc previously introduced
for handling noncritical systems. Now, in the case of a noncritical sys
tem, it is also possible to write (8.222) in the form
k
k =
(1
(8-224)
+ L\B*){\ + tcB*)
Combining (8.222) and (8.224)
we obtain [cf.
Eq. (8.189)]
k
K" ~ 2
"M2
M*
(8
2^Ll j
OD
(if
Bi
For each of the elementary reactor calculations in the iteration we

This quantity is given by the
require an expression for the buckling B1.
small positive root ^ of (8.222) [cf. (8.173)].
The solution for B1 is
[Mt
225)
where we have used (8.189) for km and the relation M* = tc

L\. An
approximate expression for the two roots (which will be useful for "hand"
calculations) can be obtained by expanding the radical in
Thus, one obtains
^^.(T-Ot'-t^T"-')]
X!-rF
i(-')

and
Taylor series.
<8m)*
sec. 8.6]
REFLECTED REACTORS
481
The application of these results to a reactor calculation utilizing the

iterative procedure is demonstrated below.
As an illustrative example consider a completely reflected rectangularblock reactor of the type shown in Fig. 8.8. A cross-sectional view of
this reactor along any principal axis would have the form shown in Fig.
8.26a.
The configuration which will be used in the iteration, however,
Bare ends (and sides)
Core
Cross section through rectangular-block

(6) simplified reactor.
8.26
reactor: (a) actual cross section;
Multiplication k
Fig.
8.27
Buckling as function
flected on one pair of faces only.
of multiplication
for rectangular-block
reactor re
In the case of the rectangular-block reactor we

is shown in Fig. 8.266.
In
three
such
one for each of the principal axes.
figures,
will require
each figure the actual core width and reflector thickness will be correctly
The other two pairs of faces are assumed to be
shown for one axis only.
Figure 8.266 shows the proper sectional view for the x axis. With
bare.
these three elementary geometries as a basis, the first step in the analysis
for each axis x assuming
is to compute the multiplication constant
fc,
various values for the bucklings (i.e., bare widths) of the other two pairs
The computational procedures outlined in Sec. 8.4 are directly
to graph these results
applicable here. The next step in the calculation
The notation which has been
in a suitable form, as shown in Fig. 8.27.
of sides.
is

a ~"~d
(b)
(a)
Fig.
"
REACTOR
482
ANALYSIS
[CHAP.
used in the figure is as follows:

Axis
and the
B,
B,
= x/26
= t/2c
y
2
Bl
B\ + B*
Buckling
B,
with
B%, m
of core
Half-width
B\ + B\
r/2a
B*
B\ + B\
(8.228)
"total" buckling
B* =
Bl + Bl + B]
(8.229)

The calculations for the elementary reactors in the iterative sequence

are carried out with direct application of buckling curves of the type
shown in Fig. 8.27 and the use of the expression (8.226) for p*. The
computational procedure might be as follows:
bare core dimensions (i.e., bucklings).
1. Select a set of equivalent
Denote these Oi, bi, and C\.
(Note that this is equivalent to assuming a
set of reflector savings.)
2. Compute the corresponding values of B(, B$, and Bl,lJ, where the
index (1) refers to "first estimate."
3. From a set of buckling curves of the type shown in Fig. 8.27, find
the resulting values of k\ k(yl\ and A;'1'.
4. Substitute these three values for k into Eq. (8.226) and compute
the corresponding values of B2 (here called fi1). Note that the kx which
appears in (8.226) is a fixed number throughout this calculation and is
determined by the selected fuel loading.

5. Use these three values of B1 to obtain improved estimates for the By.
These may be computed from
[>]' =
[B<]
- B*[ky\
[B?]*
m B*[k?>]
B*[k's]
- [fi]'
- [B
- [B\*
(8.230)
where it is understood that Bf} is the second estimate of the buckling

in the direction x, and is obtained from the value of the total buckling
in step 3. It should be
recognized that each of these Bf1 implies a second estimate of the reflec
tor savings in the x, directions, since, for example, from B we can com
pute o2, and (a2 a) is by definition the reflector savings for the x
B^fc',"] based on the estimate
fcj1'
determined
direction.
With these new values of B<" the entire procedure is repeated,

starting with step 2 above.
The completion of the four basic steps in this procedure constitutes
6.
a single cycle in the analysis,
and each
cycle is expected to yield an
sec. 8.6]
REFLECTED REACTORS
483
improved estimate of the reflector savings over the preceding cycle.

The number of complete cycles to be carried out is determined by the
accuracy desired, which in turn may be measured by how well the three
roots B2 in step 4 agree.
Even a rough guess for Oi, bi, and ci based on
the one-velocity relations (8.20) and (8.22) should yield convergence to
three significant figures in B2 within three cycles.
We demonstrate this general procedure by working through a numerical

example.
As an illustration we will compute the multiplication constant
- All dimensions in centimeters

- Origin of coordinates is
center of reactor core
Fig.
8.28
One-eight section of core-reflector geometry of Oak Ridge Research Reactor.
for
a specific fuel loading of the Oak Ridge Research Reactor.

The basic
nuclear and engineering features of this reactor which are essential to this
study are as follows: (1) The standard core consists of a 4 by 7 vertical
array of MTR-type fuel elements (cf. Fig. 5.27). The half-width in the
x direction is o = 27.7 cm, in the y direction b = 16.15 cm, and in the z
(vertical) direction c = 30.0 cm.
(2) This central region is surrounded
on four sides by a 7.75-cm-thick beryllium reflector.
(3) The beryllium
in turn is contained in a 1.905-cm-thick aluminum tank.
(4) Both the
top and bottom of the core are reflected by a double layer of metal and
The first layer is a mixture of water, stainless steel, and aluminum

and is 18.73 cm thick, and the second layer is 50 cm of pure water.
water.
A sketch of the core-reflector configtiration of the ORR is shown in
section in Fig. 8.28.

Note that the origin of the coordinate system is
placed at the center of the reactor core.
484
REACTOR
ANALYSIS
[CHAP.
The principal nuclear data for the two-group model are listed below:
kK = 1.677
= 0.07992 cm-1
Vc = 107,000 cm'
= 64 cm2
=
0.269 cm
Z)2
=
3.366 cm1
L\
tc
tp =
D,
= 1.32 cm
= 67.37 cm1
14 1
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
Multiplication constant k
Via. 8.29
Buckling-versus-multiplication
(standard core loading).
curves for Oak Ridge Research Reactor
in
three-
been carried out by means of the scheme described above; however,
the present study it was not necessary to do these by hand since a

This
group three-region (3G3R) code was available for the ORACLE.
was especially convenient in this case since the system is in effect a threeregion geometry in both x and y directions, as well as in the z direction,
if one takes the aluminum container into account.

The existing code
was applied (after a suitable reduction to a two-group model) to this
geometry, and the multiplication constant k was computed for each axis
for various values of the buckling for the remaining two directions.
These results are shown in Fig. 8.29.
The next step in the calculation is the determination
of the buckling
multiplication curves. This involves the iterative procedure

The actual iteration was begun with the assumption
that the reflector savings for the first three reactors were all zero. That
in step
of the iteration were taken from
the values used for the
B1 from the
discussed above.
is,

The first step in the calculation is the computation of the multiplica

tion curves for each of the principal axes of the reactor.
These have
sec. 8.6]
REFLECTED REACTORS
485
the bare-core dimensions.

Even with this very crude guess it was found
that the value of B2 converged to four-figure accuracy in four cycles.
The results of these computations are given in Table 8.6. The figures
in the left-hand column denote the cycle number.
Note that in this
particular formulation each row represents a complete reactor calculation
so that in effect we have computed 12 reactors.
The numbers in the
extreme right are the bucklings computed from (8.226) for the value of A;
Finally, note that the B\, Bl, and B\
recorded at the immediate left.
listed
dimensions
Bl
0.001670
0.002934
0.001570
Bl.
Bl.
Bl,
=
=
0.001503
Bl.
Bl.
Bl,
002767
0.001542
0.002754
0.001805
Bl.
0.001818
Bl,
Bl.
Bl.
Bl,
=
=
0.01220
0.005958
0.01268
k.
k,
: 0 8570
=
1 037
k.
0.8350
0.004504
0.003240
0.004604
k.
k,
k.
0.004585
0.003321
0.004270
k, -.
0.004559
0.003347
0.004296
k,
=
=
=
B'
0.01387
B = 0.008963
B'
1 187
0.01425
0.006007
1 187
fla = 0.006007
1 163
B*
1 180
1
0.003216
0.009460
0.002742
183
Iteration fob the Buckling of the ORR Reactor
183
B] B',
Bl
182
182
182
B*
0.006422
0.006127
B1 = 0.006075
B*
fls
/?J
=
=
0.006075
0.006092
0.006092
0.006092
It seen from the results of Table 8.6 that to four-figure accuracy the
= 1.182; this corresponds to
multiplication constant for the reactor
=
multiplication
1.677.
To find the k (i.e., the
an infinite-medium
kx
fuel loading) for criticality, several other guesses of kx need to be made,
computed, and the criticality value obtained from
the corresponding
graph or by interpolation.
As remarked previously,
not known to what solution the proposed
There
some reason to believe, however, that the
rationale converges.
Since an "exact" solution
end result corresponds to the actual reactor.
it
is is
is
k
is
of the completely reflected reactor

not available,
check of the present
method was attempted by comparing the computed critical mass for the
Oak Ridge Swimming Pool reactor with the known experimentally deter
This calculation was carried out on the ORACLE using
mined value.
is
two-region code derived from the (3G3R) code previously

However, one important modification was introduced.
mentioned.
subroutine was added to the standard code which carried out the iterative
This modification reduced the entire study
procedure described above.
a two-group

Bl
to the bare-core
Bl
8.6
B\
B\
correspond
KB\
B\
k,
cycle
k.
Table
1
under
= 16.15, and c = 30.0 cm.
ky
a = 27.7,
k.
values
to the single task of computing the group constants.
These data for the
ANALYSIS
REACTOR
486
[CHAP.
Swimming Pool reactor, along with the principal dimensions of the system,
are given below:
Table
8.7
Two-group Constants
Flux group
for ORNL
Swimming
Core
Reflector
TC = 64.0 cm*
Fast
Pc
D,
Thermal
TR
= 1.320 cm
D,
PR
= 0.02062
Pool Reactor
cm"1
= 0.269 cm
Dt
2<o
= 32.7 cm'
= 1.19 cm
=
1.
= 0.0363
cm-'
0.1602 cm
= 0.01916 cm"1
Reactor geometry:

Core: 4 by 7 array of MTR fuel elements (40.48 by 38.450 by 61.28 cm).

For calculation
Reflector: Effectively infinite depth of water all around.
20 cm water reflector in all directions.
assume;
The multiplication constant for the system described above was com
puted by the iteration program for several values of the fuel loading (kx).
By this procedure a critical mass Mc = 3.050 kg was obtained. The
known experimental value is Afp = 3.190 kg.
The two-group tworegion model using the iterative procedure gives, then, a critical-mass
estimate 4.4 per cent low.
8.7
Feynman-Welton Method
Introduction. The usefulness of the two-group model in determin

the
ing
space and energy distribution of neutrons in reflected systems was
discussed in Sec. 8.4.
Although this model cannot give a detailed picture
of the complete neutron-energy spectrum, it is nevertheless a valuable
a.
tool for obtaining good estimates of the thermal-flux distribution and of

can be performed by hand;
the critical mass.
These computations
the labor required is not small, even when one applies the
approximate method for evaluating the criticality determinant (Sec. 8.4g).
The effort required is further increased if one wishes to account also for
however,
the effects of a completely reflected core (Sec. 8.6).

For many practical applications, such as preliminary studies, the
primary information required is a reliable estimate of the critical mass.
model is well suited to this purpose, it is clear that a
formulation of the method which eliminates a large portion of the compu
tational labor would be most useful. Such a method has been devised
Since the two-group
by R. P. Feynman andT. A. Welton. The Feynman-Welton method

draws directly from the results of the one-velocity integral theory.
The
method
effectively
separates the space and energy dependence of the
neutron-flux function and describes the spatial
distribution of each
REFLECTED REACTORS
sec. 8.7]
487
However, these simpli

energy group in terms of its fundamental mode.
fications are achieved at the expense of some accuracy; nevertheless,
numerical results obtained by this method agree surprisingly well with
more
reliable
models,
especially
in applications
to
water-moderated
systems.
The particular merits of the method are: (1) it allows a two-group

treatment of a reflected reactor with significantly less effort than required
by the standard method; (2) it is well suited to small, fast reactor assem
blies; and (3) it is especially convenient for making quick criticality
calculations.
Although it has been

The method also has several shortcomings.
that the model developed for the bare reactor and for the reactor
with nonmoderating reflector has a sound theoretical basis, there is some
question about the validity of the model developed for the reactor with
a moderating reflector.
The treatment in this case is based primarily on
a physical approach and is somewhat more intuitive in nature than the
treatment of the simpler systems.
For this portion of the development,
we draw directly from the work of Feynman and Welton.2
It should
be mentioned also that the results obtained for the reactor with moderat
ing reflector apply only to spherically symmetric geometries.
These
calculations are further limited by the requirement that the reflector be
infinite in extent. Although finite reflectors may also be treated by this
method, the computational procedure rapidly becomes involved and a

shown1
great deal of the usefulness of the method is lost.

b. Summary of Method for Application to Two-group Model.
The
method developed by Feynman and Welton is a general technique for

treating the space and energy distribution of neutrons in a chain-reacting
The derivation of the general theory is the main topic of this
system.
section.
The most interesting attribute of the method, however, is the
relative ease with which the two-group version may be applied to numeri
cal computation.
Since an alternative treatment of the two-group model
for a thermal reactor has already been discussed in detail, it will be
instructive to begin with a comparison of the results obtained by the two

methods.
This comparison will reveal the unique features of the Feynman-Welton method and serve also to relate some of the new ideas
involved with concepts previously established.
For this purpose we
require a summary of the fundamental
relations developed by Feynman
1 B. Davison,
"Transport Theory of Neutrons," National Research Council of
Canada, LT-18, Chalk River, Ontario, January, 1947.
'See, for example, U.S. Atomic Energy Commission, "Reactor Handbook:
Physics," p. 450, McGraw-Hill Book Company, Inc., New York, 1955; and R. P.
Feynman and T. A. Welton, "The Calculation of Critical Masses Including the
Effects of the Distribution of Neutron Energies," Los Alamos Scientific Laboratory,
Report LA-524, Jan. 21, 1947, declassified.
REACTOR ANALYSIS
488
[CHAP.
and Welton; these are presented in a heuristic fashion, accompanied by

The detailed derivation of all
physical interpretations and arguments.
relations is presented in subsequent portions of this section.
Consider then a thermal reactor consisting of a spherical core of radius
reflector.
For the present two-group
a and an infinite moderating
model we label the fast neutrons in the core by the index i = 1 and those
in the reflector by t = 3. Thermal neutrons in the core are labeled i = 2
and those in the reflector, i = 4. By definition, all fissions are caused
by thermal neutrons (i = 2), and all neutrons born from fission appear
in the fast group. We write the neutron-balance relation for this system
in terms of the removal density (number of neutrons of group i removed
per unit time per unit volume around r): Ai(r) = S<?^,(r), where 0,(r)
is the neutron flux for group i and 2< denotes the removal cross section
For the thermal group (t = 2, 4), 2< is the usual absorption
cross section, and for the fast group (t = 1, 3), it may be approximated by
for group i.
Then, in the core (where
denotes integration over the core)
P,(r;r') dr is the probability that
of energy group i
collisions in both core
a neutron
r',
where
is
r'
is a
is
born at
after any number of eZash'c-scattering
eventually removed as group neutron in dr about
and reflector,
in
the core. The function Pi2(r;r')dr
the probability that a group
in the core escapes removal in the core,
neutron born at
slowed down
in the reflector, and finally returns to the core to be removed as
group
The source functions Si(i') dr' give the number of

neutrons produced in volume dr' about
For
per unit time.
thermal reactor these are
S,(r)
group
r'
neutron in dr about r.
S2(r) = î(r)
(8.233)
If
2}2)
is
the fission cross section for group

neutrons.
these are
applied to the balance equations we note that the removal rate of fast
neutrons
exactly balanced by the production rate from the fissions
where
is
Clearly the function Pi accounts for

produced by the thermal neutrons.
the leakages of fast neutrons in and out of the core.
The removal rate
of the thermals, however, has two components.
due to fast
The first
is

(8.157).
neutrons slowing down to thermal in the core, and the second, to those
which have slowed down in the reflector and then diffused into the core.
In general the removal densities Ai{r)
but
represent these by rough approximations
are complicated
in the present calculation we

based on the fundamental mode from the one-velocity
functions,
model.
In
the
489
BEFLECTED REACTORS
sec. 8.7]
so-called lower approximation used by Feynman and Welton, these are

taken to be of the same functional form for all energies; thus,
A.(r)
- AWo(r)
(8.234)
where A?1 is some constant and ^0(r) is an elementary function (sin B,r/r
for a spherically symmetric system). These functions are so selected
that they provide a good representation of the spatial distribution of the
neutrons in each group throughout
central region of the core (see
Fig. 8.30). It is to be expected, then,
the
appreciable discrepancies will

appear at the core-reflector interface.
This means that the neutron cur
that
Actual flux
across the interface cannot be

well estimated from the derivative of
To correct for this deficiency
ô(r).
we introduce the second integral in
rent
Core-reflector
Fig.
It is shown in a subse
quent analysis that in first approxi
mations the kernel Pn is independent of r and r' and may be given by
Pi2(r;r') cm w/Vc, where Vo is the volume of the core and w is a constant.
Eq. (8.232).
Then
[s1(t')
PhM
dr'
y-c
dx' =
jcS1(r')
jc At(f)
dr'
(8.235)
over the core gives the total production of fast

neutrons, we can identify w simply as the fraction of all fast neutrons
produced which are removed in the core as thermals, after slowing down
in the reflector. Moreover, Eq. (8.235) states that the distribution of
Since the integral of
<Si
these removals in the core is spatially uniform.
Our immediate objective is to obtain the criticality condition appro

priate to the neutron-balance relations (8.231) and (8.232). For this
purpose we require an additional expression for ô(r) which can be used
to elminate the integrals in the balance equations.
Let us define,
therefore,
*o(r)
Mi,0
Pi(r;r') dr'
fc Mi')
(8.236)
is the number of neutrons produced in group i per removal of a

i; clearly, ntf is analogous to the physical quantity tj.
A physical interpretation of the above relation is easily established if
both sides of the equation are multiplied through by Atf. Then the
right-hand side is immediately recognized as the number of neutrons
where
is
r,
namely, jijpAfttyotr') dr', times the

and this product
removed at
is
per unit time in di' at

probability that neutron born at
produced
r' r',
neutron from group

interface
Comparison of ô(r) and

actual neutron flux.
8.30
490
REACTOR
integrated
ANALYSIS
[chap.
The result is the total number of
over the core volume.
removals per unit volume per unit time at r, namely, Aôir), as speci
This equation is evidently a direct
fied by the left-hand side of (8.236).
application of the kernel method discussed in Sec. 5. 2d.
If we use the expression (8.234), the integral relation (8.236), and the
source function (8.233) in (8.231), the result is an equation in Ap and
A'q\ A second relation between these quantities is obtained by making

similar substitutions in Eq. (8.232) ; in this case, however, it is necessary
to integrate over r so as to remove the quantity
Jc
\po(r) dr
implied in the
term (8.235).
The resulting pair of equations in the unknowns A'J> and
A^ is homogeneous, and a functional relationship must be established
between the coefficients in order that there exist a nontrivial solution.
This is easily obtained by eliminating the Atf; the result is the criticality
condition:
" =
2>*>[i
(8"237)
MM8']
To solve this relation for the critical concentration implied in 2'*' and
2J?>
we require
system.
a relation
between the ntf and the dimensions
of the
This is obtained from the following differential equation:
-A-Wo(r)
2<'Vo(r)
/TOYoto
(8.238)
where the function ô(r) is approximately related to 0<(r) by virtue of

(8.234); thus,
=
A|?Yo(r)
Z*(r)
(8.239)
Note that this equation is satisfied by the

of (8.236) by the general
for
of this property of the
a
demonstration
given
(7.256);
proof
specific
is given shortly.
If Eq. (8.238) is integrated over the reactor core, the
resulting relation may be written in the form
wrim
(8-240)
where, in the case of a spherical reactor with infinite reflector, the bucklings Bi are obtained from the familiar one- velocity relations:
B\a cot Bxa =1
Dz
j
((c3a
=1
D\
{Kta
+1)
B2a cot B2a
1)
with
k\ =
2(3)
2<>
1
~
(8241>
k] = -Jr-
It
is interesting to note that each of the B, is computed from an equation

More
which involves nuclear constants of group i type neutrons alone.
over, the ô(r) used to describe the spatial distribution of neutrons in each
energy group, although of the same functional form, have different zeros
sec. 8.7]
REFLECTED REACTORS
491
" since in general

or " wavelength
B\ 9^ B2. The coupling between the two
energy groups is achieved through the criticality condition (8.237).
The constant
w is computed from the expression
where, for the spherically
[l
^]
>
symmetric system,
1_
M(.nJ
32<<>
(,
aggVfKj
+ OK,)(l +
<M)
These relations are developed later in connection with derivation of the

reflector thermal source term given in Eq. (8.235).
The computation of the critical concentration for a given core size
These are then used in
begins with the solution for the Bi from (8.241).

(8.240) to obtain the ntf. The substitution of these results into (8.237)
The details of this model are presented
yields the fuel concentration.
at the end of the present section, and suitable formulas are developed for
a direct computation of the critical concentration.
c. One-velocity Model.
Since the one-velocity integral theory plays
an essential role in the development of the Feynman-Welton method, we
will direct attention first to a study of some of the important properties
of the integral equations which are involved.

It will be convenient in
developing some of the pertinent ideas to begin with an elementary system
with which we have had some previous experience. Let us consider,
therefore, a critical one-velocity reflected reactor in which both core and
The basic differential equation for the
reflector are homogeneous media.

flux in the core of this system is
>V2<Kr)
2<Kr)
= 2!/*(r)
We showed previously in our study of the kernel method (see Sec. 5. 2d)
that the solution to this equation could be written in the integral form
+(r) = vL,
K(t;t')
Jc *(r')
dr'
(8.242)
where the integration extends over the core and K(i;t') denotes the flux
at r due to a unit point source at r'.
It follows from this definition that
K(t;t') is the solution to the differential equation
-DV2K(i;i') +
In general K(i;i') will
S0/C(r;r')
8(r
- r')
(8.243)*
function, but one which can

An example of such a
usually be obtained by straightforward methods.
kernel is the function + (and <_) which we constructed for use in the
be a complicated
of the slab with the first-scattering sources [cf. Eqs. (5.266)].

Another, even more elementary example of the kernel K is given by the
function (5.84) which applies to the special case of the homogeneous
problem
REACTOR ANALYSIS
492
[CHAP.
Throughout most of the work that follows we will not

require a detailed definition of the kernels which are specified, but ask
only that such functions exist. Thus the introduction of the kernels is
in main part a convenient formalism which allows an especially simple
infinite medium.
presentation of the neutron-balance equations.

The basic form of the balance relations used in the development of the
In most
present method is a somewhat modified version of Eq. (8.242).
instances the cross-section factor which appears in front of the integral
will be combined into the definition of the kernel. Thus in our later
analyses, we will deal with kernel functions which have a different
For the
physical interpretation from those given the function K above.
present we will not be concerned with the detailed physical features of
these systems but will focus our attention on certain mathematical
Since the equations which arise
properties of the basic integral equation.
in subsequent analyses will be very similar mathematically to the form
(8.242), we will be able to deduce the fundamental properties of these
of Eq. (8.242) and of its
Equation (8.242) is a homogeneous integral equation1'2 and possesses

solutions only for particular values of the parameter v. Let us denote
these values v, where n = 0,1, . . . , and let the corresponding solutions
to (8.242) be the functions ^(r).
*(r)
wjif
Then
Jc *(r')
K(rf)
dr'
(8.244)*
r'
r',
if
is
and the
vn are the eigenvalues of this equation
We now demonstrate that the
corresponding eigenfunctions.
is orthogonal on the assumption that the kernel K is symmetric3
K (r;r') the flux at due to unit
variables r and r'. That
at
then this must be identical to K(i';i), the flux at
due to
is,
The quantities
source at r.
the
\f/
set
\f/n
in the
source
a
unit
Thus, we require that
The first step of the proof
is
K(r?)
K(r';r)
(8.245)*
to multiply through equation

We obtain
(8.244) by
dr = 2,
*(r) dr
fQ
Mr) Mr)
fc
fc
^m(r) and integrate over the region C.
*(r') iC(r;r') dr'
(8.246)
I,
is
A discussion of integral equations

given by R. Courant and D. Hilbert,
"Methods of Mathematical Physics," vol.
Interscience
pp. 351-371, 122-136.
Publishers, Inc., New York, 1953.
Also, H. Margenau and G. M. Murphy, "The Mathematics of Physics and
is
Chemistry," pp. 509-512, D. Van Nostrand Company, Inc., Princeton, N.J., 1943.
The symmetry property of the one-velocity kernel K
easily demonstrated in the
case of both the bare and reflected slab and sphere geometries.
For example, note
that +(x;s) = _(s;x) in Eq. (5.266).
*

equations from a knowledge of the properties

kernel.
SEC.
REFLECTED REACTORS
8.7]
By interchanging the indices n and

f0 *.(r)
Mr)
dx =
n,
= VmXf
m we have also
Jc f(r)
fc
493
dr
MO
fc MO
K(t;t') dr'
K{r;x')
fc Mr)
dr'
dr
(8.247)
where the last step follows from a change in the order of integration.
Since we have required symmetry in K, (8.247) may also be written
fc *(r) Mr)
dr = ymX,
fc
MO
dr'
fc *(r)
Jf (r'jr)
dr
(8.248)
But, the right-hand side of this equation, except for the factor
identical to (8.246) ; therefore,
fc
Mr) Mr)
vm,
is
m^n
dr = 0
If n = m, this integral will be different from

where in general vn
vm.
we can fix its value to
zero, and by a suitable construction of the set
unity.
In this
way, we find that
fc Mr) Mr)
where
5mn
dr =
(8.249)*
Smn
denotes the Kronecker delta.
The orthonormal
set ^ may be used to obtain a suitable series expan

sion for both the flux
In order that such
and the kernel K(r;r').
expansions exist, it is necessary that the set ^ be complete.1 We will
Then, we may write
assume that this is the case.
(r)
+(r)
- 2 ^"(r)
(8-250)
The
are the coefficients of the expansion and can
relation
4.
be computed from the
fcM') (r)dr
(8.251)
which follows from the application of the orthogonality properties of the

In the same way, we can describe an expansion for the kernel K;
\f/.
thus,
The coefficients K(r')
Kn(r')
f(r)
are obtained in the manner of the >,
Kn{r')
fcMr)K(r;r')dr
Courant and Hilbert, Op. tit., pp. 359-360.
(8.252)
that
is,
K(r;r')

be
(8.253)
494
ANALYSIS
REACTOR
[CHAP.
The right-hand side of this equation may be evaluated by the application

of (8.244), since by the symmetry property it follows that
jT
*B(r)
K(t;i')
dx =
K(i';i)
j^n{x)
The substitution of this result into
(8.253)
dx =
yields
Kn(t')
(8.254)
(8.255)
And, the expansion for the kernel takes the form
/_/
vnLf
There is an alternate expansion for this kernel which can be obtained

from the original differential equation (8.243).
For this purpose let us
assume that the coefficients Kn(x') in (8.252) may be written in the form
[cf. (8.255)]

tfB(r') =
and compute the quantities
(8.257)
An
from the requirement that the expansion
K{x;x>) =
(8.258)
satisfy Eq. (8.243).

yields the relation
Direct substitution of this expression into (8.243)
(fly
- S.) T
-8(r-r')
If we multiply
through by ^m(r') and integrate over the region C, applying

the orthogonality property (8.249), we obtain
W-to
Am
- h Kit)
-+m(r)
Am
(8.259)
But we know that the function {x) satisfies both Eq. (8.242) and
V*(r) +
In the
same way, the
y[/m
B*4>(x)
= 0
satisfy (8.244), and
VYm(r)
BWm{x)
= 0
(8.260)*
These properties of the eigenfunctions ^m were displayed previously in

The expres
the study of the integral-equation methods (cf. Sec. 7.4d).
sion (8.260) may be used to eliminate the Laplacian in (8.259) ; thus, we
sec. 8.7]
REFLECTED REACTORS
495
obtain the relation

Xm
= 20 +
DBl
(8.261)
and substitution into (8.258) yields
*
(8.262)
Tt
Upon comparing this result with (8.256),
^fm
we find
that
Thus the eigenvalues vn and B are related through an equation identical

in form to the criticality condition for the one- velocity reactor. This leads
us to attempt a physical interpretation of the vn. Several such observa
tions, along with other properties of the eigenvalues and eigenfunctions,
are listed below:
The eigenvalues for a real, symmetric kernel are all real.1

2. The eigenfunctions \f/ have an oscillatory character, and all (except
We
one2) take on negative values in certain regions of the space C.
denote by ^0 the function which has no nodes in C, by \f/i the function with

1.
the next fewest number, and so on. In the case of the bare-slab reactor
in diffusion theory we found, for example, that
Mx)
= cos
^|
4>i(x)
= cos
+2(x) = cos
Note that
has no nodes,
has two, and ^2 has four.
3. The eigenfunctions ^(r) represent the spatial distributions of the
flux according to these modes, and the coefficients n of Eq. (8.250) give
the weighting function for each mode, i.e., the fraction of the neutron
population distributed in each space form
4. The eigenvalues vn may be ordered; thus
vo
< vi <
v2
<
(8.264)
We are not interested here in establishing a formal proof of this statement.

For this the reader is referred to the literature on integral equations and
It is possible, nevertheless, to offer some physical
eigenfunctions.3
arguments based on properties 2 and 3 and Eq. (8.244) which can give
some insight into the basis of the inequality (8.264).
For this purpose
it will be helpful to think of Eq. (8.244) in terms of a physical system.
Margenau and Murphy, op. cit., p. 511.
Courant and Hilbert, op. cit., p. 458.
* See, for example, P. M. Morse and H. Feshbach, " Methodsof Theoretical Physics,"
vol. I, pp. 719-725, McGraw-Hill Book Company, Inc., New York, 1953; also,
Courant and Hilbert, op. cit., vol. I, pp. 451-455.
1
[chap.
REACTOR ANALYSIS
496
describes the spatial distribution of neutrons in a particular

system, then v may be interpreted as the average number of neutrons
We
per fission required to maintain this distribution for all time.
has no nodes, and the
observed previously, in statement 2 above, that
Thus if
neutron yield per fission for this system we denote by v<>. The distribution
on the other hand, may exhibit several nodes, and with each there
will be a region in which
We may think of these regions as
< 0.
occupied by certain particles, the presence of which can cancel an equal
Thus to maintain this system in steady state

The same argument may be applied to vt.
However, m is also greater than v\ since ^2 is a more rapidly oscillating
function (i.e., has more nodal surfaces) than î, and the "leakage" of
mutually annihilating particles into one another will be more pronounced.
From this point of view it is clear that increasing the number of nodal
surfaces will increase the value of v required to maintain a particular
number of actual neutrons.
yp.
v\ must be greater than
space mode
v0.
Thence follows (8.264).
\pn
given by
where
t/
(8.265)
The above
and Sa denotes the absorption cross section of the mixture.
relation follows immediately from the interpretation given for v in
statement
above.
Since
ij = average number of neutrons produced per absorption

= average number of absorptions
per neutron produced
(8.266)
Vn
dr ~
X,
Jc
fadr
dx
Z'
Total neutrons produced
Leakage from core
Then
Jc
d. Multivelocity Reactor with Energy-independent Cross Sections.

The analysis of multivelocity reactors
carried out by considering a
The systems
sequence of reactor systems of increasing generality.
studied in this section, both bare and reflected, will be limited by the
assumption that all cross sections in the core (and reflector) are energy
is
independent.
dependent
Studies
of more general reactors which include
cross sections and moderating
reflectors
energyin
are considered
subsequent sections.
In the present study of multivelocity reactors
will be convenient to
To the first cate
separate all scattering collisions into two categories.
gory we assign all nuclear scattering phenomena which cause substantial
it

Leakage =
the mode
is
5.
The probability of leakage from the core of neutrons distributed in
REFLECTED REACTORS
sec. 8.7]
497
These processes we repre

changes in the kinetic energy of the neutron.
In the second category we place
sent by the cross-section symbol 2*.
all other scattering events which yield relatively small (and presumably
negligible) changes in neutron energy, and to these we assign the crosssection symbol 2e.
Precisely which of the various nuclear scattering
processes are to be included in each category is determined, of course, by
For
the expected neutron-energy spectrum of the reactor in question.
example, if the system contains an appreciable amount of hydrogenous
materials, then in this model it would be appropriate to assign the usual
scattering cross section (2.) of the hydrogenous materials to the group
All other materials in

represented by 2*.
relatively small changes in neutron energy in
properly include the scattering cross sections
On the other hand,
group represented by 2,,.
this system would cause

scattering, and one could
of these materials in the

if the system of interest is

a fast reactor, it would be in order to associate the 2* with the inelastic
cross sections of the various materials present and the 2e with the usual
"elastic" cross sections 2,.
A formal definition of these two scattering cross sections is as follows:
average number of "scattering" collisions per unit
2*(J) = path of travel by neutron of energy E which could
cause a "substantial" change in its energy
(8.267)
average number of "scattering" collisions per unit

path of travel of neutron of energy E which cause
a negligible change in its energy
(8.268)
2()
be
The term "scattering" has
been included
in
these definitions to differenti
ate these nuclear events from other processes, such as fission reactions,
which can yield neutrons at markedly different energies.
The particular merit of this formulation of the scattering phenomena is
that it allows one to treat a large variety of reactor problems by means
of the energy (velocity) group methods developed for the present model.
Thus by a suitable choice of the scattering-cross section assignments it
is possible to study thermal, epithermal, or fast reactors.
The cross sections introduced above may be combined with the cross
sections for the absorption phenomena to obtain a new definition for the
total cross section; thus,
2t 27
+ 2/ + 2* + 2.
where 27 and 2/ have the usual meanings (2.40).

introduce also a removal cross section 2r where
so
that
2r = 2 + 2*
2, = 2r + 2e
(8.269)
It
is convenient
to
(8.270)
(8.271)
REACTOR ANALYSIS
498
Note that the cross section 2r will

lost from
neutrons
"removal"
[CHAP.
for computing the number of

since both the 2 and 2* are
be useful
an energy group,
processes.
Homogeneous Reactor. We begin our treatment of multia study of the bare reactor in which all cross sec
tions are energy independent.
Neutrons emerging from fission and 2*
1.
Bare
velocity reactors with

reactions
will, in general,
distributed according
be
to some specified
energy spectrum, however.
(t,E)
S(t,E)
we introduce
the following functions:
dr dE = total track length per unit time in di about r

due to neutrons with energies in dE about E
dt dE = total number of neutrons produced per unit
time in dr about r with energies in
PB(r;r') dr = probability' that
For this analysis
dE about E
a neutron produced at
(8.272)
r' is
is
if
r',
The function /,fl(r;r') defined here will be the kernel in the integralequation formulation of the neutron-balance equation; note that it is
independent of energy in this particular model in which the cross sections
are energy independent.
It should be noted that this definition differs
from that given the function K(r;x') used in the preceding section.
Whereas the kernel K represented the flux at r due to a unit source at
function of the type K multiplied by 2r. This
the kernel Pb implies
we consider, for example, the
relationship may be seen more clearly
quantity 2rX(r;r') dr, a physical interpretation of which
about
2rK(r;r') dr = number of neutrons removed per unit time in dr

due to one neutron released per unit
r'
time from
= probability that neutron released at
removed in dr about
r'
is
is
(8.273)
The last statement follows from the fact that the flux
due to a unit
In setting up the neutron-balance relation

in
place of the flux, the removal
it
source.
will
(collision) density
be convenient
(8.274)*
then given by
A{r,E)
J/
is
The neutron balance
to use,
A(r,E), where
A(r,E) = 2r<Kr,)
reactor
S(r',E) PB(r;r')dr'
(8.275)
is
added here to distinguish this function for the bare reactor

The subscript
given for multiregion systems.
from subsequent definitions of
1

removed in dr about r
sec.
REFLECTED REACTORS
8.7]
In terms of the diffusion-theory
499
notation, the source function S takes the
form
S(r,E)
J" (r,E')
CE)Z,
fj A(r,E') (,')
dE' + 2*
fj
4>{t,E')
h(E;E') dE'
dE'
(8.276)

$(E,E>) ^
VZ^'
+ **W>E')
probability that
(8.277)
a neutron of energy
h(E;E') dE = scattered by the process

emerges
with energy in dE about
and z{E) denotes the fission spectrum [cf. Eq. (4.153)].

of the relation (8.276) into (8.275) yields
A(r,E)
["
JO
&(E,E')dE'
J\reactor
E' when
(s) represented by
2*
(8.278)
The substitution
A(T',E')PB(r?')dr'
(8.279)
A(r,E) = A(E)*(t)
(8.280)
into (8.279) gives immediately
Substitution of this relation
complete
separation of the space and energy dependence:
I"
J.
= m
JL
A(E)
%(E,E') A(E') dE'
(a)
Pb^ dI'
reactor
(8.281)
mX =
and the condition

(8.282)
These results are entirely consistent with those of our previous analysis
In this
of the bare reactor using the Fermi age model (refer to Sec. 6.3).
is
formulation, Eq. (8.281a) describes the neutron-energy spectrum and

easily recognized as the integral equation for the collision density in an
infinite homogeneous system. If we select, for example [cf. Eq. (4.36)],
=
(1
h(E;E')
- a)E'
and let 2* = 2, so that 2, = 2,

2a = 2,, then Eq. (8.281a) takes on
the form of Eqs. (4.105) when the fission neutrons are represented by
monoenergetic source. The solution of these equations in the asymptotic
In the more general situation involving
discussed in Sec. 4.4b.
range
distributed fission spectrum z(E), the methods and general results of
is

We look for solutions of this equation which have the form
Sec. 4.6 are directly applicable for obtaining solutions in the asymptotic
ANALYSIS
REACTOR
500
[CHAP.
range and lead to the critical equation in the usual way [see, for example,
Eq. (4.162)].
Equation (8.2816) gives the spatial distribution of the neutrons
throughout the reactor at all energies. The solutions to this equation
are given by a set of eigenfunctions ^(r) with eigenvalues X [cf. Eq.
(8.244)]:
*(r)
= X
*(r') Pb(t;t')
J[reactor
dx'
(8.283)
X0
Ms')
mode.
then the one-velocity
{x;x')
is
PB(.x;x')
where
Pb(x;x') dx'
If we adopt the formalism

kernel PB(x;x') may be written
for the fundamental
(8.284)
of Eq. (5.266),
2t(x;x')
(8.285)
the flux at x due to a unit source at x'.
Substitution into
= X02,
Mx')
4>{x;x')
dx' = vXf
Ja
Mx)
f'a
Eq. (8.284) yields
Mx') (x;x')dx'
(8.286)
If
the last equality follows from the general form1 (8.242).
easily
apply the solutions (5.266) for the kernel (x;x'), then
shown that the \po{x) given by (8.286) also satisfies the equation
where B% = {vV,
i'0'(x) + Bl4>0(x) =
2)/D. Thus in this simple
case ô(x) has the usual trigonometric
form, and the condition on the
the usual criticality relation for the one-velocity model.
eigenvalue
The critical size
this reactor
obtained by computing the lowest
of
Eq. (8.281a) and adjusting the size of the reactor until
eigenvalue m0
it
we
is
where
0,
is
is
the lowest eigenvalue
Xo
of (8.286) satisfies the relation

X0 =
where in this case
Xo =
reactor size through the

1

*o0c)
|_o
For an actual system having the nuclear and geometric properties assumed
This
above, only one of the solutions (8.283) is physically acceptable.
It is clear that, for n > 0, the
solution is the fundamental mode ô(r).
steady-state spatial distribution would require that A{r,E) < 0 in cer
tain regions of the reactor.
The determination of the neutron-flux
spatial distribution in a finite system by means of the integral-equation
formulation was demonstrated in Sec. 5.7c for the case of an infinite slab.
Those results may be applied directly to the multiplying medium problem.
In the case of the infinite-slab reactor, of width 2a, Eq. (8.283) takes the
form
Note that the functions
(8.287)
v2//2, and the cross sections are related to the
B\ equation.
K
and
(x;x')
have identical definitions.
REFLECTED REACTORS
sec. 8.7]
501
As in the pre
2. Reflected Reactor with No Moderation in Reflector.
ceding case, all cross sections in both core and reflector are independent
All
of energy although the spectrums z{E) and h{E;E') need not be.
cross sections in the core 2/, 2*,, 2JP, and 2<C) are in general different
In the reflector
and 2^S) are different from zero, but
from zero.
2/
= 2i*> = 0.
For this system
we define the kernel
di
P(r-r')
'
P(r;r') such that
Probability that neutron born at r'
is removed in dr about
In this definition r' must
this kernel allows

neutrons created in the core to wander out into the reflector, be "elastically scattered" about and finally removed in either the core or the
be in the core; however,
reflector.
hcK(V ' F'\'

o
=
~~
gn(E,E')
m 0
+ 2lC)HE;E')
(8.289)
where
A(t,E)
$C(E,E') dE'
denotes integration
fc
equation is of the form (8.275), with the Pb now

replaced by P given in (8.288). The substitution of (8.289), along with
the general relation (8.276), into (8.275) yields
fo"
The neutron-balance
A(i',E') P(r;r') dr'
tion
accomplished by the separation of variables,
calculation. The results are
yf
"
m m
Jc
A(E)
%C(E,E')
(8.290)
The solution to this equa
over the core.
is
as in the bare-reactor
A(E') dE'
*(r') P(r;r') dr'
(a)
(8.291)
(b)
Equation
(8.291b) has a set of eigenfunctions ^(r) and eigenvalues A.

are defined as orthogonal over the core only.
The solution for the critical size proceeds as before.
e. Multivelocity Reactor with Energy -dependent Cross Sections.
The
basic ideas developed for the reactor with constant cross sections are
The
ypn
In the present study we allow

for the energy dependence of all cross sections and focus our attention
applied here to more general systems.
1,
on reactors which are either (1) bare, or (2) have reflectors which do not
moderate neutrons.
For all reactors in class
the cross sections of the
core 2/, 2T, 2#, and 2 will be, in general, some specified functions of
energy.
For systems described by class
2,

The sources in the reactor are given by
we allow in addition that
REACTOR ANALYSIS
502
the reflector cross sections 2^' and

zy = Si*' 0.
[CHAP.
but that
be energy dependent,
2<ft)
we study in detail reactors of class 2.

The
for
these
are
modified
to
easily
results derived
describe those of
systems
class 1. As in the study of the constant cross-section systems of Sec.
8.7d, we begin the analysis with a statement of the neutron-balance con
dition. For this purpose we require the functions (r,E) and A(r,E)
In addition we introduce the kernel
defined by (8.272).
r' with energy

dr
r
E is removed in
about
without intervening
energy loss
probability that
P(t;r',E)
dr =
a neutron born at
(8.292)
Thus this new kernel includes the possibility that the neutron of energy
created at r' may experience any number of elastic scattering collisions

(described by Z) in both core and reflector before it is finally removed in
dr about r.
A(t,E)
In the diffusion-theory notation, it
is clear
S(t',E) = ,*()
f' ME')
be defined
(T',E') dE'
f0"
A
Z*(') (r',E') h{E;E') dE'
(t',E') (,') dE'
(8.294)
A(t,E)
W) h(E;E')
ZiC){E')
these expressions
ME')
fj
The substitution of
= vz(E)
$(E,E') dE'
(8.29o)*
into (8.293) yields
fc
>(E,K
where [cf. (8.277)]
It
that the source may
[cf. (8.276)]
fQ"
by
(8.293)*
JcS(T',E)P(r;x',E)dT'
A{i',E') P(t;t',E) dr'
(8.296)*
clear from the above expressions that, since there are no "sources"
of neutrons in the reflector, the spatial integration extends over the core
volume only.
the determination of
The immediate objective of the present section
is
is
It will be convenient in under

the solutions to Eq. (8.296) or (8.293).
somewhat modified version of the
taking this analysis to introduce
basic relation (8.293), which may be derived from a different physical
a

The neutron-balance equation may be expressed in terms of this func

If as before we define a source S(r,E) according to (8.272), then
tion.
the removal rate is given by
sec.
REFLECTED REACTORS
8.7]
503
For this purpose we construct a fictitious one-velocity

system in which all neutrons have energy E. The cross sections (2') of
this system are related to the cross sections of the actual system through
the following definitions:
2i ^
2;
(2)
(3)
Z'. 2'e
(1)
= 2T
= 2,
= 27
the core:
2;
the reflector:
= 2,
= 2.
(8"297)
In
In
~L'y
point of view.
2*
is
We note immediately from these expressions that the removal cross sec
identical to that of the actual system,
tion of the fictitious system (2^)
=
2;
2;
2;
2;
+
2;
2;
since
= 2, + 27 + 2* = 2r
(8.298)
is
that
The only significant difference between the two systems, then,
" slowing-down " cross section of the fictitious system
zero.
(2*)
quantity to define the neutron yield for
Now, let us introduce also
the fictitious system; thus,
a
is
the
(8.299)*
clear from these definitions that the source S(t,E) in the fictitious
due only to the removal reactions.
We can write, therefore,
system
is
is
It
S(t,E) =
(T,E)
ul{E) 2'T(E)
p(E) A(t,E)
(8.300)
A(i,E)
n(E)
A(r',E) P(r;r',) di'
(8.301)*
for Eq. (8.301) which

=
fc
tJjJB)
where the
possible, therefore, to construct

set of eigenare solutions to the equation
a
It
present system.
unctions 4>{t,E)
is
is
is
is
It important to recognize that, in this formulation, the variable

which appears in the integral equation serves only as a parameter.
The
principal variable(s)
r. Thus the analytical machinery developed pre
viously for the one-velocity model (Sec. 8.7c)
directly applicable to the
f
4,n(r',E)
P(t;t',E) di'
(8.302)
are the corresponding eigenvalues [cf. Eq. (8.244)].

With the present point of view
easy to show that the properties
of the eigenfunctions ^(r,) and of the kernel P(i;i',E) are identical to
those of the ^(r) and K(r;r') developed in Sec. 8.7c.
Consider first the
is
it
m(2?)
is
is
is
it
orthogonality properties.
From physical requirements
clear that
the kernel P(t;t',E) must be symmetric in and r'. If we bear in mind
that
to be treated as
parameter, then all the derivation leading to
the orthogonality relation (8.249)
directly applicable to the functions

fi{E) = average number of neutrons produced per removal
REACTOR ANALYSIS
504
\f/n(i,E) and
P(t;t',E), and
we obtain
jc Mr',E)
[CHAP.
immediately the property
*m(r',E) di'
(8.303)*
Smn
The series expansion for the kernel P(t;t',E) in terms of the eigenAs in the one-velocity
ipn{r,E) is also easily demonstrated.
derivation we postulate that the ^/{r,E) form a complete set. It follows,
therefore, that we can construct an expression for A{t,E) in the form
functions
A(t,E)
An(E)
Mt,E)
(8.304)
where the coefficients may be computed from the relation

=
An(E)
In the same way

kernel
K,
we define,
PW t^'M
are given by
Thus,
(8.305)
P.W
P(r?',E)
(8-30(i)
*(r>)
it is easily shown that the coefficients
(8-307)
(8.308)*
The two basic results obtained above, (8.303) and (8.308), are gener
ally valid and may be applied directly to the physical system described
Our only purpose in introducing the fictitious system was to
by (8.296).
offer some aid in grasping the essential ideas and analytical tools which
Let us now apply
may be used to solve the neutron-balance equation.
The substitution of (8.304) into (8.296)
these results to Eq. (8.296).
M',E)
g(E,E')
dE>
P(t;t',E)
An(E)
fc
f0~
yields
An(E')
W,E')
dv'
dE>
An(E') fc*m(r,E) di
fc
g(E,E')
Am{E) =
fQ~
We multiply this equation by ^/{t,E) and integrate over the core.

the aid of (8.303), we obtain
P(t;t',E)
With
dr1
ME')
Mr',E') di'
jQ
(,') dE'
fc
J*

dx
on the basis of our experience with the
we use this expression in (8.302),
Pn(E)
A(t,E)
the series expansion [cf. Eq. (8.258)]
P(t;t',E)
If
fcMr,E)
P{t;t',E) +m(r,E) di
(8.309)
REFLECTED
SEC. 8.7]
Since the kernel
gration; this yields
is symmetric,
REACTORS
505
apply (8.302) to the r inte

Equation (8.309) then reduces to
we may
^m(r' ,E) tim(E)
the form
Am(E)
^ JJ
An{E') jjn{r',E')
%{E,E') dE'
+m(r',E) dt'
(8.310)
If
we call
Anm(E,E')
fc
Mr',E') Mi',E)
dr'
(8.311)

=
Am(E)
It
V /
^ /_/
Jo
Hm{&)
An(E') $(E,E') Anm(E,E') dE'
(8.312)*
should be noted that, for
E', (8.311) reduces to the orthogonality
relation (8.303), i.e.,

Anm(E,E)
= 5
(8.313)
The eigenfunctions \ffn(r,E) are computed from (8.302) and represent

the solutions to a set of one-velocity problems, each denned by a specific
value (or increment) of E. Once these one-velocity problems have been
solved, the Anm(E,E') may be computed from (8.311), and these in turn
used to solve the integral equation (8.312) in the single variable E.
Thus
the more difficult integral equation (8.296) involving two independent
sets of variables r and E is replaced by a system [Eqs. (8.302)] of integral
equations in r (for each E) and another [Eqs. (8.312)] in E. The solu
tion of the neutron-balance equation (8.296) is greatly simplified in this
way, especially since in many cases of interest the infinite series in
(8.312) is well represented by the first few (frequently the first) terms.
This formulation of the general space-energy problem
for
use
in the energy-group method,
as
will
is well adapted
be demonstrated
in subsequent
analyses.
f. Upper and Lower Approximations.
Various approximations have
been suggested for solving the neutron-balance equation (8.296) which
we have reduced to the set (8.302) and (8.312).
These involve the use
expressions for either of the two functions A{t,E) or
we apply the approximation to the
removal density A{t,E). Specifically, we assume that
of approximate
P(t;t',E).
A(i,E)
Thus we have set all A =
An(E)
0
Mr,E)
for n >
0.
=a
A(E)
Mr,E)
(8.314)
The corresponding expression
50G
REACTOR
ANALYSIS
_-
[CHAP.
for the flux takes the form
*(r')
If
we use the relation
A(t,E)
MET
Ap(E) Mt,E)
(8'315)
ME)
(8.314) in (8.312), we obtain the single integral
equation
Ao(E)
-J
MA) Jo["
A0(E') g(E,E') Aoo(E,E') dE'
Numerical calculations have shown that the

calculations
seen from
(8.316)*
this relation in reactor

of the critical mass.
This may be
yields an overestimate
the fact that in estimating
use of
the removal rate (effectively
the
neutron "absorption" rate) we have ignored the contributions of the

Thus the neutron production rate in the core
higher modes in (8.314).
and the resulting computation of the critical
underestimated,
has been
mass will be larger than that required to maintain the system at steady
For this reason we call (8.316) the upper approximation.
A second estimate of A o(E) may be obtained by assuming, further, that
is,
*o(r,)
^ 4">(r,E')
= *0(r)
(8.317)
that the fundamental modes of neutrons of all ener

If we use this assump
gies are in first approximation nearly the same.
tion in (8.311), we obtain
we assert
(8.318)
follows that (8.316) reduces to the form
Ao(E)
This relation
-j
Mo(.a)
J0
and
it
Aoo(E,E')
That
ME')
g{E,E') dE'
(8.319)*
is
is
it
is
called the lower approximation, since

believed to
an
underestimate of the critical mass.
The comparison of mass
yield
estimates based on these two approximations with known experimentally
determined values would indicate that this
indeed true.
Furthermore,
there
reason to believe that both approximations are not far removed
from the correct value.
The validity of these assertions
best deter
mined by comparison with experiment in each case.
is
is
Since both estimates appear to be fairly reliable, we will use the lower
approximation in all subsequent analyses in view of the fact that the
relation (8.319)
more easily applied to reactor calculations.
We con
is

state.
REFLECTED REACTORS
sec. 8.7]
507
tinue with the calculation on this basis and introduce the source function
(8.295)
into (8.319).
=
Ho{h) A{E)
vz(E)
')
+
This equation may also
dE'
zlC)(E')
be
[ 2*(E')h(E;E') A(E') dE'

WW)
Jo
written in terms of the flux.
'
Let
Since A{t,E) is directly related to the flux and has been defined by the
series (8.304), it is easy to show that
An(E)
ME) n(E)
(8.322)
We apply this relation to (8.320) and obtain

4>o(E) v<>(E)
= z{E)
f* ME') h(E;E')
ME') ME')
dE'
(') dE'
(8.323)*
i.e., we have set

(8.324)
fo~
where we have normalized the neutron production,
Two-group Model for Bare Homogeneous Reactor.

Throughout
of
of
the
Feynman-Welton
the present treatment
the remainder
method
g.
we adopt the energy-group description of the neutron-energy spectrum in

multivelocity reactor in place of the continuous spectrum previously
In practical situations wherein the Feynman-Welton model
considered.
a
is
most useful, reactors reflected by moderating materials, the two-group

is
the more easily and effectively applied.

formulation
The basic relations for the two-group model are developed by treating
a sequence of progressively
more general reactor problems.

We begin
with the bare homogeneous reactor. Since the energy-group relations
results, we summarize
are derived from the continuous-energy-spectrum
these equations below:
A(t,E)
S(i',E) =
Pb(t;i',E)
P(t;t',E)
[vz(E)
ME')
(8.325)
ME')
h{E;E')] dE'
(8.326)
the kernel for the bare reactor and comparable

defined by (8.292).
is
where
S(i',E) Pb(t;t',E) dr'
[" [

lr(E)
to
REACTOR ANALYSIS
508
[chap.
In
the two-group formulation the various functions which appear in

The index
the above equations take the special forms given below.
denotes fast neutrons, i = 2, slow.
i=l
number of removals per unit volume per unit time

=
~~
around space point r of neutrons in ith energy group
. .
. .
number of neutrons of energy group

ume per unit time around point r
born per unit vol-
probability that neutron of energy group t born at

P$(r;r') dr = r' is eventually removed as an t-group neutron in
2u
dr about r
average number of scattering collisions per unit path of
= travel by neutron of energy group 1 which reduce it to an
energy group 2 neutron
,g
,g ^2g\
(8.329)
(8.330)

Note that the group cross section 2i2 is in fact the " slo wing-down " cross
With these
section (analogous to 2*) from the fast group to the slow.
definitions Eqs. (8.325) and (8.326) may be written
Ai(r) = Jc P(r;r') Stf) dr'

Si(r) = j>Zi[-yiAi(r) + ytA2(i)]
with
S,(t)
(8.331)
(8.332)
vZl[yiA,(T) + y,A2(i)] + Zl2^(r)
denotes the number of neutrons of energy group

fission reaction and
where
z<
y<
Then, from (8.317),
WW
produced per
(8.334)
vfe
'r
As indicated previously we apply the lower approximation

First, we have from (8.314),
lation.
A<(r)
(8.333)
^W(r)
to this calcu
(8.335)
= fl,2'(r) = *o(r)
(8.336)
The application of these relations to the integral equation (8.302) yields,

for the fundamental modes [where ptf is defined as in Eq. (8.236)],
WW
If
the kernel
Ml*
nXr;r')
fc WW
dr1
(8.337)
P$(t;i') were known in detail, the solution of this equation
would yield the spatial distributions for the neutrons in group i. These
results in turn could be used in Eqs. (8.331) through (8.333), and a
simultaneous solution would give the criticality condition for the reactor.
As
a rule, the calculation
of the
P$ will
be somewhat
involved, like
REFLECTED REACTORS
BEC. 8.7]
509
These computations became even more compli

wise, that of the M(r).
It is possible, however, to
cated when dealing with reflected systems.
obtain reasonable approximations to these functions, in particular to the
and the eigenvalues (iff, by the application of the
spatial modes
diffusion-theory or infinite-medium
(7.254)].
transport equation
results [cf. Eq.
In the present treatment we consider in detail only the compu
The use of
tational procedure based on the diffusion-theory solutions.
the transport-theory results in the calculation of the eigenvalues is easily
shown and involves some minor changes in the procedure developed for
the diffusion theory.
The application of the diffusion theory to the calculation of the eigenfunctions and eigenvalues for the Feynman-Welton method requires a
restatement of the neutron-balance equation in terms of the removal
cross section 2^ and the neutron yield per removal ntf. It is obvious
that the appropriate form is [cf. equation preceding (8.242) and Eq.
(8.238)]

-DtV4p(t) +
2<= Mi20p(r)=
(8.338)
Mi20p(r)
(8.338)
In
where ^(r) is the fundamental mode for the flux in the ith group.
Eq.
(8.242)], the neutron balance would be
the kernel formulation [cf.
given by [with P$ = 2?K$]
m*)
fW fc m*1)
KSW)
dr'
(8.339)
It
follows directly from the results of Sec. 8.7c [cf. Eq. (8.263)] that the
eigenvalues p must satisfy the relation
=
(8-240)
where from (8.338)
VW(r) + BW(r)
Note that the mode index n = 0 has
Thus, as
to simplify the notation.
(8.337), we use the solution
= 0
(8.340)
been omitted from the

a first
approximation
buckling B{
to the Eq.
(space form) given by (8.340) and compute
the eigenvalues /4'1 from (8.240), keeping in mind that the B, are given
by the usual relations for the bare reactor (e.g., in the case of the sphere,
Bi
=
1.
B = t/R).
Application
to
Thermal Reactor.
A simple, direct illustration of the
use of the results obtained above is demonstrated in the analysis of the

For this
reactor in which all fissions are due to thermal neutrons alone.
example we make the following choice of group cross sections:
2>
Z</
= 0
=
2/
Zx
z2 = 0
2?>
= 0
2,2
5*
= 2'
+ 2/
(8.341)
ANALYSIS
REACTOR
510
[CHAP.
The corresponding expressions for the source functions given by (8.332)

and (8.333) are
Si(r)
= uy2A2{i)
4,(r)
St(r)
(8.342)
and by (8.331)
A,(r)
/cAi(r')
(8.343)
Pf{r;t')dT>
But, by (8.335) and (8.336),

=
A.(r)
so
(8.344)
that Eqs. (8.343) reduce to the form
Wo(r)
AWa{r)
^4?/cWr') /^(r;r') dr'

A? fc Mr') PgW) dr'
These integrals may be evaluated

relations are
by means of (8.337).
^
V
A (2)
(8.345)
The resulting
A (1)
(8-346)
The ratio of these two equations gives the criticality condition for the
jS}
reactor; thus,
= MiM2'
(8.347)*
The eigenvalues are computed from (8.240)

Mi)1'
DiB*
2+
D2B2
~^^>
(8.348)
if
where we have used the fact that the B< = B, a constant independent of
the energy dependence of the extrapolation dis
the neutron energy,
is
ignored.
If
the above relations are used in the criticality con
we obtain
[2i2
+ L>,2][2 + D2Bl\
(8.349)
tance
dition,
identical in form to the

It
interesting to note that this result
criticality condition for the bare reactor derived previously by the stand
is
is
ard two-group method of Sec. 8.4b [cf. Eqs. (8.155)], since (8.349) may
also be written
2S,2(1
We
have
notation,
+ L22)(1
+LW)
used here the identity 2^2) = 2> 4and L\ = D,/212 and LI = D2/Z.
(8.350)

A}fV0(r)
2/
2,
in the usual
REFLECTED REACTORS
sec. 8.7]
511
We develop next the general criticality
Application to Fast Reactor.

relation for the bare fast reactor.
2.
72
dr'j
A,(r') nl>(r;r') dr' +
A,(r') P$(r;t')
(8
= *,
jf
At(r)
Tti
fc
In this analysis we retain, as generally

different from zero, all the group constants for both the slow and fast
The general source functions (8.332) and (8.333) are used,
neutrons.
therefore, in (8.331). The neutron-balance conditions are then
As(r')
P(r;r')
ic
Ai"
dr'
[7iAi + 7SAi]
(8.352)
^[71A0
The first of these equations
7.Ai]+|^
gives
the simpler
between
relationship
and A0:
Al)
353
^,7Ag
"Zi7i
Mo"
And, the simultaneous solution of the pair (8.352) yields the criticality
=
2i[7.Mo2)
+ TW2<"]
"
relation
(8-354)*
2*72/4"
is
As in the case of the thermal reactor previously analyzed, an approximate

estimate of the eigenvalues may be obtained from the simple diffusion
straightforward and involves the com
The procedure
theory results.
putation of the B2 from the usual relations given in Table 5.3.
h. Two-group Model of Reactor with Moderating Reflector.
The
most general reactor problem so far considered has been the reflected
multivelocity reactor with energy-dependent cross sections but with no
slowing down in the reflector.
general method for treating this sys
tem, and other more elementary ones, has been discussed in some detail.
general
is
been mentioned that the extension of these

methods to the analysis of reactors with moderating reflectors
group model.
It has already
it
has been shown that, by introducing various approximations

Further,
into the expression for the removal density, the general results could be
reduced to
more tractable set of equations.
final simplification was
added in the case of the bare reactor with the formulation of the twoa

Ai =
Ap
P(r;r')
of (8.344) and (8.337) yields the set
^-
The application
Ax(r')
dr'j
Pg>(r;r') aY + y,
i.(r')
= wt,
jc
Ai(t)
[71
(8.351)
diffi
512
KEACTOH
ANALYSIS
[CHAP.
It is possible, however, to develop directly an approximate

treatment of this problem based on the two-group model. 1 For this pur
pose we draw from the relations derived for the bare reactor (8.351) and
cult task.
introduce
appropriate modifications in the definition of the kernel to

account for the presence of the reflector.
In addition we provide an
extra term in the slow-neutron balance equation which represents the
contributions to this group from the slowing-down reactions in the reflec
The calculation of this term is based on a direct physical interpre
tor.
tation of the neutron transport and slowing-down processes.
The analysis of the reactor with moderating reflector is carried out for
As in the bare-reactor calculation we
the case of the infinite reflector.
in
the core by the index i = 1 and the slow group
denote the fast group
=
In the reflector, the fast group is denoted by * = 3 and the
2.
by i
slow by
= 4, so as to be consistent
with the two-group
notation of
Sec. 8.4.
We begin with the neutron-balance
equations,
which we write in the

form
Ai(r)
A2(r)
(8.231)
fcS1(r')P1(i;T')di'
fc
S2(r') P2(r;r') dr' +
fc Stf) P(t^)
dr'
(8.232)
where the S.(r) are given by (8.332) and (8.333) and the kernels
are defined by
a neutron of energy group i born
at
number
of elastic collisions in both
aêr an^
dr =
(r-r'1
'
core and reflector, is eventually removed as a
group t neutron in dr about r
Pi(r;r')
probability that
P
It
should be noted that these functions correspond

The kernel P(r;r') is defined
(8.292).
probability that
P 12 (rr'f-r') drf
a fast neutron
to the
P(r?',E) of
born at r' in the
escaPes remova' m tne core, is slowed down

in the reflector, and finally diffuses back to the
core to be removed as a slow neutron in dr about r
COre
This
is the only new quantity which appears in the set (8.231) and
In determining this function it is convenient to introduce the
(8.232).
slow-neutron removal density in the reflector, At(r') and a new kernel
for slow neutrons, Q2(r';r), which we define:
Q2(r';r)
total track length of slow neutrons per unit time per

= unit volume around r' in the reflector produced by a
unit source of slow neutrons at r in the core
(8.357)
P. Feynman and T. A. Welton, "The Calculation of Critical Masses Including

Distribution of Neutron Energies," Los Alamos Report LA-524,
Ch. Ill, Jan. 21, 1947, declassified.
1 R.
the Effects of the
BEC.
REFLECTED
8.7]
REACTORS
513
Note that this function is symmetric in r and r7; i.e., Q2(r;r') is the track
length at r in the core produced by a unit source at r' in the reflector.
The removal density At(i') may be written in terms of Q2. Thus if
<S2(r) is the distribution of slow-neutron sources in the core, we can write
A4(0
=Q2(r';r)
Z< dt (8.358)
2<"= Z< fc 5,(r)
fc
The source distribution may

slow-neutron
eigenf unctions
5,(r) Q2(r';r) dt
be represented
#,2>(r) ;
S,(r)
(8.358)
by an infinite series of the
thus,
n <SW(r)
(8.359)

A(r')

It is convenient to
set Afir'); thus,
Z<<>
*(r) Qtf?)
dr
identify the individual terms of this expansion
A*W
Note that At(r')
S>2<
/c
tf<r) Q2(r';r) dr
as some
(8.360)
is then given by the sum
A4(rO
- J A<(r') - J
(8-361)
where the last definition introduces a set of dimensionless functions com

parable to the set
^"(r) defined for the core
definition (8.360) may
be
AJfVW)
[cf.
Eq. (8.359)].
With this
written
=
S<2,2 fc #(r) Q2(r';r) dr (8.362)

Q2(r';r)
fc #(r)
dr
(8.362)
These results may be used to obtain suitable expansions for the kernel
If
we define
Q2(r';r)
Gp^) #?(r)
Q2.
(8.363)
then the substitution of this expansion into (8.362), followed by the

application of the orthogonality properties of the ^!,2)(r)> yields the
following expression for the coefficients
",(r,)
Now,
Qj,2)(r')
Sgfew
(8"364)
neutron-conservation
requirements, the total number of
per unit time of slow neutrons in mode n in the reflector must
be precisely the number of these neutrons escaping per unit time from
from
removals
REACTOR ANALYSIS
514
[CHAP.
But, this is just the total number produced, multiplied by the

probability of escape 6>< [cf . Eq. (8.265)] :
the core.
<P
L A?(i)
M<2*
M<2)
dr
[
&
dr
Jc A'?(i)
Therefore,
^<41
A(r) dr
^4,(r) dr =
Tl
S<
^ j jf
- -il
"
#(r) dr
(8.365)
(8.366)
and S(B2). If this

which gives a relationship between the coefficients
result is used in (8.364), and thence in (8.363), Q2 may be written
Q,(r';r)
V [l-ilJ JC[W(r)dr)
L
Mn
/H^4,(r')*'
W(r'W(r)
(8.367)
In the analysis which follows it will suffice to use an approximation to

this series.
Now the functions ^"(r) to be used in the expansion of the
density in the reflector are nearly all the same shape; in the
limit of diffusion theory, these are in fact simple exponentials.1 Thus in
first approximation we represent the #,4,(r) by a single function ^(1)(r) :
^(4>(r)
= ty,(4>(r)
for all n
(8.368)
is some constant.
Then, we can factor out the quantities
which
on
these
depend
functions, and
(8.367)
where
~
R
*>(r)
- -L]
Q2(r';r)
|~1
**(r)
in
dr
The sums which appear in this expression are integrals over the core
volume (Vc) of
^(r')
*S?W
- r')
(8-369>
P,(r';r)
(8.370)
(r
and [analogously to (8.308)]

y
'
'
WHr)
itwCr/)
Wn
Mn
If
2?' JR^t)(r')dr'
Ubid.
Jc
P2(r";r) dr"]
Q2(r';r)=-
the integrations are carried out as indicated and the results applied to
the above approximation to Q2, we obtain
i \l

removal
(8.371)
REFLECTED REACTORS
sec. 8.7]
515
This expression may also be derived on the basis of simple physical

Since P(r";r) is the probability that a slow neutron born
arguments.
at r in the core is removed at r" in the core, the integral of P2 over all r"
gives the probability that a slow neutron born at r is removed somewhere
It follows, therefore, that one minus this quantity, the
in the core.
brackets in Eq. (8.371), is the probability that the neutron is removed
Moreover, the probability that a slow neu
somewhere in the reflector.
tron is removed at some point r' in the reflector is just ^(4)(r') divided
by its integral over the reflector volume. The product of this ratio and
the preceding probability are by definition 2<.4)Q2(r';r), thence, (8.371).
The kernel Pi2 may be given in terms of Q2. If <S4(r";r') is the distribu
tion of slow-neutron sources in the reflector produced by the slowing
down of fast neutrons arriving from the core, then by the definition of
P12 given
in (8.356),
= 2<

P12(r;r')
fRS<(j"?)
Q2(r;r")
di"
where we have employed the symmetry property of Q2 noted in the

In order to compute St we introduce
discussion following Eq. (8.357).
a fast-neutron kernel analogous to Q2:
total track length of fast neutrons per unit time per
Qi(r";r') = unit volume around r" in the reflector produced by
unit source of fast neutrons at r' in the core
(8.372)
If 234 is the slowing-down cross section in the reflector, then it follows

that 2;uQi(r";r') is the number of fast neutrons arriving in the slow group
per unit time per unit volume about r" in the reflector because of the
presence of the unit source of fast neutrons at r' in the core. We have,
then,
Pi2(r;r')
2342<2'
jR Qi(r";r') Q2(r;r") dr"
(8.373)
The kernels Qi and Q2 can be approximated by the form (8.371). By

following an analogous line of reasoning to that leading to the result
(8.371), we obtain
2 jR
Q2(r;r")
2<4'
where the function
jR
^<3)(r")
dr"
dr"
Ptfrtfdt]
P,(r';r)dr']
(8.374)
(8.375)
represents the distributions of the fast-removal

This function
reactions (and therefore also the flux) in the reflector.
is the fast-neutron counterpart to the ^(r) developed previously for
^<3)(r)
REACTOR ANALYSIS
516
[CHAP.
JL
P2(r';r) dr']
(g
f P,(r;r')
12
\l
fl
2i4<P
Pi(r^)
di]
yields
The sub
the slow neutrons and may be derived by similar arguments.
for
and
into
Eq.
of
above
stitution
the
approximations
Qi
Q2
(8.373)
376)
where we have called
Mo
(P42
J 4sl
WVo
2<<>
7*
|fl^(r)
W)
^(4)(r") dr"
dr
(8.377)
yB^(r)dr
Finally, we introduce the Feynman-Welton lower approximation for the

From (8.370) and
P which consists of setting all vtf =
kernels
(8.369) we obtain
P^rjr')
=^ (r
(TT>>
PP i2(r,r
Z34g42[l
(1M)1 _
=
2VC
Vc
^.î-^J^^,
with
It
(8.378)
,0,7CA
(-379)
(8.380)
is
is
interesting to note that the result (8.379)

independent of both
and r'. Thus we find that in this approximation the probability of a
fast neutron from the core diffusing out into the reflector, slowing down
slow
there, and finally returning to the core to be absorbed there as
a
the same for all combinations of origin and terminus.

neutron
In order to obtain the criticality condition for this system we observe
first that in analogy to (8.301) we can write
A*(r) =
m(<)
fc
is
By using the lower approximation

*o(r) =
m'o
A,(r') P,(r^) dt'
(8.381)
(8.344), this equation reduces to
fc
Mi')
Pi(rjr') di'
(8.236)
The substitution of these forms into Eq. (8.231), along with the source
Si from (8.332), yields

Mn
The corresponding
=
Mo
tô1' + "Mi2)]
expression for (8.232)
is

and substitution into (8.376) yields
- r')
obtained
(8.382)
by using (8.333)
SEC.
REFLECTED
8.7]
for
S2 and (8.379) for

may be written:
^'o2'
Pn
REACTORS
517
over all r.
and then integrating
+ iw
[7î,l>
+ 7iAp] +
fg&
The result
(8.383)
and Atf> which is identical

Equation (8.382) gives a relation between
to (8.353).
The criticality condition is obtained by eliminating Aj," and
The result is found to be
by means of (8.382) and (8.383).
The only quantities
in this expression which are as yet undetermined

are the eigenvalues pj,1' and Mo2) and the functions ^(3)(r) and ^(4)(r)
which appear through the quantity w. As in the case of the bare-

reactor calculation, the determination of these quantities is a relatively

easy matter if one is willing to accept the one-velocity diffusion theory
or the infinite-medium transport equation results for the spatial distribur
It is expected that for those

tions and the characteristic equations.
situations in which this method will be most useful, the one-velocity
diffusion-theory solutions should be entirely adequate.
We proceed on
this basis.
As an illustrative example consider a spherical reactor of core radius a
In the case of the one-velocity diffusionand an infinite reflector.
theory model we found that the appropriate characteristic equation was
the transcendental relation (8.11).
For the reactor with an infinite
reflector, this relation reduces to
Ba cot Ba
-~
(KRa
1)
(8.385)
This equation is to
be solved simultaneously with the B2 relations (8.5).

the notation of the present method, the B2 equations take the form
Thus in computing the eigenvalues p, we determine first
(8.240).
Bi and B2 from the relations
In
Bid
COt
Bid
B2a cot B2a
=1
yr-
Dn-
(/c3a
1)
(8-241)
(x4a
1)
Di and Ds denote the fast-group diffusion coefficients in the core

reflector, respectively, and Z)2 and D4 the slow-group diffusion
Also,
coefficients in core and reflector.
where
and
(8.386)
REACTOR ANALYSIS
518
[CHAP.
Bi
With the
and Bt values computed from (8.241) Eq. (8.240) may then

be used to obtain
and ntfK All this assumes, of course, that the
various D, and k, are independent of the /i'0rt. If they are not, then a
graphical solution may be applied.

The flux distributions ^(3) and ^(4) in the reflector are easily obtained
from the diffusion theory.
For nonmultiplying media these are found
to be
*(4)(r) =
Note that
these would
(8.387)
proportionality constant is not required in either

be eliminated
in any event in the calculation
case since
of
It
(P42.
can be shown that the substitution of (8.387) into the integrals of (8.377)
^(,V(4> dr
fR
fR
<
dr
fR
<
dr
(4r)2
j"
I.'
yields
g-r(,+ ,)
we use also
Vc
rer dr
k)(1
+ o3)(l +
am)
(8.388)
then for the spherical reactor with infinite
= îra',
and
Feynman-Welton
to
m'o2'
Thermal Reactor
reflector.
symmetric system
cross sections listed in Table 8.8.
with
OK3)(l
OK4)
(8.389)*
Moderating Reflector.
An
For our analytical model we select

with infinite reflector and the group
Two-group Cross Sections for Feynman-Welton

Thermal Reactor
Core
z2<y"
2<
- 2
2/
- 2
2<4
i =
2<" 2<
2-
2(3,
H =
Model of
Reflector
With these data, the criticality condition
Fast
0 0
Energy group
Thermal
<c4)(l
example of the application of the two-group

method
the criticality calculation for the thermal
a spherically
8.8
useful
reactor with moderating
Table
(C3
is
Application
interesting
48
reflector,
i.

4tt(k3
If
Jj
re' dr
fa
2,
(8.384) takes the form

(8.390)*
sec.
REFLECTED REACTORS
8.7]
An alternate expression may
be obtained by
519
eliminating the eigenvalues
with the aid of the relations

Mb11
which follow from (8.240).

=
L\B\
Here,
M<*
L\B\
= Z>,/2 and
L\
(8.391)
=
L\
Z),/2.
Then
+L\B\){\ + L\B\)
w(l + L\B\){\ + L\B\)]
Zj(l
2,(1 +
K*-6VZ)
It is of interest to note that if 2i2, Dx, and D2 are independent of the

fuel concentration (at least in first approximation) then it is possible to
obtain an explicit expression for the critical concentration directly from
the criticality
relation.
the B, given
Under these circumstances,
by
(8.241) are also independent of the fuel concentration and the values
computed from these equations may be substituted immediately into the
criticality condition.
Now, for the present system [from Eq. (8.380)],
"
[1-i']^

It
(8,393)
(rrTbi)(P42
is convenient to write also
32<<>(k3
4)(1
+ )
(8.394)
Here C denotes all the various factors in (8.389) which are independent
of the fuel concentration.
If we use these two expressions in (8.392) and
solve for the critical fuel concentration
'
vo,[{\ +
NF, we obtain
CD2L\B\Bl)/(l
+ L\B\)\
<r'a2F)
^MWv*
The symbol 2^20) denotes the thermal absorption cross sections of the
nonfuel components of the core and 2'aF\ the thermal absorption cross
section of the fuel.
Thus the present method permits a direct solution
for Nf and avoids entirely the trial-and-error computations required for
the standard two-group method of Sec. 8.4.
It should be evident from
the procedure outlined above that the computational effort involved in
the Feynman-Welton two-group model is only a small fraction of that
required in the standard method.
This fact may be more clearly demon
strated by considering a numerical example.
As an example, let us compute the critical mass for the nuclear package
power plant analyzed in Sec. 8.5.
Since the group constants selected
for the present method are defined in the exact manner as those of the
method, the data listed in Table 8.5 are directly applicable to
The only information we lack is the values for the
and the constant C. The direct substitution of the appropriate
standard
this calculation.
Bi
REACTOR ANALYSIS
520
numerical data into the relation
B\
= 0.007444
cm-J
B\
[CHAP.
(8.394) and the set (8.241) yields
= 0.007878 cm-2
= 61.9 cm
(8.396)
If
we apply these results to (8.395) it is not difficult to show that the

volume fraction of fuel corresponding to the critical concentration is
This result is to be compared to the value vF = 0.006425
vF = 0.006398.
obtained in Sec. 8.5.
The present computation yields a critical mass
of 9.512 kg, which differs from the previous computation by less than
1 per cent.
Thus the Feynman-Welton calculation allows not only an
extremely quick computation, but one that apparently agrees very well
with the results from the standard

water-moderated
8.8
two-group
model in the case of a
system.
Multigroup Treatment of Reflected Reactors
Integration over the Lethargy Intervals. In preceding sections

the treatment of the neutron population of reflected reactors based on
a.
one group of diffusing
neutrons (thermal) or on two groups of diffusing

thermal
and one fast) has been examined in some detail.
(one
is natural to examine the possibility of a more general treatment
neutrons
It
involving many groups of diffusing neutrons such that the one- and twoHow
group models would be special cases of the more general theory.
ever, it is to be noted that the one- and two-group models are really
based on fairly crude assumptions and are expected to be applicable to
only a limited class of reactors; even so, the use of either model is justified
only by previous successful application in reactor calculations.
Much
more confidence in a many-group model would result if it could be derived
from a more basic treatment of neutron problems.
At least one such
approach has been sufficiently successful to justify detailed consideration;
this approach, known as the multigroup method, is based on the age-
Of course, using the age-diffusion equations as a

starting point for obtaining multigroup equations limits the application
of the results to those problems to which the original equations are appli
cable; however, the limitations of the age-diffusion model can be clearly
brought out by the underlying approximations needed to derive them
from the Boltzmann equation (as discussed in Sec. 7.3) so that some
understanding of the range of applicability of the multigroup method is
Further, since in all reactors the neutrons must go through a
obtained.
slowing-down process, it would be expected that any method properly
diffusion
equations.
on the age-diffusion equations would yield better results than the

or
two-group methods. Attempts have been made to base multionegroup calculations on the original Boltzmann equation, but such methods
are not as yet sufficiently well developed to justify detailed presentation
here.
On the other hand, the multigroup method based on the age
based
sec. 8.8]
REFLECTED
REACTORS
521
diffusion equation has been widely used in reactor calculations

given satisfactory results.
In obtaining the multigroup equations' we consider
and has
a multiregion
reac
tor to which the following
set of slowing-down diffusion equations is to

be applied; each region is taken to be homogeneous and uniform, and the
equations apply to each region separately with appropriate constants;
proper boundary conditions (to be discussed later) are to be applied
throughout the system:
-D()V>(r,w)
+ Z.(u)
4>(t,u)
- du q(T,u)
S(r,u)
q(j,u) = f2,(u) (i,u)

-Z)uV20th(r) + 2i"0[b(r) = qfau*)
A < u <
0 <
wlh
(a)
< U*
(b)
u = utk
(6.54)
(c)
with
S(r,u)

This
[|o"'h
2,(u') *(r,u') du' +
du
2}kMr)]
(6.90)
set of equations
has already been applied (in Sec. 6.3) to a singleregion reactor, and it was possible to obtain a solution there by the
method of separation of space and lethargy variables; as has been dis
cussed previously (Sec. 8.1b) this is not possible, in general, in the
case.
Thus we are led to consider techniques of numerical
of
the
The method under consideration
integration
equations above.
here is intended to be a natural one for calculations using (6.54) and
(6.90) in the multiregion case performed on large-capacity fast computing
multiregion
machines.
the zero of lethargy such that there are (for all practical
purposes) no neutrons in the reactor whose energies are greater than
that energy corresponding to lethargy zero (e.g., 10 Mev). Consider
the lethargy range from u = 0 to u = uth as divided up into AT intervals
Consider
of (generally) unequal sizes:
AUi =
Mo
"v
i
= 0
=
= 1, 2,
(8.397)
intervals may be chosen for convenience, according to

lethargy scale and according to size.
the
position along
Now in a given region integrate Eq. (6.54a) over the ith lethargy
These lethargy
interval; thus
D(u) V20(r,u) du +
=
j"
\
2(u)
4>(t,u)
du
q(r,u) du +
'
S(t,u) du
(8.398)
original form of this method (not exactly the one given here) was developed by
the Physics Division of KAPL [see R. Ehrlich and H. Hurwitz, Jr., Nucleonics, 12
. i
(2), 23 (1954)].
1 The
REACTOR ANALYSIS
522
[CHAP.
In order to progress from Eq. (8.398) when confronted with the unknown
variation of the neutron cross sections and the flux with lethargy, the
following definitions are made :
Dt
y~
fu'
*.(r) =
(r,u)
I
Jfi-i
'
du
(6)
(r,u)
S(u)
'
4>{t,u)
(8.399)
du
du
S/(m)
J
JUi-l
u =
;
du
4>(t,u)
{r,u) du
J/ Ui-l
- 9(r,u.)
over the source term;
i(u) du
[
^
S(t,u) du
Consider the integration
gfou.-,)
+
if
+ 2~AU,:(r)
/"
-Âw.V^r)
f"t
The definition of 2/, is intended for later use in connection with the treat
ment of the term <S(r,u).
If we use Eq. (8.399) in (8.398) and perform
the integration in the final term on the right-hand side of (8.398), we
obtain
j~t
S(r,u) du
we apply (6.90),
*(r,u) du +
k-i
(8.400)
2}"th(r)]
(8.401)
^note
that from (6.88),
SAui
itAu* =
lj,
Now define
AT
j(u) du
(8.402)
Then, applying (8.399rf), we have
The thermal fissions have
IJZ,
2?
been
= xj.AUi
2/.**(r)Au*
(8.403)
formally included by defining
^7x(t)
0,h(r)
AwlV+1
S(t,u) du
JV+l
J*i

(a)
4>(i,u) du
/
V
(r,u)
D(u)
du
(8.404)
sec.
REFLECTED REACTORS
8.8]
523
The application of (8.403) to (8.400) yields
-TMw.V^r)
+ 2ÂM,^(r)
=
g(r,u,_i)
q(T,Ui)
be regarded
K&Atti
hi
ZA4>*(r)Aw*
as a restatement
of Eqs.
(8.405)*
(6.54a),
(6.54c), and (6.90), since the thermal equation can be formally included
of Eq. (8.404) and the following limits:
using the definitions
g(r,w.v+i)
= 0
(8.406)
Thus the index i in Eq. (8.405) takes on the values 1, 2, 3, . . . , N,

N + 1. Note that no approximations have been made in obtaining Eq.
At this point, however, some approxi
(8.405) from (6.54) and (6.90).
mations must be introduced to make Eq. (8.405) tractable.
b. The Simultaneous
Set of Multigroup Equations.
As a digression
from the main purpose of this section, which is to demonstrate the tech
nique for applying the age-diffusion equations to certain classes of reac
tors using fast computing machines, consider the use of Eq. (8.405) to
obtain a generalization of the one- and two-group methods.
The diffi
culty in Eq. (8.405) from this viewpoint is the presence of the terms in
the slowing-down density at the lethargy limits of the ith interval; it
would be desirable to have these expressed in terms of the average fluxes
in neighboring intervals.
As an example of such a relationship similar
to that used in the two-group method, assume that a removal cross sec
tion for the ith interval (or group) can be defined such that
=
9(r,M.)
Z**(r)Aiii
(8.407)
Then Eq. (8.405) becomes
-AV^,(r)
N+l
+ (S + Z,)*(r)
with
<fc(r)
2ft._,*.-i(r) +
ifcAWi
z7**(r)
(8.408)
(8.409)
<fc(r)Au (8.409)
is
it
set of N
equations for N
unknowns, the <fc(r)
then,
since the equations are homogeneous
to be expected that there will
determinantal equation [cf.
be a condition for solubility in the form of
This
is,

lM+i = 0
Eq. (8.188) in the two-group case] which would serve as a criticality con
dition. The average cross sections needed for Eq. (8.408) present no
more problem than in any other scheme based on Eq. (8.405) and accom
panying definitions; since these calculations will be discussed later, let us
assume here that they are known.
However, the removal cross sections
ANALYSIS
REACTOR
524
[CHAP.
quite arbitrarily in Eq. (8.407) and, as in the two-group

method, can be chosen to give good results in various cases.
As an example of the use of Eq. (8.408), consider the case of a reactor
are introduced
medium for which the materials and lethargy groups are so chosen that
fissions are caused only by thermal neutrons and neutrons from fission
appear only in the highest lethargy group, thus
jiAwi =1
2,+1 =
For this
case, Eqs.
Xf
it
= 0
for
k >
2A
= 0
for
k <
N+
(8.408) become [assuming >0(r) =
-ThV'frii)
-DlV2Mr)
(2^+
+ (2^+ 2,)2(r)
-IT^Mi)
2,)*i(r)
(2a*
-D^V^wii)
2Rk)<fc(r)
2.+1^+1(r)
(8.410)
1
0]
*Z>h.v+i(r)
= 2,<Ai(r)
(8.411)
= 2i_,<fc_l(r)
=
2fi>iV(r)
(for the thermal flux) follows from the fact that by

If Eqs.
0. (8.411)
Eqs. (8.406) and (8.407) 2,<JV = 0.=If Eqs.
are applied
to a bareare applied to a bare
(8.411)
reactor, a solution may be obtained as usual by assuming that the spatial
dependence of the flux is the same for all groups; thus let
Mr)
thus
the
Eqs.
+ B*F(i)
= 0
(8.412)
for bare reactors of vari

and the B2 are the corresponding eigenvalues.
With
(8.412), Eqs. (8.411) become N + 1 homogeneous equations for

1 coefficients
>*; the condition for solubility is seen to be
N +
(with
with
<twith V2F(r) + B*F(i) = 0V2F(r)

(8.412)
F(t) are the usual eigenfunctions
ous geometries
the
2ih =
2^, A,,
D^[,
and L*. m
DJX*)
1
+ LlB>)->
(l
"? (1
[[]
(8-413)
we compare this result with the accurate equation (6.78) for a bare
= 1), we find that the
reactor in the age-diffusion approximation (with
removal cross sections should be chosen so that
If
[no+fe+tF)]"1
i-1
be carried
step further
(8-4i4)
if
The comparison may

The last equation
the ratios 2,/2, and
REFLECTED REACTORS
sec. 8.8]
525
are all small compared with unity; then
[nO+S+^r~[n(i+)n(i+^)r
i
i=1
If
JV
êxp(-2fc)exp(-*22S)
~
<
Thus the removal cross sections might well
Vs.
_-/Z*
i-l
va
Z/2.
i-l
be chosen to be consistent
with
(u,h S.(tO du
&,{u)
Jo
fu"D(u)du _
jo
*2,(u)
V / xA
Z/V?2<A
=
YTjjT
Z/VV<
i=l
where we have used the definitions

D(tt)
C?M
(8.417)
Further, if the lethargy intervals are sufficiently small and the cross sec
tions sufficiently slowly varying over the intervals that an average over a
product or a quotient of cross sections is approximately equal to the
product or quotient of the separate averages, both statements of Eq.
(8.416) are consistent with the choice for each interval:
2*
(8.418)
where (2(),- is an appropriate average of 2,(u) over the ith lethargy

It is not intended here that Eq. (8.418) should be regarded as
the best method for choosing the removal cross section for any actual
calculation.
In the actual case it may be an advantage to have the
interval.
removal cross sections as adjustable parameters (within limits) so that

as experience is gained various classes of reactors might be subject to
fairly accurate calculation.
Note that Eq. (8.418), along with the basic
assumption of Eq. (8.407), leads to the result
q(T,Ui)
pjA(r)
(8.419)
which seems to be consistent with Eq. (6.546), at least for very small
intervals Aw,; however, Eq. (8.419) is by no means a necessary conse
quence of (6.546). Indeed, the assumption of Eq. (8.407) is only an
526
REACTOR
example of the many
possible
ANALYSIS
relationships
[CHAP.
which
may be used.
As
example, which will not be considered further,

any lethargy might be related to the flux in
at
density
another more complicated

the slowing-down
all lethargy intervals at lower lethargies by
(8.420)
where the Aki may be chosen for consistency with an appropriate slowingdown kernel.
It is even possible that a set of Aki might be so chosen
that the resulting multigroup equations gave better representation of the
slowing-down process than the original age-diffusion theory which basi
cally assumes a continuous slowing-down process.

Let us now consider the application of the set of Eqs. (8.408) to a
reflected reactor, assuming that all cross sections and constants have been
chosen.
For a two-region reactor, for each lethargy interval there are
two sets of differential equations, one for the core and one for the reflector.
An appropriate set of boundary conditions must be chosen, for example,
analogously to that for the one-velocity reactor model; thus if #C)(r) and
^'(r)
are the core and reflector fluxes, respectively, we might take the set
(1)
(2)
(3)
(4)
<#(r) is nonsingular
=
*f>(R)
and nonnegative
#fi)(R)
-Dfv^'(R)
#>(R') = 0
-Z)v#*>(R)
(8.421)
where R defines the core-reflector interface, R' the extrapolated outer

boundary of the reflector, and D[C) and Z>{S) are the average values of the
diffusion coefficients in core and reflector, respectively, as defined by
(8.3996).
One of the difficulties
in using the set (8.408) is that they are simul

For any number of groups larger than
taneous equations in the i(r).
three, the equations become so cumbersome to handle that recourse is
usually had to numerical integration in combination with iteration tech
Thus
However, this usually implies large computing machines.
niques.
attention is focused throughout the remainder of this section on a direct

approach to such a numerical integration-iteration solution of Eqs. (8.405)
which is not dependent on the restrictive assumption of Eq. (8.407).
The analytic solutions to Eqs. (8.408) are so involved for any large
number of groups and of such restricted use that a detailed discussion
of these methods is not justified; however, brief mention will be made
of two such methods.
1. The Harmonics Method.
set of normalized
functions,
The functions
<,(r)
which are orthogonal
ing the core and the reflector;
may be expanded in a
over the range includ
then substitution into Eqs.
(8.408) and
sec. 8.8]
REFLECTED REACTORS
527
application of the orthogonality property give linear simultaneous equa

tions in the coefficients of the expansion which may be solved algebrai
Along with the matching conditions at the core-reflector interface,
cally.
these results will then yield a criticality condition.
2. The Method of Characteristic Functions.
This is the generalization
of the method used above for the two-group calculation.
For purposes
of simplification of the presentation we will confine our attention to the
system described by Eqs. (8.411), a special case of Eqs. (8.408).
The
flux in each group, in each region, is expressed as a linear combination
of functions
appropriate
for the particular geometry:
the functions Z and U are given in Table 8.3 for a few geometries.
The
parameters /i* are obtained by assuming that the core fluxes are propor
tional to a space function, as in (8.412). Then the solutions for B2 are
(>
L\
where
'
fe)
found from
- g(i
fum
LI
and
Ri
(8.424)
â
The
(S-
Zb<)r
x\
Note that the solutions to (8.423) for B* will, in general,

The xk of Eq. (8.4226) are given by
be complex.
(8.425)
of Eq. (8.4226) are in part determined by requiring that the fluxes

set of equations similar to those of Eq. (8.411) but with source
terms appropriate to the reflector; e.g.,
the reflector contains no fission
able material, all the equations (8.411) may be taken over and applied to
aik
if
satisfy
the reflector (with the proper constants) except that there would be
zero source term on the right-hand side of the equation for group one.
In this special case the solutions for the reflector would be of the form:
0i(r)
= a07(xi,r)
*.(r)
aiU(xhT)

N+l
(4^F)
\*2
*1/
-UM,
etc.
(fU26)
REACTOR ANALYSIS
528
[CHAP.
Equation (8.422a) contains (N + l)2 unknown coefficients; substitution

of (8.422a) into (8.411) yields relationships among these (N + l)2 coef
ficients such that there remain only (N + I) independent coefficients.
Along with the (N + 1) coefficients from the reflector fluxes, this gives
2(N +
unknown
The 2(N +
coefficients.
homogeneous equations
obtained by matching flux and current in each group at the core-reflector
interface allow for a complete solution except for a scale factor in the
The condition for solubility of this set of 2(N + 1) equa
flux, as usual.
1)
1)
tions is of the form
A = 0
(8.427)
where A, a 2(N + l)-rowed determinant, is the criticality condition for

the reactor system.
This brief outline is sufficient to indicate the difficulties inherent in
analytic solutions to the multigroup equations.
Two group calculations,
but can be systematized so that para
metric studies of a series of reactors is feasible by hand calculations.
Three-group calculations can be and are done by hand, but beyond this
the analytic methods become so cumbersome that it is advantageous to
as discussed above, are laborious
directly to numerical computations.

Multigroup Equations Adapted for Numerical Computation.
We
return to the main interest of this section by considering Eq. (8.405),
=
which is now written in the form
. N, N
1):
= 9(r,M,_,)
.
.
2,
- 9(r,uf)
-AAM<V2^(r) + ZÂu^r)
1,
(t*
c.
xi,Aw,F(r)
(8.428)
is
Here the total number of fissions per unit volume per unit time caused
by neutrons of all energies at space point has been designated by F(i),
This
replacing the summation over lethargy groups of Eq. (8.405).
done for convenience in treating the multigroup equations by the iterative

From the beginning, in the present method, the multigroup
set of simultaneous equations in the
equations are not regarded as
a
technique.
set of equations which describe the flow of neu

neutron fluxes, but as
given initial distribution in
trons downward in energy, starting from
Thus, for
space and energy.
given source distribution F(t), the Eqs.
is
it
Then,
(8.428) give the complete history of one generation of neutrons.
necessary to apply the condition that the spatial
for the reactor case,
distribution of fissions calculated from the results of applying Eq. (8.428)
must be the same as the source distribution with which the generation
is
was started.
In reactor calculation, then, certain source distribution
which we shall call F0(i)
chosen (somewhat arbitrarily) to start the
calculation supposing that all the constants in all regions and groups are
This
known, then the spatial distribution of fissions may be calculated.
as
distribution
be
called
and
used
source
distribution may then
/*\(r)
This process may be repeated until the calcu
to start new generation.
a

go
SEC.
REFLECTED REACTORS
8.8]
529
lated distribution of fissions agrees with the starting source distribution

This iterative tech
for that generation within the required accuracy.
nique makes good use of a general property of fast computing machines,
namely, that they are most efficiently used by a calculation which requires
a repetitive operation.
With the source distribution chosen, actual calculation from Eqs.
(8.428) depends on a choice for the relationship between the slowingdown densities at the lethargy limits of the group and the average flux or
Either the average flux or
average slowing-down density in the group.
the average slowing-down density may be used, since by virtue of Eq.
In the multi(6.546), Eqs. (8.428) may be written in terms of either.

is devel
form
of
coupling
equation
the
group formulation the appropriate
oped from #,(r) and g,(r), where <k(r) is given by (8.399a), and
2((u)
(r,u)
du
(r,u)
Ui-l
and apply (8.399a) for the denominator,
(r)
({2^
Some of the choices for this relationship

attention are given below:
q<(r)
(r)
(8.431)
which have received considerable
(a)
tt(i,Ui)
+ +
o(r,M,) = 2((w,)^(r)
<fc(r)
we call
(8.430)
du
/*g(r,uO
,(r) = i[q(i,Ui)
(8.429)
2((m) <p(x,u) du
J/
(iZ().
(6)
?(!,,_,))
(c)
(i,Ui-i)}
with
q(T,Ui)
In Eq. (8.432a) the quantity
= 2,(w.)
(r,u,)
(8-432)
(d)
is
it
fm
an adj ustable parameter to be deter

mined in part by iteration of the group calculation under certain criteria
for selection of
value of
and in part by experience in calculations to
In this sense
give good results.
somewhat resembles the choice of
Eq. (8.407) discussed above. Method (a) has been carefully tried at
is
KAPL, and some discussion

Methods
given in the reference report.1
with adjustable parameters such as
in (8.432a) are particularly useful
in calculations using a fairly small number of intervals for the range of
certain amount of
energy above thermal; however, they all involve
1

The last equality follows by the application of Eq. (6.546).
If
q(t,u) du =
j"'
3<(r)
Ibid.
530
REACTOR
ANALYSIS
[CHAP.
trial and error, and for fast computing
machines, it is more convenient

to have a direct connecting equation such as (6), (c), or (d) of (8.432).
It has been demonstrated that at least (c) and (d) of (8.432) can give
For the
good .results for a finely divided lethargy range below thermal.
present discussion we will assume (8.432d), a method which has been
well tested in multigroup calculations at ORNL (especially).
Equation (8.432d) may be used in Eq. (8.428) to eliminate the slowingdown density out of the ith group, g(r,Mi) ; thus
- A-A^V^r)
[IÂw,-
+ 2(2,)u,]&(r)
1
(^rj (r>u<-i)
+ "iAtaF(T)
(8.433)

Here ({2,)Ui designates the value of the quantity 2( at lethargy u,-; as

discussed in an earlier chapter (see Sees. 4.5b and c), since it is really the
quantity 2, which is to be calculated at each lethargy for a homogeneous
mixture of materials, it is not proper to choose one particular value of {
for the entire lethargy range.
We may now describe in general terms the procedure for the calcu
lation based on Eq. (8.433). The lethargy range is divided up into
intervals as described above; the constants required in Eq. (8.433) are
A starting source dis
calculated for each interval and for each region.
is
this
be
called
Then
chosen;
suppose that lethargy
tribution
let
F0(i).
Mo is so chosen that
= 0
q(T,u0)
for all regions of the reactor
begins with the first
The calculation
lethargy
(8.434)
A differential
group.
equation of the form
-D'1'"Aw1VJ*r(r)
[S'.I'Am,
+ 2({2()l^'(r)
^"'Au^to
(8.435)
is written for each region of the reactor, where the superscript (n) desig
Proper boundary conditions are chosen to
nates the region considered.
connect the 0in)(r) and their derivatives at interfaces between regions,
and the set of space-wise differential equations is solved numerically by

With ^"'(r) known, the slowing-down
methods to be described later.
density out of the first group may be calculated; thus, from (8.432(f)
qM(r,Ul)
(SBjfffWfr.uO
= (Z,)i?[2ft-(r)
0]
(8.436)
Then this slowingis taken to be zero from (8.432d).

may be used as a source term for the equation for the
since the flux at

down density
second group:
-D2">AM2V2&">(r)
u0
[2>Au2
=
[l
2({2(r)
jjjH!]
8(">(r,0
xS^Fnr)
(8.437)
sec. 8.8]
REFLECTED
REACTORS
531
The 0"(r) are then found for each region, again by numerical integration
The slowing-down den
and application of proper boundary conditions.
sity from the second group may be calculated from
(pyS^r.ud
g(">(r,) =
= (*2,)S>
|W>(r)
^ff]
(8-438)
The process thus proceeds from group to group until the (N + l)st group,
For the thermal group Eq. (8.433)
i.e., the thermal group, is reached.
does not really apply and we must return to Eq. (8.428), along with
Eqs. (8.404) and (8.406), to write
-S&iV'fl&xW
with qM(r,un) to
g(r,uw)

As
be calculated
(*2()(r,Ujv)
(8.439)
from
= (2,)
a result of this calculation
qw(r,uN)
2^'(r)
- ^g=^j
(8.440)
the average fluxes in each group in each
region, 4>^(t), are known at every space point [the additional subscript
zero indicates that these fluxes are derived from the assumed source dis
tribution F{,n)(r)].
The comparison of the derived source distribution
with the original one may now
be made; therefore compare
N+l
i
- 1
2>#;>(r)Au,~n")(r)
and, if certain standards of comparison have been chosen, it may be

decided whether the original source distribution was a sufficiently good
choice.
This is con
However, there exists a more valuable criterion.
nected with the calculation of the multiplication constant for the system,
which should be one of the results available from the computation.
Sup
pose the following normalization had been chosen for F0B,(r):
y+i
v
L4
j.Aw,
f
J reactor nn)(r)
dr =
di
J[reactor F^(i)
(8.441)
where the appropriate form for F'0n)(r) is to be used in each region when
performing the integration.
Equation (8.441) states that one neutron is
introduced into the reactor to start the generation.
At the start of the
next generation, we have n0 neutrons, given by
N+l
n0 =
y J[reactor
2V>W>(t)Atti
11
0
dr
(8.442)
>= 1
By definition,
the multiplication constant of the system is the ratio of
REACTOR ANALYSIS
532
[CHAP.
fco
the number of neutrons in one generation to the number in the preceding

generation; therefore,
= n0
(8.443)
is
is
put on the multiplication constant to indicate

the estimate of the multiplication constant obtained from the
Here the subscript zero
that this
original source distribution F^(i). Of course, the number of

neutrons in
second generation will, in general, depend on the spatial
distribution in the original generation.
The true multiplication constant
for reactor system will be obtained from that source distribution which
repeats itself in the next generation.
The calculation described above
for the reactor fluxes may now be repeated, choosing
tribution ^"'(r) and normalization as follows:
use of the
new source dis
As
result of this reactor calculation, the fluxes $"'(r)
'
(8.445)
dr =
(8.444)
reactor
S^'^'WAm,-
t-i
FP(i)
1
J[
FV>(r)
AT-+-1
are obtained,
fc,
w+i
Wl
t-l
/reactor
(8.446)
W*< *
(8-44?)
be repeated until the values for the multiplication

in succeeding iterations coincide to the desired accuracy.
In
found that the multiplication constant has converged upon
general,
is
constant
it
This process may
final value before that stage in the calculation

distribution has converged acceptably.
at which the source
the direct lethargy average
JUi-l
Except in so far as the spatial dependence
2(u)
h~rir.
JUi-i
du
(8-448)
du
of the average cross section
is
2(u)(r,u) du
s hhî
4>M(r,u) du
["'
%H)
and most often used
is
is
it
is
is
is
As shown in the treatment of the multigroup equations using the defi

no approximation involved in the use of
nitions of Eq. (8.399), there
average cross sections for a group so long as the flux-weighted averages
Of course at the outset of
of Eq. (8.399) are used.1
calculation the
not known; thus, unless some estimate
lethargy dependence of the flux
of the behavior of the flux
available from previous calculations on simi
necessary to begin the calculation with cross sections
lar systems,
The simplest
obtained by some arbitrarily chosen averaging process.

and the multiplication constant may be computed by
neglected.
REFLECTED REACTORS
sec. 8.8]
533
Once a first estimate of the fluxes is obtained it is appropriate to perform

the flux-weighted averages; however, this is a laborious, time-consuming
operation and is usually avoided by choosing the lethargy intervals so
that the flux has only slight variation over any one interval.
This not
only calls for small lethargy intervals, but, since only a limited number
of intervals may be used in a practical calculation, it also calls for the use
of more intervals in lethargy ranges over which the flux may be expected
to be more rapidly varying.
Such ranges are, of course, those over which
the cross sections themselves vary most rapidly.
(Note that the method

of obtaining the multigroup equations implies that the macroscopic cross
section, using the microscopic cross sections for each material present,
should be obtained as a function of lethargy, and then averaged over an
interval.)

There remains the discussion of the spatial integration of the multiSince

group equations and the treatment of the boundary conditions.
there are numerous satisfactory methods of solving the spatial problem,
no attempt will be made to give an exhaustive treatment here.
How
ever, there are a few interesting points in the special problem at hand
which may be brought out by a brief description of one method which

has been successfully used in the treatment of spherically symmetric
It is convenient (but not necessary) to divide the radial dis
systems.
tance into uniform segments of length Ar in each region such that all
medium-medium interfaces occur at integral multiples of Ar.
Then the
differential Eqs. (8.433) are converted into difference equations in terms
of the values of the functions at various radial points.
If
the following definitions are used
mr) s
q^r.Ui-x)
f
Eq. (8.433) may
be
(>(r)
Em
T
T/(n)
m
(8.449)
S_r ^
written in the form
*L
jpj>(r)
oJ"W5>(r)
In (8.450) the following definitions have
af s
(r\
frj^ [22'
CJ>(r) =
V(SHr)
+ 2({21)]
-CJ->(r)
(8.450)
been used:
and
<>
- |g
REACTOR ANALYSIS
534
[CHAP.
Let the values of these functions at the chosen space points

by a subscript ; thus
W\*\m Ar)
#(m Ar)
V^(m Ar) = [V^]m
[W<p]m
be denoted
[#]
(8.452)
f[W-+
Then the difference equation may
[WT'Wi
2
where
be approximated
be
kFWPU
a(Ar)s
- 2[lF!n)]-
by (for example)
(8-453)
written in the form
- [TPM - [SP)m
(8.454)
(Ar)'[C],
(8.455)
W'"){r)
(aV
The second derivative of (8.450) may
[^">]m
Note that, from the definitions above and the assumption of (8.432d),
In (8.454) the fc$"> and [<S;n)]m are known from the calculated constants,
the assumed source distribution and the results of the calculation of pre
may proceed, for example,
assumed that
$"(0)
[Wp]t
[W?U
W]t
~mh
k?WPh
- [Slh
(8.457)
etc.
used as a recursion formula to generate the W's of

Eventually an interface between two regions
reached,
and
necessary to invoke the conditions of continuity of flux and
current in order to cross the boundary and proceed with the recursion
formula (8.454) as applied to the second region. Let the interface be at
=
Ar; then, using the definitions above, we have
radial position
is
Thus Eq. (8.454)
is
is
[inn>]m
it
The calculation of the
from the center of the innermost region where

not singular; thus we have
is
is
it
higher index m.
Z)f>
jfl.Ar
(8.458)
[WT>]*
[WT'Ib*
';2,]Âr
[W?>]**J
(8.459)
Ibid.
is
is
it
The application of the current condition (8.459) involves a choice for

the representation of the first derivative in terms of the difference quanti
very useful
ties Wm. As has been pointed out by the KAPL group,1
preto make this choice in such
way that detailed neutron balance
1

vious groups.
REFLECTED REACTORS
sec. 8.8]
535
For example, let all integrals of continuous func

served in the group.
tions over any volume of the system be approximated by a summation
of difference quantities as follows:
mi
(Ar)2
+ 6*,
r+(r)dr
m Mj Ar
~2
i
/ra
miAr
*.)}
(8.460)
where
This approximation
for the current at an interface
may be used
by
recognizing that
V#1!(r)-dS
Vs#(r) dx
-DÂw.4*
p-^r"
rfr
[wrk-o
ark+i
may be
[wf]B.
WW. - [WTk-i)
(8-462)
|
[HT>U
A"
[I
we use (8.453) for the second derivative in the final integral,

shown that the current condition (8.459) takes the form
(8.461)
For consistency the source term must

is,
be normalized using the form

the integral in the condition expressed by Eq. (8.441)
(8.460); that
to be replaced by the corresponding summation as obtained by apply
The advantage of this method
ing (8.460).
then, realized at the end
Integrations of
of
group calculation.
continuous-group
equation
such as (8.433) over
given region or over the entire system give an
equality whose physical content may be stated in the form
total neutrons thrown

into group from sources _
[region (or system)

With the consistent
(8.463)
total neutrons
total neutrons
lost by
degraded into
next group
[absorption
total neutrons lost

by leakage from
neutrons degraded
from previous group
[Total
is,
is

If
it
volume
-DÂiti
J/
aurf ace
-DÂu,
JI
J<il)(R)4irR2 =
use of Eqs. (8.460) and (8.462), the relationship

in terms of summations over constants and the
be
written
may
(8.463)
IWî*1]^; because of the care taken in obtaining the choice (8.462), this
REACTOR ANALYSIS
536
[CHAP.
equality holds rigorously if the calculations are absolutely accurate.

Thus such summation equalities (for a given region or for the entire
system) may be used as a valuable check on the numerical accuracy of
the calculation.
Eventually, the process of using the recursion formula (8.454) in each

using (8.458) and (8.462) 1 reaches the
very outside boundary of the system R0 = BoAr; here the condition must
be applied that
region and crossing boundaries
= 0
[Wp].
(8.464)
where index
denotes the final, outermost region.
The resultant value
of [Jiy]/)0 may be seen to depend on the choice for
of Eq. (8.457)
which is otherwise unspecified.
Any general choice of [Hr!"]i will not
result in a zero value for [JT/'Ib,.
One simple way around this difficulty
is to define two functions
Thus
and
(B
such that
- [an
=
= o
[],
[}]. = o
W%
(o SRz,
{*Abd)
(8-466)
is a particular solution of Eq. (8.454), and Ct is a solution of the

Both a and (B are required to meet
homogeneous part of Eq. (8.454).
the same boundary conditions as W. The function W is then expressible
(B
of
a and
[Wf]
(B;
thus
[af]m + cfffii-'l,
that once a and (B are known at every point W may

The constant
every point if the constant c is known.
so
by the outside boundary
(8.467)
be calculated
c
at
is determined
condition
c[<&f)B, = 0
(8.468)
is,
While the scheme just presented for handling the outside boundary
condition has been used successfully, it is not the most convenient or
the most accurate method.
The presentation of improved methods
however, deemed outside the scope of the present discussion; the
KAPL
reference gives some details of such an improvement.

The results of multigroup calculation give very detailed information
Note that (8.462) requires the calculation of values of the function Win) at points
outside the nth region (e.g., [Wyja.+O; this
a convenient mathematical device, and
is
should be
Further,
values are not regarded as having physical meaning.
noted that one of these
never really calculated; the quantity [WJ"]b,-i >s related to
through the recursion relationship of Eq. (8.454) for n
[W'J2)]fl. and to
and so does not need to be computed explicitly.
it
these
is

[av>]+i
Ct
sec.
REFLECTED REACTORS
8.8]
537
about the reactor system. Besides the multiplication constant, the dis
tribution of flux in energy and space is given, from which may be com
puted, for example, the distribution of neutron absorptions or fissions in
energy and space or the current of neutrons of various energies emerging
Of course, the information input
from the outside of the reactor system.
10"1
8
6
r~
/ 2,
i
fE
rJ
c
o
r"
i
3
<nirT2
10
w
M
u
Q.
lJ
pJ
o
O
1, i
r--J
102
L_J

r-
(0
'
10
r1
+-2 a
10"
10
12
14
16
18
20
Lethargy
Fia.
8.31
Average cross sections in the core used in multigroup calculation.
is also very detailed.
for
However, once the multigroup equations are coded

machine, the major effort required for each system
a fast computing
is the collection of nuclear data and the computation of the average cross
sections for the various groups.
Even this chore has been eliminated at
many of the major computing facilities wherein "libraries" of subroutines
routine codes not only provide up-to-date

data of all the elements from the thermal to the
range, but also incorporate various energy-averaging procedures for
are now available.

nuclear
Mev
These
cross-section
rapid preparation of group constants.

Various special modifications of the multigroup equations have been
538
REACTOR
[chap.
ANALYSIS
10"
io
\"
"
:,-ic
.
1
[
r1

-2.
10"
10"
10
12
14
16
18
20
Lethargy
Fig.
8.32
Average cross sections in the reflector used in multigroup calculation.
Table
8.9
Nuclear Densities in Core
and
Reflector
(In units of 10" nuclei/cmJ)

Material
Core
0.0003946
0.0000835
0.0004274
0.0005713
0.05827
Fluorine
Oxygen
0 05742
Uranium
Inconel
(235)
Reflector
0.0002365
0.0004096
0.000256
0.0002365
0.06577
0.06526
0.00004418
0.01546
0.00127
sec. 8.8]
REFLECTED KEACTORS
539
2.2
2.01
1.8
1.6
E
..jr"
IQ
D
c
1.
1.0
S 08
3
n
5
re
- -i
0.6
Reflec to
0.4
0.2
10
11
12
13
14
15
16
17
18
Lethargy (origin at 10 Mev)

Fig.
8.33
Diffusion coefficients in the core and reflector used in multigroup calculation

1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
Core radial coordinate
Fio.
8.34
10
12
14
(in units of 2.206 cm)
Assumed and resultant source distributions over the core.
considered; three of these are: (1) the extension of the method to cylindri
cal geometries, (2) the modification to give improved results for hydrog
enous systems, and (3) the improvement of the equations to include
more information about the angular distribution of the neutron flux so as
to gain an improvement
in the treatment
of the slowing-down
process.
0f9
aoxovaH
sisaivnv
dVHO]
8*0
1
! I
L.
so
L_
L.
T
-1
|
fr'O
1
j
isej xnu le
_
.18)1183JO 8J03
i L.|
90
'0
*1
....

'T
U
0
'oij
"
L.
xn|| 03 8
38|}<JO UlSJJ81 83
ro
r J
Z"0
E"0
dsziibuijom uoissii uirupods

suoj)nau) jad jiun (XSjeq)8|
c,{vg
uoissij
fr
01
uiruôads pun a3jaAB
21
81
91
xng-isBj uoiinqujsip ut jCSjBiiiai
r
r
n
'oij
gg's
aAijtqoy jaqtunn jo 'suoissg
01
iod
%\un
ZT
'S2iv\\%o\
frt
9t
anp 0} suoa-jnau
8t
jo
U3A13
REFLECTED REACTORS
sec. 8.8]
541
The average cross sections and results for a typical multigroup calcu
The particular assembly
lation are shown in Figs. 8.31 through 8.39.
chosen is a spherically symmetric system consisting of core and reflector.
The fuel is U2S6 and the principal moderating material in both the core
0.5
1
nl
S3 0.4
0.3
0.2
if
0.1
-44-1
,
8
10
12
14
16
18

Lethargy
Fio.
8.37
Relative number of neutrons absorbed per unit lethargy.
0.6 r
10
12
14
16
18
Lethargy
Fio.
8.38
Relative number of neutrons escaping from outer surface of reflector per
unit lethargy.
and the reflector is BeO.
Various coolant and structural materials are

also present in both regions; the appropriate nuclear densities are listed
in Table 8.9. The core is 30.88 cm in radius, and the reflector is 30.88 cm
thick. With the zero of lethargy selected to correspond to 10 Mev, the
fast range is divided into 25 lethargy intervals of various widths.
The
REACTOR ANALYSIS
542
[CHAP.
thermal base is chosen at a lethargy of 18.4 corresponding to 0.1021 ev

Figures 8.31, 8.32, and 8.33 show the average values for the
or 1672F.
cross sections and diffusion coefficients chosen in each group for the core
The thermal cross sections are shown as points when

and the reflector.
the
value of the immediately preceding group.
they differ from
The
normalized source distribution in the core which was assumed to start
1
1
1
1
1
1
= 1 Mevflux
(0.821-1.35
u=
1/10 Mevflux
(0.009-0.183 Mev)
100 ev flux (33.7-112 ev)
m=
0 E = 1 / 10
\a
v
102-0. 125
ev f lux (0.
_ (' ottc
\sar ne sc ale)
F efle<:tor
\\

<i
\
l
Cor e-
12
Radial coordinate
Fig.
8.39
ev)
*th = Therm al flu x(0. 102 ev)
'V
I'm
1
1
Mev)
'
\
V
~T
16
20
24
\
28
(in units of 2.206 cm)
Spatial distribution of average flux in four fast groups and in thermal group.
is shown in Fig. 8.34, along with the resultant neutron

production distribution after one complete calculation.
(In this case the
calculation was not iterated to convergence of the source distribution.)
the calculation
The originally chosen form was based on previous experience with similar
Figure 8.35 shows the lethargy distribution assumed for the
assemblies.
source neutrons [the j, of Eq. (8.402)] along with the average neutron flux
in each group at two special space points, the center of the reactor, and
the core-reflector interface.
The thermal flux is not shown in Fig. 8.35
since the fast flux and the thermal flux are not dimensionally equivalent
quantities.
For each neutron introduced into the assembly (in the distribution
described in Figs. 8.34 and 8.35) the following final results are obtained:
sec. 8.8]
REFLECTED REACTORS
543
0.1563 neutron absorbed in the entire assembly as fast neutrons

0.4882 neutron absorbed in the entire assembly as thermal neutrons
0.0355 neutron escape from the assembly as fast neutrons
0.3200 neutron escape from the assembly as thermal neutrons
Oooo
and
neutron enter into fission processes
0.28139 neutron enter into fission processes
0.08837
as fast neutrons
as thermal neutrons
Thus,
76.1 per cent of all fissions are caused by thermal neutrons; 75.7 per
cent of all neutrons absorbed are absorbed as thermal neutrons; 90.0 per
cent of all neutrons escaping from the assembly escape as thermal neu
Figures 8.36, 8.37, and 8.38 show the distributions in lethargy
trons.
of the fissions, absorptions, and escapes, respectively (all per unit leth
argy)) m the fast range. Figure 8.39 shows the spatial distribution
through the core and reflector of the average fluxes in four particular
fast groups plus the thermal flux distribution.
The assembly was found to have a multiplication constant k = 0.924.

PROBLEMS
8.1
Determine the critical equation for a one-velocity spherical reflected reactor
which has the following specifications: (1) a homogeneous active core of radius a;
a; the neutron properties of which are,
(2) a spherical shell reflector of thickness b
in general, different from those of the core; (3) an infinitesimally thin spherical shell
of absorbing material at the core-reflector interface; this absorber has the property
that it transmits a fraction or of all neutrons incident on either face.
Show that the critical equation for this system reduces to the usual relation for the
one-velocity reflected reactor when the transmittancy a 1. Obtain an expression
for k(a) and compute
Sk
*()
"
k(l)
k(a)
and 0.80 using a 50 cm, 5 a = 25 cm, xr = 0.02 cm-1, Dr = Dc =

Lq = 128 cm'.
Derive the complete expression for criticality in a spherical reflected reactor
8.2
which has the following properties:
Active core: (1) Uniform distribution of fissionable material and moderator through
out a sphere of radius a; (2) neutron cross sections arc energy dependent, except for
the condition given in (3); (3) diffusion coefficient Dc(u) = Dc, which is energy
independent; (4) neutrons from fission appear monoenergetically (at u = 0).
Reflec
tor: (5) spherical shell of inner radius a, outer radius 6; (0) contains no fuel; (7) neutron
cross sections of all materials are energy independent; (8) neutrons suffer no change
in energy from elastic-scattering collisions with nuclei of reflector material.
In both the core and the reflector regions of this reactor, the neutrons enter a
The coupling relation q(r,u) = J2i(u)<>(r,u) may be
diffusion process at u = uth.
assumed to hold in the fast range.
A reactor consisting of a homogeneous core contained in a spherical pressure
8.3
shell is operating at steady state. This reactor is to be analyzed on the basis of the
for o
0.95
1 cm, and
544
ANALYSIS
REACTOR
[CHAP.
following model: (1) The active core is a homogeneous sphere of radius R.

(2) The
neutron-reflecting properties of the pressure shell can be described by means of the
relation
4>'(R,u)U(R,u)
where <P(t,u) is the neutron flux in the core and a some constant independent of
lethargy (which can be determined from the albedo of the shell material).
(3) The
Fermi age model applies.
(4) The fissionable material has a zero fission cross section
at all energies except thermal energy. (5) One of the components of the circulating
fuel has an (n,2n) reaction cross section which is different from zero for all u < uu.
(6) All neutrons produced from fission and (n,2n) reactions appear at u = 0.
o. Obtain expressions for the slowing-down density g(r,u), the fast flux (r,u), and
Use the relations
the thermal flux 4>u,(r).
- 2(u) + 2.(u)
S.(u) - 2/(u) + 27(u)
2,(u)
=
q(T,u)
Z2t(u)(i,u)
with
+ 2(u)
d.
is
is
6.
is,
where 2/ and 2T have the usual definitions and 2 denotes the (n,2n) reaction cross
section of the circulating fuel.
b. Find the critical equation for this system.
c. Derive an expression for the multiplication k in this reactor.
d. Evaluate any arbitrary
constant which remains in the results of part (a) in
terms of P, the total power output of the reactor. Note that the energy release during
a fission reaction will be, in general, different from that released from the (n,2n) reac
tion. Also, the energy released by the (n,2n) reaction may be dependent on the
energy of the incident neutron.
bare spherical reactor consisting of the materials
8.4 a. Consider a homogeneous,
Compute the critical radius Rc of this reactor for Nu/Ny =
described in Prob. 4.8.
10,000 on the assumption that the resonance escape and fast nonleakage probabilities
are unity, that
the one-velocity model applies.
On the basis of the information given in part (a), compute and plot Nu/Ny for
criticality as function of reflector thickness
by using the above sphere of radius Rc
For these calculations assume that the same material
as the active core.
used in the
reflector as in the moderator of the core and that Dc Dr.
c. What fraction of the decrease in fuel mass obtainable with an infinite reflector
realized with
reflector of one diffusion-length thickness?
Plot the ratio of the decrease in fuel mass to the reflector mass as a function of
reflector thickness.
e. Plot, as
function of reflector thickness, the fraction of neutrons produced per
second in the core which are (1) absorbed in the core, (2) absorbed in the reflector, and
(3) escape from the reactor.
for
reflected spherical reactor which has an active core radius
8.5 Compute
37.2 cm and a spherical shell reflector of thickness 20 cm.
For this calculation
use: (1) the material composition described for the reactor in Prob. 4.8, in the case
Nu/Ny = 15,500; (2) nuclear cross sections for fuel and moderator given in Prob. 4.8:
(3) the two-group method.
Assume that the moderator and fuel are uniformly distributed in the reactor core
and that the reflector consists of pure moderator.
Compare the value of
obtained
from this calculation with the value obtained for this reactor on the basis of a onevelocity model (as computed in Prob. 8.4).
8.6
right cylinder of beryllium containing some Um has been left
homogeneous
The cylinder
standing upright on
exposed in an unattended area of a laboratory.
is

REFLECTED REACTORS
545
one of its flat surfaces.

In its present state, it is not critical. The radius of the
The macroscopic cross section of the U"* it
cylinder is 20 cm and its height 40 cm.
contains is 2(aJi) 0.01297 cm-1. One day a new employee of the laboratory who
had been working nearby took a moment's rest and sat on the cylinder.
The resulting
What are the conse
configuration (in first approximation) is shown at the right.
For the analysis of this problem use the data:
quences of this experiment?
Beryllium:
= 23.6 cm
= 0.70 cm
Za
= 0.0013
p = 1.84
Water:
L
D
I/Employee
cm"1
g/cm
= 2.85 cm
= 0.16 cm
Beryllium
cylinder

and the following idealized model: (1) One-velocity diffusion theory applies in all
regions.
(2) All surfaces and diffusion media, which may be near the cylinder (such
as floor, etc.) may be ignored, except the presence of the employee.
(3) Employee is
composed of pure water.
CHAPTER
REACTOR KINETICS
In this treatment the problems of reactor kinetics are grouped into

three categories.
Each category deals with a basically different type of
nonstationary neutron phenomenon.
To the first category we assign
those problems which involve rapid transients in the neutron population
arising from various disturbances.
These disturbances are conveniently
specified as changes in the multiplication constant or the reactivity.
The
study of these problems is based on the methods and results of the onevelocity diffusion and the continuous-slowing-down
diffusion
theory

models.
In the first few analyses the effect of the delayed-neutron
groups is omitted so as to focus attention on other physical features of

The delayed-neutron groups are introduced in subsequent,
more general, treatments, and various analytical models are discussed for
the problems.
describing
them.
Finally, several of the analytical tools developed in
these studies are applied
to the analysis
of the pulsed neutron-beam
experiment previously mentioned in Sec. 5.5.

The kinetics problems of the second category are concerned with the
influence of the reactor temperature on the time-dependent behavior of
the neutron flux and on the power generation.
The phenomena of inter
est here are again short-term effects, and the importance of the delayedneutron groups in establishing dynamically stable systems is discussed.
The analytical models used in these treatments
draw principally from
the so-called Stein model.1

The problems of the third category deal with phenomena that cause
These include the pro
long-term variations in the neutron population.
duction and decay of the principal fission fragments that appear as poisons
The effect of these substances upon the reactivity of the
in the reactor.
This treat
system is examined and their practical significance noted.
ment is concluded with a study of the use of burnout poisons for con
trolling these long-term changes in reactivity.
Nearly all the physical situations considered in these analyses refer to
This restriction allows solutions to be obtained by the
bare reactors.
Thus the kinetic features of a specific
method of separation of variables.
'J.
Reactor Theory Lecture Series," Lecture III, Kinetic

Oak Ridge School of Reactor Technology, ORNL CF 51-5-113, June 4,
M. Stein, "Simplified
Equations,
1951.
546
SEC.
REACTOR
9.1]
KINETICS
547
A fur
problem can be examined independently of the system geometry.
ther restriction which is made is that the transient problems of the first
category arise from nuclear changes which are introduced uniformly over
the entire reactor.
Transients in the flux (and power and temperature)
due to local changes are best treated by numerical methods or by the
application of perturbation
theory.
Time-dependent Behavior of the Neutron Flux

with Delayed Neutrons Neglected
9.1
of General Problem in Terms of the Fermi Age Model.

In the present treatment the most general neutron-balance relations
developed for a nonstationary system are given in terms of the Fermi
Although this formulation allows some improvement over
age model.
the very elementary models used in Sees. 3.2 and 5.4a, it
nevertheless,
is,
a. Statement
is
cal methods such as the one described in Sec. 8.8.

Apart from this limitation, the principal restriction
in the use of the

model
due
to
the
number
of
conditions
which must be
large
age
satisfied by the physical system
the model
to be valid (see Sec. 7.3i).
However, even with this shortcoming, the model
still very useful for
is
is
if
Fermi
is
preliminary studies and for gaining insight into the essential features of
many reactor problems.
The Fermi age or continuous-slowing-down
diffusion theory, then, offers some generality without introducing exces
sive computational
difficulties.
The appropriate time-dependent differential equations for the combined-slowing-down diffusion theory may be obtained by recognizing
that the time rate of change of the neutron density
given by the differ
is

is
still a somewhat limited tool. Its greatest usefulness

realized in the
study of bare homogeneous systems. Closed-form analytical solutions
for multiregion systems are not possible with this model because of the
difficulty in satisfying boundary conditions which involve simultaneously
a dependence on space, lethargy, and time.
This difficulty was previ
ously encountered in the application of the continuous-slowing-down
diffusion model to
bare system at steady state [see discussion following
Eq. (6.56)]. The only way in which this model can be applied to the
study of nonstationary multiregion systems
through the use of numeri
ence between the rates of disappearance and appearance of neutrons from

sinks and sources. Thus we write
dt
n(t,u,t)
S(r,w,<)
Slh(r,<)
+ D(u,t) V20(r,)t,O
D.(0
V->lh(r,<)
S0(w,<)
- SÔ
*(r,w,0
-~
for u >
q(x,u,t)
ulh
(9.1)
u = ulh
(9.2)
<MM)
for
548
REACTOR
The coupling equations for this

n(r,,t) =
ANALYSIS
set are taken
JL 0(r,u,i)
to be
n,h(r,t) =
g(r,w,<) = $2,(ti,t)
and
[CHAP.
J-
*,(r,<)
(9.3)
(T,u,t)
(9.4)
The various functions which appear in these relations are defined by

n(r,M,0
= number of neutrons per unit volume per unit lethargy around

(t,u) at time
(T,a,t) = neutron flux per unit lethargy around r at time
S(r,u,t) = number of neutrons produced by sources per unit volume per

unit lethargy per unit time around (r,u) at time t
q(r,u,t) = number of neutrons slowing down past lethargy u per unit
volume per unit time around (r,w) at time t
nth(r,<) = number of neutrons of thermal speed per unit volume around r
at time
= thermal-neutron
flux
<th(r, = thermal-neutron
flux around r at time
/
The set of equations (9.1) through (9.4) define the distribution of neu
trons in space, time, and lethargy for a homogeneous system (note that
all neutron constants are independent of r, but are, in general, functions of
u and
t).
The solutions to the above equations must satisfy an appropriate

For these we choose:
of boundary conditions.
(1)
(2)
set
The flux (or neutron density) must be symmetric about the cen
tral axes of the reactor; or the solutions must have no singularities
at points not occupied by localized sources.
The flux must be continuous through interfaces between dissimilar
media.
(3)
The neutron net current
J(r,,0 = -D(u,t) V(T,u,t)
(9.5)
must be continuous through interfaces.

(4) The flux must vanish at the extrapolated outer surfaces of the sys
tem (or the return current from the surrounding space must be zero).
Two special cases of the one-velocity

problem have already been discussed in connection with
The results from these calculations
other methods of reactor analysis.
are summarized below as an introduction to the more general one-velocity
In the first of these cases we examined the problem of the
problem.
b. One-velocity Bare Reactor.
time-dependent
to steady-state
operation
in
For the thermal system the appropriate
multiplying slab
differential
(Sec.
equation
is
approach

around r at time /
Sth(r,0 = number of thermal neutrons produced by sources per unit vol
ume per unit time around r at time t
5.4a).
(9.2),
SEC.
KINETICS
REACTOR
9.1]
549
where the source term is now [cf. Eq. (5.128)]

S,(r,0 =
rzy*tt(r,0
(9.6)
The general solution for th(r,<) is given by Eq. (5.144).
cos
<fch(x, = (?=
y (Ane-D^"
cos anx + Cne-D^" sin
yKx)anx
(9.7)
(?
(Ane-D^"
+ Cne-D^" sin
yKx)
(9.7)
where the a and y are given by (5.137) and the X and to by (5.139).
The coefficients An and C are obtained from (5.143).
The second special one-velocity problem previously considered dealt
with the spreading out of neutrons through a nonmultiplying infinite
medium after an initial burst at t = 0 from a plane source at x = 0
In this case, Eq. (9.2) is used everywhere except at
(Sees. 6.2a and b).
x = 0, and the source condition is given by [cf. Eq. (6.7)]
"
with lim f
S*(x,t) m 0
(9.8)
nlx.t) dx = q0
nih{x>t)
As
(9.9)
(4rZW)
a somewhat
more general statement of the one-velocity problem

consider the time-dependent behavior of the thermal-neutron
flux in a
bare homogeneous multiplying medium which is operating as a chain
reactor at steady state up until time t = 0. At t = 0 some uniform
change (over the entire reactor) is made in the concentrations of the
materials in the system so that the reactor is no longer critical. If we
assume that no changes are made in the fast cross sections or in the size
of the reactor, and that the slowing-down time is zero (with no fast
fissions) but with fast-neutron losses by escape and capture, then the
appropriate equation is (9.2), with the source Stk "Xfp^g^k] thus,
p^,
e_B'T'h
- J *h(r,0
(9.10)
probability to thermal, and the fast
is the resonance-escape
nonleakage probability
0,h
= y2fptigth4>*(T,t)
is given by
with
B2 =
al
fr
where
+ S?*,h(r,0
(j^j
- AhV'^frQ
slaD reactor
(9.11)
Note that Eq. (9.10) implies that all neutrons are prompt.
The solution th(T,t) to (9.10) must satisfy the conditions
(1)
*th(r,0)
(2)
nonsingular
<th(r, is nonsingular
(9.12)
(3)
Mft.t)
*0(r)
is

The solution is given in (6.32) ; in the present notation, this is
(9.12)
REACTOR ANALYSIS
550
[CHAP.
The
where ft denotes the extrapolated
outer surface of the reactor.

initial flux distribution #o(r) satisfies the equation
= 0
V2(r) + 52*o(r)
(9.13)
where B is given in (9.11) for the slab (see also Table 5.3).
The solution to (9.10) may be obtained by the method of separation
of variables used in the case of the slab problem of Sec. 5.4a [cf. Eq.
(5.128)]; therefore, take
*th(r,<)
0(0
*0(r)
(9.14)
Clearly, all conditions of (9.12) are met by this relation since <o(r) is
chosen to meet conditions (2) and (3). However, we require also
0(0)
(9.15)
The substitution of (9.14) into (9.10), using also (9.13), yields the
equation
d
with
L2 = Dth/Zla
that
is,
v^{\
+ L2fi2) di
^"Ua+w)-1]^0
(916)
Z?(l+L2fl2)
= JoPnl
(9.18)
thermal neutron in an infinite medium [see (5.221)]

the nonleakage probability for thermal neutrons [see (5.219)].
we call
6k =
(9.19)*
then, with the above identifications,
(9.16) reduces to
G(t) =
6k
G(t)
(9.20)
Note that these results depend upon the assumption that no change
made in the dimensions of the bare reactor.
The solution to (9.20)
is is
the life of
Further,
and pNL
+ LIB2)
reactor (cf. Sec. 5.4g).
where
lo
^"(1
the average neutron lifetime in the one-velocity

It follows immediately from (5.223) that
1.
Note that the cross sections which appear in this relation were derived
from the noncritical system of Eq. (9.10); therefore, by definition
^
The factor on the left-hand side of Eq. (9.16) may be recognized as
if is is

.
The first term at the right may be recognized as the
effective multiplication constant of the system after the change is made;
seen to be
0(0
= e**'"
(9.21)
SEC.
HEACTOR
9.1]
KINETICS
551
The thermal flux

=
<fch(M)
is given
by
*o(r)e"'
(9.22)*
Thus we find on the basis of this simple model that the flux rises and falls
exponentially in time, depending on whether the change made at t = 0
results in a k which is greater or less than unity, respectively.
The period of the reactor is defined as the time required for the flux to
times its value at I = 0 (in the case of an increase in k).
the above model
rise to a value
By
The period = T =
(9.23)*
OK
It
should be borne in mind that this result is based on the assumption

that there are no delayed neutrons in the system. Thus all neutrons
from fission appear "immediately."
The expression (9.23) should be
compared with the more general relation obtained in Sec. 9.3.
c. Fermi Age Model of Multivelocity Bare Reactor.
The general
approach developed for the one-velocity bare reactor may be applied to
the multivelocity system with the aid of the Fermi age model.
Although

a complete
analytical solution is possible for the multivelocity bare reac
tor, the present treatment is confined to a study of the time behavior at

long times after the system has been disturbed from an initial steadystate condition.
transients
Thus the analysis which follows
does not include
the
which
appear immediately after the perturbation has been

but gives only the asymptotic time variations in the neutron
Further, this calculation is limited by the assumptions that
population.
all fission neutrons are prompt and that the changes which disturb the
reactor from steady state are introduced uniformly over the entire system.
introduced
The specific details of the model used to describe the multivelocity

bare reactor are:
(1)
The differential
(9.1) and (9.2) describe the fast and
equations
thermal fluxes, respectively.

(2) All neutrons born from fission appear (without delay) monoenergetically at u = 0, and fissions are due only to thermal neu
trons; therefore, the fast-neutron
S(r,u,t)
source is
vZf 4,^,1)
(u)
(9.24)
(3)
The thermal sources are due entirely to the slowing down of fast
neutrons; thus,
Sth(r,0 = g(r,W)
(9-25)
(4)
The boundary conditions

4>(t,u,t)
and
0(R,m,/) and
<ti,(r,0
are
are everywhere nonsingular
ih(&,t)
are zero
(9.26)
(9.27)
where R denotes the extrapolated outer boundary of the reactor.
REACTOR
552
(5)
ANALYSIS
[CHAP.
The initial conditions are

*(r,u,0)
**(r,0)
where
o(r,u) and
<o(r)
o(j,u)
(9.28)
*.(r)
(9.29)
denote the fast and thermal fluxes, respec

as it operated at steady state prior to
tively, in the bare reactor
= 0.
(6) At t = 0, a change is made in the thermal cross sections so that the
reactor is no longer critical, and these new cross sections remain
t
fixed for all
> 0.
The appropriate differential equations indicated in

in the present
IT ^ **(r>0
*2,,(r,0
- 9(r,uth,0
(9.30)
= ~ O,hV2,h(r,0
+ Zf*fcfr,fl
We know from the analysis of the one-velocity bare reactor that the
solutions may be written as the product of a space function and an expo
nential in time.
Let us look, then, for solutions to the above equations
in the form [see also (6.59) and (6.60)]
*lh(r,<)
- F(r) ***
nr)
Wliere
h
V
and B2 = (ir/^)2, etc., chosen to meet conditions

tion of Eqs. (9.31) into (9.30) yields
Q(0)
^*
= (AB2 +
(9.27).
Q(uth)
[(')B2
/0"
afu p j -
(y"^;
The substitu
(9.32)
(6)
with <x(m) and /3(m) given by (6.58). The solution

satisfies the condition (9.326), is seen to be
q() =
'W
(c)
to (9.32a), which
^yk(y)]
dM')
q(T,u,t) = F(x) Q(u)e
(9.33)
(D*B* + S*)**
(9.34)
exp
[(')*'
(') +
fr(u,;Z|(M>)]
rf
we use this result to compute Q(w,h), (9.32c) takes the form
If

written
case
9(r,0,0
~
(1) may be
SEC. 9.1]
KINETICS
REACTOR
This result may

[cf. Eq. (9.18)]:
lr
be simplified
we introduce the
du
Uth
M)
Jo
if
553
Ub
2.()
following identifications1
(9-35)*
+ A52)
and recognize that the first term of the exponential is simply gth, the fast
nonleakage probability to thermal, and that the second term is pth, the
Thus, (9.34) may be written
resonance-escape probability to thermal.
her*1'
1
+ Jit*
"_
(9.36)*

v2/P">g">
k =
(9.37)
2* + DlhB*
is,
Equation (9.36) may be regarded as the defining relation for n and must
be satisfied if the forms (9.31) are to be solutions to the set (9.30).
It should be borne in mind that the solution obtained in this way
at times long after the initial pertur
applies only asymptotically, that
a
is
0.
if
0)
exp
density prior
du'\
,2jrW(r)
|^
<z(r,u,0)
(9.33) does not satisfy the expression for the slowing-down

to the perturbation:
{^(u')
(9'38)
We may conclude, therefore, that the solutions (9.31) describe only the
long-term behavior of the system and omit entirely the transients which
arise immediately following the introduction of the changes in the cross
sections.
is
lF
is
The separability assumed in Eqs. (9.31)

not rigorously valid. For
example, the number of neutrons slowing down into the thermal group
will not change appreciably until time of the order of
after the change
made in the thermal cross sections; thus q(i,uth,t) cannot rise as e"'.
The problem treated in this section could be as well represented on the
basis of the time-dependent thermal equation (9.30) with the source term
taken as
Eq.
- lF) .h(r,< -
lF)plhe-Blrth
only an approximation.
function
of time
The proper definition
given in
is
for
is
The expression
(9.40).
vZfit
may be obtained for the thermal flux as
lr
Solutions
g(r,^h,)

0;
=
bation has occurred at
however, the result given may be
good
=
It evident that complete solution
approximation very near
has not been developed since (9.33) does not satisfy the initial condition
Thus,
the nuclear properties of the reactor at steady
on the reactor.
are denoted by the superscript (0), then we find that
state (for
<
554
ANALYSIS
REACTOR
[CHAP.
this section using this form of

the source function ; these will show different analytic forms in each time
interval of length lF after / = 0. Asymptotically, the exponential rise of
Eqs. (9.31) may be observed with a time constant l//i to be obtained
from an equation very similar to (9.36).
More general problems may be
treated by observing that Eq. (9.1) implies a delay time in the fast range,
under the conditions
(1) through
(6) of
i.e., the slowing-down
density at time
by repeated application of the form
q(v,t) = q(u
if
Am, t
- At) -
q{u
and lethargy
Am,
u is to be obtained
- At) ^~^r
Am
is small and related to Au by
A<
Am =
2,(m,0 v(u) At
It should be further noted that an important part of the actual physics

of the time-dependent behavior in the fast range is not included in the
basic equations of this section, (9.1) through (9.4).
Neutrons introduced
at a particular time at a lethargy u will slow down past a higher lethargy
Equation (9.1) actually implies
(m + Am) at different subsequent times.
a continuous-slowing-down
rate, representing the average behavior; the
spreading out in energy and time cannot be handled on this basis.
The expression for the life of
a thermal
neutron
ith
previously
is
If
is
obtained
is
given in (9.35)
in (9.18).
The function lF,
on the other hand,
an entirely new quantity and needs to be examined
in detail.
It evident from the form of this function that
may be
precisely the relation
identified as the slowing-down time for fast neutrons.

This identification
easily demonstrated by making a direct calculation of the time required
neutron to slow down through
by
lethargy interval Am; thus [cf. Eq.
a
is
(4.47)]
Average time
to slow down
through
average number of
collisions needed to
slow down through An
Am
Am
average time between

collisions at lethargies
around the interval Am
X,(w)
(9.39)
lF
if
To compute the time required to slow down from m =

to u = Mth we
need only to sum the slowing-down times through all the intervals along
we call
Then,
the way.
the average time required to slow down to
thermal, we have
du
lv{uyz.(u)
As
approximately
if
the relation given in (9.35),

S,(m) ~ 2,(m).
simple illustration of the use of the general result (9.36), consider
is
which

SEC. 9.1]
the case when nlF

approximate form
1
so that,
<K 1.
REACTOR
KINETICS
Equation
(9.36)
+ n(U + h)
=1
555
may then be written in the
where
y.
(9.41)
y-
+ If
I* ^
for example, the time behavior of the thermal flux is given by

0lh(r,O = F(r)*tt exp
(rA^j-)
(9.42)
In comparison to (9.22) it may be seen that in this approximation the

effect of including the time-dependent behavior in the fast range is to
add the average slowing-down time to the average thermal lifetime.
Reactor.
The computational
d. One-velocity Reflected
difficulties
in
the
time
behavior
of
neutrons
in
determining
encountered
multiregion
media were mentioned previously, and it was pointed out that problems
of this type require the use of numerical (multigroup) methods.
This

very general approach is not treated here; we consider instead a simple

description of a reflected reactor based on the one-velocity model which
will display some of the essential physical features of the general problem.
The reactor to be examined consists of a multiplying core surrounded

by a single nonmultiplying reflector. It is assumed that this system is
critical with the cross sections and neutron parameters listed below :
Core
Reflector
Diffusion coefficients:
Cross sections: 2f,z<0)
Buckling: B\ =
Z)'
2?
- Z>]
l>
Diffusion length:
L\
- - - -JL
1
/)(0)
Now suppose that an amount of absorption 52 is added uniformly

the core and reflector at time t 0; the one- velocity
throughout
flux (r,t) must now satisfy the following time-dependent differential
equations
Core:
-D>Wc(r,0
[sc<o,
J*,(r,0
Reflector:
-2>jfW(r,0
and the boundary

(1)
(3)
(r,t)
[Zf>
J*,(r,0
_IA0c(M)
2<0c(r,O
(9.43)
- \jt4>n(j,t)
conditions
is nonsingular
= >k0,VMR,<)
DV*C(R,<)
(2)
(4)
*c(R,0
*(R',0
s(R,0
= 0
(VM)
where R denotes the core-reflector interface and R' the extrapolated outer
boundary of the reflector.
REACTOR ANALYSIS
556
[CHAP.
As solutions to Eqs. (9.43) try

*c(r,0
- F(i)e-'S1-'
M*,t)
= G(r)e-*2-'
(9.45)
V2G(r)
- x?G(r)
(9.46)
where
= 0
V*F(r) + *F(r)
= 0
and
(1)
(3)
and G are nonsingular

=
DffVFQL)
(2)
D$VG(R)
(4)
F(R)
G(R')
G(R)
((
= 0
.
}
Thus all the conditions (9.44) are met, and 4>c(r,<) and >(r,<) reduce to
the proper space forms at t = 0.
Direct substitution of (9.45) into (9.43)
shows that the assumed solutions for the fluxes satisfy the required differ
ential equations.
We determine next the value of ve which would be required to maintain

the system in the steady state after the extra absorption has been added,
i.e., the value of vc which satisfies the following differential equation and
boundary conditions:

(1)
->>V2*c(r) + [Z< + 520]*c(r)

-Z>5?>V'**(r) + [Zf0' + 52a]*(r)
*(r) is nonsingular
(2)
= 2>S>V*,(R)
DV*C(R)
(3)
(4)
*zy0c(r)
= 0
*,(R)
**(ft')
- MR)
^*yj
,Q ,Q>.
Equations (9.48) and (9.49) lead to the criticality condition which will be
of the type given in (8.11) for a spherical reflected reactor
BR cot BR
- ^ [KR coth x(R' -
R)
1]
(9.50)
62al
(9-51)
where
B* =
4?
["c:S}0>
2aCl<"
5Zo]
*'
"
^ [2afi
Thus, for known reactor dimensions and original core and reflector cross
sections, and a given value of 6S0, a value of vc may be computed from
Then 52 may be related to Sk, which is given by
(9.50) using (9.51).
Sk =
--
(9.52)
Vc
According to Eqs. (9.45) (which are now seen to be acceptable solutions

to the problem), the flux throughout the reactor in this special case falls
exponentially, preserving the space form of the flux which existed during
the steady-state
9.2
a.
operation
(for
< 0).
Reactor Parameters
The Experiment.
by Pulsed Neutron Beam
The measurement of the fundamental
parameters such as the diffusion
coefficient and the diffusion
reactor
length, as
sec. 9.2]
REACTOR
KINETICS
557
well as the determination of the critical size and the fast nonleakage
probabilities for a specific reactor geometry and internal arrangement of
nuclear components, is an essential step in the design and development
Unfortunately this phase of a project is
of every basically new system.
This is especially
of the critical size esti
invariably extremely expensive and time consuming.
true in the case of the experimental verification

Measurements of this type
mates of a complex heterogeneous structure.
carried
out
with
the
aid
of
and criticality
customarily
exponential-pile
are
The purpose of the exponential-pile experiment is the
experiments.
determination of the critical fuel concentration for a specified arrange
ment of fuel and moderator from measurements of the thermal-flux dis
tributions in a subcritical assembly of the actual system. Since such a

system cannot support a steady-state neutron population on the basis of

its fission sources alone, a steady-state distribution is achieved by add
ing a steady external neutron source. The criticality experiment, on the
other hand, utilizes a test system which is designed and operated as a
critical assembly. The desired nuclear information (such as the critical

fuel concentration) is obtained by direct measurement and test. As a
rule, the test system is a crude mock-up which serves primarily to simu
late the essential nuclear features and geometry of the actual reactor.
Thus many important engineering features of the reactor may be omitted,
but from the viewpoint of the neutrons, the test system is nearly indistin
Even then, the complete experiment can be
guishable from the reactor.
quite costly in time and facilities.
It is clear that although the exponential pile and criticality experiments
would be required ultimately in any large-scale reactor development pro
gram, it would be desirable to obtain some preliminary experimental
verification of reactor calculations by means of other more modest tests.
One experiment which appears to be eminently suited to this purpose is
based on the pulsed neutron-beam technique.
This technique has been
applied by several investigators to the determination of the thermaldiffusion coefficient and macroscopic absorption cross sections of reactor
materials.1-4
More recently, it has been used by E. C. Campbell and
P. H. Stelson in the study of short-lived isomers and for the measure
ment of reactor parameters of multiplying media.6
The experiment con1 G.
von Dardel and N. G. Siostrand, Phys. Rev., 96, 1245 (1954).
F. R. Scott, D. B. Thomson, and W. Wright, Phys.
Rev., 96, 582 (1954).
A. V. Antonov et al., "A Study of Neutron Diffusion in Beryllium, Graphite and

Water by the Impulse Method," Paper 661, Proc. Intern. Con}. Peaceful Uses of Atomic
Energy, Geneva, 1955, vol. 5, pp. 3-12.
4 R. Ramanna et
al., "On the Determination of Diffusion and Slowing-down
Constants of Ordinary Water and Beryllium Oxide Using a Pulsed Neutron Source,"
Paper 872, Proc. Intern. Conf. Peaceful Uses of Atomic Energy, Geneva, 1955, vol. 5,
pp. 24-27.
6 Oak Ridge National Laboratory, Physics Division, Semi-annual Progress Report
for Period Ending Mar. 10, 1956, ORNL Report 2076, pp. 37-39.
a
REACTOR
558
ANALYSIS
[CHAP.
of irradiating a sample of the reactor material with a

very short burst of neutrons and measuring the decay constant of the
The reactor parameters of
fundamental mode excited in the specimen.
interest may then be deduced from a knowledge of the dependence of the
decay constant on the shape and size of the specimen (i.e., the buckling).
sists essentially
This experiment is especially useful for determining the thermal-neutron

However, it is not necessary in this case to
properties of the material.
In fact, it would be prefera
provide a pulsed beam of thermal neutrons.
ble to expose the specimen to a burst of fast neutrons since in this way it
Test specimen

Neutron
burst
-Triton target
Pulsed neutron-beam experiment.
would be possible to obtain also some information about the slowing-down

The pulsed beam of fast neutrons
properties of the system in question.
can be obtained by means of certain charged-particle reactions.
Two
useful examples are the deuteron-deuteron
1H2
+ ,H2
2He3
2He<
and proton-triton reactions:1
+
+
on1
0n'
(9.53)
Briefly, the experimental procedure consists of bombarding a target of,

say, triton with a beam of protons from an accelerator (see Fig. 9.1).
The ensuing reaction, as indicated in the above equation, produces high-
If the proton beam is applied

energy neutrons (in the order of 1 Mev).
intermittently, the target will yield bursts of fast neutrons, and some of
these in turn will produce neutron-flux transients in a test specimen
The time variations in neutron
placed in the vicinity of the target.
density at various stations within the specimen may then be measured
E. Fermi, "Nuclear Physics," rev. ed., pp. 180-181, course notes compiled
by Jay Orear, A. H. Rosenfcld, and R. A. Schluter, University of Chicago Press,
Chicago, 1950.
1 See also
sec. 9.2]
REACTOR
KINETICS
559
by means of suitable counting devices, and from these one obtains the
The
decay rates of the spatial modes of the neutron-flux distribution.
determination of the various reactor parameters from a knowledge of the
decay constant for the fundamental
which follows.
mode is demonstrated
in the work
b. Flux Transients in Subcritical Multiplying Media Irradiated by

Neutron Burst. As the first step in the analysis of the pulsed neutronbeam experiment we derive the expressions for the neutron flux in the
Since our primary
specimen as a general function of space and time.
purpose is to obtain the relationship between the decay constant of the
fundamental mode and the reactor parameters, we will assume that cer
tain preliminary information about the neutron slowing-down process in
In this way we can focus attention
the specimen is already available.

on the study of the space and time distribution of the thermal neutrons
in the specimen. For this purpose we introduce the following model of
the neutron processes in the system: (1) A short burst of fast neutrons
occurs either outside or within the test specimen at t = 0.
(2) These
fast neutrons give rise, ultimately, to a thermal-flux distribution a.

(3) It is assumed that the flight time from the fast-neutron source to
the specimen is negligibly small, and (4) that the slowing-down time
within the specimen may also be neglected so that the distribution 4>0
appears effectively at t = 0.
(5) The leakage and capture of fast neu
is
taken into account, but fast fissions are
trons during slowing down
ignored.
Thus our immediate problem is to determine the thermal-flux distribu
tion <Kr)0 m an arbitrary geometry, given the distribution >0(r) at t = 0.
For simplicity, both analytical and experimental, we assume that the
specimen is a bare homogeneous system and of some elementary shape
such as a parallelepiped, cylinder, or sphere. The differential equation
which contains fissionable material is given by [cf.
for such a system
'
Eq. (9.10)]
-Z>V0(r,O
+ 2*(r,()
= vXfPtbg^(T,t)
- ~ *(r,t)
denotes the fast nonleakage probability.

equation must satisfy the conditions
The solution to this
where
4>(R,t)
= 0
(r,0) *o(r)
(9.55)
where ft represents the extrapolated outer surface of the specimen.

convenient to write (9.54) in the form
V**(r,0 + B'*(M)
with
2 = ~ (vS/Pthff,
- S.)
0(r,O
"^g^
(9.54)
It
is
(9.56)
*
(9.57)
REACTOR ANALYSIS
560
The solution may
(i,t)
Equation (9.57) may then

V2F
written
be
G'
The solution is given by
is some constant.
p2
Thus take
be obtained by the separation of variables.
= F{i) G(t).
where
[CHAP.
(7(0 = G(0)e-^'<
V2F(r) + (B2 + p2)F(r) = 0
(a)
ôy;
(6)
The solutions to the wave equation (9.59b) were developed previously in

For our present purposes it is convenient to write
Sees. 5.4a, b, and e.
these solutions in the form

V2Fmn,
+ ain,Fmn.
= 0
with Fm.(ft)
= 0
(9.60)
where the Fmn, are the various eigenf unctions (modes) of (9.596), and the
am, are the corresponding eigenvalues.
If the results of Sec. 5.4a are
extended to the case of the parallelepiped, it is easy to show that
fimy cos
y,z
where
(2n
+1)
pm
(2m
- i.
"
dimensions,
(2s
(9.62)
1)
of the
respectively,
(9.63)
takes on the values

P2
and in general,
p2
i.
/32
and 2a, 26, and 2c are the x, y, and

Note that in this case
medium.
7. =
1)
(9.61)
7.2
an =
-* Pi..
(9.64)
If
is
The general solution to (9.56)

obtained by summing the contribu
tions of all the modes [cf. Eq. (5.141)].
we call
- G(0) Fmn.(T)e-Dna'~"-B,)'
(r,0
(^..^^(r.O
.)
0,
1,
(9.65)
s
given by the triply infinite series (m, n, and
then
is
4>n,n.(i,t)
4>
(9.66)
These
where G(0) has been absorbed into the arbitrary coefficients amn,.
Thus, from (9.55)
are of course obtained from the initial condition.
*(r,0)
- *o(r)

Fm.(r) = cos ax cos
aFmn.(r)
(9.67)
sec. 9.2]
KINETICS
REACTOR
561
In the usual way, we compute

amn. =
fo(r)
F(r) di
(9.68)
exp [
Dv
*2
m,n,
- ^g^)
^
X F.(r)
t]
*(r,<)
over the entire
volume of the specimen.

Finally, we can write the general solution in the form
integration
the
(L
to extend
fy
ô(r')
FM(r') dr'
(9.69)
where we have introduced the definition for B2 from (9.57).

c. Reactor Parameters from the Decay Constant.
At long times after
each neutron burst from the target plate, the time behavior of the thermal
is
well represented by the fundamental

flux in the specimen
Thus the expression (9.69) reduces to the form
4>(r,<)
aooo
exp [-(t>20
+ vDa\w
mode F0oo(r).
i>v2/p,h^)<]F0oo(r)
(9.70)
is
X000
vZa
+ vDalo
multiplying medium
- wE/pag^
will
It
i'2
+ vDalM
vl</\\
Then,
in the
(9.72)
be convenient
the fuel and nonfuel components of the system.

usual notation,
X2ooo
is
exponential factor.
case of the
()'
of the relation (9.70)

the coefficient of
We give this the symbol X00:
decay constant
In the
(I)'
it
The
()'
- -
to separate
we write, in the
- vPlhgJ
(9.73)*
is
is
knowledge of the decay constant

evident from this relation that
function of the buckling a0oo can yield considerable information
Xooo as
about the various parameters which appear in (9.73). The principal
task of the experimenter, then,
to obtain measurements of Xooo for
This can be accomplished, however, only

the higher
aooo.
modes have virtually disappeared so that the relation (9.70)
valid.'
Thus the first step in making the measurements
to record the neutron
it,
is
is
if
various
The interaction due to higher modes can be reduced appreciably

at a nodal point of the first higher mode.
placed
if
is
as function of time.
When
density, or some quantity proportional to
plotted on semilog graph paper, the curve reduces to
this function
is

is
the eigenvalue corresponding to the fundamental

The quantity a0oo
the
case
of the parallelepiped
and
for
given by [see Eq. (9.63)]
mode,
the detector
REACTOR ANALYSIS
562
[chap.

straight line over that portion of the time scale wherein the approxi
An example of such a curve is shown in Fig.
mation (9.70) is satisfied.
These data were obtained by Campbell and Stelson from measure
9.2.
ments performed on a 5-in. cylinder of pure water.1 The decay of the
2,000 1
10'
0
1
100
200
300
400
I
500
Time, Msec
Fig.
Decay curve of thermal neutrons diffusing out of H20 cylinder as detected

with LisI(Eu) scintillation counter.
9.2
higher spatial modes mentioned above has also been studied, and a typi
cal set of thermal-flux distributions at various time instances is given in
Fig. 9.3. These measurements were made on a rectangular block of
beryllium. The curves in the figure show the shapes of the flux distri
The presence of the higher
bution at successive 100-jusec time intervals.
modes, which decay more rapidly than the fundamental, can be seen from
From these
the gradual shift to the symmetric fundamental mode.
within
mode
is
established
about
that
the
fundamental
it
appears
results
one millisecond after the initial burst.
It is evident from the relation (9.70) that the decay constant of the
1 Oak Ridge National Laboratory, Physics Division, Semi-Annual
for Period Ending Mar. 10, 1956, ORNL Report 2076, p. 156.
Progress
Report
sec. 9.2]
REACTOR
KINETICS
563
fundamental mode is simply the slope of the decay curve, and measure
ments of Xqoo may be readily obtained from data of the type given in
Fig. 9.2. The second and final step in the evaluation of the reactor
parameters, then, is the measurement of Xqoo for various alw, and this
10
13
Distance
9.3
(9
Fig.
block
at,
the decay curves for specimens of
16
19
22
in.
Time-space distribution of thermal neutrons diffusing out of a beryllium

by Hi by 23 in.) following 1,000-psec bursts of fast neutrons.
is
plotted as a function of the

various size (and/or shape). When Xqoo
buckling al00, the resulting curve will be of the form shown in Fig. 9.4.
These data are then easily interpreted in terms of the reactor parameters.
Consider first the case in which the specimen contains no fissionable
(1)
Nonmultiplying media: 2^' =

X?oo
[2<0)
material

may be accomplished by determining
(9-74)
Dal00]
REACTOR ANALYSIS
564
[CHAP.
which follows immediately from (9.73). This function is shown in the

It is clear that the alw = 0 intercept yields the absorption cross
figure.
section of the material of the specimen, and the slope of the curve, the
diffusion coefficient.
When the specimen contains fissionable material, the Xjoo curve is no

longer a linear function of ajS00, but has the general shape indicated in
A specific case of this type is represented in Fig. 9.5, which
the figure.
(multiplying
media)
p[x+i*4.]
(nonmultiplying

E
to
T3
media)
Buckling a2000
(increasing
size)
"Critical size
Fio.
9.4
Decay constant of thermal neutrons in a multiplying
medium.
presents the decay constant for thermal neutrons in a cylinder of U2,6-H20

solution.1
The interpretation of the results for the general multiplying
medium is as follows.
Consider the case when
= 0
(2)
(9.75)
obviously corresponds to the critical system, since the timeThus a second

dependent function in the expression (9.70) now vanishes.
experiment
neutron-beam
fact
to
be
obtained
from
the
pulsed
important
Another fact to be learned
is the critical dimensions of a given geometry.
This
case
coefficient of the mixture and the

cross section of the fuel, since for small specimens
from the experiment

absorption
(3)
1
a00->
is the diffusion
oo
From the work of Campbell and Stclson reported in
(9.76)
ORNL
Report 2076, p. 38.
BEACTOR KINETICS
sec. 9.2]
565
This function is represented by the broken-line curve in Fig. 9.4. Note

that the \lQ0 = 0 intercept yields the total absorption cross section of the
and the slope of the curve, as before, the diffusion coefficient.
these results for D, 20(", and 2<,0) the g+ corresponding to each value
mixture,
With
0.1
0.2
Buckling,
Fig.
0.3
0.4
0.5
0.6
ao00(cm-2)
Decay constant of thermal neutrons diffusing out of cylinders of U,"-H20

solution as function of buckling.
of
9.5
a0oo is
now obtained directly from the known values of
the intermediate
(4)
(a0Oo)erili<ml
Xooo-
Thus for
range,
<
"000 <
00
yields the product
pthg+
If
the resonance-escape probability has been obtained by some other

means, then a knowledge of X00 yields g^.
It is interesting to note that,
for very large specimens, the value of g obtained in this way should agree
well with the Fermi age expression for the fast nonleakage probability,
is,
that
(5)
For small
a0oo
ff*
It
g,
= e-wu,
(9.78)
perhaps evident from these observations that the relatively modest

and inexpensive pulsed neutron-beam experiment can yield
great deal
of information about a particular reactor material.
The outstanding
a
is

45
566
ANALYSIS
REACTOR
feature of this experiment
[CHAP.
is,
of course, the fact that this wealth of data

may be obtained without the use of
complete reactor facility.
9.3
Effect of Delayed Neutrons
is
a. Delayed-neutron Groups.
Most of the neutrons produced by the
fission process appear, effectively, at the instant of fission.
However,
small fraction of the neutrons produced in fission
emitted by daughter
nuclei from the primary fission fragments when these nuclei subsequently
Neutrons produced by this mechanism do not
undergo beta decay.
appear at the instant of fission but at some later time.
The delay period
is
is
principally determined by the decay constant of the preceding beta

emission, since the neutron
produced immediately after the beta decay.
These tardy neutrons are conveniently classified by groups, each group
being defined by the characteristic delay time between the instant of
fission and the appearance of the neutrons.
Tables 9.1 and 9.2 list some
of the properties of the principal delayed-neutron
groups for various
fissionable nuclei.1
a
being
is
the order of 100 sec or longer), the number of delayed neutrons

produced per unit time at any instant
given by
X,C<(r)
(9.79)
Number of delayed neutrons

produced per unit time per
unit volume in the steady state
replaced by an emitter of the same type from fission; that
XiC.W
is,
Xi
where C,(r) = concentration of beta emitters of the tth type which are
precursors of delayed-neutron emitters
= decay constant for the neutron emitter of the tth type
N = total number of delayed groups
In the steady state each emitter which decays must, on the average, be
f.Z^r)
(9.80)
is
the fraction of fissions which yield precursors of the tth type.

Then the total number of neutrons (prompt and delayed) produced per
where
f,
1,
'See also G. R. Keepin, "Delayed Neutrons," in R. A. Charpie et al. (eds.),

Progress in Nuclear Energy, Scries
Physics and Mathematics, vol.
pp. 191-225,
Pergamon Press, New York, 1956.
I,

is
The presence of the delayed neutrons

not very important to the
operation of
stationary-fuel reactor.
In such
reactor
at any time instant the beta emitters are located essentially at the point
of the original fissions, and these nuclei are present in concentrations
which reflect the rate of fissioning at some previous time.
However,
since the flux and fuel concentrations are constant in time (over times of
steady-state
CO C
a o
'i
3 ^
cc
O
?
U
CU
r*
5?-
iO 00 Ci M
N
M 5
-h in
8-;o
o o
d d
p o o
5
5 o S
o ci o
CU CQ.
r.
ii
CO O
Cf &
PL, 2
n
o
o
o
f
o n
a)
2C
SO M
3 -a
_,
^
in
s a
OU o
o o as
o o
o
1S
lO 'f Tj< H
iO CN
t
m
J5 oi
cu
cu II

n
n qo oo e
CI N O N N
O O "i o o
o o o o o
o o o o
d d d
CD
t5
53
co 3
c
.2

c o
S 3
o
0
IO iO
bo
s~ S
n 'bo
&
&
03 Ci
HOO
_ w
n h
o o o o o o
o o o o o
eo N oo
CO CO O CO CO
53
n n
h
(i
t
h N
N
O
od
8
3
U5 to co
M IO
in io
o o o o
co
53
>> o
a
3
ii
n,
-S
h N
CO
667
(S. -a
WQ
^ o a
JJ
Sod.
cu b
d T3
cu .3
be
"3
ca q
0,
03 o
X
If S
a"
CO
XI Q
CO
c3 -= a
53
to O
cu 2 -i
S pi
J*
-o
c
i 13^
iO
-ill
o o o
Si '
s
o
- >
9
2 =3 -a
m co co on o>
N iO ^ co *f
Q S
"H CM eg m
m m
m in *
m in
BC 22 o
O O
O
O O O O
O O O O O O
ss
5 c
e
s
o3 cn
Q Ml
t~ 00 CD CO
oo -h
3
.2
eg m
o o o o o o
1 S.
'6q
iO
O
N
t-
J3
<
I
"
2 (
53
REACTOR ANALYSIS
568
[chap. 9
unit time is
Total neutrons produced
per unit volume per
unit time
where
/3
1\
Wr)
(1
- P)2/+(t)
(9.81)
is the fraction of neutrons from fission which are delayed
is
and
j8,
i= 1
Table
the fraction of neutrons from fission which are produced by the

Delayed Neutrons"
9.2
from Fast Fissions6 of Th!" and U,m
ThJ
Group
54.0
22.0
6.2
2.1
0.52
0.18
0.00208
0.00838
0.01083
0.02890
0.01068
0.00221
53.0
22.0
5.30
2.00
0.51
0.18
'
1
2
3
4
5
0
Total...
Neutrons
per fission
0.00035
0.00445
0.00695
0.01796
0.01048
0.00383
0.044
0.063
4,
G. R. Keepin and T. F. Wimett, "Delayed Neutrons," Proc. Intern. Conf. Peace

Atomic Energy, Geneva, 1955, vol.
P/831, pp. 162-170, 1956.
Fissions produced by neutrons from fission spectrum.
Uses
of
ful
1
ith precursor,
The substitution of (9.80) into (9.81) yields, for the
f,-2,0(r) +
(1
Total neutrons produced per

unit volume per unit time
steady state,
/3)"2/<Kr)
(9.82)
"2/cKr)
[ifc+a-]
*2/*(r)
The last step in the equation follows from the fact that each precursor
decay yields one neutron, so that
N
=
(9.83)
'-
We may conclude from this calculation that in the steady state the
total number of neutrons produced from fission
not altered by the dis-
is

Neutrons
per fission
Half-life,
sec
Half-life,
REACTOR
sec. 9.3]
KINETICS
569
tribution of the delayed neutrons; however, it is true that the energy

spectrum of the fission neutrons is dependent upon the properties of the
Thus, if the average energy of neutrons from pre
delayed neutrons.
cursors is different from the average energy of prompt neutrons, an accu
Actually, there is
rate calculation must take this effect into account.
in the average energies of the prompt neutrons and the

Table 9.1), but the difference is not significant from the stand
some difference
delays (see
point of the calculation
of the fast leakage and absorptions
in a thermal
In the analyses which follow, we shall disregard this effect.

reactor.
The presence of the
b. Application to One -velocity Bare Reactor.
in
important
an
effect
determining the neutrondelayed neutrons has
The essential features of
flux distribution in a nonstationary system.
this behavior are amply demonstrated by means of the one-velocity bare
In this case the appropriate time-dependent differential equa
reactor.
tions for the thermal flux and the concentration
-DV'*(r,<) + Z.*(r,t)
~ *(r,J)
(1
of precursors are
- P)y2fPg*+(T,t)
AT

jtCi{T,t)
-X,Ci(r,0 + U2,(T,t)
P.h0h
2, WM)
(9.84)*
(9.85)*
= 1, 2,
where all the neutron constants which appear refer to the thermal neu
trons.
The second and third terms on the right side of (9.84) account
for, respectively, the source neutrons which are prompt and those which
are released from the delayed-neutron
represented by the concentrations:
n.fft)
'
concentration
precursors.
of delayed-neutron
The precursors are
precursors
of the tth type at space point r and time
The quantities pih and gtk in (9.84) denote the resonance-escape and fast
nonleakage probabilities to thermal in the usual way [(4.260) and (6.79)].
Finally, note that since the reactor described by (9.84) is not critical the
multiplication constant will be different from unity and may be obtained
from the relation
k =
(f
with B*
(i)2for
a sPhere- etc"
<9-87)
In the system under consideration (a noncirculating-fuel bare thermal

reactor), it is physically reasonable to suppose that the spatial variation
of the concentration of delayed-neutron precursors is proportional to that
of the neutron flux and that this mode persists even though the magni
570
ANALYSIS
REACTOR
tude of the flux changes in time.

(r,Q
Thus
- F(r) *(Q
(Dfl +
- i (0
so)4>(0
we assume
- F(r)
C,(r,0
where V2F + BW = 0 and F(ft) = 0.

tions into (9.84) and (9.85) yields
[CHAP. 9
The substitution of these rela
- P)2fP*g*4>(.l) +
(1
(9.88)
C(0
V X.î(0
p,hff,h
i= i
C\(0 = -X,C,(0 + f<2/*(0

This set may
<
(9.89)
JV
1, 2
further reduced by introducing the definitions
be
Gi(t) =
along with the relation (9.87)
- ff)k -
-[(1
and
= yfr
(9.90)
thus
-W(0
1]*(Q
fc
^(0
(9-91)
t=i
t =
(9.92)
0(0 =
y-i
N'
and
^-e-'-'
G.(0 =
y-i
IV*
(9.93)
to be chosen so that these relations are solutions to the

The substitution of (9.93) into
equations (9.91) and (9.92).
TV'
is
The number
N'
We attempt
differential
JV
1, 2
given by (9.35).
solution of Eqs. (9.91) and (9.92) in the form
is
Z,h
where
-XA(<) + ft0(t)
(9.92) yields
Since the functions
e"'' are
nfc*
linearly independent,
y-i
2,
1,
X,)e"><
y-i
I\,(ay +
JV'
AT'
TV
(9.94)
(9.94) must be met by
r<y(y
requiring
X.) =
vfr
(9.95)
By similar set of operations the substitution of (9.93) into (9.91) yields,

with the aid of (9.95),
a
-1]
- La,
-k
- 0)k
j=l,2,
...
,N'
(9.96)
[(1

0,(0
Thus, Eqs. (9.93) are acceptable solutions of (9.91) and (9.92) provided
SEC. 9.3]
REACTOR
KINETICS
571
that the quantities a, are chosen to be the solutions (N' in number) of the
following characteristic equation for a:
[(1
This equation
- fi)k - - La
ously denned by (6.143):
We have
/tha
i-l
Ztllar
0 =
9.6
o>
previ
a +
X,
Equation (9.98) is
ft-
a con-
if
the indicated
-X2
X3
functions
are to
a=-\i
Characteristic function /"(a).
X,
ltb
satisfy the relations (9.91) and (9.92), then the a, to be used are the
solutions to Eq. (9.98).
Thus given the neutron lifetime
and the
and j8,, (9.98) yields the set a,.
delayed-neutron properties
Figure 9.6
constant in time (and,

actually dimensiona
The reactivity
is
constant of the system (thus also the reactivity)
in general, different from unity).
is
is
shows a sketch of the function F(a), given by the right-hand side of (9.98).
The situation under discussion here
one where, over the period of
time in which Eqs. (9.93) apply [and therefore (9.98)], the multiplication
is
is
it
less number, but on the basis of (9.98)

sometimes measured in inhours.
An inhour
unit specific to given reactor)
defined as that amount of
reactivity which will allow, as solution to (9.98),
(a

dition on the solutions (9.93); that
is,
Fig.
(9.97)
used here the definition
written in terms of the reactivity
is conveniently
-k
1]
ai
doosec"
REACTOK ANALYSIS
572
It
[CHAP. 9
is easy to verify that the function F(a) has the following properties:
F(0)
lim F(a)
= 0
=
a*
-* o
lim
lim
e-0
F(-\i
F(
()
t
Hh
= +oo
= +
(9.99)
00
From these facts it may be seen that for a given w (positive or negative)
Here N is the number
there are (N + 1) solutions for a of Eq. (9.98).
If we denote the roots ay in the order shown
of delayed precursor types.
in Fig. 9.6 we find that <*i is positive if co is positive; if w is negative, <*i is
also negative, but it is also, then, the smallest root in absolute magnitude ;
i.e.,
< || <
Evidently in this
for
|a*fi|
w < 0
case the roots am (where m = 2, 3,
(9.100)
.
N +
1) are
all negative.
In
oi(0) =
G.(0) =
(9.101)
Giu
(t)
The problem is to determine the functions
and G,(t) for all subse

quent time {t > 0). A solution is now possible since the system has been
Note that the conditions (9.101) along with the
completely specified.
Eqs. (9.95) provide a sufficient number of relations from which to com
pute the coefficients <p,-and Tijt since the a,- are now known ; thus,
ff+i
y-i
with
JV
AT+ l
ipj
Tij(aj +
X.)
<pfii
(9.102)
'
'
t =
1, 2,
1, 2,
N
N, N +
is,
give (iV + l)2 equations for (N + l)2 unknowns.

The time-dependent behavior of the flux
then, quite complicated in
the immediate time interval after the addition of the reactivity to; how
(t)~e'*t
Gi(0~e
1)
sufficient time the terms of (9.93) involving ea>' (for

>
will be negligible in comparison with the first term e"1'. It follows, there
fore, that after some time
ever, after
(9.103)
An approximate solution for the root ai may be readily obtained when

In that case we may expand F(a)
small and positive.
the reactivity w
is

the general solutions (9.93) will be required to

For example, the reac
satisfy certain initial conditions on the system.
tivity oj may be introduced at t = 0, at which instant
a specific problem
sec. 9.3]
KINETICS
REACTOR
in a Taylor
series about a = 0; thus,
for small
~F(0) + F'(0)a, +
It
573
a>,
(9.104)
is easily shown that
oi =
(9.105)
1= 1
We find, then, that after sufficiently long time the effective reactor period
would be
i-l
+ Pi
(l*
(1
(l*
fii
A result similar to (9.106) may be obtained heuristically by calculating

an average lifetime for all neutrons; thus,
)+"
+'*(i*+)-,*+2
(9-io?)
is,
(1
since the prompt neutrons [of which there are

0)] have lifetime lth,
while the delayed neutrons in each group have the lifetime ltb, as thermal
neutrons plus the average delay time before they appear in the system,
that
1/X. Of course, (9.106) holds only for very small reactivity; in
that
Stable reactor period =
reactor period;
is,
is
more complicated dependence on

general, the form of 1/ai will have
The coefficient in the exponent of the leading
the parameters involved.
term after long time
usually referred to as the reciprocal of the stable
=
(9.108)
c.
c*i
is
to be obtained as the appropriate solution of (9.98).

with One Group of Delayed Neutrons. The
complexity of the flux time behavior [as given by (9.93)], which results
from the large number of important delay groups, tends to conceal the
where
Time Dependence
It
picture of the time-dependent behavior of the reactor.
possible, however, to display the essential features of the system by intro
simplified model based on the concept of single group of delayed
ducing
a
is
physical

We have used here the approximation Sk ~ u, which is valid for small

By comparing (9.106) with (9.23), which gives the reactor
reactivities.
period in the absence of delayed neutrons, it may be seen that the effect
of the delays is to increase the effective mean lifetime of the neutrons.
REACTOR ANALYSIS
574
[chap. 9
The foregoing results may be applied directly to this model

= 1. The properties of this single group are represented by
the decay constant X and the fraction of all neutrons delayed by 0. These
quantities may be chosen to make the results of the calculation as nearly
correct quantitatively as possible.
For this purpose, it is convenient to
neutrons.
by setting
choose
/3
for
U236 (see
= 0.00755
also Table 9.1 for values of

X is given by
(9.109)
/3
for other fissionable materials).
i- it
<9i"
In the present notation
and
X,
where
/3,
i= 1
and decay constants of the actual
are the delay fractions
delay groups.
The appropriate
for the one delay-group model are easily

obtained from the more general results of Sec. 9.3b. These are
rin^
X)
ri(i
<pi
IV"'
IV"'
(9.111)
(9.112)
drx)
and G0, then from (9.102)
T2 = Go
=
r2(<*2 +
Ti
*>,/3
(9.113)
<ptP
now a
^)-0
i,h("
^V
4X.
'th(w
li
IM
J
[l
i{Ml
1)|
identified as the positive root and a2 as the negative.

The particular solution for the coefficients <p\t <p2, Ti, and T2 depends
upon the initial values of the flux <o and the concentration of the single
precursor Go. These quantities, in turn, are dependent upon the state of
the reactor prior to =
[In any case, the applicability of the present
0.
is
where m
1)
2U" /,x)
co(l
(g
1)
The roots c*i and a2 are readily computed from (9.112), which
quadratic equation in a; these are
is
we denote the initial conditions
If
G(<)
X)
"
and
= Vie"1'
<K0
results
always limited to systems in which the separation of time and
space functions, as in Eqs. (9.88), may be assumed.]
For the special case
wherein the reactor was at steady state with flux <o prior to =
we
may obtain
relation between <o and Go by noting the balance for pre
cursors in the steady-state condition as given by (9.80).
In the case of
a
0,
is

relations
REACTOR KINETICS
SEC. 9.3]
a single group of delays, this relation

XGo
575
yields
0o
(9.115)
With this condition the coefficients are found to

ipi =
a2(ai +
X(a2
\)o
r-
fi
ai)
be, from (9.113),
= i(a2
X(ai
\)o
r-
(9.116)
a2)
In the case that the reactivity change co is positive, an appropriate set of

approximate expressions for en, the reciprocal of the stable reactor period,
is found to be
(9.117)
is
/3.
"
Ul "
reactor period calculated for the system in which the delayed neutrons
were neglected, (9.23),
(l^)
\L
7-^ co G(t)
(9.118)
W(D~CW
(9.119)
coy
()
this relation reduces to
co
/3,
W(0
When
model, this equation has the form
delay-group
/3,
is
is
very nearly the reactor

For large reactivities, the stable reactor period
The critical point
period computed without regard to delayed neutrons.
for the reactivity
at co =
as may be seen from Eq. (9.91) in the one
CO
is is
is
is
changing in time only because of the appearance of the

critical on the basis of the prompt neu
delayed neutrons; the reactor
trons alone.
It customary to refer to reactor system in which the
just equal to the fraction of neutrons which are delayed
added reactivity
Thus the flux
as prompt critical.
Some idea of the magnitudes of the various parameters and variables

in
transient system may be obtained from the following
numerical example based on the one delay-group model.
Let:
0
= 0.00755
co
= 0.001
1^
involved

/3
/3,
These results are obtained from Eq. (9.114) by neglecting co as compared

to /3 and unity for the case when co <K
and by neglecting
as compared
It of interest to compare these relations with the
to u when co >>>
= 0.076 sec"1
= 0.001 sec
ANALYSIS
REACTOR
576
[CHAP. 9
In this case the roots of the characteristic equation may

by the expressions
Otl
/3
at
tii
be approximated
(9.120)
may be derived from Eq. (9.114) by neglecting Xith relative

unity. With these relations and the set (9.116), we obtain
which
= 0.0117 sec-1
<*i
= 6.55 sec-1
<p2 = 0. 132#o
ar2
ip\ = 1.132(ô
The corresponding expression for the flux

4>(t)
= 1.132e0
to
0117'
(t)
is
0.132e-8-65'
(9.121)

For this problem we have assumed that the

where t is given in seconds.
reactor was in steady state (flux level d>o) up to time t = 0, when the
reactivity 0.001 was introduced.
Figure 9.7 shows the resulting time
Corresponds to stable
1.16
reactor period
PPT
1.12
X
3.0
1.08
132eo.on7c
Infinite delay time

2.0
1.041
Using:
dh = 0.001 sec
/8=0.00755
X = 0.076 sec'1 u= 0.001
1.00,
0.4
0.8
1.2
1.0,
20
_i
i
2.0
1.6
40
Time, sec
i i i i
2.4
i_
2.8
Time (, sec
Fig.
9.7
Flux transients due to addition of positive reactivity.
behavior of the flux according to (9.121).

It is of particular interest to
note that, within about 0.5 sec after the introduction of the reactivity,
the flux is rising essentially with the stable reactor period; that is (see
broken-line
curve in figure),
1.132e0
0117<
for
t>
0.5 sec
The insert in the figure shows the long-term behavior of the flux.
sec. 9.4]
KINETICS
REACTOR
577
Another approximation
of the delayed
d. Infinite Delay Time.
neutrons, which is sometimes used, is that of the "infinite delay time."
In this approximation it is assumed that the delayed neutrons, which
appear after the time when the reactivity was introduced, appear at a
constant rate taken to be the rate at which they appeared in the reactor
t = 0.
The differential equation for the time-dependent
behavior of the flux based on the one delay-group model and infinite
delay time is obtained from (9.118) by setting the precursor term equal
to its value prior to t 0. Since the present approach to the delayed-
prior to time
neutron problem will be most useful when dealing with transients in a

system formerly at steady state, the precursor concentration to be used
in (9.118) is the steady-state value from (9.115); thus
W(0
(FL)*(0+t^L- w)
\1
The solution to this equation is

(<)
= >o
a
0) p
exp
Wi - )
03
(9-122)
CO
/3
p~
(9.123)
This model would have some application in the early stages of the flux
rise after I = 0, and before the concentration of the delayed-neutron pre
cursors has been built up significantly over its steady-state value.
For
example, Eq. (9.122) is a good approximation to (9.121) when the tran
sient effects, represented by the second term, are significantly larger than
those due to the stable reactor period, i.e., when
which follows from the approximation
<pt
vHorjf
[cf.
ipse0"'
(9.124)
Eq. (9.121)]
for ait
<. 1
The solution for
a system with infinite delay time (9.123) has been applied

to the foregoing numerical example, and the results for 4>(t)/o are shown
Evidently this model gives a reasonably good representation
in Fig. 9.7.
of the transients in this example even up to 0.3 sec after the addition of
the excess reactivity.
9.4
Reactor Kinetics with Temperature
Dependence
a. General Equations for Stationary-fuel Reactor.

In the preceding
sections the primary objective was to establish the time-dependent rela
tions for the thermal (one-velocity) flux and to study various approxi
578
ANALYSIS
REACTOR
[CHAP.
mate models for describing the delayed neutron groups.

The purpose of
the present section is to explore with the aid of these relations some of
the elementary kinetics problems of stationary-fuel reactors taking into
account the effects of temperature.
The influence of temperature upon
the reactor kinetics stems from the temperature sensitivity of the nuclear
characteristics of the system.
Some of these effects were discussed in
Sec. 6.5.
For the purposes of the present study these effects are described by
means of a single parameter,
the temperature
coefficient of the reactor.
The dependence of the kinetic behavior on the temperature
coefficient
is due to the coupling between the flux level and the power generation.
It is clear that if for some reason there appears an increase in the flux
level this will result in a proportionate increase in the fission rate.
There
fore, an increase in flux gives rise to an increase in the power generation,
and this in turn disturbs the energy balance of the system.
As a conse
adjustment
there
must
an
in
be
the temperature level of the
quence
Since the nuclear characteristics of the reactor are temperature
Thus,
sensitive, a change in the flux level leads to a change in reactivity.
a realistic treatment of reactor kinetics must necessarily include this
coupling between the flux and the energy content of the system.
reactor.
The temperature-dependent
kinetics of stationary-fuel reactors is
treated on the basis of the one-velocity model.
The appropriate rela
tions for the time behavior of the thermal flux are given by (9.84) and
In addition to these we require a third equation which states the

(9.85).
If // (i,t) denotes the energy content per
energy balance for the system.
unit volume at the space point r and at time t, then the time rate of
change of H must be given by the net energy gain per unit time and
volume from the fission reactions and the reactor cooling.
The mathe
matical statement of this balance condition is1
jt
H(T,t)
2^(r,0
- (P(r,0
= average energy released per fission

macroscopic fission cross section
2/
*(r,0
thermal flux
energy loss per unit volume per unit time due to
<P(r,0
external
cooling
(radiation,
convection,
and/or
where
(9.125)
(9.126)
conduction)
It
is convenient
temperature
to express the energy content in terms of the reactor

a heat capacity c.
The following relation
T by introducing
is assumed:
H(t,t) = c[T(i,l)
1
ro(r,0]
- c0(r,i)
(9.127)
Properly, the cross sections should also be written as time-dependent functions,

but for the present treatment we assume that these time variations are negligible in
comparison to the rapid transients in the other functions (e.g., in
sec. 9.4]
where
KINETICS
REACTOR
c =
579
heat capacity of the reactor system (energy per unit volume
per unit temperature)

convenient reference temperature
The immediate objective is to obtain a solution of the reactor-kinetics
equations [(9.84) and (9.85)] which also satisfy the energy-balance rela
To(r,t)
m some
As in the previous problems in time dependence, we look

tion (9.125).
which
is separable in time and space.
This treatment is
for a solution
limited, therefore, to bare systems and to reflected systems which meet
On this basis we can use the
the requirements set forth in Sec. 9. Id.
relations (9.88) for the flux and the precursor concentrations.
tion, we select
- F(r) 0(0
- F(r)
e(r,<)
<P(r,0
CP(<)
In addi
(a)
(J-lzo>
(b)
The substitution of these relations, along with (9.88), into (9.84), (9.85),
and (9.125), yields
W(<)
= [(1
/3)fc(t)

+
C,(0
p]4>(t)
STW.
I/'
-X,C,(0 + wfiiZftffi
c0(<) = -<P(t)
+ 2,<K0
(0)
(9.129)
(&)
(c)
The first two equations are essentially the set (9.89), the only difference
being that we have included here the time dependence of the change in
multiplication &k(t). This dependence has been introduced since in the
present treatment the reactivity will be a temperature-sensitive function,
and the temperature in turn is coupled to the time-dependent neutron
relations.
In the analyses which follow, attention is focused on the transients

which arise when a given system initially at steady state (t < 0) is per
turbed in some prescribed fashion.
At t = 0, a disturbance is introduced
which affects the energy balance of the system, and the problem will be
to determine, in particular, the time behavior of the flux, temperature,
and power for all subsequent time (t > 0). The steady-state values of
the flux and precursor concentrations were previously indicated in (9.80)
The steady-state requirements on the power removal
[see also (9.101)].
can be obtained directly from (9.125); thus,
(P(r,0)
= 2,0^,0)
If
we use for determining (P(r,0) the expression (9.1286), and for

(9.88), we obtain the condition
(P(0)
2/4.(0)
(9.130)
<^(r,0),
(9.131)
which states that the steady-state value of the cooling (heat removal)
rate is proportional to the flux level.
It is convenient to introduce a
ANALYSIS
REACTOR
580
[CHAP. 9
in same
way
new function P(t) which is related to <( in the
theway
same
that (P(0) is
that (P(0)
is
The function P(t) will describe, then, the time behavior
related to <(0).
when the system is disturbed from its
of the reactor power generation
equilibrium.
Therefore let
P(t) = 62/0(0
(9.132)
Note that this function has the units of energy per unit time per unit
volume [see Eqs. (9.88) and (9.128)].
The transients in the reactor power production will be of special inter
est in the present study; therefore we write the kinetic equations (9.129)
in terms of the function P(t).
[(1
- f)]P(t)
f})5k{t)
-X.C,(0
C\(0
P{0 -
C0(O =
X WO
(a)
(9.133)*
+|ftP(0
(&)
<K0
(c)
set is not complete since the temperature
(and
is
applied to the reactor at
5Ar0)
(9.135)
shows that this linear dependence based on the constant
8fc(0
Sk0
y[T(t)
1.
An approximate expression
the product y6ko, which
is
if
reasonable
is
assumption
6T/T
be
obtained
from
this
equation
can
for 4fc(0
neglected
small in comparison to unity,
is
a
r,(0H
0
constant
- y[T(t) -
step change in
assumed to be
if
Experience
5fc(0
For
(9.134)
(1
Sk0
with
dependence
fc(0 =
k{T) = k[T(t)]
where the temperature
1
1
jfc
time) dependence of the multiplication A;(0 has not been specified.

the present case we take
r,(0] =
Sk0
- 70(0
(9.136)
is
It evident
given in (9.127).
where we have used the definition of
that
coefficient
of
the reactor
above
relation
the
temperature
from the
dependent on the sign of the constant y. The temperature coefficient
is
In this
at
case
if
negative.
the temperature coefficient
>
step change in multiplication, Sko >
given
0,
(9.137)
+ 8k(T))*
0,
~~
8k{T)Y
dT
Sk0)y
[1
k2
dT
(1
dk
[1
du
given by (6.144), and for the present case we obtain,
system
is 7
Thus
if
of reactivity
using (9.136),
is
is

In its present form this
2a
U>(t)
then
REACTOR KINETICS
sec. 9.4]
581
is,
the reactivity is initially positive and the system will "accelerate"; that
the flux level will rise. However, as the flux rises, the power output
and temperature will also rise. But because of the negative temperature
coefficient [see (9.136)], the reactivity will decrease from its initial value
a
Clearly,
and thereby retard the acceleration so as to stabilize the system.
negative temperature coefficient implies an inherently stable chainreacting system.
>
0
It will be helpful in exploring certain kinetics problems to use in place

of the basic set of equations (9.133) a somewhat simplified version based
In these cases
on the assumption of infinite delay time (see Sec. 9.3d).
we assume that at any time
the delayed neutrons which appear
t
0.
=
are those which were produced at the flux level existing before
This approximation has some practical value for analyzing situations
is
wherein the power disturbances in the reactor occur as rapid oscillations

and also for studying the time behavior of the system at very short time
introduced.
intervals after the initial disturbance
The steady-state production
of delayed neutrons may be obtained by

concentration
of the precursors.
From
X<C,-(0)
& P(0)
(9.138)
c0(t)
=
=
Sk(l) =
- W;(0 - P]P(t)
P(t)
- 70(0
Id
[1
l*Ht)
is
This expression
used as
first estimate for the precursor concentra
tions in (9.133a); then, the set (9.133) reduces to
+ Bk(t)]pP(P)
(a)
(6)
<P(l)
Sk0
(9.139)*
(c)
set will serve as the fundamental relations to be used in the study

of reactor power kinetics problems which include the effect of delayed
Note that in the present, somewhat generalized formulation
neutrons.
This
is
it
In order to
there are three unknown functions P(t), 0(<), and <P(t).
obtain a complete solution,
necessary to specify one of these func
tions.
It
most convenient to give the cooling rate (P(0 (the power
With this information and
removed per unit volume from the reactor).
the following set of initial conditions,
the formulation of the problem
is
is
<P(0)
(a)
(b)
T(0)
= To
dk
(c)
should be borne in mind that

given by y, since, from (9.135),
is
In using the relations (9.139) and (9.140)

the temperature coefficient of the system
(9.140)*
dk
dk
. ..
it
P(0)
0(0)
0(0)
complete:

computing the steady-state

(9.1336), we have
[CHAP. 9
ANALYSIS
REACTOR
582
The approximation is introduced because in most cases of interest &ko <SC 1 .

The derivative dk/dT may be computed by the methods outlined in
Sec. 6.5.
b. Constant
Power Removal without Delayed
Neutrons.
As a first
example of the application of Eqs. (9.139), consider a system which has

the following characteristics:
(1) constant power removal, <P(t) = (P0;
Assumption (2) is introduced only as
(2) no delayed neutrons, /3 = 0.
a simplification, but assumption (1) has some merit from the standpoint
In many problems the situations of interest
of a practical application.
will
be those which
involve small oscillations
of the flux (power) level,

but, because of the large heat capacity of the reactor, these small (and
relatively rapid) oscillations will have a negligible effect on the powerremoval rate. Thus the assumption of constant power removal is entirely
reasonable and should be useful for obtaining first estimates
dynamic response of actual systems.
The appropriate differential equations for this system are

c0(<)
P(t)
of the
(9.143)
<Po
should be recalled that P(l) is proportional to (t), which we previ

ously found to be of exponential form; therefore, we take, as a possible
solution [cf. Eq. (9.140a)]
It
P(0
(9.144)
<?oewW
and look for a function w(i) such that Eqs. (9.142) and (9.143) are satis
fied along with the requirement that w(0) = 0 [cf. assumption (1)].
The
substitution of (9.144) into (9.142) and (9.143) yields the nonlinear set
c<5)(0
= (P0[e"'("
w(t) =
[fc0
- 7(0Jr
(9.145)
= 0
(9.147)
1]
(9. 146)
with the initial conditions

w(0) = 0
6(0)
A general solution of Eqs. (9.145) and (9.146) will not
be presented here,
qualitative study being entirely adequate for this analysis.

We first reduce the pair of equations to a transcendental relation for
w(t) by differentiating
(9.146) and applying (9.145) :
w(t) =
- f 0)
- ^
[e<
1]
The integration of this equation yields
\ W)]*
[e<
w(t)}
+ C
sec. 9.4]
REACTOR
KINETICS
583
The constant of integration is determined by using the initial conditions

and (9.146), with the result that
1
7(Po
Thus
WW
/5ko\*
2\llh J
lihc
=
khC
[e-(
w(t)
_ ij
i^y
(9.148)
The solution of this nonlinear differential equation yields the function

w(t), and this result may be used to obtain @(<) from (9.145).
The
resulting functions would be found to have the form shown in Fig. 9.8.

Fig.
9.8
Constant power removal without delayed neutrons.
These functions represent the solutions for the case y > 0, which, as
previously noted, implies a negative temperature coefficient for the sys
tem.
This case describes a stable system that oscillates about the
steady-state condition existing prior to t = 0. The case y < 0 yields
a divergent
behavior
monotonically.
with time, the functions w(t) and
0(<) increasing
Some additional information can be obtained about the solution to this

problem without a detailed calculation of the actual functions.
In par
ticular, we can estimate the maximum temperature rise experienced by
the reactor, for a given rise in reactivity, 8k0.
The peak value of the
function &(t) may be computed from the condition 0(0 = 0. This con
dition requires that w(t) = 0 [see (9.145)]; thus we must determine the
roots of this equation.
The slope of the function w(t) at its zeros is
obtained directly from (9.148); these are
w
It
is evident from Fig. 9.8 that the positive root corresponds to t = 0

(and the corresponding cyclic values of t) and that the negative root
corresponds to the position of m.
If t0 denotes the value of t at which
the maximum temperature values occur, then the magnitude of the tem
perature rise is obtained from (9.146):
em.
= 0(o)
(9.149)
REACTOR
584
ANALYSIS
[CHAP.
This result is generally valid, its only limitations are that the power
be constant and that the delayed-neutron
contributions be
removal
negligible.
c. Linearized Solution for Constant Power Removal.

It is of interest
to examine the problem of constant power removal under more restric
A practical situation which can be easily examined in
tive conditions.
For this problem the study would
detail is the case of small oscillations.
limited to small disturbances &k0 on the steady-state operation of the

reactor.
Under these circumstances the problem can be linearized, and
closed-form solutions can be readily obtained.
We assume, therefore,
that the excursions from steady state, which are represented by the func
be
In this
tion w(t), are small [see (9.144)].

power production by the relation
P(i) ~
case we may approximate
+ w(0]
($>o[l
the
(9.150)
The differential equations for this problem are

6(<)
~ **E
w(t)
~^
[Sk0
- 70(0]
(9.151)
(9.147) as before. This set of simultaneous

equations may be solved by the method of Laplace transforms.
If W(s)
and 3(s) are the Laplace transforms of w(t) and 0(2), respectively [see
(6.22)], then the Laplace transforms of Eqs. (9.151) are
...
WW
sW(s)
[tt.
73(s)]
(9.1538)
which have the solutions

W{8)
U(s* +
7(Po/i,hC)
3(s) =
cUs(s*
Ttô/^c)
(9-153)
The inversion of these equations yields
These functions, along with Pit), are shown in Fig. 9.9.

Note that the
maximum temperature rise
indeed, 2(5A-0/7), as previously predicted.
The solutions (9.154) are for the case
If had been taken to be
>
negative, the appropriate solutions would be
w
0.
is,

REACTOR KINETICS
sec. 9.4]
585
Clearly ,/ these functions

where the absolute magnitude of y is implied.
with
these
solutions could apply
time, although
increase monotonically
=
0.
only for short times after t
d. Constant Power Removal with Delayed Neutrons.
The effect of
the delayed neutrons upon the power kinetics of a reactor may be demon
Consider a bare homogene
strated by means of the following example.
ous thermal reactor with constant power removal operating at steady

Fio.
9.9
Linearized solutions for constant power removal and no delayed neutrons.
state up until time
At
= 0.
t = 0
some excess multiplication Sk0 is

The problem is to determine the tem
added uniformly to the system.

perature and power output of the system for all subsequent time.
this calculation we introduce the following model:
(1)
One group of delayed neutrons with infinite delay time
(2)
<P(t)
(3)
Sk(t)
<P
5fc
-P{0)
7(0
For
(9.155)
The appropriate
kinetics equations for this system are given by (9.139).

If we bear in mind that ^
small '8k0 and ly,
1, then for sufficiently
and
In
unity.
are
also
less
than
this case (9.139)
appreciably
fiSko
7
reduces to the set [where we have used also (9.155)]
P(t)
[ifc(0
6(0
~[P(t)
- ^+
(9.156)
0]
(9.157)
<Po]

P(0)
(P0
0(0)
0(0)
= 0
(9.158)
As in the preceding examples, we seek a solution for the reactor power

which can be written in the form (9.144). In terms of the power func
REACTOR ANALYSIS
586
[CHAP. 9
tion w(t), Eqs. (9.156) and (9.157) take the form

i*(0
0(0
- 70(0
^[e<
r
1
[8k0
-P]+
fl#>->
(9.159)
(9.160)
1]
The essential features of the present example may be amply demonstrated

by means of the linearized solution.
Therefore, using the approximation
~
1 + w [see also (9.150)], the above relations reduce to
e"
w(t)
~ jtth
[6k0
70(<)
- 0w(fl]
0(0~pw(0
(9.162)
the parameter
V
-y
(9.163)
The pair (9.161) and (9.162) may be solved by differentiating (9.161)

and applying (9.162) to eliminate the 0 term.
The resulting expression
for w is
tf(0 +
r
tth
*(0
F w(0
tth
= 0
(9.164)
The initial conditions for this equation are

w(0) = 0
w(0) =
Me
(9.165)
It is of some interest to note

where we have applied (9.158) and (9.161).
that the linearized equation for this reactor problem has the form of the
familiar equation for an oscillating system with viscous damping.1
A
direct comparison
to a mechanical dynamics problem reveals that the

parameter y has the role of the "spring constant,"
temperature-coefficient
it
"stiffness" of the system. Thus a large tem

perature coefficient implies a rapidly responding system, and higher fre
Note that the power
quencies of oscillation, following a disturbance.
parameter p also appears in this term; since the heat capacity c is in
the denominator of this quantity, it follows that systems with large heat
capacities are "soft" systems, i.e., respond slowly to excitation and
Finally, we find that the
experience oscillations of lower frequencies.
Thus the presence
viscous damping term (?i>) involves the parameter
i.e.,
is a measure of the
(3.

(9.161)
'See, for example, Theodore v. Karman and Maurice A. Biot, "Mathematical

Methods in Engineering," pp. 130-132, McGraw-Hill Book Company, Inc., New
York,
1940.
sec. 9.4]
KINETICS
REACTOR
587
of delayed neutrons in the system serves to damp oscillations arising from

This role of the delayed groups was mentioned previously.
a disturbance.
The solution to Eq. (9.1(54) with initial conditions (9.165) is easily

obtained by the usual methods (see, for example, Sec. 9.4c).
This result,
along with the corresponding solution for 0(<) from (9.162), is found to be
\pt
where
(|
sin
py
2Z,h
The expressions (9.166) correspond to the
cos
- v!
(9.167)
for
case
Since the
0.
>
~ e~*'
yf/2
0)
sin
(9.166)
e~xl
= {5ka\
w(0
12r
"\
= 400
yr,
2.3
0=0. 00755
(P0/c
0.8
1.0
t,
Time
1.2
pc = 02/4/,7
I
0.6
.001
0.4
. r
= 0,
i*
50^-^
&
0.2
1.4
1.6
1.8
2.0
sec
Fig. 9.10 Temperature for constant power removal and one of group delayed neu
trons with infinite delay time.
of
0,
>
may take on all values
parameter
may be less than, equal to,
or greater than zero.
For each of these cases the solutions (9.166) have
a particular form.
The general results are well demonstrated by means
5fc0
of these results follows.
>
lo-yc
Consider
This corresponds to
(1)
= 0.001
=
Vc
first the case
(9.168)
^2
discussion
= 0.00755
= o.ooi sec
u0
numerical example.
Figures 9.10 and 9.11 show the temperature
difference
and the puwer function w(t) for various values of
obtained
for the case

10
> 0.
REACTOR ANALYSIS
588
where
pc
is the value of p which yields

Pc
\p
[chap. 9
= 0, i.e.,
(9.169)
The solutions for this case have the form of damped oscillations.
Thus the initial disturbance due to the introduction of Sko causes a

0.14,
-0.02,
0.6
0.8
1.0
1.2
1.8
2.0
Time (, sec
Fig. 9.11 Power function for constant power removal and one of group delayed
neutrons with infinite delay time.
rapid change in both the temperature and power output of the reactor.
The system stabilizes rapidly, however, and the oscillations decay to
It is obvious from (9.166)
negligible magnitudes within a few cycles.
stabilizes
at
eventually
that the reactor power,
(P0, as is to be expected,
and the temperature difference 0 at Sk0/y; these arguments follow from
the fact that
lim w(t) = 0
(-
lim 0() =
Sk0
(9.170)
0 is greater than zero, the system evidently

new higher (for positive Sk0) average temperature.
This
Since the limiting value of
stabilizes at a
is a consequence of the negative temperature coefficient of the system.
Note that the introduction of positive reactivity causes the power pro
duction to increase, this causes a rise in the temperature since the rate
REACTOR KINETICS
sec. 9.4]
589
of power removal is constant; but, as the temperature rises, the reactivity

of the system begins to decrease and very shortly thereafter becomes
The power level then begins to drop rapidly, and with it the
negative.
But the system overcompensates, causing the temperature

to drop low enough to produce a positive reactivity. There is then an
Eventually, the system reaches
upswing in both temperature and power.
condition
wherein
the
excess
multiplication Sk0 is exactly
an equilibrium
compensated by an increase in the temperature level [given by (9.170)].
Another interesting feature of this system is the effect of the delayed
It was previously mentioned that delayed neutrons cause a
neutrons.
More precisely, in the present
damping of the oscillations in a system.
model based on the infinite delay time, the reactor behaves like a subcritical assembly without delays which is maintained at steady state by
a constant external source (v/3(Po/e neutrons per unit volume per unit
By assumption, w(t) <K 1; therefore the present results apply to
time).
very small changes in multiplication, and the functions (9.166) are valid
only for very short time periods after t = 0 (cf. Fig. 9.7).
temperature.
(2)

In
V<
this case >p = 0, and Eqs. (9.166) reduce to the form
0)
0 [1
7
- tr{xt +
(9.171)
1)]
The limiting values of these functions are seen to be the same as those
This case corresponds to a system with "critical
given in (9.170).
damping." As the value of p decreases toward pc, the magnitude of the
oscillations in both temperature and power output diminishes until finally
at p = pc the oscillations disappear altogether (see Figs. 9.10 and 9.11).
It is helpful in understanding this behavior if one imagines the reduction
in the parameter p to be due to an increase in the heat capacity of the
Clearly, for a given disturbance 6fc0, the larger the

system [see (9.163)].
heat capacity, the less sensitive the system will be to changes in the power
Thus large changes in power (i.e., reactivity) are more easily
level.
absorbed and the temperature rise required in order to accomplish this is
At p = pc the heat capacity of the system
correspondingly smaller.
is sufficiently
large to accommodate
the change in reactivity without
"overshooting."
P < Pc
(3)
Under these circumstances
^2
< 0 and may be more conveniently
written
in the form
+
iS
with
S1
= x- ~
(9.172)
REACTOR ANALYSIS
590
The solutions for this
and x denned as before by (9.167).
case are
1 5k0\ erxl sii

sinh St
w(t)
e(<)
[CHAP.
S
=
r
1
sinh
S<
+ cosh
\i
(9ll73)
As in the preceding case, these solutions are nonoscillatory and the sys
This corresponds to very large values of the heat
tem is "overdamped."
In these cases the temperature and power are relatively loosely
capacity.
coupled functions and respond very slowly to reactivity changes (compare
curves labeled p = 50 to those labeled p = 142.3 in Figs. 9.10 and 9.11).
p = 0
(4)
This

primarily of academic interest.

model, the solutions are found to be
case is
!*! (i
,(()
0(0
= 0
On the basis of the present
e-*/th)
(9.174)
Since this case corresponds to a system with infinite heat capacity, it is

to be expected that the temperature will be unaffected by changes in
reactivity (i.e., power level).
under these circumstances the

negative temperature coefficient cannot be of aid in stabilizing the sys
tem and one would expect the power level to rise without limit.
Since
However,
the result given above reaches a finite value it is clear that this elementary
model breaks down when applied to this case; however, note that for
short times after the addition of 5k0 (i.e., fit/h* <K 1) the power function
has the correct form.
ip(<)
In this
(j^jt
case
and
P(O~Poexp0^
(9.175)
Thus the power level rises exponentially.

The various cases discussed above are shown graphically in Figs. 9.10
Again it is pointed out that these
and 9.11 for the data given in (9.168).
results are meaningful only during the short time period following t 0.
9.5
Kinetics of Circulating-fuel Reactors
The influence of the delayed neutrons and of the flux-temperature

coupling on the dynamical characteristics of a stationary-fuel reactor
These two factors are also
was demonstrated in the preceding sections.
In the
essential to the dynamical behavior of circulating-fuel systems.
present section we treat several aspects of the general problem and in
particular show how the influence of the delayed neutrons and of the
Since the
temperature coupling is altered by the circulation of the fuel.
sec. 9.5]
REACTOR
KINETICS
591
general problem is very complex, no attempt is made here to provide a

thorough study, and most of the discussion is based on relatively crude
models of both the nuclear and fluid dynamical features of the system.
Although this approach reduces considerably the analytical difficulties
in the calculations, it also tends to render the final results less useful
for computational purposes.
However, these models do allow a clear
exhibition of the essential features of the kinetics of circulating-fuel
reactors; moreover, some of the analytical techniques developed in this
treatment are sufficiently general that they may be applied to the analysis
of more complicated problems.
In any case the general conclusions and
results obtained with the simplified models provide a correct qualitative
description of these systems.
All the analyses which follow are based on two fundamental assump
tions, namely, that the neutron population can be well described by
means of the one-velocity model and that the circulation of the fuel can
The nuclear features of these systems,

then, may be developed from the general results obtained in Sec. 9.4.
The assumption of slug flow consists, essentially, of limiting the trans
port term in the fluid-flow relations to variations in the direction of flow
Thus the flow of mass (and energy) around the fuel circuit occurs
only.

be represented by a slug flow.1
along parallel channels or tubes, and no communication (cross-flow) is

allowed between adjacent channels.
This assumption greatly simplifies
the fluid dynamical aspects of the problem and thereby permits, with a
minimum of effort, a clear display of the coupling between the fluid-flow
and nuclear characteristics
of the system.
In the present treatment the effect of the delayed neutrons and the
effect of temperature-flux coupling on the reactor kinetics are handled
separately along the general lines previously developed for the stationaryAlthough in general these two factors must be considered
fuel systems.
jointly, there are many practical situations wherein the two effects can be
The intent here is primarily to simplify the analysis so as
separated.
to focus attention on one feature at a time.
a. Low-power Kinetics with Delayed Neutrons.
The effect of delayed
neutrons on the kinetic behavior of circulating-fuel reactors is studied
with the aid of an analytical model which describes a low-power thermal
reactor.
In the present discussion the expression low power
is used to
denote systems wherein the power level is sufficiently low that small
variations in it produce negligible changes in the temperature of the
nuclear constituents of the reactor.
Thus the energy balance of the
system is not considered, and no coupling is allowed between the tempera
ture and the neutron flux.
For the present, then, we ignore entirely
temperature coefficients of reactivity and focus attention on the role of
The specific objective is to derive the inhour
the delayed neutrons.
1 The expression
flow"
is used to denote a one-dimensional flow field.
"slug
REACTOR ANALYSIS
592
[CHAP.
equation for the circulating-fuel system which may be written in the

form of (9.98) obtained for the stationary-fuel reactor.
The effect of
the fuel circulation on the reactivity is then easily demonstrated.
The
analytical technique applied in this study follows the general scheme
used
by
J. A.
Fleck.1
For the sake of simplicity, the calculation
is
Somewhat more general treatments

specialized to the one-velocity case.
based on the two-group model have been obtained by Fleck and are
presented in the reference reports.

For the purposes of this analysis we assume that the reactor core is a
bare right circular cylinder of radius p0 and height h. The circulatingfuel circuit is a closed loop with the same cross-sectional area as the core.
The length of the circuit external to the core is h and the length of the
By definition, fissioning takes place
complete circuit (h + he) is A*.
only in the core region of the circuit. Although the assumption of a
uniform cross-sectional area for the entire circuit may introduce some
conceptual difficulties in establishing which portion of the circuit should
properly be termed the "core," it leads to no loss in the generality of the
results that are obtained since only the transit time 0e through the
It is clear that if a different
external loop is essential to the calculation.
cross-sectional
area
were
to
the external circuit so as
assigned
average
better to represent the piping and heat exchangers, etc., in the loop, this
would merely result in an average flow velocity through that portion
different from that through the core. This effect is easily accounted for
by adjusting the transit time
In this model we further assume that the hydrodynamics of the
system is adequately represented by the relations for slug flow and that
the neutron physics may be described by means of the one-velocity
Finally, we assume also that the fuel in the core is homogeneous
model.
and that the neutron macroscopic cross sections are time independent.
On the basis of these assumptions we can write the following pair of
differential equations to describe the circulating-fuel reactor:
-DV**(r,0 + S.0(r,O
- ~ *(r,0 + (1
jt
C,(r,l) + V
| CM
+ V
Ci(t,t)
^ C&fi
= W3.2,*(M)
=
-XA(r,<)
0) vZ/Pthgtl,(r,<)
Pt<7.h
X,C,(r,<)
V X,C,(r,0
(9.84)
0 < z < h
(9.176)
h<z(9.177)
(9.177)
1 J. A. Fleck, Jr., "Theory of Low-power Kinetics of Circulating Fuel Reactors with

Several Groups of Delayed Neutrons," Brookhaven National Laboratory, BNL 334
(T-57), April, 1955; "The Theory of Circulating Reactor Kinetics at lx>w Power,"
BNL 1933, Apr. 13, 1954; also, Kinetics of Circulating Reactors at Low Power,
Nucleonics, 12, 11 (1954).
sec. 9.5]
KINETICS
REACTOR
593
where V is the velocity of the circulating fuel and the coordinate z is the
distance around the circuit measured from the inlet to the core. All
the remaining
have the previously assigned definitions (see

equation for the precursor concentration is
9.3).
obtained from the relation for a stationary-fuel system (9.85) by the
symbols
The differential
Sec.
addition of
transport (or convective)
term.
Note that Eqs. (9.176)
forms
of
and (9.177) are merely special
the usual continuity relation of
fluid mechanics.1
a
with
We begin the calculation
of the flux equation to a

to the present problem; therefore let us define
simpler form appropriate

4>(t,t)
where
m R(P)
\pl
(II
\p
dp
dp
(9.84)
R(p)
<p(z,t)
R(P)
dt
S,(z,)
(9.178)
The substitution of
yields
+ R(p)
<r(z,t)
dz
(1
<p(z,t) \
+ ZaR(p)
- p)vX,plbgthR(p)
+ p*g,bR(p)
)
^
i
<p(z,t)
<p{z,t)
\iSi(z,t)
KSi(z,t
(9.179)
= 0
(9-180)
By introducing the relation
<p{z,t)
B^(2,oJ
= vXfpth/Xa,
+ [fc.0.(l
+
add and subtract
L2B2<p
where
M2
B2 =
m2
we call
B\
+
If
(9.179) may be
- -
1]
R(p~o)
P)
hft<f{z,0
fcM
*(*,<)
A.Si(z,0
= D/2, and
along with l0 = l/V2a,
further reduced to the form
(9.181)
^jY
B*R{p)
Jp [p Jp R{p)] +
~P
[ip
(9.182)
from the right-hand side of (9.181), and finally
'See, for example, L. Prandtl and (). G. Tietjens, "Fundamentals of HydroAeromechanics," pp. 100-104, McGraw-Hill Book Company, Inc., New York, 1934;
S. Goldstein, "Modern Developments in Fluid Mechanics," vol.
chap. Ill, Oxford
University Press, New York, 1938.
I,

C,(r,0 = R(p)
<p{z,t)
denotes the radial coordinate in the core.
these expressions into
_D
a reduction
ANALYSIS
REACTOR
594
divide through by
l*! *(s,f)
(1
(9.181) may be written
ZAB2),
*>(*,<)
[CHAP.
mV(2,o]
-/*)*-
+ [(i
fl*(,o
where we have used k = kxgth(l + JJB2)-1, k = 8k + 1, Eq. (9.35) for

Zlh, and (5.220) for pL.
We consider next the precursor equations (9.176) and (9.177).
The
application of (9.178) reduces these to
Sfat)
Iat
+ V%- Si(z,t)
oz
I- Si(z,t)
+ V
|- S.(z,0
= v2fPi<p(z,t)
- \iSi(z,t)
-X,S,(z,0
0<z(9.184)
(9.184)
h<z(9.185)
(9.185)
<
<
(a)
dff
The
v2/^(z,<).
x.(0,s)
<
*(y,8)e"+x->"'
P(z,t) ^
(b)
of P(z,t) and
< h*
<
e-<"+*.'"-
(9.186)
(9.187)
is
the transform
where *(z,s)
solutions to these equations are
x,(z,8) =
Xi(z,s)
=0
<
*(*,)
*(*>*)
f1^)
(S
(*,)
y^)
jz *(,)
+
+
(*,)
Si(z,t)e dt
of Eqs. (9.184) and (9.185) are
then the transforms
jQ"
{Si(z,t)\ =
transform.
where
<
< h*
(9.189)
(9.190)
= x<(fc,s)e-(*+Xi)(''F-*)
(9.188)
X.(2,s)
The constants of integration

the transit time through the core.
x,(0,s) and x<(h,s) are obtained by the application of continuity require
=
Thus Eq. (9.188) must equal Eq. (9.189) at
and Eq.
ments.
h,
is
=
= fc*.
must equal Eq. (9.189) evaluated at
(9.188) evaluated at
The resulting expressions for the constants when used in (9.188) and
(9.189) yield the solutions
z

These equations may be solved with the aid of the Laplace

If we take *S(z,0) sa 0 and call
REACTOR
^SSSi
for
dy
*(^,*)e('+x-)"
<
< /i*
(9.191)
(9.192)
(9.193)
the transit time through the external loop.

&e
The solution for the precursor concentrations Si(z,t)
is
t>b
<
(9.194)
<
when
b)
The second term in
=/()-**
0
\W)\
Fb(t) =
= F(t
where
is
obtained by the
These are easily determined by
inversion of Eqs. (9.191) and (9.192).

the application of the transform
and the expression (9.192) may be cast into

form by using the expansion
the appropriate
(9.191)
71=
in the denominator
In this way
it
0.
less than
may be shown that
is
is
valid since the exponential

unity by virtue of the fact that >
which
S.(z,0
vS/fr [f0Z/V f(z ~ Vy,
the inversions of (9.191) and (9.192) are
- y)e-^dy
71=0
Vy,
- ntf* for
<z(9.195)
i/)e^."
<
^_
e~M*
&(*,) = v2,ft
forO
so
(9.195)
fy
< h*
(9.196)
0,
m=
^)Z(Z)
4>m(
with
Zm(z)
{
V
>
,(!,*)3,0
is
is
This expression for Si(z,i)

to be solved simultaneously with the rela
tion for the neutron flux (9.183).
The solution
obtained by expanding
the flux <p(z,t) in a Fourier series (appropriate to the boundary condition
>
in
thus, for
z)

<
denotes the transit time around the complete circuit, i.e.,
The symbol
where
<
for
*{Vy,)eP***' dy
Z^SS^
(Vy,s)e<'+^ dy
&
X.(z,)
595
J*
KINETICS
= /3,-e-M-WF
^"
x,(z,a)
J*/V
SEC. 9.5]
lsin^z
= xm/h
(9.197)
REACTOR ANALYSIS
596
[CHAP. 9
suitable expressions for the Fourier coefficients 0m(<).

The substitution of (9.197) into (9.195) and (9.183), followed by the
multiplication of (9.183) by Zj(z) and integration over the interval (0,/i)
yields an equation for 4>j(t).
and developing
cMi(t)
c,fc)
[(1
P)&k
m 1
BO
- tf)
Ox*
<
e-V*
Zy() dz
- Vy)er^ dy\
Vdi,
n&* + Vdi,y)Zm(z
<M - n&* - y)Zm(z
- Vy)er^ dy\ (9.198)
J'/v+a.
(9.198)
If
= m
(9.199)
in (9.198) and call
Zj(z) Zm(z + Vd*
MUV)
^ m
|o*
Zy(z) Zm(z) dz ^
c,-
|Q*
Z,(z) Zn(z) dz =
we invert the order of integration
Vy)
dz
(9.200)
jo'
/s]J
y^1^6
^" dy
c-^**-
('
m(t
nt>*
- ndi, -
Mjm((/)e"x."
dj/
Njm(y)e-*-" dy
(9.201)
J#
n-0
i/)
- fi)sk -
+ [(i
y)
mj)
(m2
,)
fc
hMt)
(gf
An adequate approximation to this equation

obtained by retaining the
first term only in the scries (9.197); thus neglecting all terms for
>
and dropping
the subscripts
on
is

Note that,
Lu, Mn, and Nn, (9.201) becomes
M{S) =
{|o'
L(y)e->"* rfy
f*+'
AT
are obtained from Eqs. (9.200)
>*
<
<
the integrations
(9.203)
ir
- sin irf
(9.202)
tffo)e-**(to]}
;
*(
irf
cos
M, and
f)
?)"(7-r)
(f
US)
(l
The kernels
yield
L,
n-0
*(t
2/)
0,X,
fc
- P)5k -
= *(<)[(!
<K0
0]
mi
597
/*)
(note that
KINETICS
REACTOR
2/)
sec. 9.5]
;*'('-$)
+
??) ('"t)
+
<
<
= 2//#It may be seen that these functions

where
related; thus AT (J-) = L[f
(#*/>)],
1.0
a "reflection" of L(f)
and il/(f)
Figure 9. 12 shows
about =
V
\\
\\
0.4
appropriate to the circu

sinusoidal
lating-fuel reactor and
0.6
\\
complicated expressions, we apply

some rough approximations
that
will suffice for the present purpose
and at the same time simplify the
we introduce
0.2
0.4
0.6
0.8
Fig. 9.12 Comparison of kernel

linear aPProximation-
1.0
and
the following linear expressions for the
(9.206)
- MiS) NiS)
US)
0.2
axial variation in the neutron flux.

However, instead of using these
kernels
\\
equation
Thus
(9.205)
are all closely
through
(9.203)
in
(9.202) to
(9.205) may be used
obtain an expression for the inhour
calculation.
f.
approximation
results
The
\\\\
is
7C
'
sin ir

iV(f)
(9.204)
REACTOR ANALYSIS
598
[CHAP.
These approximations give reasonably good estimates of the set (9.203)

through (9.205) ; note that the values of these functions at the limits of
their respective intervals are properly matched and that the average
value of each comes within 20 per cent of that for the corresponding
exact function. This is the relevant comparison since the exponential
weighting factors e~XiV which appear in (9.202) are slowly varying over
the transit period # for most systems of practical interest.
The inhour
from
by
is
obtained
the
introducing
relation
expression
(9.202)
(/>(<)
= oeat
(9.207)
If
we use the set (9.206), then it may be shown, after some algebra, that
(9.202) reduces to the form
u.
- - m - * + *Z * [. 
o>,
- e-*'*Q(l - e-M)1
~
*7(1
"
e-M*)
(9-210)
The comparison of this expression with (9.98) reveals the influence of

This effect is seen more
the circulation on the reactivity of the system.
clearly after a rearrangement of the exponential factors in the sum.
We
first show, however, that this sum may be written in terms of the frac
tion of the total number of precursors in the circuit which may be found
For this purpose we require the total number of precursors

These are obtained from (9. 195) and (9. 196) ;
in the core and in the loop.
for the case of a flat axial distribution in power, <p(z,t) = >oe" and the
in the core.
precursor relations become
*(*.0
-(HSi)
7,.(f-t^l)
11
- e-y'"r
r^*^""-*'
1])
0 < z < h
~ l>ry'"V\
h<z(9.211)
(9.211)
of the first of these over the interval {0,h) and the inte
gration of the second over {h,h*) give the total number of precursors in
The fraction a, of precursors of type i in
core and loop, respectively.
the core is computed from
The integration
-'-868
then
a.() =
* i-***^
(
^-(1
- U)
WW
0-213)
KINETICS
REACTOR
SEC. 9.5]
599
Of particular interest will be the fraction of precursors in the core at

steady state; in that case we have, as a 0,
CI
and when
X,t?*
<3C 1,
a<(0)
p-M,) (i
e-W)
(9.215)

Thus for transit times appreciably less than the precursor mean life (see
Table 9.1), the fraction in the core is given approximately by the fraction
0.4
FlO. 9.13
0.6 0.81
8 10
Fraction of precursors in core of circulating-fuel
20
40
60 80 100
reactor at steady state.
of the total circuit time spent in the core.

It is also of interest to note
that for all values of a, when d > ao (i.e., no circulation),

lim a,(a) =
I?*so
(9.216)
Figure 9.13 shows the fraction of precursors in the

core at steady state as a function of \{d for various
It
that Eq. (9.213) for o,(a) may be

expression (9.210) for the reactivity; thus we obtain
is evident
co
IthOC
=
1
hwa
used
TTLraZ^[yi-XMa)]
to reduce the
(9-217)
This result has several interesting features. Consider first the reactivity
For
required to maintain the circulating-fuel system at steady state.
this condition a > 0, and (9.217) takes the form
codrc
(9.218)
600
ANALYSIS
REACTOR
[chap. 9
where a,(0) may be obtained from (9.214) or Fig. 9.13. Note that, when
o
# and de
This agrees with the
(no circulation), u > 0 by (9.218).
result for the stationary-fuel reactor (9.98).
Also, in the case of infinite
= f = constant and
circulation velocity V, d and 0e > 0. If we let
take the limit as
0, (9.218) yields
lim
Thus
o,olrc
= 0
(^
(9.219)
that to maintain a circulating-fuel reactor at steady state

reactivity varying from 0 for the stationary-fuel
system to a maximum of /S#e/tf* for the system with infinite circulation
It is obvious from these observations that, if a circulating-fuel
velocity.
reactor is at steady state at some given circulation velocity, a sudden
reduction in this velocity will cause the reactor to become supercritical
since a smaller amount of positive reactivity is required to maintain the
system at steady state at the lower velocity than at the higher.
Another interesting conclusion which may be drawn is that the kinetic
behavior of a circulating-fuel reactor (within the limits of the present
model) may be described in terms of an equivalent stationary-fuel sys
tem in which the fractions of the delayed neutrons Pfn <
the actual
To demonstrate this we determine the "excess reac
delay fractions.
tivity" for the circulating-fuel system. By this we mean the reactivity
over and above that required for criticality at any specified circulation
This is the quantity w wcirc. From (9.217) and (9.218) we
velocity.
we find
w toeiro =
^*h
obtain
[7,(1 +
Uha)ai(0)
\&i(a)
l*ay,]
(9.220)
It
is
If
eM
h^j
\idt
+ eM, _
-^_ X,i?*
X,t?
hi
where
~ o,(0)
l
a,()
(l
is
is
to evaluate this expression for the case wherein the

X,.
slow period, i.e., when a
on
For this computation
first-order approximation to a, (a).
we use
The appropriate form
easily shown to be
of interest
reactor
we
o>,,
a>
h
,
is
further impose the condition that X,d*

[cf. Eq. (9.215)], then
d/d*) both this quantity and /ttXi are much less than
^t?*X-(l
unity. Thus to zeroth-order approximation, the expression within the
bracket of the sum (9.220)
The result (9.220) then reduces to
aOi(O).
h(

we require some positive
(9.221)*
REACTOR KINETICS
SEC. 9.5]
601
>
is
it
is
/3,
This result1 is to be compared to (9.98). We find, then, that the circu

lating-fuel system responds like a stationary-fuel reactor with the delay
Thus the circulating system
fraction
the
replaced by a,(0)j3, < ft.
more sensitive to reactivity changes. Note that the proper correlation
made in terms of the so-called "excess reactivity."
follows
Also,
from the above result, and (9.216), that as & , (9.221) reduces to
(9.98).
multiplication constant of
Consider the multiplication fcoiro
we can estimate the effect on the
change in the circulation velocity.

required to maintain the system at steady state when the fuel
From (9.218),
lating at some velocity V.
is
Finally,
circu
*-
<w
a
Thus, as previously recognized,

where we have used (9.215) for o,(0).
decrease in circulation velocity causes an increase in the multiplication.
b.
Flux-temperature Coupling without Delayed Neutrons.
The effect
coupling upon the kinetics of circulating-fuel reactors

studied by means of
model wherein all fission neutrons are taken to
As in the analysis of the low-power systems,
be prompt.
assumed
that the one-velocity model applies and that the circulation of the fuel
well represented by slug flow.
is
is
it
is
of flux-temperature
J.
is
similar to that applied in the preceding

The general line of attack
A. Fleck.2 The problem of
analysis and follows the approach used by
temperature-dependent kinetics has received the attention also of Ergen,3
Weinberg,4 and Welton,6 who formulated the problem in terms of equa
tions which involve integral operations on the past history of the neutron
flux.
This approach leads directly to a set of equations in terms of the
ORNL CF
is
An alternative derivation
given by W. K. Ergen, "The Inhour Formula for a
Circulating-fuel Nuclear Reactor with Slug Flow," Oak Ridge National Laboratory,
53-12-108,
Dec. 22, 1953.
J.
A. Fleck, Jr., "The Temperature-de pendent Kinetics of Circulating Fuel

Reactors," Brookhaven National Laboratory, BNL 357 (T-65), July, 1955.
W. K. Ergen, Kinetics of the Circulating-fuel Nuclear Reactor, J. Appl. Phys.,
26(6), 702-711 (June, 1954).
W. K. Ergen and A. M. Weinberg, Some Aspects of Nonlinear Reactor Dynamics,
Physica, 20, 413 (1954).
T. A. Welton, "Kinetics of Stationary Reactor Systems," Proc. Intern. Conf.
P/610, pp. 377-388.
Atomic Energy, Geneva, 1955, vol.
Peaceful Uses
5,
of

- - r-i? - >
co
follows that
and
it
(9.222)
REACTOR ANALYSIS
602
[chap. 9
time variable alone.

The method used by Fleck begins with the usual
partial differential equations (in space and time) for both the neu
tron flux and for the energy balance.
It is shown that by expanding the
set of
for the flux in a Fourier series the resulting expressions for the
time-dependent Fourier coefficients are in fact the relations used by the
authors mentioned above.
Thus the present method offers a somewhat
more general approach and has the added advantage in that it contains
solution
implicitly the spatial dependence of the neutron flux and temperature

permits solutions for space-dependent perturbations and initial
and
conditions.
V,
All fluid particles and heat are transported around the circuit at
Po-
uniform velocity
is
a
is,
a
The calculation is carried out for an open-loop circuit, that

circuit
in which the entrance conditions to the reactor core are constant in time.
This implies that the external loop removes all the heat generated in the
The core proper
right circular cylinder of height
core.
and radius
is
it
and
assumed that thermal equilibrium and
changes occur instantaneously with variations in temperature.
assumed that no lateral exchange of heat occurs between
Moreover,
taken to be a line from the
adjacent fluid streamlines; i.e., the reactor
is
is
it
density
Thus all axial paths through
the core are identical.

On this basis the appropriate
flux
and the energy content
equations for the neutron

(per unit volume) of the fuel are seen
differential
be1
H(T,t) +
--
kZ/Pu.^M
*(r,f)
2o<Kr,0
-DV2<Kr,0
jz
to
at
= S^(r,<)
H(r,t)
(9.224)
(9.225)
mV(z,<)
1M2,<)
(9.226)*
the definitions
H(T,t)
vol.
H{z,t) = c[T(z,t)
Energy transport by conduction

II, chap. XIV.
is
to be represented by
Since we have assumed

(9.182).
line reactor, the heat generation
and axial posi
and therefore temperature
are functions only of time
Then, after (9.127),
tion
(measured from entrance to core).
that the system
is
where we have used
(k
<p(z,t)
is
the energy release per fission.

These two equations are readily
reduced to functional relations in and by the application of the assump
By following the pro
tions listed above and the relation (9.178) for
cedure outlined between Eqs. (9.178) and (9.184), Eq. (9.224) may be
reduced to the form [cf. Eq. (9.183)]
where

standpoint of heat generation in the fuel.
7'(0)]
- cV(z,t)
neglected; for reference
see
(9.227)
Goldstein, op. cit.,
sec. 9.5]
KINETICS
REACTOR
603
where To(0) denotes the temperature at the inlet to the core and c is the
heat capacity of the fuel.
Equation (9.225) may now be written
fat
0(z,t)
|az
0(z,O + V
' <p(z,t)
0<z(9.228)
(9.228)
This equation is to be solved simultaneously with (9.226) and

For the latter we take
expression for the multiplication.
8k = Sk0
a suitable
- yQ(z,t)
(9.229)
In the present calculation
we solve the temperature equation (9.228)

of
flux
and
in terms
the
use this result in (9.229) and thence in
<p(z,t)
The temperature equation may be solved by the method of
(9.226).
Laplace transforms as in the preceding analysis or by the general methods
for treating first-order partial differential equations.1 It may

that the application of the latter method leads to the results
@(z,t)
ev
where
&
is the
Vt) +
fz/r
Jo
0(2,0
and
ip(z
gV
c
Jo
Vs, I
fz/V
<p(z
Vs,t-
<p(z
Jo
transit time through the
s) ds
>Vt
0 <
<
<Vt
0 <
< d
- s)ds
- Vs, -
0(z) = To(z)
(9.230)*
t>d
s) ds
(9.231)*
core, and
T0(0)
(9.232)
is
<
0.
C-)
with T0(z) as the initial temperature distribution in the core.

Note
that the solutions for the interval 0 < t < d accommodate the fact that
different from zero only in this interval and identically zero for
it
is
&,
if
The results (9.230) and (9.231) are also easily deduced by physical
Note that
arguments.
<
i.e., before all the original fuel particles
in the core at =
are swept out, the temperature (rise) at
and
simply the initial value at

Vt plus the temperature rise due to the
Vt as
fission heating experienced by the slug from
travels along
is
is
0;
>
the core to point z. Clearly, when
all the original fuel in the core
has been swept out.
In this case, the initial condition to be used with
the entrance condition 0(0,0 =
thus the temperature rise
(9.228)
experienced by
slug which has reached at time
due entirely to the
See, for example, F. B. Hildebrand,
"Advanced Calculus for Engineers," pp. 368Prentice-Hall,

Inc., Englowood Cliffs, N.J., 1950; and R. Courant and
D. Hilbert, "Methoden der Mathematischcn Physik," vol. II, chap.
especially
pp. 51-57, Springer- Verlag, Berlin, Vienna, 1937.
1
378,
2,

= 0(z
0(z,i)
be shown
REACTOR ANALYSIS
604
[CHAP. 9
it picks up from fissions between the point
Vs (where it was
and
the
observation
z
point
present
s)
(its
location).
The solutions (9.230) and (9.231) may now be applied to (9.226).
For this purpose we first express the flux by means of a Fourier series.
heat
located at time
If
we use the form (9.197), then the solution
(9.231) for @(z,t) may be
written
0(2,<) =
'
V f
3*
^,(1
s)z(z
t>d
vs) ds
(9.233)
71= 1
The substitution of this expression into (9.229) and thence into (9.226),
followed by multiplication by Z,(z) and integration over the interval (0,h) ,
yields [cf. Eq. (9.198)]
_ l&l V
cCj
L4
m,n 1
^m(<)
z,(z)
Jo
Zm{z) dz
[
Jo
'
n(t
(9.234)
s) Zn(z
where we have used also the orthogonality properties

invert the order of integration and call
m
i J[k
i
Cj
Zj{y) Zm(y) Z(y
- Vs) ds
(9.199).
z) dy
If
we
(9.235)
Eq. (9.234) becomes
l*fc(0
*i(t)
fe
- tf)
(m2
-^ V
fc
<M0
m,n = 1
which applies for

t
>
&.
*(t)
fe
m-
(m2
*.(
- s)Kjmn(Vs)
In the same way, it may
ds
(9.236)
be shown
that when
relation is
< d the corresponding
**<mo =
Jo
/;)
skQ
m,n 1
^W
Z,(z) Z.()
rf2
where we have applied the appropriate forms for
(<
Z(2
Vs)
da
as indicated in (9.230).
REACTOR KINETICS
sec. 9.5]
G05
If
*f(0
0*2
m-
m,2)
*o]
UM)
[|l
we invert the order of the s and z integrations and note that @o(z) = 0
for z < 0, the above equation reduces to
BO
-^
- s)Kjmn{Vs)
(<
*m)
ds
(9.237)
m,R =
Equations (9.236) and (9.237) are an infinite set of integrodifferential

An approx
equations of the delay type for the Fourier coefficients (<)
imate solution to this set can be obtained when the temperature devia
adequate
are small; then, the first term in the Fourier expansion
If we drop the subscripts on
and we can discard all terms
>
and if in and note that in = n, the set (9.236) and (9.237) reduces to the
This result
is
i>
<
<
<
(a)
(9.238)
K(Vs)
ds
>
~(\
=
w) dz
for
Z\(z) Z,{z
appropriate
for
K(w)
4>(t
Vt) Z\(z) dz
s)
Sk0
J"
where
K(Vs)
-^
J,h<W) =
Jo
90(z
ds]
Jvt
s)
j*
--
[sfco
4>(t)
(b)
cos
(9.239)
to the sinusoidal axial variation in the neutron
In the analysis which follows we use, in place of the exact form

linear approximation similar to that used for L(f)
indicated for K(w),
a
flux.
(9.240)
Km(Vs)
t?
<
s)
<
>
P(t
ds
(a)
dsj
KAVs)
P{t)
[sk0
Jo
LP{t)
j*
P(t
--
~ Km(w)
used in (9.238), along with the substitution (9.132) for
"1
this relation
P(t), the result
is is
K(w)
the function K+{w), where

1
we introduce
Thus
s)
in (9.206).
If

U4>(t) =
pair of equations
0.
is
tions
(b)
(9.241)
G06
ANALYSIS
REACTOR
[CHAP. 9
by introducing an
This pair of equations is written more compactly

0{t), which we define
l
-
+ y
Jo
P(t-
s)
K(Vs)
- P{t -
K{0)
K(h) +
P(t
J*
-c
6(t) =
0)
that
K(Vs) is given by (9.239), or approximated
|V(0
where
fa
8(t) m
ds
(9.242)
by (9.240).
s)
effective temperature
K(Vs)
Note
ds
(9.243)
(9.244)
ds]
P{t
s)
-c
jT"
\p(t)
6(t)
this equation reduces to
K by
we approximate
If
<
<
dz
(j^J
P(*
Vt) sin2
--ejo
dsj
90(z
^1
P(i)
we have
J*
summarizing,
|^Sfc
(9.245)*
and
(o)
(9.246)*
f*
i>m
*(0=-|P(0-5
jo f*
8(t)
1
a/rt
-yP(t)
1
/>-s)dsJ
(6)
is
0)
is
is
It of interest
Note that the function K* has been used throughout.
to mention that the set (9.246) constitute the relations used by Ergen,
Weinberg, and Welton in their studies of the kinetics of circulating-fuel
reactors.
These studies place emphasis on the stability aspects of the
nonlinear set (9.246) and include the extension of the present theory
to account for the effects of alternate fuel paths through the core with
different transit times as well as the effects of delayed neutrons.
Since Eqs. (9.246) apply only for values of
greater than the core
transit time, the set may be conveniently called the "asymptotic" solu
In the general study of
tions for the temperature-dependent
kinetics.
this problem one should properly use first Eq. (9.245), which describes
the reactor power transients during the time interval before all the original
=
fuel in the core (at
then
The solution for >
swept out.
=
constructed from (9.246), using the results from (9.245) for
as
initial conditions.
In order to obtain
solution for the power transients during the time
a

UP(t)
8{t)
temperature
This relation and the definition (9.242) may be used to replace Eq.
(9.2416). The set (9.241) may then be written in terms of the effective
KINETICS
REACTOR
SEC. 9.5]
607
interval (0,6) we require an expression for the steady-state temperature

distribution 0o(z). This may be obtained from the steady-state flux
distribution <o(p,z) at t = 0, where
+ fi2o(p,z)
V2*0(p,z)
For the
case of a sinusoidal
= 0
(9.247)
axial flux
Z(z) =
The steady-state
= R(p) Z(z)
0(p,z)
4>o sin
(9.248)
distribution may be determined

(9.228) by setting the time derivative equal to zero; then
temperature
^0o(z)=^sin^
If
0O(2)
Eq. (9.232)], (9.249)
cos
(9.250)
-j
ds\
|)
- - s sin - - g- sin
j
*)
P(t)
(^1
UP{t)
|fe
The substitution of this result into (9.245) gives
<t (9.251)
(9.251)
is
is
This equation
to be solved simultaneously with the pair (9.246).
The expressions (9.251) and (9.246) form
nonlinear set of integrodifferential equations in P(<) and 6(t). An exact solution to this set
set (9.246) in
\ (l -j
Jo['
In P0
-___
V/
w{t).
8ln
With this
0<t (9.253)
ds
-70(0
(9.253)
(a)
<
Po
(9.252)
turn reduces to
,,A
w(t
f*
l*i*(0
-^
C
~]
sin
The
OWz
*.-
becomes
(9.251)
&
ltMt)
In P(t)
so that
s)
approximation
with P(0) = Po
w(t)\
^1
that w
and require
(t)
Pit) = P0[l
linearized equations.
We seek instead the solution to the corresponding

For this purpose we introduce
not considered here.

(9.249)
we call sS/ô = Po, and note that 60(0) = 0 [cf.
yields
from
>
(9.254)
608
ANALYSIS
REACTOR
[CHAP.
The solution to Eq. (9.253) must satisfy the initial conditions

= 0
u>(0)
-f2
tf(0) =
**!
tf(0) =
f2
where
- *C
(9-255)
(9.256)
In obtaining
the solution to (9.253) it is more convenient to work with

the corresponding ordinary differential equation.
It is easily shown that
this relation results from two differentiations of Eq. (9.253) ; thus,
$(<)
W) - ^ w(t)
/(0 =
with
= f2/(<)
for
0 <
<
(9.257)
_^__sm___8in_j
(9.258)
- w(t)]
=0
for
>
t>
[w(t
tf)
w(t) + f2u)() +
(9.259)
The initial conditions for this equation are obtained from the solution
= d.
to Eq. (9.257) evaluated at
For this purpose we require the
function w(d) and the derivatives w(d) and w(d) as computed from
(9.257).
We will not concern ourselves further with the detailed solution to
this problem but will instead examine briefly some of the physical
features of the system described by the pair of equations (9.257) and
(9.259).
Consider first the form of the power oscillations.
studied by introducing the elementary solution
This feature may
w(t) = e"*'
be
(9.260)
and investigating the properties of the quantity a*. The substitution

of (9.260) into (9.259) yields
transcendental equation for a* in terms
ai +
thus
f2a*
(e-**
t?;
1)
and
+
^
of the parameters

The corresponding solution for the asymptotic forms (9.254) is obtained

by eliminating 8(t) in (9.254a) by the application of (9.2546). The
result is a third-order equation involving a "delayed" term.
(9.261)
The roots of this equation are in general complex; therefore set

or*
= x* +
iy-k
The substitution of this relation, after separation of real and imaginary
sec. 9.5]
REACTOR KINETICS
609
parts, yields the pair of equations for the unknowns
xi
- 3x*yl
-Vl
+ f2x* +
(er*x* cos &y*
-j
f2
+ 3xly* + r-V*
e~*x*
and y*
x+
1)
= 0
sin dy* = 0
(9.262)
It
can be shown by the application of Nyquist's criterion1 that Eq.

(9.259) has no antidamped solutions since Eq. (9.261), or the pair (9.262),
has no roots in the right half of the a* plane.2
Thus we find that
x* < 0, and the general solution w(t) = exp (a+t) represents a damped
oscillation.
The introduction of a Bk0 into the circulating-fuel reactor
(with y > 0) results, then, in a series of power excursions which eventu
ally decay, bringing the system to a new steady-state condition.

The solutions to (9.262) may be obtained by the usual iterative tech
It is of some interest to mention that special roots of (9.261)
niques.
= 2mr/d
In these
is .).
pure
cases, a*
an undamped steady sinusoidal
is
a
(n = 1, 2, . .
imaginary, and the solution to (9.259)
occur when
oscillation.
is
The temperature of the fuel at the core exit

accord with our previous approximations we use only the first term
Jo
~ - ["y P(t
c
0(z,<)
The linearized approximation
s)
in the series; then,

Zt(z
Vs) ds
(9.263)
of this expression gives for the exit temper
w(t-
sin
J ds]
+
Jo
f^
[i
s)
[*
= (<) =
0(M)
[\
ature
(9.264)
it
is
is
The temperature oscillations in this system are quite mild as compared

to the power oscillations.
This trend
evident from
comparison of
Since the reactor temperature
given by an integral
() and w(t).
over some past history of the power,
will necessarily exhibit
much
smoother variation in time.
Thus even though the power oscillations,
as represented by w(t), may be very sharp, the temperature oscillations
will be much more gradual and of relatively small amplitude.
This
property makes the circulating-fuel reactor an extremely attractive device
is
it
source of power.
little danger
Not only does indicate that there
to the reactor through excessive temperature variations, but since the
as
See, for example, H. S. Tsien, "Engineering

Cybernetics," McGraw-Hill Book
Company, Inc., New York, 1954.
This calculation has been performed by F. G. Prohammer, " Note on the Linear
Kinetics of the Circulating Fuel Reactor," Oak Ridge National Laboratory, Y-F10-99,
Apr. 22, 1952.
1

In
REACTOR
610
ANALYSIS
[chap. 9
power extracted from the reactor is determined by the temperature of

the fuel, the fluctuations in actual power output will not be severe.
Finally, it should be mentioned that, when a certain 5k0 is introduced
into the reactor, the resulting oscillations in temperature decay rapidly;

however, the exit temperature eventually stabilizes at some new level.
This value can be determined from (9.264), and it is easily shown that as
This result may be evaluated in terms of the
t oo , 0OX(<) > 2dPa/irc.
If we set P(Qi) = 0 in Eq. (9.2416), then,
excess multiplication.
(9.265)
which states that if some positive reactivity is added to the system, then
because of the negative temperature coefficient of the fuel, the reactor
must operate at some higher temperature
level which just compensates
Decay of Fission Products and Burnout Poisons
9.6
In the time-dependent
been taken
problems considered so far, no account has

of the appearance of fission products or burnout poisons
is,
which serve as strong "neutron sinks," that

remove neutrons from the
reactor by radiative capture.
Nuclei with large neutron-absorption
cross sections are created during the fission of
nucleus, either as direct
products of the reaction or as intermediate stages in the decay process

of other fragments.
Burnout poisons are materials which have large
reactor
absorption cross sections and which are purposely included in
to aid in controlling the neutron population and thereby to aid in limiting
For some time after
the maximum available reactivity in the system.
is
added to the reactor, the reactivity

new charge of nuclear fuel
of the system may be far beyond the limits which the control system
Burnout poisons introduced along with the fuel (possibly
can suppress.
mixed in with it) can assist the control system in removing excess neu
is
As the fuel
is
consumed the available reactivity decreases, but

By this time, how
so also does the concentration of burnout poisons.
to
maintain
condition
steady-state
control
alone
able
the
system
ever,
trons.
realized, then,
The primary function of these poisons
during the initial stages of the reactor fuel cycle.
to examine some of the problems
The purpose of the present section
associated with the estimation of the concentration of poisons in the
function of time for periods of operation which are compareactor as
is
is
in the reactor.

for this increase.
sec. 9.6]
REACTOR KINETICS
611
rable with the time required
to reduce fuel concentrations by sizable

fractions of their initial values. The concentration of these nuclei may
be included in a study of the complete neutron economy of the reactor
system, and suitable adjustments may then be made to maintain the
reactivity within prescribed limits.
a. Decay of Fission Products.
The distribution of nuclear fragments
which are produced by a fission reaction extends over the entire mass
scale (see Fig. 1.3).
The majority of the fragments are grouped, how
The fragments which

ever, in two regions -mass numbers 95 and 140.
are of particular interest in the present problem are those which have
reasonably large neutron-capture cross sections as well as relatively large
We are concerned, therefore, with fragments having
percentage yields.
The most important of these are Xe136, Sm148, and
large values of 2.

We will consider first the concentration of 1 136 and determine its value
Iodine 135 is a first-stage decay product of Te186,
as a function of time.
which is a direct fission product.
The tellurium which appears from
fission reactions decays by negative beta emission to form I136, which in
turn decays to Xe136, again by negative beta emission. The decay proc
ess is indicated below, along with the appropriate half -lives:
62Te136
<2 mm
53I136
-4-*
6.7 hr
54Xe136
(9.266)
The yield of Te135 is 0.056 atom per fission. We note from the above
relation that the decay rate of the Te136 is relatively fast; so a reasonable
approximation to this reaction would be to assume that I136 appears as a
direct fission product of yield 0.056 and to ignore entirely the presence
This approximation has some merit since the half-life
of the tellurium.
is short, relative to the time periods of interest here.
On the basis of this assumption we can write the following balance
equation for I(<), the I136 concentration:
of
Te136
1(0
~ -X,I(t)
- e?l(t)<Kt)
+ MiZ/(fl*(0
(9.267)
where (refer to Table 2.1 for data to compute X)

Xi = decay constant for I136 = 0.1037 hr-1
cross section for I136
a",1' = neutron absorption
=
I136
of
fission
atoms
per
(0.056)
Hi
= fissions per unit volume per unit time
It should be recognized
and I(<) is given in nuclei per unit volume.
in
this
must correspond to the
indicated
equation
that the cross section
Inasmuch as the largest fissionneutron energy implied in the flux
2/<>
fragment cross sections occur in the vicinity of thermal energies, we will

confine our attention to the effects of these poisons on the thermal-neu
tron flux. Thus the cross section o^" and those which follow refer to
612
REACTOR
ANALYSIS
[chap. 9
the thermal-energy group denned by the reactor temperature.

Note,
that
furthermore,
(9.267) includes only the time dependence of the iodine
concentration.
Spatial variations due to a nonuniform neutron flux are
not treated in the calculations which follow.
These effects are properly
handled by perturbation methods or numerical analyses.
Equation (9.267) states that the concentration of I"6 nuclei is decreased
at the rate Xil(<) from decay and at the rate ff<,"I((0 from thermalneutron capture.
Iodine 135 nuclei are produced at the rate m2f(t)4>(t).
The solution to this equation may be written in the general form
where
m=e-f>)dt[lfo'Me)mJ"de
L(<) = X,
m]
+ <r(?(t)
0.268)*
(9.269)
Thus, given the time behavior of the flux, (9.268) yields an explicit rela
tion for the iodine concentration.
The balance relation for the Xe186 may now be given in terms of the
I11*
If X(<) denotes the

concentration.
differential equation is
Xe135
concentration,
X,I(<)
+ MxS/O
then the

appropriate
X(t)
-[X, + r*(0] X(0 +
(9.270)
and fix are the decay constant, the thermalwhere the quantities X
cross
and
the fission yield, respectively, for xenon.
section,
absorption
0.003 and \z = 0.0754 hr"1.
Note that
The value of the yield is /**
the first two terms on the right side of this equation represent the reduc
of radioactive decay and
neutron capture and that the last two terms give the rate of production
from the decay of I135 [the parent nucleus, cf Eq. (9.266)] and as a direct
product of fission. The general solution to (9.270) is
tion
XW
with
in the
Xe135
e-l>)d
concentration
[uw
as a result
(9)\Jem")d"
M(t)
m X,
(t)
de
X(0)}
(9.271)*
(9.272)
The third reactor poison considered in this analysis is Sm14*, which

appears as the radioactive beta-decay product of Pm149 (47-hr half-life).
The Pm149 is itself the beta-decay product of Nd149 (1.7-hr half-life),
The decay process is given by
as well as being a direct fission product.
1.7 hr
47 hr
Because of the relatively short half-life of Nd149, we will assume that

The combined yield of Pm149 is, then,
Pm149 is a direct fission product.
This number is 0.014 atom
the fission yield plus the Nd149 production.
REACTOR KINETICS
sec. 9.6]
The balance relations for the two remaining
per fission.
methium
613
Pm(<)
and samarium
Pm(0 = ^,,2/(0 (<)
S(i)
\PPm(t)
nuclei,
pro-
are given by
Sm(<),
[Xp
o(t)Wm(t)
(t)
<7Sm(<)
(9.274)
0-275)
M.2,(<) *()
e-/(9)
"
MB)
dB
|>P
P.(0
/'
Samarium 149 has a stable

where the index p refers to Pm, and s to Sm.
The
nucleus, and therefore its balance equation contains no decay term.
solution to (9.274) is
(9.276)
P(0)]
with
G(t) =
10
is
(9.277)
[X,P(9)
d0
S.(0)}
(9.278)*
The results (9.268), (9.271), and (9.278) are the general expressions
Detailed
for the time dependence of the principal fission products.
calculations for these functions can be carried out once the (t) and 2/(<)
is
are specified.
Some simple cases are now considered.
b. Steady-state
The concentrations of the
Poison Concentrations.
reactor which
operating at steady
poisons I136, Xe136, and Sm149 in
is,
is
At
state are easily obtained from the original differential equation.
stabilized to some value o- As
steady state we assume that the flux
the rate of fuel
simplification we also take 2/ to be a constant; that
is
lo
is
is
not much different

consumption
small, so that 2/ at any time instant
from its initial value. We find then, on the basis of these assumptions,
that the steady-state I,M concentration
iJ%r
Al +
0
0-279)
Xo
Sm149
X'Io
Ax
obtained from (9.270)
Xe136
^y
o
aa
The steady-state
computed from (9.279).
derived from (9.274) and (9.275)
(9.280)
concentration
of
where
of
The steady-state concentration
is
aa
is I0
is

+ M.2/(0) *(9)]e'0/^<9')'i9'
Sm(<)
\j0
8,(0
and the solution for
XP
(0)
S/
M<rj,p'fnoail
fnoail
REACTOR
(514
ANALYSIS
[chap. 9
It is of interest to determine the limiting values of these concentrations

for arbitrarily large values of the thermal flux. These are
lim
I.
&
lim
(T
X0 =
lim
S%>
&
(9.282)
Thus the steady-state concentrations of Xe and I increase with increasing

flux whereas that of Sm decreases, reaching in the limit the asymptotes
given above.
c.
Poison Concentrations
following Reactor
Shutdown.
The time-
dependent behavior of the poison concentrations in a reactor following

shutdown can be obtained from the general results given above if t = 0
is taken as the instant of shutdown and the flux is set equal to zero for
In the
all subsequent time.

yields the result
Sm(<)
case of samarium,
= S(0)
+ P(0)(1
this procedure
- rV)
easily
(9.283)
= 0.01472 hi-1.
Although the postshutdown concentration of samarium can be
a serious
consideration in some reactor situations, the control of the xenon concen

tration is a problem of more general importance in all thermal reactors.1
This problem is especially serious when the I136 concentration at shut
Under these circumstances the Xe136 concen
down is relatively large.
tration increases rapidly after shutdown and can reach appreciable mag
This buildup is important from the
nitudes before decaying away.
standpoint of control, since a large positive reactivity change must be
available if it is to be required that there be a restart of the reactor
Thus special "xenon-override"
around the time the maximum occurs.
control rods would be required in a mobile nuclear power plant if shorttime startups were to be an essential operational feature.
At the time of shutdown, there exists a particular combination of
in the reactor. This combination
iodine and xenon concentrations
depends, of course, upon the operating history (i.e., flux program) as
In any case, these can be deter
indicated by Eqs. (9.208) and (9.271).
The concentrations after

mined in detail, by measurement if necessary.
shutdown are obtained by setting 4>{t) = 0 and letting t = 0 be the
Then,
1(0 = I.e-X
X,
X;2.)=
IX'
X,
Especially solid-fuel reactors.
(X,
X*)
e-x-'
(9.284)
e-M'J
instant of shutdown.

where Pm(0) and Sm(0) now represent the concentrations of promethium

This result shows, then,
and samarium in the reactor at shutdown.
that the samarium concentration increases monotonically to the asymp
totic value Sm(0) + Pm(0). The buildup is slow, however, since
(9.285)
SEC. 9.6]
where
and
I,
KINETICS
REACTOR
and
X,
denote the concentrations
z, =
of
615
and Xe136 at shutdown,
I136
(9.286)
i*
We have introduced the parameter z, in the symbol for the xenon concen
tration since the absolute magnitude of the concentration depends upon
this ratio, as well as upon the value of I..
The expression (9.285) is generally applicable and gives the xenon
For
concentration after shutdown for a specified value of the ratio z,.
each value of z,, a separate calculation for X(t;z,) is required, and each
will yield a different time behavior.
This complete system of solutions,
The appro
however, can be combined into a single generalized relation.
priate form is obtained by applying a displacement transformation on the
Thus we define a new timelike variable 8 which is
time coordinate.
related to I by
=
6*
(9.287)
where 0* is to be determined for convenience.

relation into (9.285) yields
=
(X,
The substitution of this
Xx)]
X;z.) -> X(0;z.)
<rM-*)
e-x,(.-.*>j
The result
0.
We now choose 0* so that X(0;z)
is
(9.288)
lnd-Kz./x.Kx.-x,)]
Xi
i =
28Q)
Xx
Ai fti
be
Ai
(Xi
(e-M
e~)
we evaluate
in terms of its
"1m/<m-x,)
written more simply

The maximum
if
This result may
- X.)
f*
r-^V
X,
^
X(0;z.)
6a =
and
is
maximum value X(0m;z).
it

(9.290)
seen to occur at
ln (X[/Xj)
(9.291)
Xi Xx
follows that
(0;z.)
X(0m;z,)
- X.
XxeXl-
(c-x,
This function gives the so-called universal
c-x,)
3(0)
(9.292)*
xenon curve1 shown in Fig. 9.14.
'A. G. Ward, "A Universal Curve for the Prediction of Xenon Poison after a
Reactor Shutdown," Atomic Energy of Canada Limited, CRRP-685, Jan. 25, 1957.
ANALYSIS
HEACTOR
616
Note that it represents the xenon concentration
[CHAP. 9
after shutdown
for the

special case when the concentration at shutdown (X.) is precisely zero

and normalized so that the maximum value of the function (9.292) is
unity. Thus we find that the displacement transformation (9.287) along
with the specification (9.289) transforms the "truncated" function (9.285)
28
32
36
Timefl, hr
Fig.
9.14
Universal xenon curve.
A further obser
to a more simple form which passes through the origin.
vation to be made is that the X and X functions are related by
X;e)
X(t;0)
(9.293)
The procedure for determining the xenon concentration in a particular

situation from the universal curve is as follows.
Given the ratio of xenon
to iodine concentration at shutdown z Eq. (9.289) is solved to obtain
the "effective" shutdown time 8*. These solutions may be read directly
from Fig. 9.15. If this value of 0 is then marked on the abscissa of Fig.
9.14, the curve to the right of this point gives the relative variation in
Thus if X, is the actual
xenon concentration for all subsequent time.
concentration at shutdown and x(0) the value read from the curve of
Fig. 9.14, then X.,x(d) /x(6+) is the concentration
shutdown.
at time
0 0*
after
In the event that the reactor has
o for a long
been operating at a steady-state flux

time, the iodine and xenon concentrations will have reached
the equilibrium values given by Eqs. (9.279) and (9.280).

Xp
Io
Xl(Ml
XiMi
+ Migq'Ô
+ Miciô
Ml)
Then,
(9.294)
sec. 9.6]
KINETICS
REACTOR
617
This function is shown in Fig. 9.16.

The specific xenon curve for this
by
Fig. 9.16 for the known o.
is
obtained
from
selecting
case
X0/Io
This value is then used in Fig. 9.15 to determine the effective shutdown
time 0* and, thence, the xenon curve from Fig. 9.14.
d. Constant-flux Reactor.
We consider next the general solutions
for the I136, Xe135, and Sm149 concentrations given in Sec. 9.6a for the
12
s
o
0)
>
n
a
Using:

A/
Xi
0.1037
hr" 1
0.0753 hr"1
0
Ratio of xenon concentration
to iodine concentration
Fig.
9.15
at shutdown
z.
Effective shutdown time as function of xenon-iodine ratio.
The system
is not at steady state, however.
For convenience we assume that the
reactor is being started up for the first time, so that the initial concen
trations 1(0) = X(0) = Pm(0) = Sm(0) 0. Equation (9.268) gives the
I136 concentration
in this case to be (again taking 2/ constant)
case of a reactor which is operating at a fixed flux level
I,
1(0 = 1.(1
- e-*-')
L.
(9.295)
m X,
Note that the I* corresponds to 1 0 of (9.279) if

For Xe135, we have [from (9.271)],
we
where
AW
with
A.
X.
L^
and
_
and
LJ
j
Af
- X. +
<r
identify
(9.296)
<,
L.v,(M,
with
<<>
- L,)|
(9.297)
(9.298)
ANALYSIS
REACTOR
G18
[CHAP. 9
2.0
1.5
1.0
Using:
0.80
X,
0.1037
X, = 0.0753
0.60
hr"1
hr"1
<ri*= 3.4 X 106 barns

0.40
0.30
0.20
0.15
0.10
0.04

0.03
0.02
0.01
12
(j> X
Fig.
Ratio of equilibrium
steady-state flux.
9.16
and, finally, for
"l
Sm149
10""
20
16
concentrations of xenon
[from (9.278)],
*>-.">*.<
24
28
(neutrons/cm 2 /sec)
/>-.< 1
to iodine as function
of
2,
(9.299)
where
b- =
~ar
and
(9.300)
These results give the buildup of the fission products for a reactor
Note that
starting at t = 0 and operating thereafter at flux (t) =
for t ao these solutions reduce to the steady-state values derived in
Sec. 9.6b if we take + = 0.
The assumption of a constant flux used

e. Constant-power Reactor.
in the preceding calculations has merit from a practical viewpoint since
reactors designed and operated specifically as research tools are frequently
required to maintain uniform flux levels for prolonged periods of time in
There are other
order to accommodate various irradiation experiments.
sec. 9.6]
REACTOR
KINETICS
619
however, in which a constant-flux program would be

A good example is the use of a reactor as a source of
unsatisfactory.
It is clear that in this case the principal requirement on the sys
power.
tem could very well be that the rate of energy production remain fixed
reactor applications,
Thus, the operational condition which would

for long periods of time.
be imposed on the system would not be that the flux remain constant
but, instead, that the total number of fissions per unit time remain con
If this time interval is long in com

stant for the specified time period.
life
of
the
reactor, then the rate of fuel con
with
the
operating
parison
In
sumption must be taken into account in computing the total fissions.
this case 2/ is treated as a time-dependent function.
the appropriate condition to be imposed is
P(t) = P0
2/(0 *(0
&T/Nr(0)
For such systems
(9.301)
4>(0)
where P(t) is the power generated per unit volume by the reactor, Po is
some specified power level, NfU) is the concentration of fuel (nuclei per
unit volume), and is the energy release per fission.

As noted previously,
the calculation of the poison concentrations is

In the case of the constanton
the
time
behavior
of the flux.
dependent
power reactor, the flux must be inversely proportional to the fuel concen
tration. Thus a complete description of the operating condition of this
type of reactor will require, in addition to the differential equations for
the poison concentrations,
nuclei; namely,
a statement
NF(t)
where
-*N
denotes the total absorption
constant-power
is seen to be
relation
of the mass balance for the fuel

,(t)4>(t)
If the
the solution
cross section for the fuel.
(9.301) is applied
N,(t)
(9.302)
NP(0)[1
to this equation,
a<"(9.303)
(9.303)
and the flux is given by

=
(9"304)
above.
The substitution of this expression for (t) into
for
solutions
the fission-product poisons yields the appropri
the general
reactor.
These
ate time-dependent functions for the constant-power
results can be expressed in terms of the incomplete gamma function.
as mentioned
It will be convenient for the remainder of the present section to rewrite

Eqs. (9.303) and (9.304) in terms of the dimensionless variable z, where
z =
The symbol
(9.305)
denotes the time it takes for enough fuel to be consumed
ANALYSIS
REACTOR
620
[chap. 9
that the over-all system reactivity is reduced to zero.
so
system
alF)(0)t
<F)0)t
It
In
solid-fuel
would be related to the "life" of a fuel element.

The quantity
in the expression for the flux may be written in terms of z; thus
U,
SSfs
(9"306)
that Pok/ is the total fissions per unit volume which occur in
time k; but this may also be given by
is clear
(Fuel nuclei destroyed)
If
we call
/3/-
[NF(0)
- N,(k)] %
the burnup fraction of the fuel, where
Pr
- NF(U,)
NF(0)
(9-307)
o'aF)(0)t = &Fz
then
The fuel concentration

the form
NF(z)
(9.308)
and the flux, when written in these terms, take
NF(0)(1
- Pfz)
f. Secondary Fission Fragments.
4>(z)
j-z^
(9.309)*
In addition to the primary products
yields various secondary frag

has
can
be treated in terms of a single
shown,
ments which, experience
effective absorption cross section.
Some detailed information as to the
nature of these fragments and the sequence of their appearance following
I1'5,
Xe186, and Sm149,
a fission reaction
in the literature.1
It will suffice for the present
to consider an elementary approach for estimating the con
The fundamental assumption
centration of these secondary fragments.
will be that all the secondary poison nuclei which appear can be regarded
a fission is available
treatment
"composite" nucleus with an absorption cross section

call Nrp(t) the concentration of this composite (or effective) nucleus,
as a single
If
we
then the balance relation is clearly
Nrp{t)
= c,NF{t)
(t)
(9.310)
The cross section <y{FP) must be selected from experience. We have

omitted a term for removal by neutron capture in Eq. (9.310) on the
basis that the cross section <r(FF) represents a time-averaged quantity
which accounts for both the continuous production and loss of secondary
W. B. Lewis, "Reactivity Change Expected and Observed in the Long Irradiation
of Natural Uranium in the NRX Reactor," DR-21, Feb. 4, 1952; L. D. P. King, "Los
Alamos Power Reactor Experiment and Its Associated Hazards," LAMS-ltill,
Dec. 2, 1953; J. Halperin and R. W. Stoughton, "Some Effects of Transmutation
Products on U" Breeder Pile Operation," ORNL-1368, Sept. 29, 1952.
1
sec. 9.6]
REACTOR
KINETICS
621
The actual buildup of such fragments during reactor

poison nuclei.
in the reference reports.
is
discussed
operation
The concentration Nn>(t) is easily determined for the two flux pro
grams previously
For the
considered.
case
of constant power,
we use
(9.301) for the total fissions and obtain
5^ -[*]*where we have used the relation 0p = clF)4>(0)k, which follows from

For the case of constant flux
the fuel-concentration
equa
(9.307).
tion (9.302) gives

=
Nr(t)
iV,(0)e-'.1"^'
(9.312)
Substitution into (9.310) yields [using (9.307)]
Burnout Poisons. The use of burnout poisons was discussed in the

introduction to the present section. In the development which follows
we compute the nuclear concentration for some arbitrary poison Nj(t)
for the cases of constant flux and constant power. The results for the

g.
case of constant
strate the use of a
power are then used in a sample calculation to demon

burnout poison in controlling the reactivity of a reactor.
The mass balance for the burnout-poison nucleus
is given
by the
equation
Nj(t)
where
-o*Nffl (t)
(9.314)
cross section of the poison j.

denotes the neutron-absorption
cases mentioned are easily shown to be
The solutions for Nj(t) for the two

N,(z)
iV>(0)(l
JVy(z)
Ar;(0)(l
where
$,zy<
/8jp)'
for constant power

for constant flux
sj =
(9.315)
(9.316)
(9-317)
The result (9.315) is now applied to the criticality calculation of a bare

The appropriate
homogeneous reactor operating at constant power.
is
multiplication
for
the
constant
relation
k = T)/epth^,hpNL
(6.81)
We use this relation to compute the effect on k caused by introducing

burnout poisons which have the time dependence given by (9.315). For
convenience in computation, we choose epu, = 1, so that (6.81) reduces to
k = ij/sAhPni.
In
these calculations
we are concerned
(9.318)
primarily with time-dependent
622
ANALYSIS
REACTOR
[chap. 9
is
1,
if
is,
phenomena, and, strictly speaking, the relations (6.81) and (9.318) are
for steady-state operation.
meaningful only when k = 1, that
Thus
then the flux
we speak of
^
experiencing rapid variations in
it
is
is
shutting down or
time, so that the reactor either
supercritical.
will be useful in this analysis to think in terms of a timeNevertheless,
dependent multiplication k(t) which will describe the influence of the
is
is
various poisons and the fuel on the reactivity during the operating life
of the system.
Actually, k(t) =
all through this period, but this
accomplished only by the continual action of control system; fc(<) repre
sents, then, the available multiplication at any given instant and
the
amount which the control system must counter in order to maintain the
This concept of a
^
was previously treated
=
so
that
by introducing
Thus
a fictitious number of
vc
vc
v/vc.
neutrons which must be produced per fission in order to maintain the
In the present application we may think of a
system at steady state.
a
is
reactor at steady state.
vc(t) which specifies the neutron yield per fission at each

instant during the life of the reactor so that the system remains just
critical. The appropriate expression for
then (for gth and
inde
77
is
corresponding
If
*(0
we call
0(0
W*/(0 Pnl(0
(9.319)
- 2f^y
(9-320)
cross section of all reactor materials

2J,X) represents the absorption
other than fuel, then (for D independent of time)
It
will
t!>
WW)
.Fwrai
l9-321)
m
pM'>
where
to separate S*,1' into

component Sic) for materials
whose cross sections do not vary with time and component 2^"(<) for
those which do; thus
a
be convenient
Z(0 = Z> +
(9.322)
2?>(0
For the present calculation
is
we include in 2*)(<) only the burnout poisons

and omit the production of fission fragments.
The fuel concentration for this system
given by (9.309) and the
poison concentration by (9.315).
If these expressions are used in (9.320),
we obtain
N
o() =
y^Wz
sMl

pendent of time)
i-1
^2)"-1
(9"323)
sec. 9.6]
KINETICS
REACTOR
623
where we have assumed that the reactor contains N burnout poisons,

= 1, 2,
The symbol se is defined according to (9.317), and
, N.
...
of poison
rij is the loading fraction
j:
ArX0)
Uj =
The result (9.323) may
be used
(9.324)
N,{0)
in (9.319):
-
wth
o()
iTT^
fc(z)
[[
then,
1
^1
The reactivity
is,
(9.325)
=^ (0)(1
(9.326)*
is
is
is
fraction of the operating life of

the time variable given as
where
the burnup fraction of the fuel.
It possi
the fuel elements and 0r
ble, by making a suitable choice of the loading fractions of the poisons
Fia.
9.17
Reactivity
change
poisons
with and without burnout poisons.
rij and their types, represented by the cross section ratios 8j, to control
the reactivity to variations which lie within the capabilities of the con
trol system.
If no poison were added to the constant-power reactor, the
reactivity would vary as shown in Fig. 9.17.
The initial reactivity oj(0),
far beyond the magnitudes which can be suppressed by the

Thus by introducing sufficient poison to make co(0) =
the
here, may be
0,
controls.
reactivity can be made to take the form shown by the solid-line curve in
the figure.
Note that the maximum excursion in oi{z) has been dimin
ished and might be brought within the range of the controls.
In some
situations the maximum rise in reactivity may still be beyond the capa
of the controls; such
is
sometimes the case when BU), which has

used.
large aa,
Although enough B10 can be added to suppress the
initially large amount of reactivity available,
"burns out" too fast,
and in short time w(z) will rise to uncontrollable magnitudes.
In these
a
it
is
bilities
(e.g.,
is
possible to use
combination of B10 and some other poison
cross section.
Now the com
Li6) which does not have so large
bined initial loadings of these two poisons will suppress the large cu(0),
and yet a sufficient amount of the lower cross section poison will always
cases
it

- Without
ANALYSIS
REACTOR
624
[CHAP.
remain to keep the reactivity within controllable bounds during

In carrying out these calculations,
remaining life of the system.
useful to impose the following conditions on (z) :
w(0)
= 0
(1) = 0
These conditions will yield two simultaneous

fractions rij of the two poisons.
the
it is
(9.327)
equations for the loading

PROBLEMS
Consider a bare thermal reactor [described by the constants D, B*, 2a(0),
9.1
2/(0), pth, and to,] which is operating at steady state. At some time instant, say
I = 0, the physical state of the system is caused to vary in time, so that the macro
Such behavior can be achieved, for
scopic cross sections become functions of time.
example, by changing the pressure on a compressible fluid fuel.
a. Derive the expression for the thermal flux (p(i,t) for this system on the basis of
the following model: (1) 2(<) = 2(0)(1 + o sin wt), 2/(0 = 2/(0)(l + 6 sin wt), and
D(t) m D; (2) neglect delayed neutrons; (3) neglect temperature effects on system;
(4) include oth e~B'Tth and pthj (5) assume that both gth and pth are time-independent
quantities.
6. Determine the function Sk(t) = k(t) 1 in terms of the steady-state constants
of the system. At what values of wt does 6k(t) attain its maximum value?
c. Sketch the functions &k(t) and (r,l) in terms of the variable wt.
d. Find an expression for {t,t) in terms of the mean lifetime I = \v\DB* + 2.(0)]
and the maximum positive value of Sk(t).
e. Determine the conditions on the frequency w such that ^m, > c^min, where c is
an arbitrary constant.
9.2
Determine the one-velocity flux distribution (x,t) in a nonmultiplying slab of
width 2a for which the diffusion medium is characterized by the one-velocity constants
Apply the condition that the flux vanish at the geometric boundaries of
2 and D.
the medium and that the initial condition is
4>(x,0+)
At
t = 0, a source
= 0
function S(x,t) is introduced:

S(x,t)
4>ocos
(1
+ s'n wi)
one-velocity reactor is time dependent.

The flux in a bare homogeneous
the following two (different) physical assumptions:
(a) all neutrons are prompt; (6) a fraction 0 of the neutrons from fission are delayed
time <o- Assume that the form
9.3
Find this time behavior under
(t)
Ae"T
Compare these
applies, and find values of T which allow a solution to each case.
values for T, using the form (where a and b refer to the assumptions given above)
Tb = 7\,(1 + a)
and give an explicit form for a. In each case only one time-dependent equation is
In model 6 assume
needed; do not consider the precursors of the delayed neutrons.
that t/Tb <5C1 in order to solve the equation for 1\.
REACTOR KINETICS
625
A bare homogeneous thermal reactor is critical and operating in steady state

9.4
At time / = 0, a small positive reactivity is introduced and held
with flux 0o(r).
constant until t = <i. At time t = ti the reactivity is reduced to zero.
a. By using one group of delayed neutrons and assuming that ti is sufficiently long
that the reactor flux is rising with the stable reactor period ai at t = U, find the expres
sion for the time dependence of the neutron flux and the concentration of the precursor
for all t > 0.
6. What is the final asymptotic value of the flux long after the reactivity is reduced
to zero?
c. Compute and plot the flux and precursor concentration as a function of time
using the following data:
0 = 0.0076
u = 0.001
X = 0.08 sec"1
<i = 10 sec
ltb
= 0.001 sec
A bare homogeneous thermal reactor is operating in steady state up to time

9.6
10, when some additional multiplication Sk<, is introduced into the system.
Exam
ine the time behavior of this system
on the basis of the following model :
/S)
are present but have infinite delay time;

(1) Delayed neutrons (fraction
(2) The added multiplication may be expressed by the relation
where the temperature rise above the steady-state value T(0)
- T(t) -
7'(0)
given by
forM >
+ /*e(i)l
<P( = <P[1
<P[1
+ /*e(i)l
forM > 0
the power generated
in the reactor, and
1.
(4) Solutions of the form (Foe"10 describe

w(JL)
T0
(3) The power removal from the reactor
is
&(t)
is
- yS(t)
gXfth
sin
ft
A./th cos
ft
r/
Wth
~|
Skq
+')]
7^[1--,(fsin
ew
the specific
heat of
is
is
and l,h
the neutron lifetime in
finite reactor and
nized sample of the reactor material.
where
homoge
0.
a. Write down an appropriate set of differential equations and initial conditions

>
which describe the reactor power P(t) and the temperature rise (() for all
Use the fact that P(t) may be linearized.
Solve the equations of part a and show that the results can be written in the
form
6.

k(t) = tko
CHAPTER
HETEROGENEOUS
10.1
10
REACTORS
Introduction and Survey
Merits of Heterogeneous
Fuel-moderator Configurations.
Some
remarks were made in Sec. 1.2 about the advantages to be gained in
using heterogeneous arrangements and the basis for the choice of this
Aside from the physical
configuration for the earliest reactor designs.
separation of fuel and moderator, the most important feature of such
a.
systems is the fact that they are nearly completely "thermal" (i.e., at
least 95 per cent of the fissions are caused by thermal neutrons).
These
unique characteristics lead to significant improvements in some of the
By establishing a
which affect the multiplication constant.
thermal system it is possible to exploit the relatively large fission cross
sections of the nuclear fuels in the low-energy range, and by separating
the fuel and moderator, such quantities as the resonance-escape probabil
ity and the fast effect can be measurably increased.

factors
These advantages were well appreciated in the early days of reactor

technology, and many experiments and analytical studies were devised
to explore them.
It is important to recognize that it was only because
of the superior nuclear characteristic of heterogeneous configurations
that it was possible to sustain a nuclear chain reaction at all. As noted
earlier, the only abundant and efficient moderator materials available
in the early days were graphite and water; but even graphite, with its
significantly smaller absorption cross section, is incapable of supporting
a chain reaction with natural uranium in a homogeneous arrangement.
This fact is easily demonstrated.
To begin with we note that if a finite
system is to be made critical then kx > 1. Now fcw = r\tfplb, and for
natural uranium 7/ = 1.33. If we take < = 1.05 (a relatively large value),
then clearly we must have/pn, > 0.73.
Figure 10.1 shows a plot of the
thermal utilization and the resonance-escape probability for homogeneous
mixtures of natural uranium and graphite.
It is seen that the product
of these two quantities (the broken-line curve) is at most 0.59, which is
well below the minimum value estimated above.
Thus the homogeneous
A
system with an optimum ratio of moderator to fuel is still subcritical.
heterogeneous arrangement of these materials, however, can be made
into a critical assembly. This is achieved in main part through the
626
HETEROGENEOUS REACTORS
SEC. 10.1]
627
increase in the resonance-escape probability which is realized when the

fuel is lumped instead of being uniformly distributed.
The properties
of the graphite
systems are representative of the

general trends to be expected with other moderators.
Except for heavy
water, homogeneous mixtures of moderator and natural uranium are
subcritical; however, critical assemblies can be obtained by going to
heterogeneous
arrangements.

heterogeneous configuration
40
Fig.
10.1
60
A convenient criterion for selecting a

is the maximization of the multiplication
80 100
200
400
600 800
1,000
2,000
Ratio of graphite atoms to uranium atoms NU/NF

Homogeneous system of natural uranium and graphite.
Thus the bulk of the effort required in the analysis of

constant kx.
heterogeneous-reactor designs is directed toward the determination of the
various factors which enter into the criticality relation, particularly as
It is interesting to note in passing

given by the four-factor formula.
that much of the terminology and notation of reactor physics presently
in use originated with the early studies of heterogeneous systems.
The
four-factor formula was developed to describe the nuclear characteristics
of natural uranium lattices and is best suited to reactors with low fuel
concentration.
However, as already demonstrated in the homogeneouscalculations
of the preceding chapters, this general relation
reactor
ship is easily adapted to systems with high enrichments of fissionable
materials.
When the system of interest is to be charged with natural uranium,

each of the factors in the criticality relation must be computed with a
fair degree of accuracy since for these systems, even with the most favor
able arrangement of fuel and moderator, the resulting multiplication
REACTOR
628
ANALYSIS
[CHAP.
10
unity by only a few per cent. Under these circum

stances small errors in the estimation of the quantities /, p,h, , and 77 are
greatly magnified and can lead to large errors1 in the estimation of quan
Once
tities dependent on 5k = k 1, for example, the critical mass.2
constant
exceeds
these factors have been determined to the desired accuracy, the kK for
the system is known and suitable expressions can be obtained for the
effective diffusion length and migration area of the lattice and, thence,
estimates of the over-all size of the core and reflector configuration.
The present treatment of the subject of heterogeneity begins with an
This portion
elementary discussion organized in the form of a summary.
Attention is directed
of the treatment is given in the present section.
here to the delineation of the essential features of heterogeneous systems,
the over-all effects of "lumping," and to simplified methods for estimating
Whenever possible, experimental
the characteristics of such systems.

results and formulas are presented which can be used directly in computa
The results and discussion given in this section should suffice
tions.
for a survey-type treatment of the general subject.
In subsequent sections of this chapter we examine in greater detail

various facets of the problems mentioned above. For these portions
of the treatment we delve more deeply into the theoretical basis for
several of the semiempirical relations customarily employed in lattice
calculations.
b. Methods of Analysis. The first step in the analysis of a hetero
geneous-reactor configuration is the evaluation of the various factors
in kx. In order to obtain a reliable estimate of these quantities it is
essential that one take into account the actual heterogeneous arrange
ment of the core materials; in point of fact, most any elementary mathe
matical model which does so will yield superior results to those based
Two analytical models devised

for this purpose have gained general acceptance.
The first of these is
the so-called unit-cell model of Wigner and Seitz, and the other is the
heterogeneous method of Feinberg and Galanin.
The essential feature of the Wigner-Seitz model* is that it represents
a typical lattice element of a heterogeneous array by an equivalent unit
For example, if a heterogeneous core were to be composed of
cell.
cylindrical fuel rods placed in a square lattice of spacing a, then a reason
able choice of the unit cell would be the circle of radius a/V* concentric
with a rod. In performing a calculation based on this model the usual
practice is to specify that the net neutron current at the boundary of the
on an equivalent
homogeneous system.
1 It is easily shown that in first approximation the

per cent error in ik is inversely
proportional to ik and directly proportional to the sum of the errors in the factors
/, pth, , and 7;.
* See, for example,
Eq. (6.191) and subsequent discussion.
5 Originally conceived for the calculation of wave functions of crystal
lattices.
HETEHOOEN'EOUS
SEC. 10.1]
KEACTORS
029
The neutron distribution in the unit cell is then obtained

cell be zero.
with the aid of diffusion theory or the first-flight transport model of
Having determined both
Sec. 5.8a, depending on the "size" of the cell.
distribution
of the neutron popula
and
the
spatial
spectrum
energy
the
tion in the cell, the thermal utilization and resonance-escape probabilities
may then be computed and, finally, the multiplication constant for the
Most of the calculations carried out in the present treatment of
reactor.
heterogeneous systems are based on this method.
The method of Feinberg and Galanin, on the other hand, makes a
direct assault on the problem of computing the multiplication constant
In this
of a given configuration by treating it as an integral system.
so-called "heterogeneous method" the fuel-rod lattice is pictured as a
A set of
collection of discrete line (or point) neutron sources and sinks.
written
this
be
for
and
may
the
solu
equations
system,
neutron-balance
tion of the set leads to the criticality condition for the reactor. This
general approach to the over-all problem is both elegant and precise;
it offers great flexibility in treating complex heterogeneous
arrays involving various lattice geometries, several degrees of periodicity,
The method is also useful for
and irregularities in rod size and spacing.
analyzing control-rod effectiveness, the influence of voids and cavities,

moreover,
It should be
as well as some aspects of the problem of fuel storage.
mentioned, too, that the mathematical treatment is somewhat more
complicated than that involved in the unit-cell method, and that even
the many attributes mentioned above, its
greatest usefulness is realized when applied to small heterogeneous arrays
The analysis of large regular geometries
and irregular configurations.
though the method
possesses
by this method offers no significant
advantage
over the Wigner-Seitz
model.
The last section of this chapter is devoted to a general review and

Although several of the more important
discussion of this method.
ideas involved are considered in some detail, the method as developed by
Feinberg and Galanin is not presented here in all its generality.
The
This section
present treatment is intended primarily as an introduction.
is concluded with an application to reactors with small number of fuel
lumps, followed by a numerical example.
c. Characteristic
Parameters.
The most important parameters in
volved in the description of a heterogeneous system are the four factors
in kx, the effective diffusion length, and the age; the latter two are of
course required in the computation of the neutron leakage.
In the
presentation
which follows
calculating/,
p,h,
we consider some elementary methods for

based on the Wigner-Seitz model.
These param
eters are considered first since the transition to heterogeneous arrange
ments affects these more than the others of the group.
Although the fast
in
is
also measurably improved
going to heterogeneous configura
effect
and
REACTOR ANALYSIS
630
[CHAP.
tions, it plays a lesser role than the others.1 A discussion of

migration area, is presented in later sections.
and
10
M, the
Utilization f. If the fuel and moderator are distributed

uniformly throughout a system, then it is clear that the nuclei of both
If,
materials will be exposed to the same neutron flux at all energies.
physically
other
the
two
materials
are
then
the
separated,
on the
hand,
In the
neutron density at a given energy will differ in the two media.
in fact,
depression in the neutron
case of the thermal flux, there
population throughout the fuel region. This behavior leads to several
significant differences in the characteristics of heterogeneous and homo
Thermal
geneous systems.
One factor which
is,
1.
is
extremely
the thermal
if
is
sensitive to these differences in flux

utilization. The important trends can be
Let Fu denote
readily displayed with the aid of the unit-cell model.
cell and VF and
the volume of such
u the volumes of the fuel and
0,(r) denotes the
moderator regions, respectively, in this cell. Then
of the cell, we can define the thermal utilization
thermal flux in region
distribution
-t
(10.1)
/..
If
the fuel and moderator regions are separately homogeneous and uni
form, then the 2<, = N&F are constants.
(Here AT,- denotes the actual
The above expression may then
nuclear density of the pure material i.)
written
>,
are computed from
where the average fluxes
-w
be
fc(r) dt
(10.3)
We introduce next the ratio of the average fluxes

(10.4)*
<PF
1.
s
Z(f>
2f(Fv/7,)
/,
Note that, although the major changes appear in

by the arguments presented in Sec. 10. la.
sensitive to
,
pth,
(1-5)*
and
km
is
/"
L,
if
if
is
is
at "disadvantage"
which we call the disadvantage factor, since the fuel
less than that throughout the
the average flux throughout its region
With this definition, (10.2) reduces to
>
moderator, i.e.,

for the heterogeneous system as
still highly
SEC. 10.1]
631
Some appreciation of the significance of this result may be had from a

comparison of this heterogeneous system to an equivalent homogeneous
We define such a system as one which has the same volume
system.
Vxll as the unit cell and moreover contains the same total mass of fuel
system, but now distributed uni
=
formly throughout VM. Then Vxll
VF + VM, and if we call N?m)
the nuclear density of material i distributed uniformly over V^,, then
and moderator
as the heterogeneous
clearly
N?"VC^ =
With this definition
we can write the thermal
geneous system in the form
where
NiVi
is the actual
(10.6)
utilization of the homo
7V
cross section of material
microscopic
i.
The

application of (10.6) yields
By comparison with (10.5)
we note that the two expressions for

have
Now as a
the same form, except for the factor f which appears in /hot.
rule f > 1, i.e., the flux in the fuel is depressed; thus we find that
A.. <
/ta
(10.9)
We conclude therefore that, in going from
a homogeneous system to a
heterogeneous, keeping the mass of each material constant, the thermal
utilization is reduced.
Another way of thinking about this behavior is to assign "effective"
cross sections to the materials in the heterogeneous system such that
Win
and
[*<>]. m
<r
(10.10)
Then, /h has
a form identical to /hom, but now expressed in terms of

Thus the equivalent homogene
[ff/'l.ff, which is generally less than a(aF).
fuel
with
smaller
ous system utilizes a
a
absorption cross section than the
actual heterogeneous cell and, consequently, exhibits a smaller value of/.
The disadvantage factor is a function of the fuel and moderator cross
As the scale of the hetero

sections as well as of the unit-cell dimensions.
In the limit,
geneity decreases, the homogeneous case is approached.
as the scale is reduced to zero, the flux in the fuel region approaches that
in the moderator region, f approaches unity, and the effective cross sec
tion for the fuel approaches the actual cross section. Various methods
for the computation
of this quantity are discussed in the next section.
REACTOR ANALYSIS
632
10
[CHAP.
The effect of lumping the materials in a reactor

on the thermal nonleakage probability is difficult to compute accurately.
Certainly, the most important quantity for consideration is the diffusion
This calculation could be approached from the viewpoint of the
length.
relation between the diffusion length and the mean square distance from
source point to point of absorption, as in Eq. (5.228) for the homogene
In a heterogeneous system such a calculation would involve
ous system.
finding the flux, over the medium, arising from sources at various posi
tions, determining the mean square distance to absorption for each source
point, then averaging over source positions.1 This is a complicated pro
Further, it should be
cedure and does not easily yield useful results.
noted that there would be various directional effects in a heterogeneous
array because of nonsymmetries in the shape and arrangement of the
Thus the apparent nuclear properties would vary with
fuel lumps.2
direction, and in a bare reactor the separation of variables previously
used would not be successful; the form (1 + ZAB2)-1 for pNL would hold

2.
Diffusion Length.
only approximately.
However, it seems clear that, in the case in which the unit-cell dimen
sions are small in comparison with the diffusion length, the form (1 +
L2g2)-1 Would be a good approximation to the thermal nonleakage prob
ability, and it would also be acceptable to use the general form L2 = D/2a.
Then, if
we use
"
D
along with (where
Fcen is
3[2{* +
(10-U)
Zf>]
is the average flux
the volume of the cell and
throughout Fdi)
Zu#^-n
- VPirV, + VpiuVu
(10.12)
and similarly for Sa; then,
r,
(4>V^Y
2VM!
(1
\4>mVm)
where
LL
(10
V
131'
i +%^vTmW7t
* Ll,{l
" /hJ
(1<U4)*
We found previously that in the

with
low fuel concentration a good estihomogeneous system
Probability.
of these considerations are incorporated into the Feinberg-Galanin

Refer to Sec. 10.5.
1 Some
Resonance-escape
case of a
/het)^'f
For small fuel lumps

is given by (10.4) and fhrl by (10.5).
therefore
at
1;
fa,
,/H!*>V
and
[cf. Eq. (6.155)],
xt
4>
UPV
F at VM,
3.
model.
SEC. 10.1]
633
mate of the neutron flux was given by the asymptotic result.

systems the resonance-escape probability took the form
Pth
exp
f"
/_
2a0
() dE
For
these
(6.174)
ty.^jj.
denotes some suitable energy-averaged value for the entire fast

For the present, more
range as denned, for example, in Eq. (4.121).
general treatment of resonance absorption, we require a broader defi
nition which can be applied to a greater variety of physical situations.
Following Wigner, we define the resonance-escape probability p(E) as
the ratio of neutrons which reach a given energy E in the resonance range
where
If the source of highto the number which enter the resonance range.
energy neutrons supplies to the system one neutron per unit volume per
unit time, then the appropriate statement for p(E) is
- -
n(F> =
p(E)
F(E') Xa(E') dE'
(10.15)
%{W)
F{E) = Z(()
where
(10.16)
4>(E)
in the system.
The corresponding
by [cf. Eq. (6.174)]

=
exP["
f" F(E)
Xa(E) dE~\
L~-ME)
is
the resonance-escape prob

and (10.17)
well
pth
of practical interest,
to unity so that
(10.17)
^-77^
1,
[~
F(E)Xa(E)dE
(10-18)*
approximated
close
plh
is
For many situations

ability
Pth
is,
denotes the total collision density

of course,
expression for p,h
In the present survey of the subject
we consider only the so-called

approximations in developing the theoretical basis of the dis
cussion.
These consist of the narrow resonance (NR) and the narrow
resonance-infinite absorber (NRIA) approximations (after the terminology
of Chernick1) which were developed2 to treat particular types of resonance
As will be discussed later in Sec. 10.3, the former model
structure.
is
first-order
in describing absorptions by very narrow resonances, such as

appear at the higher energy levels in U238 and Th232, and the latter, for
the few low-lying, rather broad resonances.
It of interest to mention
is
useful
4,
Chernick and R. Vernon, Nuclear Set. Eng.,

649-672 (1958).
B. Sampson and
Chernick, "Reasonance Escape Probability in Thermal
Reactors," in D.
E. Sanders (eds.), Progress in Nuclear Energy,
Hughes and
Physics and Mathematics, vol.
Series
pp. 223-270, Pergamon Press, New York,
1
J.
1958.
2,
J.
J.
J.
*J.
I,

1i
REACTOR
634
ANALYSIS
10
[CHAP.
that this distinction is required primarily because of the crudeness of the

models used ; more precise theoretical studies indicate that in higher-order
both yield results which are in good agreement.
The criterion for using the NR approximation is that the fuel reso
nances be widely spaced and so narrow that a single collision of a neutron
with a fuel nucleus is sufficient to hurdle the resonance. In these cases
approximations
ME)
for the flux over the
-jj;
the asymptotic expression is a good approximation

entire resonance range, and we can take
4>(E)
(10.19)
is
it
normalized source function.

With this
which follows from (4.134) for
to
the familiar form (6.174).
In studying
relation, Eq. (10.18) reduces
convenient to write the integral in
the resonance-escape probability,
(6.174) in the form
Nrfo'KE)
Nm<t(E)]
dE/E
2Q)
*>(E) + *<{"() + <>]

+ NM[aw>(E) + e<;"\E)\
'
where we have introduced the resonance-scattering cross section o{JP and

cross section apF) of the fuel and the concept of
the potential-scattering
for the interference
to account
an interference-scattering cross section <r' to account
for the interference
d
<rp
is
where
integral
ioI
by
We define the effective resonance
effects between the two.1

the relation
(10.21)*
the scattering cross section per fuel (or absorber) atom
*" +
7T
(6'177)
is
the scattering cross section of the moderator (here taken to be

and a(,M)
energy independent); then,
*M'\E)
<>{P
<r(amN
m/N F.
i>2)
obtained by utilizing the fact that frequently

Thus by neglecting all terms involving <rl*\ Eq.
good approximation
+ NMe{E)/Nr\ dE/E
+ <r(E)] +cp + Nm<jM{E)/Nf
is
- W/\E)
(10.22) reduces to
o<T{E)
'J. M. Blatt and V. F. Weisskoff, "Theoretical

John Wiley
Sons, Inc., New York, 1952.
&

Eth
[N,*<(E)
(10.23)*
+ op{E)]
Nuclear Physics,"
pp. 398-411,
SEC. 10.1]
635
which is the usual form in which the effective resonance integral is denned.
It is important to point out that this result applies only to a system with
narrow resonances; as noted above, the low-energy resonances of some
absorbers are not well represented by this model.
When the fuel resonances are broad, it is better to use the NRIA
fEt
/**
The corresponding
approximation.
geneous system is
1
-l
+
expression to (10.23) for a homo
ajF\E) dE/E
v.:
v..
<ri(E)Nr/
-if/>
involved in deriving this formula from (10.17)
The approximations
are
discussed in Sec. 10.3d.
Both expressions (10.23) and (10.24) may
be
written in the form
dK
8a(E) -=r
(10.25)
The functions sa and have several properties worth pointing out. These
may be illustrated with the aid of the expression for s developed for the
NR approximation.
(Similar arguments may be applied for the case of
Using then the form given in (10.23),
we make
o
the cross section ap
[<r + aiT + <Pl
that for
Op
(10.26)
+ aiT + <Pl
a very dilute system
(N?
>
The first observation
0)
m
1
Sa
approximation.)
is
NRIA
and
(10-27)
increases, av
Second, we note that as the fuel concentration

the ratio Sa/of* approaches the minimum:
,f-:t>
resonance integral
a{p, and
(io-28)
sa/
lies in the range
it
the effective absorption cross section sa lies

and
follows that the effective
<
1,
<
Thus in the NR approximation

within the limits o<f)/o\F)
Sf
)F
<
< Sx where
refers to
the fuel material alone, and
refers to the infinitely dilute fuelWe find, therefore, that as the fuel concentration
moderator mixture.
increases, the effective resonance integral decreases from its greatest
is
value Sx. The fact that

< S may be explained on the basis of the
markedly different behavior of the total collision density and the flux in
According to Eq. (10.19), the collision
the vicinity of sharp resonance.
density
reasonably smooth function of energy, provided that the sys
a

the
REACTOR ANALYSIS
636
[CHAP.
10
The flux, on the other hand, is seen to

This behavior is seen from the expression
dip sharply at each resonance.
for the absorptions per unit energy interval:
tern is not too
highly absorbing.
ME),(E)
'
mnE)
ME)
[ME)
MEWJ
- Wr
<n
+ e,]SE
When the number of fuel nuclei per scattering nucleus is large, o'f/tj, is
varies roughly as l/2aE; however, when the fuel concen
large, and
varies roughly as 1/2?.
tration is low, o^/op is relatively small and
Thus the higher fuel concentrations result in more pronounced dips in
the spectrum of the resonance flux and thereby fewer absorptions per
fuel nucleus per unit energy interval, hence, the smaller integral.
Another interesting observation involves the physical interpretation of
According to (10.21)
the effective resonance integral.

*
(10 29)
I/Np
is the total resonance absorptions per fuel atom, and 2,, is

the moderating power of the mixture [see Eq. (4.139)]; thus we may
identify d as the resonance absorption per fuel atom per unit moderating
Now
It follows that, by the trends noted above, the resonance absorp

tion per fuel atom per unit moderating power is smaller in a system with
However, the moderating
high fuel concentration than in one of lower.
power in a dilute system is greater than that in a system of high fuel con
centration, and since ZP increases more rapidly than sa as Np increases,
the total resonance integral / of the dilute system is less than that of the
concentrated system; therefore, p(dilute) > p(concentrated).
The application of theoretical models of the sort given in (10.23) and
(10.24) is developed in Sec. 10.3f, using the Breit-Wigner line shape to
describe each of the resonances implied in the various absorption and
Although this approach
scattering cross sections of the fuel nucleus.
appears to be fruitful, the theoretical basis is still in a stage of develop
Thus until detailed information is available on each of the
ment.
important resonances in a nucleus of interest, and more accurate calculational methods are developed, one must resort principally to experi
These
ment for reliable estimates of the effective resonance integrals.
integrals have been measured for many of the materials important in
reactor technology, and some are available as general functions of the
A general listing of the available data on infinitely
fuel concentration.
dilute mixtures is given by Macklin and Pomerance1 and by Dresner.2
power.
R. L. Macklin and H. S. Pomerance, Resonance Capture Integrals, Proc. Intern.
Conf. Peaceful Uses of Atomic Energy, Geneva, P/833, vol. 5, pp. 96-101, August, 1955.
' L. Dresner, The Evaluation of Resonance Integrals, Nuclear Energy, 2, 118-127
(1955).
SEC. 10.1]
637
The effective-resonance integrals for the two fertile fuels
U238
and Th232

have been studied in greatest detail, and empirical formulas have been
obtained for computing these integrals as functions of fuel concentration.
The best available experimental data are plotted in Fig. 10.2. Dresner1
has computed from resonance parameters, by methods to be discussed
later, the appropriate functional behavior and gives, as the best powerlaw mathematical fit to his computed data, the formula
g(Eih) =
This function
3.06V"-472
for uranium
(10.30)
curve labeled "uranium" in the

These data were computed for a mixture temperature of 300K,
figure.
and should be adequate for most reactor calculations.
is given by the solid-line
Similar calculations and measurements are given for thorium

The empirical relation obtained by Hughes and Eggler2 is
g(Etk) =
Figure
10.2
8.33a"-263
shows the available
(solid line) computed by Dresner.
for thorium
232.
(10.31)
experimental points and the curve

(It is believed that the experimental
point at o = 25 barns is in error.)

The resonance integrals for infinitely dilute mixtures of the various
lh.
Dresner, Nuclear Sci. Eng., 1, 68-79 (1950).

D. J. Hughes and C. Eggler, "Resonance Absorption
Laboratory, CP-3093, July 13, 1945.
1
of Thorium," Argonne
REACTOR ANALYSIS
G38
[chap. 10
These data are listed in

nuclear fuels have also been determined.
Table 10.1.
The problem of determining the resonance-escape probability in hetero
geneous configurations of fuel and moderator has received continued
Because of
attention since the very early days of reactor technology.
the difficulty of the problem and the paucity of cross-section data, much
of the initial effort was directed along experimental lines; however,
Table
10.1
Effective Resonance Integrals for Infinitely Dilute

Mixtures of Various Nuclear Fuels
(in barns)
Th"
U9
93 1
269 X
282
69. 8
86. 2%
291
442
400
467 %
Pu"
Reference
Chernick- Vernon"
:)
520
Macklin-Pomerance'
Weinberg- Noderer*
Trubey-Lessig''
Calculated value.
Measured value.
J. Chernick and R. Vernon, Nuclear Set. Eng., 4, 649-672 (1958).
6 R. L. Macklin and H. S. Pomerance, Resonance Capture Integrals, Proc. Intern.
Conf. Peaceful Uses of Atomic Energy, Geneva, P/833, vol. 5, pp. 96-101, August, 1955.
c A. M. Weinberg and L. C. Noderer, "Theory of Neutron Chain
Reactions," vol. I,
chap. IV, Oak Ridge National Laboratory, CF 51-5-98, May 15, 1951.
* D. K. Trubey, personal communication; see also D. K.
Trubey and R. H. Lessig,
Neutron Physics Division Annual Progress Report, Oak Ridge National Laboratory,
ORNL-2609, Oct. 29, 1958.

theoretical studies were attempted from the very beginning, of particular

Until more
importance being the contributions of E. P. Wigner.1
recently
it was necessary, for the reasons mentioned, to rely principally
Most recently,
upon measured values of the effective resonance integral.
more and better cross section data have become available, and it has
become profitable again to apply theoretical methods to this problem.
As a result, some new analytical models have evolved which have proved
quite helpful in further understanding and describing the resonance
behavior of lumped fuel.

It is especially encouraging that the recent
analyses appear to be in reasonably good agreement with measurement.
Some of the more elementary models used for this purpose are presented
in Sec. 10.3.
The present discussion, however, is limited to a brief
of
the
essential features of resonance absorption in hetero
summary
geneous arrays.
1 E. P. Wigner, "Resonance Absorption of Neutrons by Spheres," Argonne Metal
lurgical Laboratory, C-4, 1942; also, E. P. Wigner, E. Creutz, H. Jupnik, and
T. Snyder, J. Appl. Phys., 26, 257-279 (1955).
SEC. 10.1]
639
Much of the resonance behavior
of fuel lumps can be explained in

of
effective
cross
section sa of the fuel in the lump
absorption
the
terms
by direct analogy with the concepts developed in connection with the
In drawing this analogy we compare two systems
homogeneous systems.
containing the same fuel and moderator materials, but in one the fuel is
lumped and in the other it is homogeneously mixed with the moderator.
If we specify that the total masses of the fuel and of the moderator in
one system be the same as those in the other, then it follows that the
moderating power 2P will be the same in the two sys
However, the fuel in the lumped system behaves
tems [cf. Eq. (10.6)].
as if it were in a mixture of high fuel concentration, and so by the argu
ments presented in the discussion following Eq. (10.25), the effective
volume-weighted
absorption cross section sa in the heterogeneous system appears to be

Consequently the effec
less than that for the fuel in the homogeneous.
tive resonance integral in the heterogeneous system is also smaller than
that of the homogeneous, and since 2pom = SJ,01, we find that the reso
nance-escape probability is noticeably
"equivalent" systems,

Phet
If
A good illustration of this point

this material were
>
improved.
As
a rule, then,
for
(10.32)
Phom
in the case of natural uranium.

uniformly mixed with moderator, then in the case
is given
of an infinitely dilute concentration the resonance integral is approxi

On the other hand, if the uranium and moderator
mately 280 barns.
were to be physically separated, but with the value of 2P the same
for both systems, then the corresponding value of the resonance integral
is about 9 barns, a factor of about 26 reduction.
Advantages of this
order of magnitude
are
generally
achieved
in going to heterogeneous
configurations.
The earliest
formulations of the resonance-absorption

problem in
heterogeneous arrangements were given by Wigner1 and by Dancoff and
Ginsburg,2 who showed on the basis of theoretical arguments that the
effective resonance integral could be conveniently written as the sum
of a so-called volume-absorption term and a surface-absorption term.
The
validity of this approach and its particular usefulness in treating thick
lumps were confirmed experimentally by Creutz3 et al. These important
advances led in the United States to the adoption of the so-called standard
formula which expresses the effective absorption cross section as

function of the ratio of surface area to volume of the fuel lump.
a linear
More
1 Ibid.
1 S.
M. Dancoff and M. Ginsburg, " Resonance Absorption in Lumps and Mixtures,"
Argonne Metallurgical Laboratory, CP-1589, 1944.

8 E. Creutz, H. Jupnik, and T. Snyder, "Effect of Geometry on Resonance Absorp
tion of Neutrons by Uranium," Argonne Metallurgical Laboratory, C-116, 1951; also
Appl. Phys., 26, 257-279 (1955).
J.
REACTOR ANALYSIS
640
[CHAP.
10
recently it has been shown1 that for many situations of practical interest
a given heterogeneous system can be described in terms of an equivalent
homogeneous system; thus many of the results developed for the homo
geneous systems may be carried over directly to applications for hetero
geneous arrangements.
In the present summary

formula,
standard
we limit attention to a discussion of the

in part because of its historic interest and in part
because of the wide usage it has received throughout the technology.

The derivation of the equivalence equations and their relationship to
the formulation of the two-term standard formula are given in Sees. 10.3e
and i.
The United States, and generally Western, usage of the standard for
mula was based for some time on the following definition of the effective
:
s(w) = o(u)
4rl
cross section
J\l
absorption
K)
(10.33)
the total number of neutrons slowing down per unit

time past lethargy u in some specified region of
heterogeneous lattice,
say, for example, in a unit cell. At steady state the change in the
slowing-down density
precisely the rate at which neutrons are removed
be
Let Q(u)
is
If these removals occur primarily in the

from the cell by absorptions.
fuel component, then we must require that
= -VrNf8a{u)4>u{u)
(10.34)
that cross section which when multiplied

where, by definition, sa(xi)
by the flux in the moderator gives the correct number of absorptions
in the fuel. In this calculation 4>m(u) denotes the space-averaged flux
in the moderator at lethargy u. The slowing-down density
related
to the flux, however, by the relation
is
is
Q(u) = ImVmNm^4>m(u)
in (10.34) and carry out the integration,
is
we use this equation
seen
(10.35)
it
If
that
Q{u)
Q(0) p{u)
'See, for example, L. Dresner, "Resonance

Reactors," thesis, Princeton University, chap.
op. cit.
Absorption
6,

is
denotes the surface area of the fuel lump and M its massThe appropriate expression for the effective resonance integral
then
developed by means of the following arguments:
where
of Neutrons in Nuclear
Chernick and Vernon,
1958; also,
I"
dw
641
= exp
So(u
,1
VFNF
_____
exp
p(M)
where
/"
SEC. 10.1]
VrNrSiu)'
^^^.^
(10.36)*
is
a{u) du
These integrals are conveniently
0
JC
U,h
b(u) du
Mr Jo
(10.37)
expressed in the form

(10.38)
'"-(fsf;)8'
(ia39)
a(u) du =
/."*
fr4-'
Jo/
is
it
Thus
follows that the effective
the resonance-escape probability.
simply the integral of sa as in the homogeneous
resonance integral
systems [cf. Eq. (10.25)]. If we use (10.33), then,
fr
p'F
for the mass

material, and for the present, we have used the symbol
In terms of the above definitions,
density.
may be written
1,
f,
we take
= 8.6 barns
now,
if
Ar
and
Br = 27-5
is
attributed to "volume" and the second to "sur

where the first term
face" absorptions.
The functions Ar and Br have been measured for
various materials and shapes of fuel lumps.
Of particular interest are
the values for natural uranium; the measured results' are
then
= 8.6
27.5
uranium metal
(10.41)
is
where AF
given in square centimeters and MF in grams.
We observe
from this relation that, as the lump size increases, AF/MF decreases and
the effective resonance integral decreases.
In the limit for very large
1

fr
where
denotes the ratio of the "average resonance flux" in the mod
the
erator to the average resonance flux in the fuel lump; thus we call
resonance disadvantage factor in analogy to the
given by Eq. (10.4)
for the thermal flux. The Nf denotes the nuclear density of the fuel
Sampson and Chernick, op. cit.
REACTOR ANALYSIS
642
[chap.
10
0, and
lump (AF/MF)
-g/dE
a
= 8.6 barns
24.9
S = 20.4
3.16
24.5
27.4
(10.42)
5U,08 + 2C (by weight)
(10.43)
U02
(10.44)
two useful relations are'
4- 4r
24.9
U308
thorium metal
0.1 <
ff
= 10.4
d
3
case of thorium,
= 8.4
Th02
0.2 <
< 1.3
(10.45)
< 1-3 (10.46)
is
An independent formulation of the problem of resonance capture in

heterogeneous systems by investigators in the U.S.S.R. led to the devel
opment of an alternative expression for the effective absorption cross
section which varied as (AF/MF)t.
This work was published by Egiazarov, Dikarev, and Madeev4 and by Burgov,6 and was summarized
later by Gurevich and Pomeranchouk6 in a paper presented at the Geneva
Conference in 1955.
The form for reported in these works
[cf. Eq.
(10.47)*
L;
(10.40)]
Although this formula differs in form from the corresponding

or Western
relation,
the constants
in standard
use are such
American
that the
Ibid.
I.
3,
E. Hellstrand, /. Appl. Pkys., 28(12), 1493-1502 (1957).

E. Dayton and W. G. Pettus, Nuclear Sci. Eng.,
286-295
(1958).
of
5,
I.
I.
of
V. B. Egiazarov, V. S. Dikarev, and V. G. Madeev, Akad. Nauk. S.S.S.R. Conf.

on Peaceful Uses
Atomic Energy, Session of Physical and Mathematical Sciences,
p. 59, English translation by Consultants Bureau, New York.
N. A. Burgov, Akad. Nauk. S.S.S.R. Conf. on Peaceful Uses
Atomic Energy,
Session of Physical and Mathematical Sciences, p. 69, English translation by Con
sultants Bureau, New York.
Gurevich and
Y. Pomeranchouk, Paper P/649, vol. pp. 466-471, Geneva
Conference.
1.

In the
= 11.0
4^
4r-
&
4^
= 12.8
which is approximately -fa of the value for the infinitely dilute homo
geneous mixture of natural uranium and moderator.
Expressions of the
have
for
been developed also
the various compounds of
type (10.41)
uranium:1,2
SEC. 10.1]
643
effective resonance integrals for uranium metal are in adequate agreement1

0.5 to 6 cm in diameter.
When cast in the
form (10.47), the effective resonance integrals for the various materials2
for fuel rods ranging from

considered above become:
d = 2.81
24.7
6 = 4.15
26.G
hj^y
(jfj
uranium metal
(10.48)
UOj
(10.49)
thorium metal
(10.50)
Th02
(10.51)
and3

= 4.1
21.1
g = 0.2
25.2 (
(Jf\f
4^ J
d. Effects of Lumping.
The distribution of the fissionable material
in heterogeneous arrays leads to three important advantages and one
The advan
disadvantage from the standpoint of the reactor physics.
tages to be gained by utilizing a heterogeneous configuration, however,
generally outweigh the principal disadvantage, namely, the reduction
in thermal utilization. This effect has already been discussed with the
aid of an elementary analytical model, and it was observed that by
separating the fuel and moderator materials the fuel is placed at a dis
advantage in competing for the available supply of thermal neutrons.
Without entering into a detailed calculation, it is easy to establish this
The basis of the argument rests with
argument on physical grounds.
the fact that the slowing-down process, which is the primary source of
thermal neutrons, occurs in a region which is physically separated from
the fuel.
Thus the moderator region surrounding each fuel lump is a
source of thermals, whereas the fuel lumps are essentially sinks.
This
large source in the moderator provides a steady stream of thermal
neutrons to the fuel lumps; the "driving force" or potential is achieved
Thus
by maintaining a higher neutron density in the moderator regions.
the moderator is exposed to average flux levels higher than those in the
fuel.
The most important of the three advantages to
be realized from hetero

geneity is the increase in the resonance-escape probability due to the
Because of height and sharpness
self -shielding effect in the fuel lumps.
of the resonance structure of the fuel nucleus, the spectrum of the reso1 E. P. Wigner, "Review of Resonance
Capture by Lumps," Proceedings of the
Brookhaven Conference on Resonance Absorption of Neutrons in Nuclear Reactors,
BNL-433, pp. 68-73, 1956.
* Hellstrand,
op. cit.
' W. G. Pettus, "Geometric and Temperature Effects in Thorium Resonance Cap
ture," Babcock & Wilcox Co., BAW-TM-203, Lynchburg, Va., Apr. 27, 1959.
G44
REACTOR
ANALYSIS
[chap. 10
Because
nance flux exhibits local depressions around each resonance.
of the lumping, the majority of the fuel nuclei in the lump are exposed
to a depressed flux much like that in a homogeneous mixture of high
As argued previously, this corresponds to a rela
fuel concentration.
tively small value of <rp and of the effective absorption cross section of the
fuel 8a [cf. Eq. (10.28)]; consequently the effective resonance integral is
reduced and the resonance-escape probability increased.
There is another, somewhat secondary, effect which serves further to
Because of the physical
enhance the resonance-escape probability.
separation of the two materials, the neutrons involved in the slowingdown process have less opportunity to encounter a fuel nucleus while
traversing the resonance range; the number of resonance absorptions is
Clearly, this effect is due primarily to geometric
reduced.
thereby
considerations.

Finally, the heterogeneous arrangement results in an increase in the

Since all fissions
fast effect. The explanation is quite straightforward.
take place in regions of high fuel density (frequently pure metal), the
resulting high-energy fission neutrons have a greater likelihood of
encountering a fuel nucleus while en route to the outer boundaries of the
Thus there is an increased probability of causing a fast
fuel lump.
before
leaving the lump.
Moreover, any fast fissions which might
fission
occur can produce additional neutrons which can cause further fast
Although some advantage

fission, thereby producing a cascade effect.
is realized from this phenomenon, the gain is somewhat limited because
In the fast range, the
of the occurrence of inelastic-scattering collisions.
process is an important channel for compound nucleus decay
Nevertheless, the net result is
and competes strongly with fission.
usually a slight gain in the available number of fast neutrons.
Some of the difficulties involved in the
e. Multiplication Constant.
inelastic
computation of the multiplication constant of a heterogeneous reactor

have been mentioned in connection with the calculation of the thermal
The principal
utilization and the effective diffusion length of lattices.
difficulty stems, of course, from the fact that there are local variations
In the
in the neutron flux at all energies in each unit cell of the array.
central regions of the reactor core this does not lead to any particular
hardship since the flux distributions will be relatively flat and variations
This is not the case however in cells
from cell to cell will be gradual.
Here, variations across the cell can
near the boundaries of the lattice.
be substantial and will influence markedly such factors as the thermal
utilization and resonance-escape probabilities by distorting the neutron
These difficulties are by no means unique to
distribution in the cells.
the method based on the unit-cell concept; similar complications arise
In the latter appli
in the application of the Feinberg-Galanin method.
cation the difficulty appears in selecting suitable kernels to describe the
sec. 10.2]
645
behavior of lumps in the vicinity of irregularities and dis

continuities of the surrounding medium.
The unit-cell method offers a relatively simple computational procedure
As already demonstrated,
for determining the various factors in kx.
the calculation of these factors in each case reduces ultimately to the
source-sink
of the neutron spatial distributions for the entire energy

For the thermal utilization we require the thermal-flux distribu
range.
tion and in particular the thermal-flux depression (i.e., the thermal dis
The resonance-escape probability, on the other hand,
advantage factor).
determination
requires a knowledge of the spatial distribution of resonance neutrons,

and finally, of course, the calculation of the fast effect involves, essen
tially, the determination of the spatial distribution of successive (cascad
ing) generations of fission neutrons.
Once these factors have been determined, the remaining problem is to
compute L and r since these quantities establish the neutron-leakage
A first approximation to L2 has already been

rates for the finite system.
given, and a few remarks have been made about corrections for direc
tional effects. Some methods have been developed for including aniso
tropic effects in the computation
of the migration area, and these are

reported in a subsequent section.
10.2
Disadvantage
Factor and Thermal Utilization
Analytical Model for Unit Cell. A practical method for the calcu
lation of the disadvantage factor and the thermal utilization is based
on the Wigner-Seitz unit-cell model.
In order that this model be appli
cable to a particular situation, it is
necessary that certain simplifying
a.
Fuel
assumptions and conditions be satis

fied.
These
Moderator
may be illustrated by
means of an example.
the reactor-core
Suppose that
of in
Cell -
configuration
terest consists of an alternating series
Fio. 10.3 Unit cell in slab geometry.

plates (thickness 2a) and
slabs of moderator [thickness 2(6 a)] as shown in Fig. 10.3.
To begin
with we require that:
of
fuel
1.
All characteristic
cell half-width
with the
of the thermal utilization may
reactor dimensions
be large compared
b.
Then the calculation
proceed on the basis of an infinite medium of this structure, since the
regions near the edges of the reactor core (for which the disadvantage
factor will differ somewhat from the fairly uniform value over the central
region) will not play an important role in the criticality calculation.
The
disadvantage factor f we compute from a knowledge of the thermal-flux

646
REACTOR ANALYSIS
[chap. 10
distribution throughout the cell, and this we obtain with the aid of the
diffusion theory.
We further require, therefore, that:
2. Diffusion theory be valid in both fuel and moderator regions of the
unit cell. It must be pointed out that this assumption may be poor
on two counts: (a) The fuel region is likely to be highly absorbing for
thermal neutrons; thus the angular distribution of neutrons in the fuel
region and near the fuel-moderator interface in the moderator will very
probably deviate considerably from the distribution appropriate to
diffusion theory [see Eq. (7.93)].
(b) The dimensions of both regions
may be of the order of a transport mean free path or smaller, that is,
much smaller than a diffusion length in either medium.
The nonisotropy
of the flux in the vicinity of an absorber-moderator interface was demon
Figure 7.16<7 shows the typical strong forward bias
strated in Sec. 7.5a.
toward the absorber region which is to be expected in the angular dis
Finally, it should be mentioned that there
tribution at such boundaries.
is also a third difficulty which is encountered in the use of the diffusion
theory, namely, the neutron-energy spectrum around thermal varies
Thus there is some
appreciably through the moderator and fuel slug.
uncertainty in selecting the "one velocity" for the model.
In spite of all these shortcomings we will assume that diffusion theory
So the results obtained below will not be useful for treat
is applicable.
absorbing fuel plates (or pins) but could be well applied
thin,
highly
ing
to large natural uranium lumps in a good moderator such as graphite.
A suitable model, based on the "first-flight-transport" concept developed
in Sec. 5.8a, is applied in a later analysis to handle these cases.
It should
be mentioned, however, that disadvantage factors calculated by diffusion
theory can be low by about 20 per cent, giving an error in the multiplication
constant of about 2 per cent in practical cases for natural uranium with
graphite or heavy water.1 The success of the diffusion-theory treatment
during the Manhattan Project was achieved by adjusting the value of jj.
When more accurate estimates of the thermal utilization are required
than possible with the methods discussed here, it is necessary to resort
to the direct use of the spherical harmonics or the so-called Sn methods.2
Another essential question which must be considered in selecting an
analytical model is the specification of the thermal-neutron sources in the
cell. In the actual situation, neutrons are born with high energy in the
fuel plates; during the slowing-down
1
process they diffuse away from the
Personal communication, L. W. Nordheim.

for example, E. Bj0rnerud and L. VV. Nordheim, "Adaption of the S* Method
to Cell Calculations," General Atomics, GA-126, San Diego, Calif., May 22, 1957;
M. Johnson, "Code for a S4 Cylindrical Cell," General Atomics, GAMD-205, Aug. 28,
1957; and I. Bornstein, "S Calculations of a Proposed Chalk River-Ontario Hydro
Reactor Lattice," General Atomics, GAMD-242; these documents are company
internal publications.
2 See,
sec. 10.2]
647
original plate and may finally become thermal at a distance many slab
thicknesses removed from the source point at which they appeared as
fast neutrons.
In fact (neglecting the difference between the slowing-
properties of the fuel and the moderator), the distribution of

on the basis
neutrons becoming thermal from a source plate at x = 0
If the half-width of the

of age theory, just the Gaussian, exp
x2/4Tth).
Gaussian, or the slowing-down length (both of which are proportional
(
is,
down
VniO,
is
large in comparison with the moderator slab thickness

a), the number of neutrons becoming thermal will be roughly
2(6
Actually, the usual
uniform over a moderator slab (see Fig. 10.4).
to
10.4
Slowing-down
Fermi age theory.
density at thermal from plane sources on the basis of the
is
is
even more favorable for the assumption of uniform

physical situation
thermalization over the moderator region. It will be recalled that the
actual distribution of thermal neutrons
better approximated by means
a
This
of age theory when first-flight corrections are taken into account.
leads to
greater spreading out of the neutrons in most moderators
Thus for
because of the low scattering cross section at high energy.
is
is
3.
the analyses which follow we assume:

uniform source of thermal neutrons over the entire
There
moderator region, but no neutrons become thermalized in the fuel region.
The latter statement
reasonable because most fuel nuclei have large
mass.
In the infinite medium of alternating fuel plates and moderator slabs,

a
is
fuel plate
we may concentrate on a typical region which includes half
This region
moderator slab, as shown in Fig. 10.3.
adjacent to half
sufficient for the present purpose since we are interested only in the
ratio of the average fluxes in the fuel and moderator regions and since the
is
is
is
flux throughout the remainder of the medium

obtainable by symmetry
from that in the region considered.
This region
called a unit cell;
in general geometry the unit cell established by requiring that (a) the
fuel-moderator volume ratio in the cell be equal to the average ratio for
a
is
the entire core and (6) the net transport of neutrons across the surface
of the cell be zero.
of interest to mention that
In concluding the present discussion
it

Fig.
ANALYSIS
648
REACTOR
some of the assumptions
and approximations
[CHAP.
10
selected for the unit-cell
are very nicely developed in the general method devised by

It is especially noteworthy that the method
Feinberg and Galanin.
clearly displays how and when these simplifications may best be intro
model
duced into the model and the limitations which

system.
b. Disadvantage
they impose on the
Factor by Diffusion Theory.
We demonstrate the
calculation of the disadvantage factor on the basis of the diffusiontheory model outlined above for the special case of the slab configuration
This derivation is followed by a summary of the
shown in Fig. 10.3.
general results for the elementary lump geometries.
The mathematical problem may be stated in terms of the general
For the present application these
relations (5.349) through (5.354).
may be written
In
the
fuel (0 < x < a):
-DF^(x)
In
2/V,(x)
= 0
(10.52)
(a < x < b):
the moderator
-DM^(x)
+
+
2<Mx)
= So
where So denotes the uniform distribution of thermal neutrons

The appropriate boundary conditions are
lated in 3 above.
(1)
Ma)
(3)
*',(())
The solutions
<M (2) DF4>'F(a) = Du'M(a)

(2)
= 0
(4)
=
DF4>'F(a)
= 0
4>'M(b)
distributions
for the thermal-flux
Du'M(a)
4>f(x) and
postu
UUiWJ
4>m (x)
are
easily shown to be
.
cosh xf
- ID,*,*
,7a
- x)
cosh
- jTtfj f - rf/Wsinh
So
*'(X)
, v
.So
<M*)
where
xM{b
c"th
= a)
and
.\f(x) dx
r
J--ra
(b-
This result may

and moderator
a)
F(x)
dx
r
aS^F)
**
in
The disadvantage factor for this system is
- m ~- aW*
6V>
,.,
*(&-a)J
6 =
(10-4)
-sinh
)-
(10-55)
(10.56)
(10-57)
be checked by taking the limiting value of f as the fuel

From (10.56) we
regions become infinitesimally thin.
have
JSt
aZ<"
2(&
- a)
sec. 10.2]
649
Substitution into (10.57) yields, in the limit,
aZ
where we have kept the ratio a/b constant.

be expected, since to make the dimen
sions of both media smaller while hold
the
volume
ratio constant
This result
of course, to
= constant
to
is
ing
is,
lim
o,60
approach the properties of the homoge

neous system.
The disadvantage fac
tor, as calculated by (10.57), has the
form shown in Fig. 10.5 when plotted
It
b;
a)
(10.59)*
are defined, in slab geometry, by
= axr coth xFa
Vmx.m
coth
x.v/(6
a)
and
The functions
block of unit cross
written
be
(10.58)
so that
volume of moderator in unit cell =
Equation (10.58) may then
- -
(b
coth x,v(b
be taken as
section extending from x =

to x =
=
volume of fuel in unit cell =
Vf
a)
xM(b
In the slab geometry, the cell may
Vu
10.5
Disadvantage factor as
function of cell size.
- a)Z
xfQ. coth Kpa
+
(6
- a)2
Fio.
(10.60)
is
is
is
of interest to note that the quantity

in fact the ratio of the flux
at the surface of the fuel plate to the average value of the flux in the
This relationship
basic and applies to other geometries as well.
plate.
The form (10.59)
also useful in expressing
for two other principal
is
and E.
The two cases are:
geometries, with suitable expressions for
fuel
rods
in
lattice
and
spherical
lumps of fuel
cylindrical
square
(2)
(1)
in cubic lattice.
The appropriate expressions for the unit-cell dimensions
by requiring that the cell volume be equal to the volume
"element"
of the lattice; thus,
a
the
of an
the lattice spacing,
is
are computed
if

against size.
For convenience in presenting the cor
responding results in other geometries,
we rewrite Eq. (10.57) in the form
ANALYSIS
REACTOR
650
[CHAP.
10
lump radius, and Rm0 the radius of the unit cell, then the cell charac
teristics are given by:
V,
Lumps
Cylindrical
*R'
Spherical
frfi'
VM
a1
a3
Ke.ll
- t*
- |t3
ay
a'
The calculation
of the thermal-flux distributions for these geometries,

thence
the
disadvantage factor, proceeds in the usual way, starting
and
with the basic relations (5.349) through (5.354) and the requirement that
the net current vanish at the boundary of the cell. The results for the
three basic geometries are summarized in Table 10.2.

Table
10.2
The Functions
F
E
F for Various Cell Geometries
and
= atcF coth xpa

(6
a)\M coth
x.w(6
a)
Cylindrical
F
p
PfxfI o(xfpf)
21
i(xfpf)
., ,
*.v(p5#
pV>
, ,
2pr
h(x\IPF) Ki(xupsi) + K0(xnpr) I\(x!tpu)

Ii(xmpm) Ki(xmpf) Ki(xmpu) Ii(xmpf)
Spherical
r\xF tanh xftf

tanh
xfTf)
3(xrTr
i
4) r
1
x.vr.w
.
xjjr.urF
coth xu(ru
xm{tm
rr)
rr) coth xu(rM
ESlFuel
I
| Moderator
The function B 1 for the

Table 10.2 is plotted in Fig.
case of the
10.6.
cylindrical geometry shown in
For this purpose
we have used the
sec. 10.2]
definition
"{y'Z)
2y
y = x.MpF
with
[h(z) K,{y)
and
651
- K,{z) h{y)\
= xmPm
The diffusionFactor for Thin Fuel Plates and Pins.

results obtained above are not expected to be accurate when
c. Disadvantage

theory
0.02
0.03
0.04
0.05
Fig.
10.6
0.06
Graph of function
0.07
0.08
0.09
0.10
*mPf
for cylindrical fuel lump.
the fuel is present in the form of thin, highly absorbing plates or in the
form of fine pins.
For these situations it is better to apply the results
of the first-flight transport model developed in Sec. 5.8a.
The applica
tion of this model is demonstrated for the case of the thin plates, followed
by a summary of the corresponding results for the pins.
Consider, then, the slab geometry of Fig. 10.3 and suppose that the
fuel plate is very thin, so that we may assume that the plate either
absorbs a neutron incident
upon it or it does not; scattering
collisions
This formulation is in
are considered to have negligible probability.
accord with the definition preceding Eq. (5.275).
Let us assume further
ANALYSIS
HEACTOR
652
[chap. 10
that the diffusion theory is valid in the moderator and that neutrons are
being thermalized uniformly throughout the moderator region (with no
The boundary condition at the edge of the
thermal sources in the fuel).
unit cell is as before. The fuel-moderator interface condition on the
other hand is now conveniently given in terms of the transmission coef
ficient a of the plate [see Eq. (5.275)], since the plate is considered too
thin to allow the use of diffusion theory.
The quantity a may be com
puted in terms of the plate dimensions, its absorption cross section, and
the angular distribution of incident neutrons.
The calculation of the thermal-flux distribution in the moderator is
carried out in the usual way by applying (10.52) and (10.53), condition
For the interface condition we use the partial current relation
(4).
j+(a) = aj_(a)
cosh XM(b
a)
+ 2Dmxm
x)
tanhxAf(6
is
The disadvantage factor
In this application
is

convenient to compute 4>y from the relation
=
ZWM(a)
it
(10.62)
(10.63)
coth xM{b
1
1
xM(b
a)
<A^[(^:)'H
o)
- a)V>
(6
since the net current into the plate must be the rate at which neutrons
It follows that
are lost by absorption.
(10.64)
Note that the only properties of the fuel plate involved in this result
cross section and the quantity a. Thus, so long as
valid in the moderator (particularly in the vicinity of
the plate), the only new information we require in using this approach
is
are its absorption

diffusion theory
the diffusion-theory expression
is
is
it
is
an estimate of a, and this can be obtained by any one of a number of

It should be noted also
methods, e.g., transport theory or experiment.
easily shown that
that Eq. (10.64) implies the result (10.58), since
if

cosh xm{o
a)
vm(x) =
is easily found to be
^jj
The solution for this system
(10.61)
used for a, namely,
SEC. 10.2]
then the function
653
is given by [cf. Eq. (10.60)]
= 2a2;
(10.66)
The analysis for the system with thin fuel pins proceeds in the
Without further comment we list the results below:
way.
Hp)
S,
4>m(p)
2D,
where
- L(pf)
L(p) = Io(x\ip) Ki(xmPm) + Ii(xmPm) K0(xmp)

L'(p) = xm[Ii(xmp) K\{x.mPm) I\(x\lpsf) Ki(xsip)}
The expression for
has the form (10.59), with the
same
(10.67)
(10.68)
given in Table 10.2,
and
Prvr
(10-69)

The calculation of the transmission coefficients a in the
Fio.
Transmission coefficient for thin fuel plates.
10.7
members was carried out in Sec. 5.8a.

(5.331) and (5.332):
Thin fuel
Thin fuel
case of very thin
-plates:
-pins:
1
1
The results
- 4aSi"'
-
2PjP2/>
are given in Eqs.
(10.70)
It
is of some interest to compare the results for a using the diffusiontheory and first-flight transport models for a thin fuel plate.
These are
found to be, with 2a2^>
1,
First-flight transport:
Diffusion
theory:
These functions
- 4a2 +
+ 2[2a2]2
[2a2^]2/!;1[2a2^]
(10.71)
AaVaF)
are shown in Fig. 10.7.

sketched
in
Fig.
10.8.
are
The corresponding results for f
REACTOR ANALYSIS
654
[CHAP.
10
Thermal Utilization.
The calculation of the thermal utilization for

the heterogeneous reactor is carried out in first approximation by means
As previously mentioned (Sec. 10.1c), the thermal utili
of Eq. (10.5).
zation is decreased in going to a heterogeneous arrangement.
Moreover,
d.
Fig.
10.8
Disadvantage factor for thin fuel plates.
a given volume ratio VM/VF,

is further decreased with increase in
cell size.
This behavior is due to the increase in f , as may be seen from
Fig. 10.8. The comparison of for a homogeneous and a heterogeneous
system with the same volume ratio gives

for
/het
AIt
is understood
2<"
(10.72)
Z{VK/Vr)t
that the cross sections indicated here are for the pure
materials.
The general shape of the
ratio of values is shown in Fig. 10.9
as a function of a characteristic cell
dimension.
Fio. 10.9 Thermal utilization as func

tion of cell size.
is seldom achieved.
e. Engineering Considerations in
Unit-cell Analysis. The unit-cell
configurations discussed so far have
consisted simply of a region of
moderator
practice,
and a region of fuel.

In
of course, such simplicity
Actual heterogeneous lattices will contain
variety
of foreign materials and voids or passages whose presence is dictated by

For example, if the reactor of interest is to
engineering requirements.
produce power, flow passages must be provided to carry coolant for
The presence of the coolant must of course be con

extracting the heat.
sidered in determining the flux distributions in the unit cell.
Further,
in many instances the fuel and/ or the moderator materials may be sensi
tive to corrosive attack or oxidation
by the cooling fluids and therefore
sec. 10.2]
655
must be clad or plated with special materials.

Thus additional neutron
poisons are introduced into the system which may have a significant
effect on such parameters as the thermal utilization.
Since an enormous number of materials and passage geometry combi
are encountered in engineering practice, it is not feasible to
attempt a general treatment of the subject.
Thus in the present analy
sis we study a particular configuration
which includes two important
features frequently
engineering
required in practical designs. The
nations
general approach to this specific

problem will display several phys
ical concepts that are applicable

to other systems as well.

Consider then the cylindrical
geometry shown in Fig. 10.10.
In this system the fuel rod (radius
pf ) is clad with
a dissimilar
of thickness t.
the rod is an
Concentric
metal
with
annular passage
(outside radius p) for the flow of
coolant gas.
Beyond this is the
moderator
region, and the outer

boundary of the unit cell is taken
as the circle of radius pM.
For
,i
i .
r a'
.
the analysis
of this system we
of ceil
Fl- J0-10
Unit
cel1
for cylindrical
fuel
rod with cladding6 and coolant gas

passage,
B
H
6
assume that (1) diffusion theory applies in both the fuel and moderator
regions, (2) the cladding material has a transmission coefficient a, and
(3) there is a uniform source of thermals S0 in the moderator region.
On
this basis the pair of equations (5.349) and
are
applicable.
These
(5.350)
equations are to be solved with the boundary conditions
= 0
(1)
4>'F(0)
(3)
aprjr(Pr) + 0pjl"'(pa)
(4)
PFjLF>(pF)
(2)
(1
4>'m(pm)
= 0
f>j?>(pt)
- H)PaV'\pQ)
(10.73)
where
0
It
55
fraction of neutrons leaving the moderator at

the surface p heading inward which miss the
fuel lump
(10.74)
that relations (3) and (4) account for both the absorption
losses in the cladding material and for the fact that with a gap there is
always a nonzero probability that some neutrons will pass from moder
ator surface to moderator surface without encountering the fuel rod.
is seen then
This probability is the quantity
10
[CHAP.
CL(P) +
and
which follow from the application of conditions

stants
The con
(1) and (2).
are to be determined from conditions
(3) and (4)
and the
written in the form
be
vrvr*,
is
defined by (10.68).
function L(p)
The thermal utilization for this cell may
(10.75)
4>m{p)
A/,(x,p)
4>f(p)
/3;
some expressions for computing

Newmarch.1
derived
by
have been
The general solutions for the fluxes may be written in the form
/3
REACTOR ANALYSIS
656
nn
7(V.
is
is
ratio of
the average flux in the cladding to the average flux in the fuel lump.
For
the moderator and fuel regions we obtain the following expressions for
the average flux:
V^Z
FXF
2
So
â
2ttpL'(pii)C
V~7*
ttxM
*M ~
as well as the
'
2irpFh{xFpF)A
*'
Thus we require the usual disadvantage factor
(10.78)
Pf(1
If Eq.
(10.79)
is
7,
where
- 0)(
P,(1
and
are evaluated next.
From the boundary con
The constants
ditions (3) and (4) we obtain two equations involving these quantities.
Equation (10.73), condition (4), yields
7/0
lT^[CLM+w]
yM =
(10-79)
(10.80)
used in Eq. (10.73), condition (3), the result may be
D. A. Newmarch,
with
J.
-fi)(G
+ yF)
T+w+w+wr,
(10-81)
\Zt
(10.82)
Nuclear Energy,
2,
y* _
(1
written in the form

it
it
refers to the clad region.

Note that
assumed here that no
where
neutrons are absorbed by the gas in the coolant passage.
For compu
convenient
to
write
this
expression in terms of I//.
tational purposes
52-58 (1955).
SEC. 10.2]
657
Equation (10.81) is an explicit statement on C, by virtue of Eqs. (10.75)
If
and (10.80).
"
these relations are applied, we obtain
- gXg
(1
2DM[l +PG +
- - PKG + T'Wa)
PhrWM
(G
+ y,)S9/V>
(1

=
fJ)L'(p,)DS0
2ap,(l +(?)(!
PFZ^h{xFpF){2DM[l + 0G + (G + fihrWM
- - PKG + yF)L{p,)\
(1
(10.84)
GF
P,2<"j
l\ i
it
p
(10-85)
/in0M
(10-86)
(1-87)
are defined in the usual way (refer to Table 10.2)
P = pFXFIg(xFpF)
7rGp
/3G
i+
E* ~
2Jl(*,p,)
~ VMxlfL(pg)
(10-88)
arp,L'(p.)
If
Trp^l
0,
(10.89)
0)
'-?(rb)'
>
These results may be checked by considering the limiting cases wherein
In the first instance

a
and p pjr.
and
>
>
0,
/3
pP, then
/3
and F>F and 2?, E.

The miss probability has been calculated by Newmarch1 on the basis
of the diffusion-theory2 expression for the angular distribution of neu
In the case of round fuel rods, he finds
trons at pg [refer to Eq. (5.29)].
we also let p
P0
/(r) =
--
sin-1
IT
- IT
(1
where
+ W(r)
+ 1M
that
(10.90)
r2)*
/Wd.
H. Tait has performed similar computations using transport theory.
The
Calculation of the Fine Structure of the Thermal Neutron Flux in
Pile, by the
Spherical Harmonics Method, Proc. Intern. Conf. Peaceful Uses
Atomic Energy,
Geneva, 1955, P/433, pp. 417-422.
of
J.

vMx<m
T^y-wJ
and
Po
The functions
vFz<p)
i^g [tw^J
where
4>,
may be shown, after some

If these results are used in the relations for
algebra, that the disadvantage factor may be written
ANALYSIS
REACTOR
r>.-)8
[CHAP.
10
Newmarch performed his analysis for the rod without cladding, i.e.,
a = 1 and G = 0. The above expression for /3 can be taken over directly
for the more general case with cladding (a < 1), however, by simply
applying the appropriate expression for yM, namely, (10.81), Then, /S
takes the form
0 =
It
- r)(l + JrG)
(1
+ yrG) +
(1
(G
+ (G + yr)Kr)
+ 7*0/(r) + r(G + yr)
(10.91)
that this expression reduces to Newmarch's result when a 1.

When the coolant-gas passage is very narrow so that pp ~ p, then the
1 yields the approximation
substitution of the form r = 1 5 with 5
is seen
+ yfG)5
+G)(1 + yF)
(1
(1
(10.92)
Similar results have also
been obtained by Newmarch for the case of

fuel rods.
Of some practical interest, too, is the special case when the cladding
Then a ~ 1, and we define
on the fuel rod is very thin.
a =
For this
case
Gâ,,
PF
where a
and we obtain the approximations

1
Pi
+
1
El
/3a
- 0
I + _F
E +
(g
aJF^'
-
(10.93)
*P.(1
and (10.85) becomes
p7
W(i
P,(1
0)
xp^jj
p^t \E
0)J
+ a*
Pa
10-94)
J)
PF
+ E
p,
pF
Similar results obtained from
reported by other investigators.
ggggjg
irp(l
Trpr(l /3)J
(10.95)
0)
Po
+ _
V.,,2
irp(l
-0))
ipS
P.
Vm2?F
F,Z/>(1
0)
I [
The computation of the thermal utilization proceeds directly from

By substituting (10.94), we obtain
(10.77).

hollow-tube
slightly different viewpoint have
It
been
may be shown, after some algebra,
sec. 10.2]
659
that the above result agrees term by term with the earlier treatments.12
The only apparent difference between this work and the previous calcu
lations is in the handling of the boundary condition at the gap and the
The
description of the flux distribution through the cladding material.
of
is
the
more
subtle
consideration
the
two.
latter
The dis
perhaps
tinction between the two approaches on this point is easily demonstrated.
In the earlier work it is assumed that the flux distribution at the outer
surface of the cladding is given by the linear function
Mpf
t)~ c(pr)
U'cM
Mpf)
^(Py)
This
The last equality follows by virtue of continuity requirements.
quantity may also be obtained from the condition on the partial current
through the cladding,
appj^M
If
we use a =
Pc??(pp +
t)
pc,
a* in this relation, and the approximations pF
as small compared to unity, then it

4>'c(pf) 'c(pr + 0 and neglect a*

may be shown that
c(pf
t) = <Pf(pf)
+ 2aifDF'F{pF)
Comparison with the expression given above for

a* must be defined by
a* =
For
a very
If
the relation
(10-96)
2lTc
thin layer of cladding it would

a* =
c(pp + 0 reveals that
be appropriate
to use
2Sf'i
(10.96) is used in (10.95), the agreement is complete, as
previously mentioned.
There are several approximations which may be generally applied
in the analysis of composite unit cell structures of the type considered
here: (1) Narrow gaps and channels containing air or other gases are
most easily treated by requiring that the flux and the net current be
continuous
through the gap; thus,

4>f{pf)
DrVF(PF)
1 See,
M{p)
= DMV<t,(pg)
nno/n
UU-y/;
for example, E. A. Guggenheim and M. H. L. Pryce, Uranium Graphite

Lattices, Nucleonics, 11(2), 50-60 (February, 1953).
1 D. J. Littler and
J. F. Raffle, "An Introduction to Reactor Physics," pp. 90-95,
Pergamon Press, Inc., New York, 1955.
REACTOR ANALYSIS
660
[chap. 10
These are simplifications of the partial current conditions given in (10.73).

If the cell is multiregional, then (2) diffusion theory may be used in
each, provided, of course, that the regions have reasonable widths, and
If a
with continuity conditions to be satisfied at the interfaces.
region is very thin, then (4) it is convenient to use the transmission
For composite regions of
coefficient to describe its neutron properties.
different moderating materials it is necessary (5) to take the thermalneutron source in each as proportional to its value of 2, ; however, (6) if
the region is thin (and especially if it is "metallic"), then the thermal
(3)
source in the region may be ignored.
Without delving into any further detail it is possible also to draw a

few general conclusions about the effect on the thermal utilization of
various engineering materials in a cell: (1) In cylindrical geometry the
presence of gas gaps between the fuel region and the moderator tends to
decrease f and thereby increase /.
(2) Metallic sheaths on fuel lumps
tend to increase f since the metal1 will generally have a large scattering
cross section and thus serve as a resistance to the flow of neutrons into
the lump, thereby giving rise to a further decrease in the flux available to
fuel.
materials
containing
cause a
hydrogenous
(3) Channels
decrease in
because the neutron-absorption
losses tend to exceed the

the
producing)
properties of the material.
(thermal-neutron
In the case of water the situation is further aggravated because of the
small diffusion coefficient which creates a large gradient in the neutron
density.
moderating
10.3
Resonance-escape
Probability
Neutron-energy Spectrum in Unit Cell. It was mentioned previ

ously that the use of a heterogeneous configuration can lead to substantial
increases in the resonance-escape probability.
This trend is explained
in terms of the neutron-energy spectrum in the unit cell.
Figure 10.11
shows the spatial distribution of the neutron flux in a typical slab geom
etry as a function of neutron energy. It is seen that at neutron energies
a.
far removed from the values corresponding to the resonance-absorption

peaks in the fuel (and also away from the thermal range) the fuel region
However,
is exposed to flux levels comparable to that in the moderator.
when the neutron energy is in the range where the absorption cross section
of the fuel is relatively high, the flux distribution displays a depressed
region in and around the fuel lump.
The appearance of this depression
gives rise to a significant reduction in the removal rate of neutrons from
the slowing-down process. This is ea.sily demonstrated if we compute
the change in the total number of neutrons slowing down throughout
1
An important exception
is aluminum.
SEC. 10.3]
661
the cell over the lethargy interval Am; thus,

Q(u + Au)
Q(u) =
=
q(x,u +
- Jf
cell
Clearly, if the fuel cross section
Am)
dx
q(x,u) dx
2a(u)(x,u) Audx
(10.98)
is greater than that of the moderator,

<Kx,E3)
(.x,Et)
*(*.,)
i,h s
Fio.
10.11
Flux
E4
Ea
E3
depression
E,
as function of neutron energy.
the change in Q will be smaller in the case of the heterogeneous geometry

than in the homogeneous, since in the latter the fuel nuclei are in as
high a neutron flux as the moderator nuclei, for all neutron energies.
As
a consequence, the resonance-escape
heterogeneous system.
probability will
be higher in the
REACTOK ANALYSIS
062
[CHAP.
The increase in the resonance-escape probability has
10
attributed
to two effects, namely, the self-shielding of the fuel lump and the phys
ical separation of the two materials (refer to Sees. 10.1c and d).
The
in
study
of
section
is
to
these
effects
detail.
the
present
The
purpose
treatment begins with an elementary analysis of the latter effect, which
This study is followed by
arises from purely geometric considerations.
an outline of an analytical procedure (based on nuclear-resonance param
eters) for computing the integrals which appear in the expression for the
resonance-escape probability [see, for example, (10.25) and (10.37)].
Application is made first to homogeneous systems, and the results are
later extended to heterogeneous systems.
b. Geometric Effect.
The influence of the geometric effect on the
resonance-escape probability is demonstrated for the case of slab geom
Consider therefore an infinite heterogeneous array of the type
etry.
Previously, we considered the diffusion of thermal
shown in Fig. 10.3.
in
with an assumed uniform source of thermal
this
system
neutrons
been

neutrons in the moderator; we consider now the slowing down of neutrons

This problem is similar to that of the reflected reactor
in this geometry.
and, as before, introduces the difficulty of applying an energy-dependent
In order to simplify the analysis we eliminate
boundary condition.1
this difficulty by proposing the wholly artificial model based on fuel and
cross sections.
moderator materials having energy-independent
We
=
u
that
fast-fission
neutrons
slowingfurther assume
0) begin the
(at
down process in the moderator, and not in the fuel slab.
On this basis
we apply to the system in Fig. 10.3 the following set of equations for the
fast neutrons in the moderator:
-DM
, |
4>u{x,u)
along with the boundary

(1)
q-x
4>m(x,u)
2<
vytf(-K,M)
- ^ qtt{x,u)
9Q)
conditions
= 0
(2)
j^\a,u)
af^(a,u)
(10.100)
Here, i!^"(a,w) denotes the partial neutron currents in the moderator

at the fuel-moderator interface, and a is the transmission coefficient.
For the partial currents we use the diffusion approximation and assume
the separability of the space and lethargy dependence of the slowingdown density:
qM(x,u)
1 See
F(x) qM(u)
(10.101)
Sec. 0.3a for discussion of this point in connection with the extrapolated
boundary condition.
sec. 10.3]
663
Substitution into (10.99) yields the requirement that

= 0
F"{x) + \*F(x)
and the solutions
F(x)
qu(u)
= cos
=
g0
\(b
exp
x)
(10.102)
(10.103)
u\
where we have used boundary condition (1) in (10.100).

Note that the
distribution of source neutrons for the slowing-down process is given by
qM(xfl)
F(x)
= F(x)q0
qM(0)
(10.104)
The quantity
X resulting from the separation of variables

in (10.99)
by the application of boundary condition (2) in (10.100).
For a independent of lethargy, the appropriate relation is found to be
is determined
= 2
tan X(6
(y^T^)
(10.105)
a)
In the case of thick fuel plates, a may be computed from diffusion theory,
and is given by (5.308); for thin plates the result (5.331) may be used.
The calculation of the resonance-escape probability proceeds from the
If we compute the total number of neutrons
expression for qm{x,u).
reaching thermal throughout the entire cell, and divide this by the total
number of neutrons entering the slowing-down process, we obtain
Pw(tO
[%M(x,uth)dx
- ^s
qM(x,0)dx
Ja
This result is to
= exp
|-)
X2
UmK
S(.vn
Mf'~
fjfZJ
uJ
(10.106)
with the corresponding relation for the

equivalent homogeneous system. We noted previously in the discussion
following Eq. (10.57) that the results for the homogeneous system could
be derived from those of the heterogeneous by taking the limiting case
Then, using
of very thin fuel plates (holding the ratio a/b constant).
for
we
find
from
a,
(5.331)
(10.105)
be compared
- n ik?
DM{b/a
10.107)
1)

Phom(Mtb)
= exp
"^(.w)^
- a/b)
""'}

(10-108)
REACTOR ANALYSIS
064
[chap. 10
A comparison of this result with (10.106) shows the effect of the scale of
the heterogeneity on the resonance-escape probability (see Fig. 10.12).
The principal conclusions to be drawn are: (1) as the cell size becomes
larger the resonance-escape probability increases monotonically from its
initial value corresponding
to that
in the equivalent homogeneous systern; and (2) for large cells the
'
Piwt(",h)
resonance-escape probability
proaches a limiting value.
It
Phom("",)
is
important to point out that

model used in the
the elementary
present
8
Fio.
ap-
^_
Resonance-escape Probabillty as function of cell size, with b/a held

constant.
10.12
calculation
was
selected
only to display the influence of the

geometric effect on the resonancee
e probabiiity
and it is not
,
.
intended that it be used as a pro
cedure for computing this quantity,

Calculation of the Collision Density. The general statement of the
resonance-escape probability given by Eq. (10.18) was denned in terms
of the total collision density throughout the resonance range.
An impor
tant approximation to this expression was the formula (6.174) which
applies in the narrow resonance approximation wherein the collision
In the present treatment we
density takes the asymptotic form 1/E.
require somewhat better approximations for the collision density in order
to develop more generally useful expressions for the effective resonance

c.
In fact, the calculation of the energy spectrum and spatial

dependence of the collision density in systems containing substantial
amounts of fuel (or strong absorber) is the essential problem in resonance
absorption.
In the case of homogeneous systems, the various approximations to
the collision density may be developed from the general expression (4.1 19)
integral.
If we limit attention to a system

derived in the slowing-down theory.
containing an absorber, denoted by subscript F, and a moderator, denoted
by M, then the appropriate form
F(u)
fu
F(u')
r-^ /
+
where
is
2(F)(V)
TJ
is denned by (10.16).
e'- du'
/
If Eq.
W^r^-V-d*'
(10.109) is solved for
(10. 109)
F(u), then
formula (10.18) to
This calculation may be
compute the resonance-escape probability.
carried out with whatever precision is consistent with the requirement
the resulting
expression may be used in the general
sec. 10.3]
665
In this development
we derive various first-order approximations

Nordheim1 has prepared a direct numerical procedure appropri
ate to fast computing machines which may be used to achieve greater
on p.
to F.
precision.
A first approximation to Eq. (10.109) may be obtained by introducing

the condition, of considerable practical importance, that the resonances
in the fuel (or absorber) material are narrow on the energy scale relative
to the average energy loss suffered by a neutron in a scattering collision
atom and, furthermore, are widely separated.
In this
case we can use the asymptotic expression for the function F which
appears in the collision integral for the moderator material and note that
it is also consistent to take 2i3 ~ 2, over the interval of integration.
with
a moderator
From the analysis of Sec. 4.4b

that for a normalized source
we have q(u) = Z((m)
F()
and it follows
(10.110)

which is a good approximation for a system with light absorption.

The
use of the asymptotic form in the second integral is permissible since,
as previously noted, the collision density is a relatively slowly varying
function, even through a strong but sharp resonance

With these simplifications, Eq. (10.109) reduces to
F(u)
^5 /
F{u>)
(see
Fig. 4.21).
(10.111)
e'- du' +
or
In the event that all resonances of the fuel
are narrow relative to the

average energy loss in a collision with a fuel atom, the asymptotic form
(10.110) may be used throughout in evaluating the effective resonance
integral.
Similar approximations
may be developed for the collision density in
In the development which follows
we represent
heterogeneous systems.
the heterogeneous system by a unit cell of volume V consisting of a seg
ment V r which is occupied by the fuel material and a segment V m by
As a zeroth-order approximation we impose the flat-flux

namely, that the neutron flux throughout the resonance
range, including the intervals occupied by the resonances, is spatially
uniform in both volumes V? and V m. This simplification is not valid
the moderator.
assumption,
when: (1) the fuel elements (represented by Vr) are closely packed com
pared to the mean free path \\m in the moderator material, or (2) when
the spacing of the elements is of the order of the slowing-down distance
(i-jy) in the moderator, or (3) when the characteristic dimension of the
fuel elements is much larger than
1 L. W. Nordheim,
"The Theory of Resonance Absorptions,"
GAMD-638, San Diego, Calif., Jan. 6, 1959.
General Atomics,
666
ANALYSIS
REACTOR
When the requirements
[CHAP.
10
for the flat-flux assumption are well satisfied,
the total number of neutrons slowing down in the cell Qo may be approxi
mated by the asymptotic form; thus we take
*s(U)isrr
where one neutron
We define
with
is introduced
Z7P =
N pa'-p,
and
Qo
(io.ii2)
per unit time in the cell volume
i,2>V, + ZmV^Vm
V.
(10.113)
denotes the spatially uniform resonance flux

In order to compute the total collision density
4>m
in the moderator region.

throughout the fuel volume, we introduce 4>F(u), the spatially uniform

(or averaged) flux at lethargy u in VF from all sources, and (m), the
spatially uniform flux at lethargy u in VF arising wholly from neutrons
of lethargy u entering by transport through the surface AF of the fuel
element from the moderator region.
We require also the quantity P.(w) ,
the probability that a neutron appearing in VF (from some source) at
lethargy u escapes without collision from Vr- For the present analysis
we take P*K to be independent of u and defined by Eq. (7.234); in this
application the appropriate form is
gpj ft [1
- e-*nHs)
ds
(10.1 14)
with s given by (7.232). It should be recalled that h(s) is the normalized

distribution of chord lengths in Vr, each chord length weighted by the
sum of the cosines of the angles it makes with the normals to the surface
of VF at each end.
A cosine weighting is applied for the neutrons at the
surface of VF since this gives a reasonable first approximation to the
angular distribution of the neutrons when the resonance absorption of
the fuel lump is small (as implied here) and the net transport of neutrons
is low.1
With these definitions we write the total collision density in VF as con

sisting of two parts: one due to fast neutrons in VF which slow down into
the unit lethargy interval about u, and another due to the net transport of
neutrons of lethargy u through the surface AF from the moderator; thus,
*,() Zj"()
\I
aF
/
Ju-\n
,
**(')
du' + 2<r\u)
(10.115)
Following (7.227) we take

4>'F{u)
P.
4>t,{u)
1 S. M. Dancoff and M. Ginsburg, "Surface

Resonance Absorption
packed Lattice," Argonne Metallurgical Laboratory, CP-2157, 1944.
(10.116)
in a Close-
sec. 10.3]
667
and using
FF(u)
m 2<"()
(10.117)
and the expression (10.112), Eq. (10.115) reduces to the form
"
lT~7
fLi F>M WW)
(ian8)
e""u du' +
The solution to this equation which applies in the asymptotic
If we take also
obtained by setting FF(u) = constant.
be consistent, then it is easily shown that
FV(u)~J^L
and
{2.7
4>F(u)
at
range is
^<r,^to to
=L
2.7
(10.119)

The results developed above may be used to obtain the first-order

estimates for the collision densities mentioned previously in Sec. 10.1c.
When the resonances are narrow relative to the average loss in energy
due to collisions with fuel nuclei and in addition widely separated, we
may use the narrow resonance
F(u)
(NR) approximation:
NR for homogeneous
(10.110)
for the homogeneous
The corresponding expression for the

system.
heterogeneous system is obtained by using the result (10.119) in the
integral of Eq. (10.118) and taking at ^<r, over the integration interval.
The solution is
FF{u) =
JLL
{2.7
seen to be
Wlp,(u)
(1
NR for
heterogeneous
(10.120)
Another interesting approximation is obtained when the fuel resonances

are narrow compared to the average energy loss in collisions with the
moderator, but wide compared to that in collisions with the fuel.
These
constraints lead to the narrow resonance-infinite absorber (NRIA) approxi
In the case of the homogeneous system the appropriate form for

the collision density is obtained from Eq. (10.111) by withdrawing the
factor F2J')/2i from inside the integral sign and solving for F. The
solution is
mation.
F{u) = -^-rT-^TWTT-^
[2,(u)
2/(u)]
For the heterogeneous system
we use
NRIA for homogeneous
(10.121)
Eq. (10.118) and find, by making
REACTOR ANALYSIS
668
similar approximation
=
F,{u)
Z.V
Pm
_____
NRIA for
aIFHu)
-(i-P._)^y
10
to the fuel integral,
-r|i
F
[CHAP.
heterogeneous
(10.122)
Formulas for Resonance Integrals. The calculation

of the resonance-escape probability (10.18) requires the evaluation of the
d. Approximate
In terms of the lethargy variable this
resonance integral.
is
2a(w) du
F(u)
The effective resonance integral may
be computed by means of the

relation (10.21).
In this development we apply the various approximations for F(u)
derived in the preceding analysis to the calculation of d for both homo
We
geneous and heterogeneous systems, using Eqs. (10.123) and (10.21).
In the narrow resonance approxi
(10.123) yields the result previously

notation, we have
J_
Substitution into
(10.110).
given in (10.20); in the present
is
^>(M) du
fuih
&P J0
j +
_- KFKu)
Op
(10.124)
+ <#'()]
#>()
and [cf. Eq. (10.23)]

N
<P(u)
11
for homogeneous
,
,Tr>
/ml
k/>()
vfKu) du
a<f\u)]
(10.12o)*
Again we note that these formulas apply

rp given by (6.177).
when the resonances are narrow and far apart.
The criterion of narrow
ness
discussed in
subsequent section.
a
is
with
is
Similar forms may be obtained in the NRIA approximation. The

in this case
form for
given by (10.121), and application to (10.123)
yields
-t:
- 2<(u)]
2(w) du
f[2((u)
which may be written
Z_">
cri"(M)
_ 2i"'
jo
2<(w)
Jo
2/>(u)
2__u) du

begin with the homogeneous system.

form for
mation, the appropriate
*{u)NrMM,NM
(10.126)
sec. 10.3]
069
in comparison to X'/K With this result,

the effective resonance integral becomes [cf. Eqs. (10.24) and (10.21)]
if
we neglect everywhere
= frmI =
Jo
2">
"
)Z
+ olP(y)/(rm
NRIA for

=
crm
<r
homogeneous
(10.127)*
(10.128)
formulas may be derived for the resonance integral in

heterogeneous systems by applying the definition
Corresponding
[ FF(u) *<f\u)
du
M0
12QU
(10.129)*
*r(u)
"Jo
which is given for a normalized source of fast neutrons throughout the

as specified in (10.112).
This definition implies, of course, that
essentially all resonance absorption occurs in the fuel element here repre
sented by the volume Vp.
The expressions for Fp{u) in the NR and
unit cell,
NRIA

tively.
approximations are given by Eqs. (10.120) and (10.122), respec

Both forms involve the escape probability which may be defined
P^
- Pe
(10.130)
where Pc is the probability that a neutron born in VF makes a collision

before escaping.
The quantity Pc has been tabulated by Case-de Hoffmann-Placzek for a variety of geometric shapes.1
It is sometimes more
convenient to write Pe in the form
P.
- ^~
9m
and
f[l- e-^r)]h(s)
ds
(10.131)
where we have used (10.114) for P,; note that the penetrability 9 is
related to the transmission coefficient a of Chap. 5 by [cf. Eq. (5.275)]
9 =
In general,
-a
(10.132)
as mentioned
previously, P, may be computed from the

integral relation (10.1 14) ; however, this formula is somewhat complicated,
and the resonance integrals will be difficult to calculate.
A considerable
simplification is possible if we use in place of the exact form Wigner's2
rational approximation
which we will designate
- r+W1
K. M.
the definition is
(10-133)*
Case, F- de Hoffmann, and G. Haczek, "Introduction to the Theory of

Diffusion," Los Alamos Scientific Laboratory, 1953, Tables 2 through 4,
(available through Government Printing Olfice, Washington, D.C.).
* E. P.
VVigner, /. Appl. Phys., 26, 255-279 (1955).
Neutron
REACTOR ANALYSIS
670
This formula is
[chap.
10
function when
Some values of
sX(tF) is either much greater or much less than unity.1
P*k are given in Table 10.3 for various elementary shapes, along with the
corresponding values of P. These results are also plotted in Fig. 10.13.
It is of interest to note that the dependence on geometry is not strong;
P*,. for spheres and cylinders are very close, with slabs differing only
The rational approximation is seen to be quite good.
slightly.
Table
10.3
a good
to the actual
The Escape Probability for Various Geometries and the

Rational Approximation
0.04
0.1
0.2
0.3
0.5
1
2
3
5

approximation
10
Sphere
Cylinder
Slab
Rational
0.978
0.946
0.896
0.850
0.767
0.607
0.411
0.302
0.193
0.099
0.974
0.939
0.885
0.819
0.753
0.596
0.407
0.302
0.193
0.099
0.952
0.902
0.837
0.785
0.701
0.557
0.390
0.295
0.193
0.100
0.962
0.909
0.823
0.769
0.667
0.500
0.333
0.250
0.167
0.091
In the present formulation of the resonance integral for heterogeneous

systems, it will suffice to use the rational approximation for P in the
expression (10.129) for I. In the NR. approximation this formula reduces
to
I=
UW
(1
- P-X'l
(10.134)
and
If
"
1 =
h [P^r
(1
"
(1-135)
Pe,c)<>1
we use now (10.133) for P, we obtain

g
=/Jo
+ &'t
'
The appropriate expression in the
NR for
NRIA
heterogeneous
approximation
(10.136)*
is seen to be
Nrf, f^P^Tdu
2.F jo
It
follows
1 The
op. cit.
/ +
PrfS"
that, in the rational approximation,
the effective resonance
errors involved in the use of this relation are discussed
by Dresner, Thesis,

10
Characteristic dimension
Fio. 10.13 The escape probability and the rational approximation as functions of the
characteristic dimension S2<1,,).
integral is [using the definition
f uth
* =
given in Eq. (10.135)]
(F ) flu
Jo
NRIA for heterogeneous
of Homogeneous
e. Equivalence
and Heterogeneous
(10.138)*
Systems.
The
approximate formulas derived above for the effective resonance integral

in homogeneous and heterogeneous systems in both the NR and NRIA
approximations can be shown to be mathematically identical.
first the corresponding expressions in the NR approximation:
$ =
/u"
/
Jo
I
jo
If
<rjF)
du
[1
aP
[v!P +
quantity
homogeneous
(10.125)
heterogeneous
(10.136)
<" +
+ sVpyaF) du
1 + sZ'
we define a new
Consider
a* such
that
(10-139)
672
REACTOR
ANALYSIS
[CHAP.
10

3 =
a(F) du
"">
a/ok."
i +
Jo
air + <n
(10-14)*
in form to (10.125).
Thus we find that a hetero
geneous system is equivalent to a homogeneous system in the NR approx
imation if (tp for the homogeneous system is the same as the a* for the
which
is identical
It
heterogeneous system.
is of interest to note also
that different hetero
geneous systems with the same <r* have the same effective resonance
This equivalence relationship has been recognized by P. Bakintegral.
by L. Dresner.2 3
Similar relationships can
shi1 and
approximation.
also in the case of the

We compare, therefore, the formulas
3 =
be established
/""th
du
u
j _L_(f)/,
r/> du
1
we introduce
homogeneous
(10.127)
heterogeneous
(10.138)
the symbol a* for the second term in (10.139), so that.
<j%
(10.141)
then
(10.142)
and (10.138) may be written
uth
CT(^>du
is
mathematically identical to (10.127).

which
It should be borne in mind that the rational
approximation
(10.133)
was used in establishing the two equivalence relations (10.140) and
These relations are valid only so long as the width of
(10.143).
a
resonance can be considered small as compared to the average energy

loss from collisions with
moderator atom at energies around that of the
Generally the results obtained above will yield estimates

resonance.
for the heterogeneous integral which are in the order of 10 per cent too
small.4 More precise analyses than those possible with the present
to the collision density indicate that the equivalence
approximations
relations are more generally valid6 and hold independently
6.
L. Dresner, Thesis,
1,
Chernick and Vernon, op. cit.

op. cit., Chap.
L. Dresner, Nuclear Set. Eng.,
501 (1956).
Nordheim, op. cit.
L. Dresner, unpublished.

If
+ 52/ >
NRIA
of the treat-
sec. 10.3]
ment
of energy
degradation
673
collisions with fuel atoms.1

needed in the application of these
in neutron
Thus the principal modifications
relations are those arising from the geometric properties of the hetero
These become important when any of the conditions
geneous system.
imposed by the flat-flux assumption are not satisfied.
The equivalence relations may be summed up in terms of two very use
ful practical rules: (1) All fuel lumps containing the same fuel (absorber)
material, but with different amounts of other moderator materials, and
having the same value of <r*, have the same effective resonance integral
regardless of shape or size.
(2) The effective resonance integral for a
lump is the same as that for a homogeneous mixture of the moderator
and lump materials in which the potential-scattering
cross section per
fuel atom is equal to <r* [as given by (10.139)].
An interesting interpretation of rule 1 is that data on the resonance
integral for a series of fuel lumps of one material can be used to com
pute directly the value for a related series of lumps utilizing the same
"absorber" but in an entirely different geometry.
Thus, for example,

data on uranium metal rods can be used to predict the resonance integrals
for UO2 spheres. It is of considerable importance to note that these
rules do not depend on any knowledge whatever of the resonance param
eters. They can be used both for predictive purposes and as a check
on the theory.
f. Evaluation of Resonance Integrals Using Breit-Wigner Line Shape.
The objective of the present calculation is to determine the contribution

$i to the effective resonance integral by a particular resonance at energy
Ei, using the various models previously developed for estimating 8.
In this treatment we assume that each resonance is well represented by
the single-level Breit-Wigner formula.
The appropriate relations for each
of the energy-dependent microscopic cross sections appearing in 8 are :2
(10.144)
(10.145)
relations apply when the nuclei are at rest in the laboratory

frame of reference (L).
The symbol 2?, denotes the position of the reso
nance, and is the kinetic energy associated with the relative velocity
These
between neutron
and nucleus.
Chernick and Vernon, op. cit.
It
is shown
op. cit., pp. 391-394.
in a subsequent
analysis
674
ANALYSIS
REACTOR
that, in the
case of
stationary nuclei,
8 =
[CHAP.
8 is given
10
by
ny
(10.147)
where v is the neutron speed in the laboratory system of coordinates, and

n^, the reduced mass, is denned by
* - Yâ
(1(U48)
with m and A denoting the mass of the neutron and nucleus, respectively.
and rn are the radiative capture and neutron widths, and
The symbols
I\
ff0
T m
r +
(10.149)
T7
Note that
plus scattering)
<ro is the peak value of the total (absorption
All the relations indicated here apply to the center-of-mass
resonance.
system of coordinates, and X, the reduced neutron wavelength, is given
by1

The symbol
equation
statistical
35
2i
=
2t(2m*S)
J'
(10.150)
Planck's constant, and the last equality in this

by virtue of (10.147). The quantity gj denotes the
h denotes
follows
weight and is defined
9J =
where
2^
is the
2(21' +
(10-151)
1)
total angular momentum of neutron and nucleus and
/'
is the spin of nucleus.2
The contribution to 3 from an isolated resonance at Ei is obtained by

substituting the relations (10.144) through (10.146) into the appropriate
form for & and integrating over the resonance (i.e., the entire energy
For these computations it is convenient to introduce the variable
scale).
x =
and the parameter
b
Consider
1
It is
| (8 - Ei)
(10.152)
(10.153)
(Tpl
now the application of these functions
to the evaluation of
shown in Eq. (10.176) that in the case of stationary nuclei the relative energy=
E. There
^
fore, the relations (10.144) through (10.146) have the same form in (L) if the widths r
are also measured in (L).
* Note that
/'
-J for neutrons having zero orbital angular momentum
note
UiS8 and Th*", gj = I.
waves);
also
that
for
(s
E is
related to the energy in the laboratory system (L) by
/'
sec. 10.3]
Si on the basis of the
NR
If
approximation.
we use the form (10.23),
which is Eq. (10. 125) in terms of the energy variable

for Si using the Breit-Wigner formula is
,' _
~
Jre, cp +
<r(l
675
, then the expression
+ s')]-Hr/2s) dx
+ x2)-1 + 2x(<r0b)Kl +
U.io*J
a2)"1
the total width r is appreciably less

than , and it is a good approximation to ignore the energy dependence
of (To, T, and 8 over the range of integration and simply evaluate them at
Ei. Moreover, the limits of integration on x can be extended to +
In the
case of narrow
resonances,
Then, with some minor rearrange
without introducing serious error.
ment, (10.154) may be integrated to yield
J-.
dx
0r2 +
2(0b)x
Sf
(1
N
0
for homogeneous
(10.155)*
(10.156)
CO
if
and
(10.157)*
is
Two limiting cases of this result are of special practical importance.

When the mixture
very dilute, <rp o and
Si^ST
(10.158)
b,
is
is
is
is
it
j3
is
This result
the so-called infinite-dilution effective resonance integral.
The sum of " over all resonances has been measured for many materials1
At the other extreme, when <3C
(see, for example, Table 10.1).
as often happens in the low-energy resonances of uranium and thorium
or other absorbers in fairly concentrated mixtures,
seen that Si
pro
roughly verified by experiment.2
portional to a\. This rule
We found previously that in the NR approximation the effective
resonance integral for heterogeneous systems could be written in the form
mathematically identical to (10.125) or (10.23); thus
(10.140) which
It
fol
=rr
NR for
+i - b)T
heterogeneous
(10.159)*
Si
is
the result (10.155) may be adapted to heterogeneous systems.

lows that in this approximation 3,
given by
Macklin and Pomerance,
op. cit.
J.
See, for example,
J.
A. C. G. Mitchell, L.
R. Pruett, and E. D. Nering,
Brown,
"Resonance Absorption of Uranium in Mixtures," Argonne Metallurgical Laboratory,
CP-1676, Mar. 31, 1944.
'

where
[l+J(l-b)J
It
2Ei
/S)
9*
REACTOR ANALYSIS
676
[chap.
where
(TO
or
|8
and a* is denned by (10.139), and
10
(10.160)
by
b*
(10.161)
When
/S*
<5C 1,
condition nearly
always
fulfilled for the important
resonances in most materials,

The factor in the brackets is plotted in Fig.
10.14 as a
function of
/3//3
t>= 1
4,
Fio.
10.14
Geometric dependence
56789
of resonance
10
integral.
for various b. It is seen that all the curves are asymptotic to (/3 //?)*.
Thus with increasing /3//8, the dependence of the effective resonance
integral
upon the interference between potential
and resonance scatter

for the heterogeneous
This square-root dependence

disappears.
systems is analogous to the o\ dependence mentioned above for the homo
ing
geneous systems.
The evaluation
procedure.
If
of
using the
we use the form
NRIA approximation follows
the same
(10.127) for the homogeneous system,
then
l<tm
SEC. 10.3]
677
The substitution of the Breit-Wigner formula (10.144) for

of variable
formation
6i =
<r andand
trans
trans
from S to x according to (10.152) yields
X.
(i +
*2)
[i + VorX-r)(i + *')-]
(iai64)
applied in going from Eq. (10.154) to

Then Eq. (10.164) reduces to
We use again the approximations
Eq. (10.155).
Si =
=rrr
r[I+.
(<Tol y/(Tml
NRIA for
)]'
homogeneous
(10.165)*
withjJf given by (10.157).
Since theform (10.127) is identical to (10.143)

the result (10.165) may be used to obtain
heterogeneous
system,
for the
Obviously,
the corresponding expression for the heterogeneous system.

Si = n j 7
NRIA for
'
7 m
heterogeneous
(10.166)*
It should be noted that both results (10.165)

<r* given by (10.141).
are independent of inter
and (10.166) in the NRIA approximation
ference effects between resonance and potential scattering, whereas the
results for the NR approximation are not; specifically, the quantity in
the parentheses of Eq. (10.159) is a correction factor contributed by the
interference effect.
The heterogeneous system formulas (10.159) and (10.166) for the con
tribution to S based on the NR and NRIA approximations may be used
to deduce the dependence of the effective resonance integral upon the
surface-to-mass ratio (AF/MF) of the fuel lump.
In the cases wherein
the NRIA formula applies, it frequently occurs that (<r0Ty/a*Y) ~ 1.
Then the form (10.166) is well approximated by
since by (10.141)
*^K^y~(^y
*
m Mfr,
=
(ioi6?)
(10-168)
r[i
is
If
is
p'P
the density of the fuel-lump material.

where
similar dependence
found also for the NR approximation.
we write out the various
terms in Eq. (10.159), then
For the two very interesting materials
+ *:/<>;
and Th232, rn/r <3C over

the most important part of the resonance range;1
we note also that in
if
U238
and
2.
Sampson and Chernick, op. cit., Tables

with
REACTOR ANALYSIS
678
then the above expression may be approximated
r<7(F,i
ST
-M
<r*
by
(10.169)
thin,
s
When the fuel lump
is
f),(NR)
10
"i*
2> 1 ,
general oo/o'f'
[CHAP.
and [cf. Eq. (10.162) and Fig. 10.14]

(10.170)
*(NR)~^Y
Thus for many situations
is
of practical interest the effective resonance

integral varies as the square root of the surface-to-mass ratio of the fuel
It also of interest
lump in both the NR and NRIA approximations.
to note that in mixtures when om/o{p > 10 both approximations approach
the asymptotic value Sf [see also (10.158)].
Line. The
g. Effect of Nuclear Motion: The Doppler Broadened
reaction-rate
relation
w(v) =
|ya-(|v
- V|)|v - V|2fl(V,0)
dV
(4.210)
for the cross sections in the resonance integrals in place of the natural
line shape given by the Breit-Wigner formula.
and use the Breitwe base the calculation on the NR approximation
is
Wigner formulas for the a" which appears in (4.210) and the MaxwellBoltzmann distribution (4.213) to describe the motion of the nuclei.
This general problem has been examined by
number of investi
In
the
treatment
which
follows
we
draw
principally from the
gators.1
The calculation
analysis of L. Dresner.2
developed from an approxi
mate expression for the relative energy between the neutron and the
For this purpose we require the velocities w and W (for
nucleus.
neutron and nucleus, respectively) in the center-of-mass system of coordi
nates (C). These are [cf. Eqs. (4.6)]
w =
iTa(v_V)
rr~A(V~v)
(10171)
H. A. Bethe and G. Platzek, Phys. Rev., 61, 450 (1937); E. Creutz, H. Jupnik,
Appl. Phys., 26, 257-259 (1955); E. P. Wigner, E.
T. Snyder, and E. P. Wigner,
Creutz, H. Jupnik, and T. Snyder, J. Appl. Phys., 26, 260-270 (1955); E. Creutz, H.
Jupnik, T. Snyder, and E. P. Wigner, J. Appl. Phys., 26, 271-275 (1955); E. Creutz,
H. Jupnik, and E. P. Wigner, J. Appl. Phys., 26, 276-279 (1955); G. M. Roe, "The
Absorption of Neutrons in Doppler Broadened Resonances," KAPL-1241, 1954.
68-79 (1956); also, Thesis, op. tit.
L. Dresner, Nuclear Sci. Eng.,
1,
J.

results of the preceding analysis apply strictly to the case of stationary

nuclei.
More properly we should take into account the relative motion
between neutron and nucleus and therefore should use the generalized
sec. 10.3]
Thus the
relative kinetic energy in system (C) is given by [cf.
and (10.147)]
iw! + pfW*
E, --
in,
(v
679
- V)2
Eqs. (4.9)
(10.172)
Now, if n0 denotes
where m* is the reduced mass defined in Eq. (10.148).
the cosine of the angle between the velocity vectors v and V, then the
expansion of (10.172) yields
Eni
j-p^- (E
+ imV*
- mvVno)
(10.173)
Further, let the

where E = \mv"-, the neutron energy in system (L).
relative velocity between the two particles be g = v V, where
=
p2
Clearly,
Enl
fr.p-

l^V*
= 8
6 =
with
In.v*
(10.174)
2vVn0
- (2M8)Uô
E
(10.175)
(10.176)
Note that, if the nuclei were stationary, 6 would be the relative energy
This was mentioned previously in connection with
of the neutron.
Now, in most situations,
Eqs. (10.144) through (10.147).
(10.177)
(^)S~^1
Thus (10.175) may
be approximated
Eni
by
P =
[l
"
(10.178)
(2M8)FMo
We have also
~L)]1
=* v
Fmo
(10.179)
These results may be used to obtain an estimate of the reaction rate

If we note that a" varies
which appears in the integrand of (4.210).
rapidly in the resonance range whereas |v V| varies slowly, it seems
reasonable to introduce the approximation
|v-V|~y
Then, Eq. (4.210) reduces to
r(i>)
Jv<r-(|v
- V|) 9JJ(V,0)
(10.180)
dV
(10.181)
This result may also be written in terms of the variable 8, in place of v.

Let us define V = V(Vx,Vt,V,) and take the direction of the component
V, to correspond to v; then Vfi0 = V and
Enl~&-
(2M.S)*F,
(10.182)
REACTOR ANALYSIS
680
[CHAP.
10
V|, and is related to E, through Eq. (10.175), and to

Since p
|v
V, through (10.182), we can write am = <r(p) = oK{V,).
Substitution
into Eq. (10.181) and integration over Vz and Vv yield
Using (10.182) to replace V, by E\,
(i^rj
^Mf] *
-p
[
/:.
obtain
we
(10.184)
directly in the resonance integral for the various

For the cross sections (r(Ei) which
cross sections a'/\ o-'f, and a'/K
appear in (10.184), we use the single-level Breit-Wigner forms indicated
in Eqs. (10.144) through (10.146). In the present application the
defined [cf. Eq. (10.150)]
reduced neutron wavelength
be used
is
This result may
2,(2^)l
(1-185)
f
P
- Enly]
4/crM.s
the total width
Generally speaking,
(î\
('
(&)
M(S
Only when
Ex
cross section the
relation
dEnl
(10.186)
is
is
is
less than Et.
is
of an absorption resonance
near Ei
there
significant
contribution to the integrand of (10.186) therefore, as before, we assume
independent of energy and set its value to be that at the
that aorT/T
If we call
resonance.
appreciably
(8
'
- Et)
- Ei)
ym (Enl
(10.187)
y)t]
{x
exp
dy
the Doppler broadened line shape.

of the natural line shape
- *
It
is
This function
rh
7$t
/_"
is

These substitutions yield in the case of the absorption
(10-188>*
the counterpart
sec. 10.3]
given in (10.144) and obtained

* oo
) ; thus,
(#
by taking the limit of ^(t?,x) as

=
lira
681
*(,x)
=
1
-r
* 0
(10.190)
xz
In applying (10.188) it is important to note that i? depends on 8 through

In the remainder of this analysis it will be assumed that this depend
A.
unimportant by virtue of the narrowness of the resonance, and
in A is replaced by Ei. An important property1'2 of ^(#,x) is that
ence is
8
^(#,x) dx =
t.
This result leads to the interesting conclusion that
in an infinitely dilute system the area under an absorption resonance is

independent of temperature, even if one takes into account Doppler
broadening.
lim
is
which
(10.193)
x(*,i)
x(,x)
2r
^r
the generalization
be used to obtain
of the effective resonance integral in both the NR and

to the case of Doppler broadening.
The calcula
developed from the resonance-integral formulas
approximations
which follows
is
tion
(10.194)
the natural line shape given in (10.146).
The functions (10.188), (10.191), and (10.192) may
NRIA
(10192)*
0,
and in the limit as
= 2x*(tf,x)
dy
Note that
x(*,x)
"
(*
(10.191)*
/_".
&
^4=
rrp exp "
for the homogeneous
0.
system and does not include the interference

=
This
between resonance and potential scattering; thus, we take
for
two situations of practical
result will be a good approximation
The various properties of

have been studied by
number of investigators;
for example, M. Born, "Optik," J. Springer, Berlin, 1933; G. M. Roe, "The
Absorption of Neutrons in Doppler Broadened Resonances," KAPL-1241, Oct. 15,
1954; M. E. Rose, W. Muanker, P. Leak, L. Rosenthal, and J. K. Hendrikson, WAPUA. Wheeler, "Effect of Temperature on Shape
SR-506 (1954); and A. T. Monk and
of Absorption Line," Argonne Metallurgical Laboratory, C-66 (1954), declassified.
'See also F. T. Adler, G. W. Hinman, and L. W. Nordheim, "The Quantitative
Evaluation of Resonance Integrals," General Atomics GA-350, San Diego, Calif.,
September, 1958.
a
see,
J.

Expressions of the form (10.188) can also be obtained for the resonance
These are easily shown to be
and interference scattering cross sections.
REACTOR
682
ANALYSIS
[CHAP.
10
importance, namely, a purely absorbing resonance (r

r) or a very
dilute mixture [a'/' <. <rp]. The latter case occurs most frequently in
It is also of interest to mention that there is a third situation
practice.
term can be neglected, namely, in the case of
strong Doppler broadening.1
We begin with the NR approximation.
The substitution of (10.188)
and (10.191) into the formula (10.23) written in terms of the variable S
wherein the interference
yields [upon omitting the a',f ' term]
Now
[We have used here also (10.149) for T and (10.156) for /3.]
=
r dx/2&, and if as before [refer to Eq. (10.155)] we consider
dE/8
narrow resonances and ignore the energy dependence of T and 8 over
the range of integration and simply evaluate them at Eit then (10.195)
may be approximated
S*-
by
2E, }-2Ei,r
&
E<
Jo P + +(0,x)

A similar result is obtained for the NRIA approximation.
uulwo;
In this
case
we use the form (10.163) and find
where we have applied the same approximations

5
=
<70l
The results (10.196) and (10.197) are

fore, if we define
as in (10.196) and called
(10.198)
seen to be of the same form.
WA- f;*4&
There
(10.199)*
then
Si = -1=--
J{d,fi)
Si = -TT- /(#,$)
in
NR
in
NRIA
approximation
(10.200)*
approximation
which gives the contribution to the effective resonance integral by the

resonance at E{. These formulas may be used, then, to evaluate the
effect of the various resolved resonances on the basis of the resonance
The necessary parame
parameters and the temperature of the system.
ters are available in the standard sources.2
1
Dresner, Thesis, op. cit.

for example, Hughes and Harvey in United States Atomic Energy Commis
sion, "Neutron Cross Sections," McGraw-Hill Book Company, Inc., New York, 1955.
* See,
sec. 10.3]
683
The function J(d,&) has been tabulated and graphed by Dresner,1 and
analytical expressions are given by Roe.2-5 Dresner also lists some of
the more important properties of the function and various series expan
sions suitable for computation.
Of particular interest here is the limiting
form of J{&,&) when t> , which corresponds to T 0. This is the

then reduces to the expres
case with no Doppler broadening; clearly,
sion for the natural line shape (10.189), and
takes the limiting form
Also noteworthy is the fact that, for both
J(#,j3) ~ ir[/3(l + (3)]-*.
limits f} > 0 and jS oo ,
approaches the value for the natural line
shape. Note that at either limit J(&,P) is independent of t>; thus the
This property
resonance integral becomes independent of temperature.
was previously mentioned in Sec. 6.5g in connection with the temper
Figure 10.15 shows
ature coefficient of the resonance-escape probability.
It is of interest
a plot of J{&,&) as a function of 0 for various values of d.
=
by the
to mention that, when d
1, J(d,fi) is again well approximated
asymptotic form given above for t? > oo .
h. Evaluation of Resonance Integrals.
In the early days of reactor
technology the determination of the resonance integral was based pri
marily upon measurement and on the use of rather approximate analyti
This approach was favored because at that time the knowl
cal models.
For
edge of neutron cross sections was limited in scope and precision.
approximation
was con
homogeneous systems the narrow resonance
sidered generally applicable.
For heterogeneous systems both the early
theory and experiments suggested that the resonance integral could be

expressed as the sum of a term proportional to the fuel-lump volume

and another term proportional to the surface area.
Although it was well
in
that
the
available
was
a
crude
state
recognized
theory
indeed, the lack
of good cross-section data lead soon to the adoption of the so-called twoterm "standard formula" which enjoyed general popularity until about
the time of the first Conference on Peaceful Uses of Atomic Energy held
at Geneva in 1955.
The subsequent publication of large amounts of
reactor information by both the Western countries and Russia revealed
a substantial difference in the analytical formulation of the resonance
A careful comparison4-5 of these
integral for heterogeneous systems.
showed, however, that over the practical range of fuel-lump
sizes these differences were minor.
Since then, many new measurements
have been made of the detailed resonance structure of the important
reactor materials, and it is now possible to calculate the effective reso
nance integral for homogeneous systems to within a few per cent.
Heterformulas
>L. Dresner, Nuclear Sci. Eng., 1, 68-79 (1956); also Thesis, op. cit.
' Roe, op. cit.
' See also Sampson and Chernick, op. cit.
* VVigner, op. cit., BNL-433.
6 Sampson and Chernick, op. cit.
ANALYSIS
REACTOR
684
[CHAP.
10
ogeneous systems, however, are still analyzed by approximate but gener

ally more accurate models than previously used and by the Monte Carlo
method.1
The viewpoint and various analytical models derived in the preceding

of this section represent the essential features of some of the
portions
ioJ,
5
1
-< ==0.1
-0.2
,1 =
*.
1.0'

i\\X
ll
4
I
6
10
*(where(3
Fio.
10.15
I
12
I
14
I
16
I
18
= 10"6 X 2*)
The function J{d,fl).
in the calculation of resonance integrals.

In
particular, the so-called NR and NRIA approximations may be applied
to obtain first estimates of the contribution to the effective resonance
In the case of the important fuel
integral from the resolved resonances.
more recent developments
materials
U238
and Th232, the resonance structure up to about 500 ev and

has been resolved.
The error due to oversimplifying
400 ev, respectively,
D. Richtmyer, " Monte Carlo Study of Resonance Absorption

Reactor Lattices," NYO-6479, May 1, 1955.
1 R.
in Hexagonal
sec. 10.3]
G85
the slowing-down problem is minimized if a choice is made between the

formulas on the basis of Wigner's practical width for the individual reso
nances. This quantity Tr is defined as the span on the energy scale over
which the Doppler-broadened
resonance cross section including both
and scattering is equal to or greater than the background
Note that
cross section <rp as defined by (6.177).
potential-scattering
the practical width therefore depends in general upon the material com
position of the system.
Values for the practical width for the principal resonances of U238 and
The comparison of the
Th232 are given by Sampson and Chernick.1'2
maximum energy degradation apEt in a fuel-atom collision with the
absorption
practical width TT gives a measure of whether a single collision can be

expected to remove a neutron from the neighborhood of the resonance.
If otpEi is considerably greater than rr, it may be expected that the neu
tron will be removed by such a collision, and the NR approximation is
If, on the other hand, aÊi is small compared to 1%, the
applicable.
scattered
will remain
in the region of the resonance,

should apply.
On this basis Sampson
and
and
approximation
Chernick show that the NR formula is applicable to all but the lowest
several resonances in the case of the pure metals U238 and Th232.
How
the

neutron
NRIA
ever, these particular resonances account for a considerable fraction of

the total resonance captures, and for these the NRIA formula is prefer
able.
Specifically,'-6
these are,
for
U238,
the resonances at 6.7, 21, 36.9,
66.3, 103.5, and 192 ev, and for Th232, those at 22.0, 27.7, 70.1, and 174 ev.
In the
case that the fuel material appears in oxide form, it might be

expected that the larger value of aoxEi, where ax refers to the oxygen
atom, would cause the NR approximation to apply for all resonances;
however, this does not prove to be the case because well within the
practical width the oxygen cross section cannot compete with the reso
nance cross section, and so the approximations appropriate to the pure
metal apply to the oxides as well.6
In the absence of strong Doppler broadening, i.e., at relatively low

temperatures,
the general result (10.155)
may be used to estimate the
contribution of resolved narrow resonances in homogeneous systems, and

The corre
(10.165) may be used for the resolved broad resonances.
sponding formulas (10.159) and (10.166) may be used similarily for
heterogeneous systems on the basis that the equivalence relationship is
applicable to the case at hand.
The accuracy which one can achieve in
1
Sampson and Chernick, op. cit., Tables 1 and

Hinman, Nordheim, op. cit., Table I.
Ibid.
4 Chernick and Vernon,
op. cit.
' Hughes and Harvey, op. cit.
' Sampson and Chernick, op. cit.
1 Adler,
2.
REACTOR ANALYSIS
68(5
[CHAP.
10
using these relatively crude models depends, of course, upon a judicious

By exercising some care,
choice in the application to specific resonances.
it
is possible1
to obtain good estimates for the contributions of the resolved

N
resonances
0<
(where
denotes the number of these for the particular
=i
material).

NR
and
More precise computations
NRIA
formulas
and various
based on modified forms2 of the

second-order theories3-' for the
of the slowing-down problem indicate that there is good agree

ment between the two models except in the case of the very broad, lowFor all other
energy resonances where the NR formula is less accurate.
treatment
resonances the more precise versions of the two models are found to yield
similar results.7 Such a comparison is shown in Fig. 10.16 for U238.
1 Chernick
*
and Vernon, op. cit., Figs.
and
K. T. Spinney, "Resonance Absorption
2.
Mixtures," Proceedings
of the Brookhaven Conference on Resonance Absorption of Neutrons in Nuclear
Reactors, BNL-433, pp. 103-109, 1956.
3 Sampson and Chernick, op. cit., pp. 236-240.
4 K. T. Spinney, J. Nuclear Energy, 6, 53
(1957).
5 Chernick and Vernon, op. cit.
'A. M. Weinberg and E. P. Wigner, "A Second-order Correction to the Usual
Resonance Escape Formula," Proceedings of the Brookhaven Conference on Reso
nance Absorption of Neutrons in Nuclear Reactors, BNL-433, pp. 125-126, 1956.
W. Rothenstein, Nuclear Set. Eng., 7, 162-171
(1960).
in Homogeneous
sec. 10.3]
687
In the event of appreciable Doppler broadening, the general results

may be used to determine the contributions of particular reso
nances. It should be noted, however, that the usefulness of these
formulas is somewhat limited because of the omission of the scattering(10.200)
interference term (which is important for thick lumps1) ; when this term
is negligible, however, the extensive compilations of the function J(&,(i)
are generally and directly applicable.
The unresolved resonances in the fuel materials account for roughly

third of the resonance absorption and are therefore generally impor
Estimates of the contributions from these higher-lying resonances
tant.
have been based on statistical methods, using average values for the
widths Ty and r and the resonance level spacing D derived from the
A detailed dis
resonance parameters of the lower-lying resonances.
one
cussion of the essential ideas involved in such computations is given by

Sampson and Chernick,2 by Chernick and Vernon,3 and by Dresner.4
For our present needs it suffices to indicate simply the general form used
for this purpose.
If

En denotes the energy of the last resolved reso

nance, then the contribution 3U to the effective resonance integral of all
resonances above En may be computed from the relation
!L JM T
'-w
JM 7
(ia201)
where Pn has the value of or/a0 at the resonance En] the averaged param
eters I) and
are discussed by Dresner.6 The approximation
sign
in the range of interest, T ~ 2F7, or greater.
applies when
I\
rI\;
Note also that when T

pH =
Also,
that
1%,
'^s^ (\ ^ â^
it is assumed in Eq. (10.201) that

Cp/cro
varies in first approximation
En
FT is independent
(10 202)
of energy and
like E.
In order to complete the theoretical development of the effective reso

nance integral we require also the contribution of the l/v portion of the
The appropriate result was obtained previously in Sec.
From Eqs. (6.182) and (6.187) we have (refer also to Fig. 6.8)
resonance range.
6.5g.
fE' (coA) dE,

jEih 1 + Co/fO
/
Dresner, op. ext., Thesis.

* Chernick and Vernon, op. cit.
*L. Dresner, Nuclear Energy, 2, 118-127
* Ibid.
*
(1955).
(10.203)*
REACTOR ANALYSIS
688
With
[CHAP.
these results the effective resonance integral may be
10
written
* = ay>(ifc)
* + ^ /" Jim f
(io.204)*
=i
where the function di is to be selected from the two forms given in (10.200)
according to the characteristics of the particular resonance represented
by the index t. This general form has been used by Dresner to compute
the resonance integrals for various homogeneous mixtures of both U238
and Th232 in moderator.
His results are summarized in Table 10.4 for
Table
Nucleus
Calculated Contributions
Integrals of U"8
10.4
T, K
TJHB
barns
Resolved
Unresolved
resonances
resonances
7.9
20
200
20
74.
8.1
80
17.
200
27.
86.

2,000
Th"
12.
2,000
nances.
the
Figure
3.0
6.6
9.4
12.
33.
2.2
6.3
14.
20
4.0
80
8.9
4.0
8.6
200
15.
12.
2,000
42.
20.
system temperatures
computation
11 .
3.7
20
200
300
1.7
4.8
24.
2,000
300
to the Effective Resonance

Tns"
and
NR
10.17
of
0K and T
l/v
1.4
1.4
1.4
1.4
Effective resonance
integral, barns
11.0
30
86.
12.5
14
25.
1.4
1.4
38.
99.
3.5
3.5
3.5
3.5
3.5
3.5
3.5
22.
51.
11.5
21.0
31.
66.
= 300K.
In this particular
form was used to describe all the resolved reso

shows a plot of these results as functions of
iAF/MF)*.
The models developed in this section are especially useful for under
the principal features of resonance absorption, and by careful
standing
choice of formulas,
first-order estimates based on the equivalence rela

can be obtained for the effective resonance integral of both
In order to achieve greater
homogeneous and heterogeneous systems.
tionship
precision, it is necessary to utilize the higher-order approximations and

the more sophisticated models mentioned previously.
Additional refine
ments of the theory may be achieved by introducing the Dancoff-Ginsburg
and Pershagen-Carlvik corrections; the former1 accounts for the mutual
1
Dancoff and Ginsburg, op. cit.;
J.
A. Thie, Nuclear Set. Eng., 5, 75-77 (1959).
sec. 10.3]
689
shielding of fuel lumps in close-packed arrays and amounts, essentially,

to an effective increase in the average chord length s for the fuel lump;
the latter correction1 is used in treating complicated fuel-element geom
etries, including clustered configurations.2

60 1
0.4
0.8
(AFIMr)
Fig.
10.17
1.2
2, (cm2/gm)
Doppler corrected resonance
i. The Standard
tematic
Formula.
and consistent
1.6
2.0
V2
integrals for U"8 and Th,3a.
The preceding analyses constitute
development
from
a common
a sys
of
useful
basis
for the resonance absorption in both homogeneous and

An essential consideration in this approach was
heterogeneous systems.
the derivation of the equivalence relationships and their use in the
analysis of heterogeneous systems.
Important limitations in this treat
expressions
ment were the flat flux assumption and the application of the rational
This general treat
approximation for the fuel-lump escape probability.
ment represents the more recent approach to the problem of evaluating
resonance integrals.
It is expected that in time, as more precise measure
ments are obtained of the resonance parameters, there will be greater
incentive for further development of these general methods and models
in order to achieve wider applicability and accuracy.
Although
these recent calculational
methods have proved to be useful
I. Carlvik and B. Pershagen, "Calculation of Effective Surfaces for Resonance

Absorption in Tubes and Parallel Plates" (Aktiebolaget Atomenergi), AEF-49, 1955.
* Hellstrand, op. tit.
1
REACTOR ANALYSIS
690
[CHAP.
10
in understanding
the phenomenon of resonance absorption and moreover

show promise of high reliability, it is still of considerable interest to
The comparison
compare the present models with the standard formula.
is drawn with the aid of the narrow resonance approximation for the
collision density.
For this
case the resonance integral
is given
by
(10.135)
If
we separate the integrand into its two terms as written above, then
comparison with Eq. (10.37) reveals that we can identify the first integral
as
(10.205)
and the second as

(10.206)
We have used here Eq. (10.131) for P.K, and MP = VFp'F.
When these
expressions are compared with (10.38) and (10.39), it is seen that the
NR approximation for $ leads rather easily to a statement in the standard
form.
Note, however, that because of the flat flux assumption the
present derivation does not contain the resonance disadvantage factor1'2
This quantity is customarily computed using a one-velocity model
fr.
to represent the entire fast-neutron population.
that this point of view is crude and somewhat
hand, when used with the
It
is well recognized
unclear.
On the one
NR
approximation, it may be observed that

only one collision is required to remove a resonance neutron from the
vicinity of a resonance; thus the spatial distribution would be very
On the other hand, if the NRIA approxi

nearly uniform and isotropic.
mation is valid, then the absorptions are necessarily very strong and the
use of a disadvantage
factor based on diffusion theory is not well justified.
For these reasons it has been omitted in this treatment as an unwarranted

refinement not in keeping with the precision of the over-all model.3
Another point of immediate interest concerns the nature of the integral
(10.206); this term in the standard
surface absorption.
1
It
formula
is
customarily called the

lump
is seen to depend upon the geometry of the
A discussion of the
resonance
disadvantage factor and its use in estimating

resonance
absorption is given by A. M. Weinberg and E. P. VVigner, "The Physical
of Neutron Chain Reactors," pp. 663-668, University of Chicago Press,
Theory
Chicago, 1958.
'See also S. Glaastone and M. C. Edlund, "The Elements of Nuclear Reactor
Theory," pp. 258-263, D. Van Nostrand Company, Inc., Princeton, N.J., 1952; and
3 Chernick and Vernon, op.
cit.
sec. 10.3]
691
the factor AF/MP and through <P; thus the surface term
proportional to Ay/ My, as implied in the construction of the
Although this dependency was known in the early
standard formula.
days of the technology, the (P factor fell into disuse; it was reintroduced
both through
is not really
more recently by Wigner.1
The physical interpretation of the standard formula is that resonance

absorption consists of two parts, each arising from a different mechanism
for the supply of resonance neutrons to the fuel lump.
The first term in
is
to
fast neutrons
called
volume
due
absorption,
,
(10.205)
(or mass)
which diffuse into the lump at energies between resonances and then
through a series of scattering collisions with fuel nuclei, are eventually
slowed to an energy interval overlapping a resonance (i.e., the practical
The second, or surface, term
width Tr), and subsequently captured.

(10.206) represents resonance captures arising from neutrons having

energies that already overlap the practical width of a resonance and
which have appeared in the lump from direct transport through its
surface.
This viewpoint of resonance capture originated with Wigner2
and Dancoff and Ginsburg.3
The interpretation is given in terms of the energy spectrum of the
neutron flux.
Throughout the
Consider first the moderator region.
moderator, and at all energies, the flux is assumed to have the usual
In the fuel lump, between resonances, the spectrum

asymptotic form.
is again l/E. This flux serves as an internal source within the lump of
neutrons fed into the resonances and leads, therefore, to the volume
term in the standard formula.
When the resonances are narrow and
far apart
(so that there is no mutual interaction), their effect on the

spectrum is highly localized (see Fig. 4.21), so that the supply of neutrons
to each resonance is essentially l/E and spatially uniform.
This gives
rise to a uniform distribution of fast-neutron sources in the lump, and
the corresponding term is therefore independent of lump geometry.

The
surface term, on the other hand, must be dependent on the geometry
since it represents absorptions in the lump arising from an inward current.
In contrast to the spectrum in the lump, this current is an undepleted
l/E over the entire resonance range. Thus, when the lump is thin, this
flux from the moderator is essentially l/E throughout, and
when the lump is thick, it is
at least within the outer layers of the
resonance
l/E
lump.
1
E. P. Wigner,
J.
J.
Chernick,
Confer
Physics of Reactor Design, vol. 5,
Appl. Phys., 26, 255-279 (1955)
; see also
"The Theory of Uranium Water Lattices," Proceedings of
the work of
the International
the Peaceful Uses of Atomic Energy,

P/603, pp. 215-228, 1956.
! E. P.
Wigner, "Resonance Absorption of Neutrons by Spheres," Argonne Metal
lurgical Laboratory, C-4; and J. Appl. Phys., 26, 255-279 (1955).
3 Dancoff and Clinsburg, op. eit., Argonne Metallurgical Laboratory, CP-1589.
ence on
092
REACTOR ANALYSIS
[chap. 10
This point of view12 places emphasis on the volume term, and treats
the surface contributions as a secondary effect. Although this approach
is useful for handling most of the resonances in the important fuels, it is
inadequate when dealing with the low-lying broad resonances wherein the
volume term actually makes a relatively small contribution.
was recognized by the early investigators, but its importance
This point
was under
because the rough resonance parameter measurements avail

able at that time gave low estimates of the practical widths of these
resonances.
It is expected that the use of the NRIA approximation
in the present analysis will yield a somewhat more realistic representation.
estimated
We conclude the section with a last remark on the approach used

by the Russian investigators in their treatment of the resonance absorp
tion problem.
These analyses, in contrast to the Western approach,
place emphasis
on the surface absorption
phenomenon
and treat
the
volume absorption as a secondary effect. The calculations are based

on a neutron-flight-path analysis, and in the case of thin lumps, scatter
ing and slowing down within the lump are neglected.3 Thus this viewpoint
NRIA model, and it is not surprising

that
both
yield expressions for the effective resonance integral
therefore
which show a square-root dependence on Ap/MF [see Eqs. (10.167) and
(10.47) and the discussions of Sampson and Chernick and of Wigner and
Weinberg, who compare the Russian and Western work].

bears some resemblance to the
10.4
a. Significance
Fast Effect
of the Fast Effect.
The role of the fast-fission
effect
in
a neutron chain reaction was first demonstrated in connection with

the design and operation of heterogeneous natural uranium reactors.
Since the multiplication constants of these systems are not much different
from unity, the few per cent increase in A; to be realized by taking the
This factor is important,

fast effect into account was quickly recognized.
of course, not only to the design of marginal systems, but to any system
which is dependent on the fast-fission phenomenon as a principal source
Good examples are converter and breeder reactors which
maintain their chain reactions on fast fissions alone.
Nonthermal fissions are due to two different sources. There are first
the resonance fissions which occur in the thermally fissionable fuel (such
of neutrons.
For further discussion of the arguments leading to the use of the standard formula,
Weinberg and Wigner, "The Physical Theory of Neutron Chain Reactors,"
pp. 681-689.
* Also, Sampson and Chernick, op. cit.
3 Gurevich and Pomeranehouk, op. oil.; see also Sampson and Chernick, op. cit.,
see
p. 247.
sec. 10.4]
U!",
693
and Pu239), and second there are the fast fissions which
U238 and Th232, which have fission thresholds
in the Mev range. Once the fast-fission threshold1 has been exceeded,
these sources can become an important factor in the over-all fission rate.
as
U236,
occur in materials such as

In the
natural uranium the relative abundance of U238 is some

two orders of magnitude greater than that of U236; thus, the fissions in
IT238 represent the bulk of the fast-fission reactions in these systems.
The calculation of the fast effect in homogeneous systems was discussed
in Sec. 6.3a.
The method outlined is generally applicable to any system
which satisfies the requirements of the Fermi age model.
When the
system is multiregioned, the methods of multigroup analysis may be
Finally, one can also modify the
employed, as discussed in Sec. 8.8.
standard two-group model to take fast multiplication into account.
In
any case, the approach is straightforward, and the analytical machinery
The calculation of the
previously developed is adequate for the task.
fast effect for heterogeneous systems, however, requires some additional
The method generally used consists of two parts: (1) There
effort.
is the calculation of the probability that a fast neutron born in the fuel
lump makes a collision before escaping. (2) There is the calculation of
the contribution this neutron makes to the fast effect, as well as the
contributions of subsequent generations of neutrons produced by cascade
case of
effects.
b. Cascade Effects on the Fast-fission Factor.
Fast neutrons appear
in the fuel lump as a result of both fast and thermal fissions.
This supply
is continually depleted, however, by two processes which reduce the
energy of the fast neutrons to values below the fast-fission threshold.
Once so degraded, a neutron can no longer contribute to the fast effect.
The first removal process is the inelastic scattering in the fuel lump.
Since the inelastic-scattering cross section is a sizable fraction of the total
cross section in the Mev range, this process is a powerful loss mechanism.
The second process
is the leakage into the moderator.

This is a removal
process since a fast neutron which reaches the moderator has small
likelihood of returning to the lump with sufficient kinetic energy to excite
a fast fission.
The supply of fast neutrons from thermal fissions occurs at a steady

rate, and the source is distributed within the fuel lump in the shape
The supply due to fast fissions, however, varies
of the thermal flux.
from one generation to the next because of cascade effects. This phe
nomenon can give rise to a slight augmentation of the fast-neutron popu
The cascade effect
lation, depending on the properties of the fuel lump.
may be described in terms of the collision probabilities and the fast1
Approximately
0.9 Mev for
U*".
REACTOR
G94
ANALYSIS
[CHAP.
For this purpose we
neutron cross sections of the fuel-lump material.

introduce a quantity (P, where
(P
It
10
probability that a fast neutron produced in

= the nth generation fast fission makes a collision before escaping from the fuel lump
(10.207)
that the sequence of n fast-fission generations was

With
single fast neutron produced by a thermal fission.
is understood
initiated by
the function
(P, and a suitable set of fast-neutron cross sections, we can

define several pertinent processes which describe alternate histories for
each neutron appearing in the nth generation.
These are:
(1)
(2)
Probability of escaping from the fuel lump:

Probability of making
a collision
i:
of type
<$*
2<Pn
i denotes any one of the following nuclear reactions:

scattering 2in, radiative capture 27, fission 2/, and elastic
The symbol 2 denotes an appropriate "total" cross sec
scattering 2,.
tion for the fast neutrons; it is defined by [cf. Eq. (2.50)]
Here, the subscript
2 = Zy +
2, +
2in
+ 2.
(10.208)
A suitable approximation for 2
is given in Sec. 10.4c.

Consider now the history of neutrons produced in the first generation
of fast fission.
The number (Pi gives the fraction of these which make
For each such
some kind of collision in the lump before escaping.
a certain
fraction yields additional fast neutrons;
thus,
for
("2/
neutron
S-)
collision
each first-generation
<?i
(10.209)
cfP!
produced; note that this expression implies that an

negligible effect in reducing
elastic-scattering reaction in the lump has
defined [cf Eq. (7.230)]
The quantity
the energy of the fast neutron.
are
is
fast neutrons
y2' + 2'
(10.210)
The first generation also suffers some losses by escapes, inelastic collisions,
Of these,
certain fraction enter
and radiative captures, as noted above.
2in
process:
1
the slowing-down
1

inelastic
~~
fraction of first generation neutrons

entering slowing-down
process
The histories of subsequent generations may be traced in the same way.

The neutrons entering the
Some of these are indicated in Table 10.5.
sec. 10.4]
695
slowing-down process may still produce fast fissions in the thermally

fissionable materials; however, they cannot, once slowed below the fission
threshold, produce further neutrons from such materials as U2JS and
Th"2. The contribution of the thermally fissionable materials to the
previously
process is neglected in the present analysis,

cited.
Table
Fast-effect
Fast neutrons
available
Neutrons
participating
in cascade
c(P,
C(Pi
e3>i(P,
c-'(P,
. . . <P_,
Cascade
Neutrons entering
slowing-down process
C(Pi(Pa<P,
c"(P, .
Histories
Genera
tion
:'>,,
c-'(P,
(P.
. . .
- ft +
-
|\
>_,
a>,
|\
<ft
Y<Paj
(PB_,
Y <*]'+
- +y +
(5>t
(Pn
(Pnj
+ c"-'(Pi
+C(P1Jl
i-ff*!
[^1
t=
The calculation of the fast-fission factor proceeds directly from results

In order to obtain we need simply sum the num
given in Table 10.5.
ber of neutrons supplied to the slowing-down process from each fastfission generation.
These are the terms indicated in the last column.
Thus we obtain
'
(10.212)
The collision probabilities (P given here vary from one generation to the
next since the spatial distribution of the fast-fission neutrons changes
The original, or source, distribution F0(r) of
with successive cascades.
is
depressed toward
the first generation varies as the thermal flux, which
the center of the fuel lump.
The spatial form of the next generation
is
given by the distribution of first collisions experienced by the

The next generation in turn
neutrons originating from the source F0(r).
obtained from Fi(r), and so on. With each succeeding generation, the
Fi(r)
is
distribution becomes more peaked toward the center of the lump. The
asymptotic form corresponding to n does not differ much, however,

10.5
for the reason
""-J
fast-fission
REACTOR
696
from the forms for n =
or 2.
ANALYSIS
Figure
[CHAP.
10.18 shows a
10
typical sketch of
It may be
the spatial forms Fn of the first few fast-fission generations.
from these remarks that (Pi < (P < 1, but because of the
small variation in the Fn for n > 0, a reasonable approximation is to take
=
Conceivably it might be
some fixed number independent of n.
3>n
concluded
(P0)
better to take the actual (Pi for the first generation and then some other
number for all n > 1 ; but this much care may not be entirely justified.
If
we assume the same number
for all n, then Eq. (10.212)
is
(Po
readily
summed, giving
2,(P0
2(1
c(P0)
(10.21.3)*
It
if
that this sum exists only
is clear
<
c(P0
1.
c. Computation

^Surface of fuel lump
Fig. 10.18 Spatial

cessive generations
Cross Sections and Collision Proba

bilities. The cross sections to be
used in the relations derived above
should be averaged over the normal
ized fission spectrum j(u).
This is
distribution of suc
of fast-fission neu
trons.
required also for the collision probability

Thus we take
energy dependent.
(Po
of Fast-neutron
(Po
since this function too is
= /(Po(m) 3(u) du
(10.214)
it is understood that the integration extends down to the fastA consistent definition for the various cross sections
fission threshold.
where
2( is then
If it
is assumed that
I -s/^ *
(P0
is energy independent,
f 2,(u)
2,-
then
, , ,
(10.216)
This definition of the 2, is useful for estimating the fast-neutron radiative

It is not so useful for determining the
capture and fission cross sections.
corresponding inelastic cross section because of the dearth of information
on the details of the actual inelastic cross sections and of the spectrum of
the neutrons coming out of these reactions.
This quantity is best deter
mined from the measured value of the reaction cross section,1 2m<:t; thus,
2in
1
Values for
arct are
2t
(2/ +
27)
given by Hughes and Harvey, op. cit.
(10.217)
sec. 10.4]
697
For the total cross section 2 of these fast neutrons we take the transport
This choice is suggested by certain physical features
cross section S,r.
of the problem, the most important being the strong forward bias of
scattered fast neutrons.
Because of this effect, the fast neutrons have a
The net
greater tendency to move outward from their point of origin.
result is as though the supporting
medium had a mean free path some
what larger than its actual value. Moreover, in calculating the func
tions Fn and (P we will assume that the scattering process, as well as the
By combining this assumption with the use of the
others, is isotropic.
transport cross section based on a value of
appropriate to the forwardscattering tendency of the nucleus in question,1 it is expected that a
fairly good representation of the fast-neutron "diffusion" in the fuel lump
can be achieved.
Thus we take
S = 2tr = 2V + 2, +
2in
+ 2.
(10.218)
Table
cross sections which have been used in the
10.6 lists the various
past to relate the theory to actual measurements on the fast effect.
Table
10.6
UM9
Cross Sections fob Fast-effect Computation

Reaction
a barns
Radiative capture <ry

Fission <r/
Inelastic scattering ain
Elastic scattering a,
"Total"
0.04
0.29
2.47
1.5
4.3
<rtr
dr' about
r',
The corresponding value for v to be used with these data is 2.5.

A formal definition of the collision probabilities (P may be given in
terms of the transport kernel and the normalized spatial distributions
F(t). Thus to compute the number of first collisions which occur in
dr about r due to the F(i') dr' nth generation neutrons produced in
we take
e-2t,|r-r'|
-rri ^n(r') dr'
(10.219)
I1
Jv,
|reTr
dr
Jvr
<P,+1
r'
is
Since Fn
normalized, to compute (P+i we simply integrate the above
and
thus
expression over all
S.
J.
Carter, "Neutron Cross Sections; Angular Distributions,"

D.
Hughes and R.
Brookhaven National Laboratory, BNL-400, June, 1956.
1

Note that if the relation (10.217) is used to obtain 2in then Eq. (10.218),
The approach
along with a computed 2tr, can be used to determine 2,.
outlined here for estimating the fast-neutron cross sections is found to
give reasonably good agreement with experiment in the case of U!M.
REACTOR ANALYSIS
098
It
[chap. 10
follows that
fWi(r)
The functions Fn may
- r'|*
2*
/"
FnCrQe-^lr-
4r<JWi
Jvr
[r
(10.220)
of F0.
Some of these
be developed successively from a knowledge
which is obtained from the shape of the thermal flux.
functions have been computed for the standard fuel-lump shapes.1

It is
of interest to mention in this connection that, when the Fn are uniform,
the collision probabilities (P are simply the quantities Pc defined previ
ously in Eq. (10.130).
10.5
Effects of Cavities and Fuel Lumps on Migration Area
The first-order correction to the migration area to account for the

effects of heterogeneity appears through the diffusion length.
A simple
was
in
relationship
given previously
Eq. (10.14) for computing L2 in a

system consisting only of fuel lumps and moderator.

When the nature
of the heterogeneity becomes more complex, additional corrections are
Some of these were considered in Sec. 10. 2e in connection with
required.
of the disadvantage factor and the thermal utilization.
From the standpoint of its effect on the migration area, the most impor
tant correction is due to the presence of gas passages and cavities.
Of
particular interest, from an engineering standpoint, are power-reactor
designs utilizing parallel arrays of cooling holes or pebble beds and
research reactors with irradiation holes and tubes.
In each case, the
presence of the cavity, whether regularly or randomly distributed, serves,
effectively, to lengthen the total path traversed by a neutron from point
of birth to point of capture (or escape from the reactor).
The problem
of computing this increase in path length is further complicated by con
siderations of anisotropy, as occur in designs involving parallel rows of
the calculation
holes.
will
In these situations the average path length transverse to the holes
shorter than that in the direction parallel to the holes,

since a neutron entering a hole at a shallow angle will enjoy a long free
ride before encountering any solid materials.
be noticeably
An elementary approach frequently used for estimating the increase in

the average path length consists of applying a uniform correction factor
to all macroscopic cross sections to account for the effective reduction in
density due to the presence of the holes. On this basis, it is easy to show
that the corrected diffusion length LH (with holes) is related to the diffufor example, G. \V. Anthony, "The Average Escape Probabilities for Onceand Many-times-scattered Neutrons for a Slab," Hanford Laboratories Operation,
HW-49188, Mar. 22, 1957.
1 See,
sec. 10.5]
699
sion length L of the reactor materials [computed from Eq. (10.14)] with
out the holes through the equation
L%
where
vH
L*(l + vy
(10.221)
\"s
(10.222)
and Vh denotes the total volume of holes in the reactor and V s the
volume occupied by the solid materials.
It has been demonstrated by D. J. Behrens1 that the estimate given
by (10.221) is the first-order isotropic correction term for small holes in
into account the distribution and shape

The results obtained by Behrens for the diffusion length
are summarized in the work which follows, and extension is made to the
calculation of the age, using a procedure suggested by the work of Kaplan
Finally, the estimates of the age and diffusion length
and Chernick.2
a more general result which takes
of the holes.
obtained in this way are used to compute the migration area and thence
the multiplication constant of the reactor.
a. Corrections to the Diffusion Length.
The general relations obtained

by Behrens for computing the diffusion length in a system containing

uniform distribution of holes are summarized by the expression
where
Vg,
LH, and
have the definitions
given above, and rfI is the
hydraulic radius of the holes (twice the volume of the hole divided by
the surface area), \m is the mean free path in the moderator,' and
G = s2/(s)2 for the holes; here denotes the average passage (chord)
The
length through a hole and s2 the average square passage length.
quantity G has been computed by Behrens for various shapes of holes,
It is of interest to note that
and the results are given in Table 10.7.
=
1 for the annular geometry corresponds to a twothe limiting case x
This is an unfavorable arrangement
dimensional gap in the reactor.
from the standpoint of neutron leakage and should in general be avoided.
Such a gap, however, has been incorporated in the design of the BrookThe effect
haven graphite reactor as a part of the gas cooling circuit.
of such gaps on criticality is discussed by Kaplan and Chernick in the
reference paper.
J. Behrens, Proc. Phys. Soc., 62, A, 607 (1949).

Kaplan and J. Chernick, "Uranium Graphite Lattices The Brookhaven
Reactor," Proc. Intern. Conf. Peaceful Uses of Atomic Energy, Geneva, 1955, P/606,
pp. 295-308.
1
D.
'I.
May
be estimated from \m
REACTOR ANALYSIS
700
[chap.
10
When both the size and spacing of the holes are small compared
to the mean free path in the moderator, the expression (10.223) is well
approximated by (10.221); moreover, it may be shown that in any event
Table
The Quantity
10.7
for Various Cavity
Geometry
11 25
Sphere
Infinite cylindrical hole with:
Circular cross section
Hexagonal cross section
Square cross section
Equilateral-triangle cross section
Cylindrical annulus
Spherical annulus
x
1.333
1.397
1 .487
1 .648
1
.333f,(i)J
1 125F
,(x) t
ratio of inner radius to outer radius.
F.(x)
0.1
0.2
0.3
0.4
0.5
1.001
the value of
F.(x)
F.(x)
Fc(x)
0.6
0.7
1.11
1 .0258
1 .005
1 .0536
0.8
1.2023
1.2604
1.3422
1 .016
1 .0839
0.9
1.18
1.27
1.43
1.04
1.07
1 .1178
1 .0
00
Lj,/L2 obtained from Eq. (10.223)
oo
is
at least
as large as
Thus, properly
jjf
that from Eq. (10.221).
1 .53
1 . 1567
>
(1+ vy
(10.224)
In the event that the holes are far apart, Eq. (10.223) may
(1
mated by
vHy +
be
approxi
(10.225)
Some representative values of L2/L2 are given in Table 10.8 for various
hole sizes and spacings so as to display the relative accuracy of Eq.
(10.221).
is
is
is
directionally non
When the distribution of cavities in the reactor
uniform, some modification must be made in the general result (10.223).
affected by the directional
The only term in this expression which
the last, and this dependence appears through
properties of the cavities
particular type of anisotropy which
frequently
the shape factor G.
in the form of parallel passages of cooling
encountered in reactor design
In this situation the corrected diffusion length in the direction
holes.
is
is

Shapes
SEC. 10.5]
701
of the passage axes will be different from that in the direction perpendic
ular to the passages.
Behrens has shown that in this case the appropriate
expressions for the diffusion lengths in the two directions are given by
=
2rt,/vH\M
v%
+ 2v +
v%
2oB
exp (2ri,/v\M)
2rH/vH\si
{2rH/v\M)
exp
3GrHvn
These relations apply for any cross-sectional

is obtained from Table 10.7.
Table
10.8
3Grv
2\M
shape.
(10.226)
(10.227)
The quantity G
Typical Values of L*h/L* fob Square Array of

Cylindrical Holes
\.
^\Lattice
spacing
Hole
diameter
1.016
1.000
1.066
1.000
1.200
1.000
1.548
1.000
1.019
1.003
1.079
1.012
1.234
1.028
1.609
1.039
1.026
1.010
1.108
1.039
1.319
1.099
1.815
1.172
1.047
1.194
1.066
1.574
1.312
2.466
1.593
10
^\

rB/\u
th/am = $
r///X. =
tbI\u
Lower
figures
1.031
computed from
(1
+ vh)*-
Finally, it should be mentioned that calculations of the diffusion

length for systems with holes formed as interstices between random
arrangements of solid bodies in point or line contact have also been
These
carried out by Behrens and are reported in the reference paper.
results may be applied in the study of reactor core designs utilizing
pebble beds and other similar configurations.
b. Corrections to the Age.
The corrections
outlined above may
be extended
to account
to the diffusion
length
for the streaming effect
on the neutron age.

The correction consists of two parts, one to account
for the effect of cavities (as in the preceding analyses) and another to
The latter correction is neces
account for the effect of the fuel lumps.
sary because of the difference in the slowing-down properties of the fuel
In first approximation the effect of the fuel
and moderator materials.
lumps is to give a neutron a "free ride" in the slowing-down process,
since a collision with a fuel nucleus produces no significant change in the
REACTOR ANALYSIS
702
neutron energy.
[chap.
A second effect of the fuel lump on the
10
age is to lower
the initial energy of a fast neutron entering the slo wing-down process;
this reduction results from inelastic-scattering collisions in the lump.
An estimate of the combined effect on the age of fuel lumps and cavities
flWCu^tO]
(PioWCUthjO)
(PiaTif(uu,;,.)]
(l
yA f1
(2
^f)
jf)
v"]
<Pi)Tjtf(wth;0)
(10-228)
v]
Tt,.. = [(1
(2
[(1
[l
(l
rw
from the following
may be obtained
passages
in the form of parallel

relations :
(10-229)
gives
.,)
(10.230)
^p?)
is
On the basis of these

given by the usual relation (7.234).
seen that the first bracket in Eqs. (10.228) and (10.229)
weighted age for fission neutrons in the moderator which accounts
definitions
it
with
-P
^n
(1
it
is
quantity (Pin
fission neutron makes an inelastic
collision before escaping from the fuel lump;
may be estimated by
is
for the fact that not all neutrons enter the slowing-down process at
u = 0. The first term in the bracket
a population-weighted
age for
=
The second term, on the other
those neutrons which do start at u
0.
hand, accounts for the age of fission neutrons which begin the slowingdown process in the moderator after they have first suffered an inelasticThese substantially lower-energy
scattering collision in the fuel lumps.
neutrons enter the slowing-down process at u = win and thereby acquire
Note that the age r,tf(uth;?*tn)
ages TM(uth',uiD) different from Tjtf(Mth;0).
is
weighted
by the fraction of neutrons
inelastic-scattering
(Pin
which
have experienced
collisions.
a
first-order
The second bracket in Eqs. (10.228) and (10.229) gives
correction to the moderator density to account for the free ride acquired
This factor corresponds to
fuel lump.
by the neutron in traversing
the isotropic term (10.221) derived for cavities.
The last bracket in
the two expressions for t takes into account the presence of holes in the
form of parallel passages and includes the directional effects discussed
a

where rj/(u,h;0) denotes the age to thermal of fission neutrons in pure

moderator and TM(utb;uin) the age from lethargy uiQ to thermal in pure
moderator.
The symbol Mia denotes the average lethargy of a fission
neutron after an inelastic-scattering
collision in fuel material.
The
sec.
10.5]
703
These factors
are approximations
to expressions (10.226)
and (10.227) given in the form of (10.225).
c. Migration Area and Multiplication Constant.
Estimates for the
migration area which include streaming-effect corrections for cavities
previously.
and fuel lumps on the thermal-diffusion length and the age may be
obtained from the general relations (10.226) through (10.229).
The
appropriate expressions for A/2 are seen to be
M2axial =
-r
iwA
Taxial
1/2
iw
trana
= rTtrana
AI
i
I
T 2
âxial
7 2
/Ifi
OQ1"\
"traoe
It
is evident from these relations that the buckling parameters required

in the criticality calculation will differ for the various axes of symmetry
An idea
of the reactor.
as to how this
anisotropy might
in the calculation
of the multiplication constant

model of the homogeneous reactor.
migration-area
relation for this system is
-DlhV2*th + 2?*th
V2</Hh + B2tb

or
be included
is suggested by the
The neutron-balance
fc.ZLWu,
(6.87)
= 0
where [using Eq. (8.84)]
which yields for B2

B2
with
M1
+ V-
Tlh
Substitution into the diffusion equation yields

M2V24nh
+ (kx
l)th
which applies for the isotropic system.

On this basis we take, for the anisotropic
MlM
*J
M?_
(10.232)
system,
+ (*.
The multiplication constant for this system

k =
= 0
L)*
= 0
(10.233)*
is seen to be
A;
i + mLmbi+"mi.u.{BI + bd
(10-234)
where Bx, Bv, and Bt are the bucklings in the x, y, and z directions, with
x denoting the axis parallel to the gas passages and y and z the axes
perpendicular to the passages.
REACTOR ANALYSIS
704
10.6
[CHAP.
10
Feinberg-Galanin Method of Heterogeneous -reactor Calculations
a. Principles of the Heterogeneous

Method. If a heterogeneous
reactor contains a small number of fuel lumps, its critical size cannot be
well computed, in general, by means of the unit-cell model of Wigner
and Seitz.
In these cases the detailed arrangement of the fuel lumps
is an essential feature of the nuclear configuration which must be included
Thus, it becomes necessary to discard

the "homogenized" model based on the gross properties of the unit
cell in favor of a more detailed picture which accounts for the discrete
form and distribution of the neutron sources.
in the criticality considerations.
Such a model1 has been developed by Feinberg2 and Galanin, and in this
approach the general heterogeneous reactor is treated as a collection of
line or point sources in
matrix of moderator
material.
The fast
(fission) neutrons produced by these sources slow down in the moderator,
and it is assumed that the resulting spatial distribution of thermal neu
a
by the Fermi age solution appropriate to the

source geometry.
The usual one-velocity diffusion equation is then used
to describe the distribution of the thermal group in the moderator.
The
source term in this relation is obtained by superimposing the contributions
be represented
of all the fuel lumps in the reactor.

It is necessary, however, to provide
an additional absorption term which accounts for the fact that each fuel
lump acts also as a sink for thermal (as well as fast) neutrons.
The
results obtained from diffusion theory for line and point sources are used
Throughout these calculations the properties of the
for this purpose.
fuel lumps as neutron sources and sinks are described in terms of the
so-called thermal constant (effectively,
flux at the surface of the lump).
the logarithmic derivative of the
The heterogeneous method is especially suited for analyzing reactors

with small numbers of fuel lumps.3 Because of the general formulation
of this model it is also useful for determining the effects on criticality of
various lattice geometries.
Feinberg has treated specifically the cases
of the square, rectangular, rhombic, and hexagonal arrays.
He has
further used the method to study the effect of superimposing several
application
is,
lattices of fuel lumps, each lattice having a different module and size and
nuclear property of the fuel lumps.
The practical interest in this
of course, the use of "seeded
cores" in heterogeneous
5,
of
First described by E. Teller, " Influence of the Lattice Structure on the Exponential
Pile Experiment," Axgonne Metallurgical Laboratory, CP-165, 1942.
S. M. Feinberg, "Heterogeneous Methods for Calculating Reactors: Survey of
Results and Comparison with Experiment," Proc. Intern. Conf. Peaceful Uses
Alomic
Energy, Geneva, vol.
1955, P/669, pp. 484-500.
A. D. Galanin, "Critical Size of Heterogeneous Reactor with Small Number of
Rods," Proc. Intern. Conf. Peaceful Uses
Alomic Energy, Geneva, vol.
1955, P/663,
pp. 462-465.
5,
of

trons may
sec. 10.6]
705
This general approach has also been applied by Galanin to

It is perhaps worth while
determine the effectiveness of control rods.1
to point out that the analytical approach used in treating these problems
arrays.
could be extended to study the effect of other kinds of heterogeneity

well, for example, cavities.
as
The present treatment of the heterogeneous method develops the

principal concepts obtained in the general theory.
The analysis begins
with a derivation of the neutron-balance equations.
These results are
then used to obtain the criticality condition and to demonstrate the basis
of the homogenization method used with the unit-cell model.
The sec
tion is concluded with an application of the method to a reactor with four
fuel rods and a numerical example.
The general theory of the hetero

geneous method is developed on the basis of the following assumptions:
flux in the vicinity of a fuel lump possesses
(1) The thermal-neutron
spherical or axial symmetry so that the individual lumps may be treated
as point or line sources and sinks.
(2) Elementary diffusion theory
the
thermal
flux
in
for
the
moderator
regions between the lumps.
applies
properties of the fuel lumps are adequately
(3) The thermal-absorption
b.
Neutron-balance Equation.
by means of the logarithmic derivative (thermal constant)

of the thermal flux at the surface of the lump.
(4) There is no resonance
absorption in the fuel lumps.
The first of these assumptions is valid if the distance between lumps
is much larger than their dimensions.
Because of the symmetric con
described
figuration of most reactor lattices, this condition tends to
be favored
and the theory is applicable even in the case of quite large lumps.
The
second assumption requires that the absorption in the moderator be low
and that the diffusion length for the moderator material be small in
to the lattice spacing.
The third assumption is essentially
convenience, and its applicability is determined primar
ily by the accuracy of the model by which the thermal constant is com
A treat
The last assumption is made to simplify the analysis.
puted.
ment in which resonance absorption is included is given by Galanin.2
With the above assumptions as a model, consider an infinite hetero
comparison
a computational
lattice composed of cylindrical fuel rods of infinite length (i.e., a

Let the radius of these rods be pf
configuration).
two-dimensional
absorption be ij. The
and their neutron yield per thermal-neutron
geneous
thermal constant 7* for the fuel lumps is defined as
ratio of total net current of thermal

=
jh
neutrons into the lump to the value of the
thermal flux at the surface
(10.235)
D. Galanin, "The Thermal Reactor Regulator's Efficiency," Proc. Intern. Conf.

Peaceful Uses of Atomic Energy, Geneva, vol. 5, 1955, P/668; pp. 511-514 and P/663.
1 A.
*Ibid.
REACTOR ANALYSIS
706
[chap. 10
On the basis of the diffusion-theory model the thermal-neutron

field point r in the moderator is given by
-OtfV'Mr)
20.w(r)
S(r)
Arod(r
rk)
flux at a
(10.236)
The two func
where the vector r* denotes the position of the fcth rod.

tions at the right-hand side are defined by
S(i) = number of thermal neutrons appearing at r per unit

volume per unit time from all sources in the system
-Arod(r
>
ik)
absorptions per unit

time due to the fuel rod at r*, if the field
point r is located at one of the rod positions
number of thermal-neutron
For the present, the effect of the rod size
is ignored
and all rods are
treated as line sources and sinks.

The thermal-source function S is easily computed in terms of the ther
If J(Tk) is the net current of thermal neutrons at the
mal constant 7*.
surface of the rod at r* (taken as positive when directed outward), then
-r,2TPFJ(ik)

S/(r) =
(10.237)
is the total number of fast neutrons produced per unit time per unit
But, by (10.235),
length of rod from fissions in rod rt.
7*
(10.238)*
a
(r \
<PM(Ik)
where \i(Tt) denotes the thermal flux at the surface of the rod.
Note
that this expression is for unit length of rod and that the proper units of
7* are length squared.
Sf(tk) =
im<Mrt)
(10.239)
gives the source strength of fast neutrons in terms of the local

To obtain the thermal source at field point r due to this
thermal flux.
source of fast neutrons at r*, we simply multiply (10.239) by q(r;Tk), the
number of neutrons slowing down to thermal at r per unit time per unit
which
volume due to a fast source producing one neutron per unit time at r*.
The function S(r) is then simply the sum of such contributions from all
the rods; thus, we compute
S(r)
ss Vyk
y q(i;Tk)
(10.240)
<pM(rk)
noting that 7* and
tj
the two-dimensional
are the same for all rods.
matrix of rod locations
The sum extends over

by the two
represented
coordinates of the vector k.
The rod absorption
(sink) term may also be expressed in terms of 7*.
sec. 10.6]
An appropriate relation
^rod(r- r*) =
is seen to be (cf.
- 2TPFJ(Tk)
(r
Prob.
r*) = 7*
6.4)
^ 5(r
707
r)
(Mr*)
(10.241)
where 5(r r) denotes the Dirac delta function.

this result, along with (10.240), into (10.236) gives
-DmV*4>m{t) +
2u(T)
= V7k
9(r;r*)
4>u(tk)
7a
(r
The substitution of
r)
*(r)
(10.242)*
for the neutron-balance equation at field point r.
This equation defines

flux in the moderator, given a suitable expression for the
function g(r;rt).
On the basis of the Fermi age model, the appropriate
relation for the line source is obtained from (6.37).
In the present appli
cation, this takes the form
the thermal

9(r;r*) =
(10.243)
where r denotes the age to thermal in the moderator material.

The solution to Eq. (10.242) may be obtained with the aid of the twodimensional Fourier transform.
In the case of the flux we define this
to
be
ff{<Mr)}
ff *u
*(s)
(')*** dr
(10.244)
in accord with previous work [cf. Eq. (4.216)].

With this general form
we compute the Fourier transform of Eq. (10.242) and solve for the trans
form function *(s). The result is found to be (where x2 2'aM)/D\f)
where we have used Eq. (4.229).

The inversion of this result yields the
thermal flux in the moderator.
For the inversion formula we use the
two-dimensional form of Eq. (4.226).
The first term in (10.245) yields
ff exp[-sV
2 J J
V_
+ is (rt
s' + x2.,
- r)] ds _
"
["
2tJ0
se-'r ds
s' + x^Jo
2'
pf,.(rt_r)
(10.246)
where s = s(s,6) and 6 is the angle between the vectors s and z, with
z = r* r.
The 6 integration is recognized as the definition of the
REACTOR ANALYSIS
708
10
[CHAP.
&,""tde
The inversion of the second term leads to
00
2tJ0(sz)
similar result
"^TîT
Jo
JJ
(10.247)
;
d8 =
J**
eW*-*)
J*'
Bessel function1 of the first kind and yields
zeroth-order
thus,
(10-248)
se~'',Jo(s\I
is
- n\)ds
<r
f"
yh
a.
is
is
the last step follows from an integral expression for the Ko function.8
If
substituted into (10.246) and the result, along with (10.248),
(10.247)
used in the inversion of (10.245), the solution to the flux
found to be
(10.249)*
It
will
be convenient
for
analyses to write the above

For this purpose we define a function
use in subsequent
result in a more symmetric form.
2r
,-/<r
The solution to this differential equation yields (with
/3
v2
XmU
9r
follows then that

HQ
It
txJ,)
^]*
(10.251)
K^)m |o-exp[-(S +
^)]|
(10.252)
G(r)-^/;p[-(.
where we have used the fact that
and noted that G(0) = /o(x.\fz).

(10.249) yields
The substitution of these results into
- r*|V4*L|,)]
(10.253)
Copson, op. cit., p. 319.

E. T. Whittakcr and G. N. Watson,
p. 383, example 36, Cambridge University
[-
|r
exp
(a
/,"-/.")

G(t) such that
"A
Course in Modern Analysis," 4th ed.,

New York, 1950.
Press,
sec. 10.6]
709
with the application of the relations z = |r r*|2 and L\, =

The present notation may be further simplified if we call
where Vu denotes the total area of a

may then be written in the form
c0<Mr) =
with
co
H(\*
DM/Z(aM).
"unit cell."
Equation (10.253)
- r*D<Mr*)
(10.255)*
yiw
(10.256)
7*
Thermal Constant.
It was noted that the thermal constant could
be computed with various degrees of accuracy, ranging from an estimate
Since
based on diffusion theory to one based on the transport relations.
the diffusion-theory estimate will suffice in many instances, the appropri
The compu
ate relations will be developed on the basis of this model.
tation is straightforward and utilizes the unit-cell concept in the usual
In general the relations (5.349) through (5.354) will be required.
way.
By applying the continuity conditions for both the net current and the
flux at the surface of the fuel rod, it is necessary to consider only the
This is given by the usual differential
flux distribution on the inside.
In cylindrical geometry the solution to this equation
equation (5.350).

c.
is seen to be
Mp)
= AU{xFP)
(10.257)
wherein we have applied the requirement that the flux be finite within
The net current is obtained from (10.257); thus
the rod (5.354).
J(pf)
DF'F(pF)
ADFxFr0(xFPf.)
Substitution of these expressions into (10.238) yields

yh =
toefPUxrPr)
Io(xFpF)xF
(1Q 258)
It
is of interest to mention that the thermal constant is related to two

other quantities previously defined; these are yF and F. The relation
ship is
yh = wpFyF =
where yF is given by (10.80) and
F by (10.88).
(10.259)
d. Criticality Condition.
In the case of the infinite heterogeneous
lattice it is clear that there can be no gross spatial variations in the
These local vari
neutron flux, only the local variations between lumps.
REACTOR ANALYSIS
710
[CHAP.
10
ations are of course repeated with the periodicity of the lattice.

The
flux at each fuel lump must be the same, however, for a completely sym
Then, the flux at any rod r must be
metric and singly periodic lattice.
some constant, and the neutron-balance
simple form
Co
relation
(10.255) reduces to the
^ #(r*)
(10.260)
where r* now denotes the distance between some reference rod and any
This result is in fact the criticality condition
other rod in the lattice.
for the infinite lattice, since given any three of the four basic parameters tj,
the lattice spacing, the rod size, and the reactor materials, Eq. (10.260)
will yield the value of the fourth which will make the system critical.
The result (10.255) may also be used to obtain the criticality condition
for the reactor with finite fuel lattice but with infinite reflector (com
It is necessary to impose the

posed of the same moderator material).
latter condition because otherwise expression (10.243) for the slowingdown density would no longer be valid and new functions satisfying

boundary
conditions
at the surface of the reflector would be required.

in terms of the flux at the
The appropriate neutron-balance relation

various fuel-rod locations is then given by
c4>m(X)
H(\X
- v|)*jf()
(10.261)
The components of the vectors X and y denote the coordinates of any

two fuel rods in the finite array, as located from some arbitrary reference
The result (10.261) is a system of homogeneous equations for
point.
A solution to
the thermal fluxes at the surfaces of the fuel rods >a/(^.).
this system exists if the determinant is identically zero. By imposing
this requirement the determinant reduces to an algebraic equation, and
As in the
the roots of this equation yield the condition for criticality.
case of the infinite lattice, all four of the basic parameters will be involved
in this equation, and by specifying any three, the determinant will yield
the fourth. The calculational procedure involved here is demonstrated
in a subsequent analysis, and an example of a four-rod reactor is com
puted in detail.
In as much as the homogenie. Proof of the Homogenization Method.
zation method based on the unit-cell model is entirely adequate for com
puting most heterogeneous-reactor configurations, it is of interest to
determine the error acquired in using the method and to delineate the
To study this problem we postulate
conditions under which it is valid.
a reactor with a finite array of cylindrical fuel rods and an infinite reflec
tor composed of the same material as used for the moderator in the lattice.
We further specify that the lattice is a square array with spacing a.
sec. 10.6]
HETEKOC.EN'EOUS
REACTORS
711
The general solution for the thermal flux in this reactor may
Thus
be given
of the elementary solutions e'Br (cf. Sec. 7.4d).
we take for the flux at the surface of the rod at r
Mr) =
elB-r
(10.262)
The substitution of this relation into the thermal-flux equation (10.255)

yields
Co
H(\i
- rt|)eiB-("-'>
(10.263)
In order to demonstrate the relation

ship between the present model and the
method, it is necessary
homogenization
to reduce Eq. (10.263) to a form which is
readily
results
identifiable
with the established

This is accom
Fio.
of the latter.
plished by deriving suitable approxima

tions to the sums indicated in (10.263).
Si
We begin with the first term Su
by the definition of H in Eq. (10.254)
I 4r X
cxp
iBi{akl
~x) + iB*{akl
- [(at, - x)* +
- yY]
(afc2
(10.264)
wherein we have defined the components of the vectors r, ik, and B by

(refer to Fig. 10.19)
r = r(x,y)
The
ik = r*(afc,,afc,)
B = B(B,,B2)
series given in (10.264) may be summed
summation
(10.265)
with the aid of the Poisson
formula,1 which is given by

M
^ f(D
(2t)
go
SF{/} m
F(2irm)
(10-266)
Morse and Feshbach, op. cit., p. 467.
- -
(afcl
xr~*
;
TS,(afc,
x)
= exp
L
/(*
fci
where f(l) denotes some function of the summation index I and F is

its Fourier transform.
In the present case, there are two summation
variables k\ and fc2, and for
we have

as given
Position vector of fuel
10.19
rod.
4s
REACTOR ANALYSIS
712
[CHAP.
10
"
*)
e"P
(lr
/--
-5
follows that the double sum in (10.264) may be written in the form
+-(5-')-(5-),4]*
>
and
it
/(*0
The application of the summation formula yields
">
The integration
with u = u(i,u,)
In this way we obtain
/_>[- v +*(>> +*?)]*
2L*(r) exp
(rt
- -
(rt
r)
^]
r)
exp
[iB
is
If
evaluated in the same way.

The integral involving u2
these expres
sions for the integrals are used in (10.267), the double sum takes the form
'^
-^X-K?)-'m
s,
The substitution of this result into Eq. (10.264), followed by the integra
tion over the range of the variable
yields the following double sum for
Si:
>[(*)"-(+)']
+
*r-L
where we have used Vû = a2 for the square lattice.

The result (10.270) takes on an especially simple form
if

and completing the square in the exponential.

for the U\ component
(10.268)
us
easily carried out by introduc
(î
where w = w(u>i,w2).
ing the new variable
is
(10.267)
the following
sec. 10.6]
713
conditions are satisfied:
4e-4"'"**l
^1
and
(10.271)
The first is satisfied if the lattice spacing
is small compared to the slowingIt is interesting to note that under

down length in the moderator.
these conditions the Gaussian distribution of the thermal neutrons is
It is a
broad compared to the unit-cell dimension (refer to Fig. 10.4).
that
the
thermal
in
to
assume
source
the
then,
good approximation,
moderator is uniform; of course, this is precisely the assumption made
Thus we have established one condition

in the unit-cell calculation.
which must be satisfied if the homogenization method is to be valid.
The second condition in (10.271) simply requires that the reactor be
If these conditions are satis
large in comparison to the lattice spacing.
fied, then a good approximation to the sum
first term only (wii = m2 = 0) ; thus
Si
is obtained
by taking the
(10-272)
which leads to the interesting observation that in first approximation

Si is independent of r and the lattice geometry (i.e., shape of the unit
Thus in the present computation it would have sufficed to take
cell).
the field point r at any rod location i\; this would have reduced (10.263)
to the form [cf. Eq. (10.261)]
Co
- v|)e<-M
H(\X
(10.273)
(IkDc''8-
- -
(r*
- I)2t
(10-274)
(r>
r)
exp
eJir
[*
aS
s> =
/."
where y and X denote two rod positions and y X = ok; note that
r\ as X. The evaluation of (10.273) is a somewhat easier task than the
evaluation of (10.263).
The second sum 6'2 which appears in (10.263) is
comparison of this function with (10.264) reveals that S2

simply Si
divided by minus
and evaluated at r =
Thus we obtain from
n
0.
is
(10.270)
Xei(2x/o)mT
(2TWa)s[m +
m
(aB/2^)]2
As in the Si computation,
(10.275)
and
aB
the above result can be reduced to

simpler
form by imposing some restrictions on the system.
If we require that
*

s^rrxpr>
(10.276)
ANALYSIS
REACTOR
714
where
ci
with m = m(mi,m2)
(dfe)2
10
by
then the series (10.275) is well approximated
r+74F=
[CHAP.
"
(10.277)
coS[(2r/a)m.r]
(1027g)
m\ + ml
the prime means that the first term (mi =
m2 =
0)
is omitted.
The comparison
with the homogenization method is to be made on

criticality
basis
of
the
condition.
To obtain the appropriate relation
the
for the present method, we require that the field point r be taken at r*,
This corresponds to the procedure outlined in
the fuel-rod locations.
Then, collecting the results (10.272) and
connection with Eq. (10.261).
(10.277), we obtain for (10.263)
Co
YTTJTF*

where Ci(0) is (10.278) for the case r > r*.
and
co
=-
y~
Cl(0)
(10 279)
Finally, if
we call
+ c,(0)
(10.280)
(10.281)*
we obtain from (10.279) the relation

1
=
1
vfe-"*
+ L2B*
where L2 is defined by
Thus, if we identify as the thermal utilization and L as the diffusion

length for the equivalent homogeneous system, the relation (10.263)
yields the usual criticality condition (without resonance absorption)
obtained in the homogenization
method using the unit-cell concept.
But to obtain this result we introduced several restrictions on the system,
(10.271) and
(10.276);
these are the conditions
then which
must
be
fulfilled in order that the homogenization method yield reliable results.

It should be noted that the function Ci(0) accounts for the nonuniformity in the thermal-neutron distribution through the unit cell. Galanin1
has shown that for a square lattice
C^=44(lnS
where In 4x
1
t/3
= 1.48.
Galanin, op. tit., P/666.
^-ln4T + l)
(1-282)
A careful calculation using the Wigner-Seitz
SEC. 10.6]
715
Other
model leads to the same result, but with $ in place of the 1.48.
lattice geometries have also been computed by Galanin, and the appropri
Another detail
ate results for Ci(0) are given in the reference paper.
of
should
take into
that should be mentioned is that the computation
c0
account the finite size of the fuel lumps; therefore, in applying Eq.
(10.256), one should use in place of Vu, V m = Fu irp%.
Finally, it is of interest to note that the quantity c defined by Eq.
(10.280) is essentially the disadvantage factor f used in unit-cell theory.
Since,
!-i [wJf
(10-283)
(10.284)
f. Reactors with Small Number of Fuel Lumps.

Inasmuch as the
Feinberg-Galanin method is especially useful for analyzing reactors with
small numbers of fuel lumps, it is of interest to develop the criticality
The analysis which follows
relations in a form suitable for computation.
of
is carried out for the case
a square lattice geometry with spacing a
which constitutes the core region; the core is surrounded by an infinite
reflector.
The fuel lumps are cylindrical rods of radius pF and infinite
As in the preceding analyses, the resonance absorption in the
length.
fuel is omitted in this calculation.
The criticality condition is obtained
*i<
t)
t
1*
f;
- - Kt(xM\* -
(10.285)
denote the position of various fuel rods.

The notation
somewhat simplified by introducing the definitions
dXx
= pf
whenX^M
rji'
2irL\,
j\xp(-s-
Then
H-fl-dx,
and
where
of function H
[*
- 2-4
3.
i>
is
(-
from Eq. (10.261), with the

function H given by (10.254). For the present application it is con
venient to write this function in the form
- Koidû)
(102g6)
(10.287)
useful approximation to this expression can be obtained in the case

that rxlf =
<3C
which will apply in many practical situations.
To
1,
/3

we obtain from (10.281)
REACTOR ANALYSIS
716
[CHAP.
10
derive this form we write the first integral in (10.287) as
j~ exp(-swhere
A(dx>)
d^fS)
pg
Ko(dû)
A(dXll)
j\xp(-s - ^g^) ^
(10.288)
(10.289)
A(d) ~
|[(l
(^j
Next, the factor e~* is expanded in a Taylor series and the integration
carried out term by term; the first two terms yield the approximation
Ex
0e-w]
(10.290)
wherein we have used Eq. (5.306) and the recursion relation
\.e~*
- xE^[x)\
n >
Finally
Kl*
ff(dx)
- ^4(dv)]
l)ô(rfxÂ/)
(10.292)*
with A(d\M) given by

In the event that
be used in the present calculation,
1,
(10.290) and the d\ defined according to (10.286).

the Bessel function may be computed by
d\XM <C
H***
or
(10.293)
derived from Eq. (10.261),

= H(d\^)]
form [with
~
c0*x =
we write in the
.r
Co5x")*"
The criticality condition for the reactor

which, to simplify notation,
with
+ In
(o.5772
is
K0{x)
(10.294)
and for convenience we have labeled each rod with

As previously indicated, (10.294)
a set of homogene
single index.
nontrivial solution exists only
ous equations in the unknowns
identically zero, that
the determinant of this system
4>?
4>m(v):
is
is,
is
Here,
\H*
c05Xm|
(10.295)
if
In many cases of practical

This equation gives the criticality condition.
interest, the lattice configuration will exhibit several degrees of symmetry.
As
result, various elements of the determinant will be repeated, and
this in turn can lead to high orders of degeneracy in the determinantal
Thus the complicated equation symbolized
(characteristic) equation.
by (10.295) will in actual practice reduce to considerably simpler forms.
well demonstrated in the example which follows.
This point
To illustrate the computational procedure, we consider
heterogeneous
Each rod
at the corner of a square
reactor consisting of four fuel rods.
is
is

This result will
(10.291)
-4rr
n 1
En(x)
sec. 10.6]
with sides of length
a (see
Fig.
717
The distances
10.20).
d\ between the
various rods are
dii
^23
du
du
= a
(fu
d2i =
y/2
ggg)
and we find that the determinant
(10.295) can be written in terms of

three numbers, namely, bu, bu, and bu, where
An,
Hn
#12
Hu
(10.297)
o^rr [(^
iT^2a2
2^2"
- i?eMn]
l)Xo(axv) - ijeMu]
l)Ko(V2 ax*) - vefAu]
1)K0(pfxm)
(10.298)
Au
-4i2, and
may be computed from (10.290).

(10.295) is a simple 4 by 4 determinant
The relation
coK
values of H\ are [with A\? = A(d\)]
The appropriate
and
shown to yield the equation

(6n
- MWn
+ bnY
46?2]
which is easily
= 0
(10.299)
which has four roots:

bw = &i3
a double root
bu + bu
Only one of
=
physically acceptable, however.

require that
^
situation
impossible
4>t,
since
an
since all the
they
must be identical by
Likewise the
symmetry requirements.
tfy
-(/)i
third root, corresponding to the negative sign in the second equation, must
be rejected since it would require that
= 2Only the root correspond-
ing to the positive sign is acceptable

inasmuch as it allows the solution
= constant.
This relation is in fact
4>n
the criticality condition.
As previously
noted, this result can also be obtained
by imposing the symmetry requirements
directly onto the determinant.
Since
4>
The first two
these roots is
are rejected
<pt
(10.300)
2f>i2
= constant,
then
X/
/
/
4^
r*
FlG-
\s
10 '20
sw'
/ \s
//
~0~~
t
s\
\j>.
~
<
Four-rod
heterogeneous
in the present case we have

each of the four equations implied in (10.294)
REACTOR ANALYSIS
718
[CHAP.
10
must be identical; moreover, they must each yield the relation
#11
Hu + 2Hu
(10.301)
Co
which is precisely the second of equations (10.300), using the plus sign.
For computational convenience we write (10.301) in the form
ijW-u + *i8 +
2^12)
xii + Xn +
Hx = #x
where
2Xi2
Co
(10.302)
(10.303)
xxM
and
l^xji
a2gfl
2vli
[ô(d\xM)
x\n
A\\
a2
2^2 Ko(dixu)
(10.304)
The results obtained in the preceding analy

criticality requirements of a four-rod reactor
used
to
determine
the
ses are
using enriched uranium and a graphite moderator and (infinite) reflector.
The present results are presented in the form of curves of the multipli
g.
Numerical Example.
constant versus rod spacing for various U23S enrichments.

A
The computed multiplication con
single rod size is used throughout.
stant for two different heterogeneous arrangements is then compared to
the values predicted by the homogenization method using the two-group
The nuclear data and other parameters used in these compu
model.
tations are as follows:

cation
Lu
= 52 cm
PF = 3 cm
a28 = 2.75 barns
= 364 cm2
= 2.08
1726
Nv
= 0.0482
a?
1024
DF
nuclei/cm'
= 0.000355 cm"1
= 687 barns
= 0.702 cm
where (28) refers to U2M and (25) to U235.

The multiplication constant for the four-rod reactor is obtained from
Thus if tj is the average number of neutrons produced per
(10.302).
absorption in the fuel rod (taking into account the concentration of U23i
and U238), and ije is the usual fictitious yield required to achieve steady
state, then
k =
Ve
For
Co
we use [cf.
+ 2^)
^n+^,
+
Xll + Xl3
2X12 +
Co
Eq. (10.256) and discussion following Eq. (10.282)]

Co
= (a2
TpF)h{pFxr)'ZiaM'>
7-7
t^ttt
MnQfiR^
(10.306)
Equation (10.305) has been applied for the enrichment ratios Nn/Nu =
The
0.05, 0.10, 0.20 as well as for natural uranium N/Nv = 0.0071.
results are shown in Fig. 10.21.
Since the four-rod lattice is a near minimum
heterogeneous configu
which is incurred in
determine
the
error
it is of some interest to
To obtain a com
analyzing the system by the homogenization method.
ration,
sec. 10.6]
719
parison we define an equivalent homogeneous reactor which has the same

moderator and reflector as the heterogeneous configuration.
Since the
heterogeneous system has an axial symmetry we define the core of the

l-2i
30
40
Rod spacing a, cm
Fia.
10.21
Multiplication constant
as a function of rod spacing in a four-rod graphite
reactor.
homogeneous reactor to be a cylinder

IL,
where n is the number of fuel rods.
of infinite length and radius:

(10.307)
(v')'
The nuclear characteristics
of this
system are computed with the aid of the two-group model of Sec. 8.4.
The principal parameters required for this calculation are the infinitemedium multiplication constant A"'i", the roots of the characteristic equa
tion (8.171), and the constants ai and a2 (8.175), as well as the cylindrical
functions listed in Table 8.3. The infinite-medium multiplication con
stant is computed from
1
using Eqs. (10.280), (10.282), and (10.306).

teristic equation we use the approximations
(10.308)
For the roots of the charac

In the
(8.226) and (8.227).
REACTOR ANALYSIS
720
present case
m2
[CHAP.
is given by
'i(-0[-SK->)]
where
the L2
10
is computed
L2 = L%(1
from
/).
(10'309)
The symbol
Arhom
denotes the multiplication constant of the equivalent homogeneous sys

Finally, the coupling constants are obtained from
tem.
01
(1-310)
2'*^
and 2?' = D,/L\

This general scheme has been used to compute the three configurations
All three reactors have been taken from the
indicated in Table 10.9.
Table
10.9
CoMPARisOiN
of the Feinberg-Galanin
Method with the
HoMOGENIZATION METHOD
cm
cm
20
22.5
33.8
45.1
0.978
0.934
0.868
1,

30
40
*sr
&hotn
fchet
1 .890
1.198
1.220
1.060
0.954
1.005
0.978
1 .802
1.680
Error
+26
21
8.4
results given in Fig. 10.21 for the 3-cm fuel rods with 5 per cent enrich
The three cases correspond to a = 20, 30, and 40 cm. The
ment.
parameters common to the three systems are:
v = 1.935
^
IS u
= 0.05
Di
2tf>
D,
D,
= D4 = 0.909 cm
= 2g = 0.00249 cm"1
along with the nuclear constants given previously.

Similar calculations performed by Galanin1 for a number of lattice
schemes indicate that the error due to the use of the homogenization
method is generally a few per cent, the actual magnitude depending
strongly upon the nuclear properties of the system, as well as on the
geometry and number of rods.
1 Galanin,
op. cit., P/663.
CHAPTER
CONTROL-ROD
11.1
11
THEORY
Central Rod in One -velocity Model
The present treatment of the problem of computing the effectiveness

of a control rod is carried out on the basis of two analytical models.
In the application of the first, the one-velocity model, attention is directed
toward displaying

some of the elementary
physical ideas involved in the
primarily for purposes of

illustration; it is not intended that it be considered an accurate tool for
The second method utilizes the
computing control-rod effectiveness.
is
two-group model, and it
expected that the results obtained from this
formulation will be useful for the solution of practical problems.
In order that the present analyses remain relatively unencumbered
general problem.
The model
is used here
of mathematical complications a single elementary reactor configuration

has been selected, and all control-rod calculations are carried out for
The geometry used is the bare cylindrical reactor.
that system alone.
It should be recognized, however, that this choice does not necessarily
limit the results obtained here to this particular configuration.

If the
actual reactor of interest has a markedly different shape, it is still possible
to determine an "equivalent" bare cylindrical configuration by matching
Also, if the actual reactor is reflected, a suitable bare cylinder
bucklings.
can be computed by applying reflector savings corrections to the various
Some care should be exercised in dealing with this
reflected surfaces.
latter situation, however, if the control-rod locations are near the coreUnder these circumstances the bare-reactor calcula
reflector interface.
will
tend to underestimate the rod effectiveness in certain cases
tions
since in the actual reflected reactor the rods may be exposed to sub
stantially higher thermal fluxes than in the bare model because of pos
sible peaking in the reflector (see Fig. 8.23).
In any case, it is expected
that estimates based on the bare cylindrical reactor will give a reliable
indication of the characteristics of a design configuration, very possibly
within the limitations of the two-group model.

In the various control-rod problems considered here it is always
assumed that the rods have circular cross sections and that they are
inserted with axes parallel to the axis of the reactor.
Calculations
of rod effectiveness are carried out only for the case of the fully inserted
721
REACTOR ANALYSIS
722
[CHAP.
11
Although methods have been developed for treating the problem

of the partially inserted rod, these either require more sophisticated
analytical techniques (perturbation theory) than available here or are
semiempirical, and calculations must be correlated with measurements
(the "window-shade" method1 used by H. L. Garabedian).
Finally, it is pointed out that, unless otherwise stated, control-rod
effectiveness is measured in the present work by the reactivity of the
reactor without the specified rod configuration,
In performing
but with the same fuel loading.
these calculations it is assumed that the reactor
with fully inserted rods is just critical; the reactor
without rods contains no "holes" corresponding to
rod.
the rod locations.
Cylindrical Reactor with Gray Control Rod.

elementary example of a control-rod
calculation consider the case of a bare cylindrical
thermal reactor operating at steady state with a
fully inserted central gray control rod. By "gray
rod" we mean a rod composed of a material which,
a.

As
a first
although a reasonably good neutron absorber, is

not so strong an absorber that it cannot support
Also, let us assume that diffu
some neutron flux.
sion theory is applicable within the interior of this
-H26
Fig.
11.1
Cylindrical
reactor with central control rod.
It should be noted that these assumptions

rod.
not unlike those imposed for fuel lumps in the calculation of the ther
mal utilization for heterogeneous lattices.
The general geometric configuration of this reactor with control rod
The radius of the reactor is R and its height 2h.
is shown in Fig. 11.1.
The radius of the rod is b. At steady state the appropriate one-velocity
neutron-flux equations are
are
+ 2Lc'4>c(r)
DbV*Mt) + 2fl,B(r)
->cV2>c(r)
= v~2fP*g*4>c(r)
= 0
(11.1)
where C refers to the reactor and B to the control rod and the remaining
The neutron fluxes given by the
symbols have the usual definitions.
above equations are to satisfy the following boundary conditions:
(1)
c(R,z)
Mp,fi)
*(P.)
= 0
finite for all r
(2)
4>b
(3)
The flux and net current

reactor interface
are continuous
at the rod-
(11.2)
Reactor Theory,"
H. L. Garabedian, "Boundary Conditions in Multi-region
Westinghouse Atomic Power Division, Pittsburgh, Pa. WAPD-TN-516, July, 1956.
1
THEORY
CONTROL-ROD
SEC. 11.1]
723
The solution to these equations may be obtained by the method of sep

It is easy to show that, with the application of the
aration of variables.
first two conditions (11.2), the results may be written
*c(p,2) =
B(p,z)
A[Y0(\iR)
ô(Xip)
= CI0(X2p) cos
- Jo(Xi#)
ro(Xip)] cos
TTZ
where
*B
DB
(11.3)
(11.4)
The criticality condition for this system is obtained by the application

condition (3), in (11.2), above.
This is readily found
of the interface
to
be
Dc\i[Y0(\iR)
J^b)
- JoixM)
F^X.b)]
F0(Xi6)
The usual procedure in carrying out control-rod
P^/t^b)
70(X2b)
calculations
is to deter
mine the rod radius
required to overcome a given amount of excess

multiplication. Thus the reactor configuration and nuclear parameters,
including the fuel loading, are generally known, as are also those of the
rod material (in the case of the gray rod).
This means that all quantities
in the above equation are determined,
except the rod radius b. This

equation is used, then, to compute 6; since the relation is transcendental, a
trial-and-error procedure is required and the usual practice is to select
various values of b and plot the b-dependent portion of the function
For this purpose the relation

against the /2-dependent portion.
above is more conveniently written in the form
Ko(X^)
Ja(\iR)
= Y1{X1b)
Ji{\ib)
- DgXihiXj)) F0(X,b)/DcXi/o(X2b)
- DeXj^b) y0(Xib)/Z)cX,/o(X2b)
given
'
is,
The effectiveness of the control rod determined by this equation is

measured in terms of the excess multiplication of the reactor prior to the
insertion of the rod. If we assume that the omission of the rod leaves no
"hole" (e.g., a fluid core reactor), then the multiplication constant of
As a rule it is useful to describe
the reactor is obtained from Eq. (6.81).
the
cent
it can suppress, that
reactivity
the rod effectiveness by
per
by
w
= bk/k.
b.
Cylindrical Reactor with Black Control Rod.
In the present exam
ple we consider the reactor configuration of the preceding analysis, but

black control rod.
The "black" rod differs from the
now containing
a

Fo(X,) y0(X,6)
- Jo(X.tf)
REACTOR ANALYSIS
724
[chap.
11
Thus the black

gray rod in that it can support no neutron flux at all.
rod serves effectively as an "absolute" sink of neutrons.
From this
rod
like
just
the
black
behaves
the
space surrounding the
standpoint
outer surface of a reactor; any neutrons which enter this region are
lost to the chain reaction.
The analogy is in fact precise,
permanently
and one is easily lead to the use of the extrapolation-distance
concept
for describing the characteristics of the rod.

Figure 1 1.2 shows the local flux distribution in the vicinity of a centrally
located black control rod.
The dimension b is the actual radius of the
rod, and e is the extrapolation distance into
the rod (cf. Fig. 5.10).
The dimension 6,
defined by the relation
b, =
computed in terms of the nuclear proper

ties of the reactor materials (and not of
the rod).
However, inasmuch as b is fre
b >

(11.6)
As in
is the effective radius of the rod.
the case of the extrapolation distance
for the outer surface of the reactor, t is
6.
1>
Fig. 11.2 Extrapolation distance

for black control rod.
quently of the same order of magnitude

as the transport mean free path in the
Xjf, or very much
reactor
less, the usual
(5.71) is not generally applicable.

In fact, in the limit for very small values of b, a proper estimate for t is
For intermediate values1 of the rod radius, e may be determined
4X[f >/3.
relation
from Fig. 11.3.

The neutron flux in the vicinity of the rod may be estimated with the
aid of the extrapolation length from the relation [cf. Eq. (5.65)]
If
K&,)~
/>(&)
we specify
- *'(&)
(11.7)
<Kh)
(11.8) *
Q>)
(11.9) *
then [cf. Eq. (5.66)]

4>'(b)
In most of the subsequent analyses
we
will
use the
extrapolation-length
For the present problem it is more suitable to use the

the latter is the form in which it was first used by Scalettar
condition (11.8).
relation
(1 1 .9) ;
and Nordheim.2
Davison and S. Kushneriuk, "Linear Extrapolation Length for a Black Sphere
and a Black Cylinder," MT-214, Mar. 30, 1946.
R. Scalettar and L. W. Nordheim, "Theory of Pile Control Rods," MDDC-42,
1 B.
June
17,
1946.
SEC. 11.1]
THEORY
CONTROL-ROD
725
The appropriate differential equation for this system is given by the

first of the set (11.1), and the boundary conditions by (1) in (11.2) and
which satisfies condition
by (11.9). The general solution for the flux

1-4 1 I
r-i
i i
i I i
0.7
0.6
2
Control-rod
Fig.
11.3
Extrapolation
(1) is given in (11.3).
condition.
3
radius
b/\u
length for black cylindrical control rods.
The application of Eq. (11.9) yields the criticality
(11.10)*
This equation may
be solved
for the required rod
size b for a given set of
reactor parameters by the procedure described previously.

c. Cylindrical Reactor with Central Hole.
In some practical situations
(solid-core reactors) the withdrawal of a control rod leaves a hole in the
reactor.
The present example demonstrates
a procedure
for computing
the effectiveness of a control rod in such systems; however, in order to

we assume that the reactor is of infinite length.
simplify the calculation
ANALYSIS
REACTOR
726
[chap.
11
With the application of the outer boundary condition, the solution

given by the form (11.3), which we now write in terms of X3; thus
0c(P,z) =
A[Y0(\,R)
- Jo(\iR)
70(Xp)
F(X3P)]
cos^|
XjsB!_C*Y
with
and B2 as given by Eq. (11.4).

criticality relation
The
Y0(\3R)
use of
is
Mp'2)],-b
[fp
The flux distribution for this system is given by the differential equa
tion (11.1). In addition to the extrapolation boundary condition for the
outer surface, (1) in (11.2), we require for the rod-reactor interface1
(11.12)
(n.13)
condition (11.11) yields the
YÛb)
(11.14)
a gray rod we apply Eq.

(11.14) to determine first the critical fuel concentration for the reactor
with hole, given the core size and rod radius.
[Note that the solution of
(11.14) yields X3, which may then be used in (11.13) to compute the con
6,
we have
z;f>
+ Dc{\\
(*M)2]
v;
then
thus,
2hy-\
then given by the relation
\\
of the rod
is
worth,
The reactivity, or
Dc[\\ +
xi
s<C)
The reactor with rod inserted, and the same fuel concentration,
subcritical, and for steady-state operation we require some vc >
(1L15)
is
vlH^
just critical,
rod
is
is
is
The next step

to use this concentration, along with the
centration.]
other nuclear parameters and the values of
in Eq. (11.5) to
and
determine
value for Xi.
The final step
to compare the values of
Since the reactor without
required to keep each system at steady state.
+ {v/2Ky
(n.17)
The appropriate interior boundary condition for a finite cylindrical reactor with
central hole
discussed by E. L. Zimmerman, "Boundary Values for the Inner Radius
of a Cylindrical Annular Reactor," Oak Ridge National Laboratory, ORXL-2484,
June
5,
is

In order to compute the effectiveness of
1958.
CONTROL-ROD
SEC. 11.2]
THEORY
727
It
is of interest to mention that one of the important effects of a control

rod is the increased leakage it produces at the outer surface of the reactor
and that this, rather than the extra neutrons absorbed, may be the pre
dominant factor in determining the worth of the rod.
A similar computation to that given above may be carried out to

determine the reactivity of a black control rod which leaves a hole in
For this case Eq. (11.10) is used in place
the reactor when withdrawn.
of Eq. (11.5).

11.2
Central Rod in Two-group Model
The one-velocity model is generally inadequate

a. General Solution.
for obtaining reliable estimates of control-rod effectiveness. A more
accurate method which is customarily used in design practice is based
The application of this model to the
on the standard two-group model.
of
a
calculation of the effectiveness
single centrally located round rod in
a bare cylindrical reactor is a straightforward extension of the general
results obtained in Sec. 8.4 for the cylindrical reactor with side reflector
(refer to Table 8.3). The only modification required for this application
is a restatement of the rod-reactor interface condition in terms of the
particular properties of control-rod materials.
The primary purpose of the present section is to exhibit the various
interface conditions that have been proposed for use in control-rod calcu
lations.
It is of interest to mention that the accuracy of the entire
computation is very much dependent on the proper choice of the inter
In the analyses which follow we con
face condition for the fast flux.
sider a number of such choices and show how the appropriate criticality
relation is derived for each case for both a solid and a hollow control rod.
This section is concluded with a numerical example whicji compares the
effectiveness of these two types of rods when fully inserted in the center
of a bare cylindrical reactor.
We begin the analysis with a summary of the general solutions derived
for the two-group model. The appropriate differential equations for the
fast flux
and the thermal flux (<2) in the multiplying region of the
reactor are given by (8.147).
In the present problem it is convenient to
write the solutions for these functions in the form [cf. Eqs. (8.164)]
(i)
*(r)
*,(r)
= Z(r)
=
+ W(r)
o,Z(r) + a,W(i)
condition on the outer surface of the
These must satisfy the boundary

reactor :
*,(ft)
uli8J
*,(ft)
= 0
(11.19)
The functions Z and W are solutions to the differential equations

V2Z(r) + M2Z(r)
= 0
T-W(i)
- \W(t)
= 0
(8.174)
728
REACTOR
ANALYSIS
[CHAP.
11
and the parameters ps and X2 are the two roots given by Eq. (8.171).
The coupling constants Oi and a2 are defined by Eqs. (8.175). For the
reactor with a single central rod, the flux distributions are axially sym
metric and we may define
Z(r) = H(z) R(P)
W(r)
H(z) 5(p)
(11.20)
If
these relations are used in Eqs. (8.174), it is readily shown that for the
bare cylindrical reactor the indicated functions are given by the expres
sions (refer to Fig. 11.1)
=
H(z)
R(p) = OJ0(yp) + (&Y0(yP)

S(p) = 3R/,(/9p) +
and
fHKM
(11.22)
X- +
(11.23)
. - (gf , .
where
The neutron-flux
*1(p,2)
2(p,z)
(g
relations (11.18) may then be written
H(z){dJo(yp) + Fo(7p) +
H(z)\ai[aJ0(yp) +
(&Yo(yp)}
+ yiKoipp)]
+ a2[3n/0(/3P) + HK0{fip)]\
3Tt/o(/3P)
(11.24)
The coefficients a,
rod-reactor
(B, 3E,
and
31
are evaluated
by the application of the
interface condition.
Solid-absorber Control Rod. For the first example we consider

the case of a fully inserted central control rod which is black to thermal
neutrons only; thus the thermal-neutron properties of the rod material
are similar to those specified for the rod in the one-velocity problem of
Sec. 11.1b.
For the fast-neutron properties of this rod we allow two
different possibilities: (1) the fast flux remains finite everywhere within
b.
the rod, and (2) the net current of fast neutrons is zero at the surface
of the rod.
Consider first case 1. Specifically, we will impose the conditions:
2
=
with
02(6,
6,
(2)
31
,z)
(1);
1
(11.25)
e2
0.
The first of these follows from the fact that the expression for the fast
=
flux
contains two functions F0 and K0, which are singular at
Since we have that
is
limp!g=_?
it

(1L21)
cos(i)
(11.26)
clear that the choice given in (1) of (11.25) provides that the fast
CONTROL-ROD
SEC. 11.2]
flux remains finite for all values of

dition (2) requires that the thermal
THEORY
729
The second con

flux vanish at the effective radius
which is determined from the actual radius b by the appli
p
inside the rod.
of the rod 6* ,
cation of an extrapolation length correction based on the thermal-group
properties of the core [cf. Eq. (11.6)].
The combination of control-rod boundary conditions (11.25) is based
on the notion of representing a control rod by a line singularity. It

should be recognized that condition (1) in (11.25) serves essentially to
keep <pi finite, and this is achieved by relinquishing control over other
As will be seen shortly, this condition
physical features of the system.
leads to a large net current of fast neutrons into the rod, which is incon
sistent with the premise that the rod is a strong absorber of thermal
neutrons only.
This discrepancy can result in serious overestimates of
rod effectiveness, and for this reason the method is not recommended
for general usage.
It has been included here primarily for reasons of
historical interest, being
formulations of the control-
one of the earliest
rod problem.
The application of condition
Ja/0(7P)
3HJoG3p)
H(z)
+ l^lY0(yp) +
joJo(7P)
Jo,
*,(p,z)
9lF0(7P)l
+ 9ltf
(11.24) yields
(&>)[
a2[3TC/o((8p)
(/3p)]J
(11.27)
obtained
if
is
between the coefficients 9TC and 91 and between

and 3d
we use next the boundary condition for the outer extrapolated
surface of the reactor (11.19).
These are
and
3lff(2)
YojyR)
- M^
J0(yp)
Jo(yR)
+
a2
|-^|
(11.28)
yields
r(7P)
ZoOSp)
KoOSp)
(11.29)
j
into the expression for
<p2(p,z)
these results
jâ,
The substitution of
-aZtfrg
relation
ft
is
The application of condition (2) from (11.25) to this equation gives the
criticality condition for the reactor. This
found to be
2q2r
Y0(yR)

*,(p,) = H(z)
(1) to the general solutions
K{ffi)
11
(11.30)
REACTOR ANALYSIS
730
In general f)R >
and K0{pR)/ Io(pR)

to the above result is
1,
Yo(yR)
J0(yR)
_ Jo(yK)
1
<C 1;
[CHAP.
11
then, a good approximation
Ya(yK) +?^Ko(f)h)]
(11.31)*
It is of some interest to note that this result may also be written in

terms of the extrapolation distance 2 in place of the effective rod radius
6,. For this purpose we require the Taylor series expansions about
In general, we can
p = b for the various Bessel functions involving 6* .
apply the functional form (11.7); then, for example,
J 0(ybt) ~ J
and so on.
The
use
of this
0{yb)
+ e2yj i(yb)
form in (11.31) yields
the alternative
expression
Y0(yb) +
wY^yb) +
Jn
J0(yR)
^ [Jf .056)
us-^-n
n
+ e2yJ i(yb)
+ crftf ,(/)]
(H-32)
o(yb)
prefer this form to (11.31).

From an analytical stand
point this implies that the thermal-flux condition of the type (11.9) was

Some authors
used instead of (11.8).

Criticality analyses may be performed with the aid of Eq. (11.31) and
the procedure outlined for the one-velocity model of the preceding sec
tion.
As before, the criticality relation may be used to compute the
effective rod size 6* required to maintain a reactor of given size and fuel
concentration at steady state.
The rod effectiveness may then be deter
the multiplication constant of the reactor for the
same fuel loading, but without the rod (and without a residual "hole").
Figure 11.4 shows the local flux distribution in the vicinity of a central
mined by computing
control rod as determined from the condition that the fast flux be finite
inside the rod.
Note that, although i(0,z) is finite as required, it dips
to large negative values inside the rod and produces a large gradient at
the rod-reactor
For
interface.
case 2 we take the

(3)
following pair of boundary conditions:
!"*i(p,)l
=0
(4)
2(6,,z)
= 0
(11.33)
The coefficients in the general solutions (11.24) are evaluated by first

applying the boundary condition for the outer surface of the reactor.
From the condition for the surface p = R we obtain the relations
and
(11.34)
SEC. 11.2]
CONTROL-ROD
THEORY
731
Rod reactor interface
Fig.
Flux distribution around a central rod using the condition that fast flux is
11.4
finite everywhere in the rod.
If
these results are used in Eqs. (11.24) and conditions (11.33) applied,
two equations relating, say, CI and 911. From condition (4)
we obtain
we have,
for example,
a \j0(yb.)
- ig. F(76,)l
- J [7,096.) - Mjg- KMb.)]
311
(11.35)
If we
use also condition (3) and divide the resulting expression by (11.35),
the following criticality equation is obtained :
0
Y0(yR)
Jo(y&)
Y0(yb*)c3(b) + (a2/al)Y1{yb)c1(K)
I Jo{yK
(11.36)
)C3(6)
+ (0s/0l) J1(76)c1(6.
where
Ci(x) = i0(/3x)
The indicated
Ko(pR)
approximations
KoiPx)
apply for R
KoW)
>5> x.
(11.37)
For most practical
732
ANALYSIS
REACTOR
[CHAP.
11
situations ft^>b, and a good approximation to (11.36) is
"<3
J*(yb.)
AA Kt(flb)
"
K,(7b)
,,M
YojyR)
(11.38)*
is
The flux distribution for this case

shown in Fig. 11.5.
c. Hollow Control Rod.
In most practical situations control rods are
is
it
not fabricated from solid sections of high thermal cross section materials.
Since the mean free paths in such materials are very short,
clear that
very few thermal neutrons can ever
penetrate far inside
rod of several
centimeters width. Nearly all the
neutron captures occur close to the
surface of the rod, and the interior
serves no real purpose.
Thus the
one
is
In practice
therefore led to the

configurations
Rod reactor interface
I,
ft
use of cross-sectional
Fio. 11.5 Flux distribution around

central rod using the condition of zero
net current of fast neutrons at the rod
surface.
having large perimeters, such as

blades, shells, and tubes.
In the treatment which follows we
analyze the effectiveness of control
rods in the form of hollow tubes of
is
absorber (thermally black) material.

carried out in detail for the round tube,
the results may be applied in first approximation to tubes of any cross
section by simply matching perimeters.
Two cases of considerable practical interest are considered: (1) the
Although the present study
I.
it
is
is
hollow "control tube" and (2) the control tube containing moderator
material.
The first case
essentially the same as an absorber tube filled
In both situations the tube stops the
with high-atomic-mass material.
Fast neutrons, however, easily pass
passage of all thermal neutrons.
black to thermal neu
through the tube (since by assumption the tube
trons only) and, moreover, are little affected by the interior of the rod,
In either case the fast
whether
be empty or filled with heavy nuclei.
neutrons emerge with much the same angular distribution as that with
which they entered and without any significant degradation in energy.
From an analytical viewpoint both problems are equivalent to the solidcontrol-rod problem with zero fast-neutron net current at the surface.
Thus the results (11.36) and (11.38) apply directly to case
In case
the moderator material within the control element serves
2

is
surface area, or cross-sectional perim

eter, of the rod
rough measure of
its effectiveness as a neutron sink.
CONTROL-ROD
SEC. 11.2]
THEORY
733
Since the tube itself is transparent

effectively as a sink of fast neutrons.
to fast neutrons, these have access to the interior and some are removed
from the fast group by being scattered to lower energies by the moder
Configurations of this type are frequently encountered in

ator material.
A good example is the control-rod design appropriate to the
practice.
The core of this reactor
nuclear package power reactor of Sec. 8.5.
The
consists of a lattice of MTR-type fuel elements immersed in water.
control "rods" are simply hollow rectangular tubes of absorber material
which fit into regular fuel lattice points (see Figs. 5.27 and 5.28). When
the rods are inserted, the interior regions
fill with water, and the resulting fast-neutron
Rod-reactor interface
within the rod yields a substan

tial increase in the worth of the rod by vir
moderation
tue of its properties as a fast-neutron sink.

(Note that thermal neutrons produced by
this moderation cannot escape from the rod

and thereby contribute

tion in the core.)
to the chain reac
A first estimate of the effectiveness of a

moderator-filled control element can be ob
tained from the application of the diffusion

11.6
dis
Extrapolation
theory to the interior region of the rod.
Thus in the present calculation we treat the tance for the interior of a hollow
control rod.
rod as though it were an internal reflector,
from the viewpoint of the fast-neutron group.
From the viewpoint of
the thermal
neutrons, the absorber tube acts as a curtain for neutrons

passing both in and out.
The appropriate
equations for the flux distribution within

the rod are given by (8.149).
In the present application fa and fa
denote the fast and thermal fluxes, respectively, in the reactor core, and
fa and fa, those in the control rod. The boundary conditions to be
satisfied by the fast flux fa are the usual requirements of continuity
two-group
and finiteness:
(1)
fa(b,z) = fa(b,z)
(2)
^*3H,-=^*iH,
(3)
fa{p,z)
finite within rod
and for the thermal flux fa
*4(W)
= 0
(11.39)
or
fa(b,z) =
-u
*4(p,z)l
(11.40)
distance based on the transport mean free

4 is the extrapolation
path of the material contained in the rod and (refer to Figs. 11.3 and
where
11.6)
+ u
(11.41)
REACTOR ANALYSIS
734
[CHAP.
As before, we require that fa satisfy Eq. (11.33), condition (4).

solution for the fast flux can be written in the form
,(p,z) = H(z) T(p)
where H(z) is given by Eq. (11.21).
condition (3) in (11.39), is
T(p) =
11
The
(11.42)
The function T(p), which satisfies
eloiw)
y,3>
/ ir
(11.43)
\2
(11.44)
Note that the first term in this relation may be approximated by 1/r,
where r is the age to thermal for the central region of the rod.
These results may be used to determine the flux distribution within
the core and the criticality condition for the reactor.
The general
solutions for the fluxes in the core are given by Eqs. (11.24).
If we apply
boundary condition (11.19), i.e., the relations (11.34), the solutions may

be
written
- ff(2)
*,(p,*)
(a p,(w)
K(W)1
+
*,(,2)
;/(z)
L,a
- ['*>- 8 *HI
(11.45)
r,(Yp)J
The expression for the fast flux must satisfy the interface conditions
(1) and (2) in (11.39), and the thermal flux, condition (4) of (11.33).
The relations for the fast flux we write in the form
"3
_
where
dp
r
"
.
01(P'Z)
41
n i i
k<
(11.4b)
Jp
ti denotes the extrapolation distance for the fast group in the core.
Note that this quantity may be computed in terms of the fast-group

properties of the interior of the rod by virtue of the continuity relations
(11.39).
Equation (11.46) together with (4) from (11.33) constitute the critical
ity condition for this system. From the latter, we obtain Eq. (11.35)
BEC. 11.2]
relating
CONTROL-ROD
and
Ct
THEORY
735
A second relation is obtained from (11.46)
911.
0311
7,036)
Kxifib)
(11.47)
Equation (11.35) may be used in this expression to eliminate a and 3TC.

The resulting expression is the criticality condition, which we write in
the form
Yo(yR)
Jo(yR)
with
c\ and
c3
(a,/ai)ci(b.)c(fr) + F0(76t)[gf,c3(6)
(o/oi)ci(6,)c,(6) + y0(76*)[/3tlC3(6)
- c,(b)]
- ci(6)]
defined by (11.37) and

c2(6)
J0(76) + 7*1^1(76)
+ 7iF,(76)
(11.49)

c4(6) ^ F0(76)
When R
~2> 6,
a good
approximation to (11.48)
^0(76,) A7(6)
7o(7^)
wherein we have used the approximations
The result (11.50) has also

distance
e2,
(11.48)
at
is given by
c2(6)
from (11.37) and defined
rgilg1(|6MôW]
(n
51)
been developed1 in terms of the extrapolation
where
*2(6,z) =
(H-52)
*[^*(p.*)]^
It
is of interest to determine the limiting form of the criticality con

dition (11.50) for very small rods (6 > 0). The application of the
appropriate forms for the various Bessel functions in the case of vanishing
argument
gives the approximations
M(b)
~ j3ai6i7Li(/36)/a2/C0(i36)
7iFi(76). Also since c2 approaches a finite limit as

neglect it in relation to the M term in the denominator.
also the limit
c4 =^
Um^fg=-|
7*1(76)
fc-.o
1 R.
L. Murray and
J.
W. Nieetlie, Nucleonics, 13(2), 1&-22 (1955).
and
> 0, we may
Then, using
(11.53)
ANALYSIS
REACTOR
736
[CHAP.
11
it is easily shown that the result (11.50) reduces to the form

Yo(yR)
_ Ya{yK)
/oA
{orb_^0
Ra(fiK)
(u
54)
This approximation may be checked by comparing it to the limiting

form of the relation (11.36) or (11.38), the criticality condition for the
solid rod with zero fast current at the rod surface.
The comparison
is easily drawn with the aid of the approximations mentioned above
for b * 0. It is then seen that (11.38) reduces to (11.54), as is to be
expected, since for the very small moderator-filled control tube, the effect
of the moderator as a fast-neutron sink becomes negligibly small and
the tube behaves like a solid rod.
The thermal-flux
control rod (tube)
distribution within the moderator contained in the

is obtained from the differential
-Z>V**(r)
ZS*>*4(r)
equation
p^Mr)
i-
a,T(p)]
(11-55)
where pj denotes the resonance-escape probability for the moderator

material inside the rod.
The solution for
we take in the form
Up,z)
= H{z)
[&U(P)
with H and T as given in (11.42).

tion to the homogeneous equation
VV
x4V
The function
with x\ =
= 0
HU
(11.56)
denotes the solu
(11.57)
The function H(z) was previously given by (11.21), and

U(P) =
where k\ = x\ +
Zo(p)
Q^Y
(11.58)
We have used here the requirement that the solution remain finite inside
If we use (11.56) in (11.55),
the rod and that Eq. (11.19) be satisfied.
along with (11.42), we obtain
o,[Vs(ffD
- x]HT]
But, by Eq. (8.149) for
wherein we have used (11.57).
V\HT)
- x\HT
Thus (11.59) yields the relation
HT
= 0
[cf.
with x\ =
(11.59)
3,
we note that
(11.60)
Eq. (8.184)]
(11.61)
CONTROL-ROD
SEC. 11.2]
THEORY
737
The constant 8 indicated in (11.56) may be related to 6 from (11.43)

by applying the boundary condition (11.40). The result is
8 =
If
fl/o(*ib)
this expression is used in (11.56), along with (11.61),
we obtain
for the
thermal flux inside the rod

<(P,z)
It
n ?'e
^(Xj
*l)
V,
[{'{"H
L'o(*4t*)
/.(/)
/.(*)! H(z)
should be noted that the constant 6 may be related to
CI
and
911
(11.62)
through
(2) in (11.39).
Figure 11.7 shows the flux dis
tribution around
rod containing
d.
tubular control
moderator material.
Numerical Example.
sults
developed
The
re
in the preceding

analyses are used to determine the

reactivity of a single centrally lo
cated control rod for both the solid
absorber and moderator-filled tube
For the sample re

configurations.
actor we use the package power
plant of Sec. 8.5. The appropriate
two-group constants for the core
region of this reactor are listed in
Rodreactor interface
Table
8.5.
ample
we
obtain
B2 = 0.006125 cm-2, which corresponds to a square
For the present ex

Fig. 11.7 Flux distribution
around
select the fuel volume
rod
moderator
containing
a control
fraction vF = 0.00622, which gives
material.
=
1.312.
con
Since
the
actual
fcM
figuration is a completely reflected square cylinder, it will be necessary
to determine the size of an equivalent square unreflected cylinder with the
same fuel loading.
For this purpose we use Eq. (8.224) with k = 1, and
cylinder with
= 36.7 cm.
Some indication of the relative effectiveness of solid absorber and

water-filled control rods is obtained by computing 8k/ k as a function
of rod size for the 36.7-cm-radius bare cylinder.
The results of such
a computation are shown in Fig. 11.8.
For both types of rods, the
problem was solved by selecting various fuel loadings (corresponding
to various values of 6k/ k) and computing the required rod radius to
make the system just critical.
The zero fast current criticality equation
was
for
solid
used
the
absorber rod, and the approximation (11.50)
(11.38)
was used for the water-filled rod.
It is of interest to note that the com-
REACTOR ANALYSIS
738
[chap.
11
puted results indicate that the water-filled rod is approximately twice as

effective as the solid rod in this application.12
These results should be
most reliable for the larger rods, since the requirements of the slowingmodel, as represented by the two-group
down diffusion

Water filled
contrc Irod
equations,
are
^- Solid cor itrol

rod using zero
f ast current :ondition
!
Control
Fio.
rod radius 6, cm
Comparison of water-filled and solid-absorber control rods.
11.8
not well satisfied when the characteristic dimensions of the rod are in the
order of one diffusion length.
11.3
Eccentric Control Rod in Two-group Model
The calculation
of the effectiveness of a single eccentric control rod in a bare cylindrical
reactor on the basis of the two-group model requires the general solution
to Eqs. (8.147) which includes the azimuthal dependence.
The formal
solution for the fast and thermal fluxes
and d>2, respectively, we writ*
a.
General Solution
in the form (11.18).

and W by
for the Two -group Equations.
In the present application,
Z(x)mH{g)Q(p,9)
where
p,
z, and
>p
however, we define Z
W{t) = H(z) F(P)V)
denote the usual cylindrical coordinates.
(11.63)
If
these
1 Some additional examples

and experimental data are reported by Murray and
Niestlie, op. ext.
* See also P. F. Gast, "Experimental
Checks of Control Rod Theory," Proc. Inlern.
Conf. Peaceful Uses of Atomic Energy, Geneva, 1955, vol. 5, P/612, pp. 389-392.
SEC. 11.3]
CONTROL-ROD
THEORY
739
expressions are used in Eqs. (8.174), it is easily shown that the function
H{z) is given by Eq. (11.21), for the bare cylinder of height 2h, and the
functions G and F by
op'
(92F
+ JiS
1
p' dtpp op
dF
d2F
p op
p'
dtp'
op-
72G = 0
with 7 and /3 defined as before (11.23).

equations in the form1
G(p,<p)
*(>)
We seek solutions
- (*) S(p)
*W)
(p)
(11.64)
to these
(11.66)
Substitution into the differential equations yields the eigenfunctions

*(<p)
-*(<?)
n = 0, 1, 2,
e*"*
(11.67)
and [cf. Eqs. (11.22)]

(p) -> Rn(P) = anJn(yp) +

S(p) -> Sn(p) 3Tl/n(/3p) +
<&nYn(yp)
(11.68)
3lnXn(/3p)
(11.69)
The general solutions are then

G(P,<p)
f(p,v)
[-^.(7P) + ny(7P)]ein-
(11.70)
l3R"/s") + tt"^(/3p)]ein*
(11.71)
f| a 00
3C
n ao
The substitution of these results, along with (11.63), into (11.18) yields
for the fluxes in the core:
4>x(p,v,z)
= H{z)
|[-^-(7P) + ffinFn(7p)]
+
*(P,V0 = H(Z)
[3K/03p)
2_ aC |ai[n./n(7p)
+ HJCCftO]
}*-
(11.72)*
+ nF(7p)]
n ,
a2[sm/03P)
3iB/C(/JP)]}e<*
(11.73)*
(B, 3TC, and 9l are arbitrary constants to be deter

mined from the boundary conditions at the outer surface of the reactor
The application of these requirements
and at the rod-reactor interface.
will also yield a criticality condition for the reactor with inserted rod, as
in the analyses of the preceding section.
The coefficients (J,
1 See also
Chemistry,"
H. Margcnau and G. M. Murphy, "The Mathematics of Physics and

p. 226, D. Van Nostrand Company, Inc., Princeton, N.J., 1943.
REACTOR ANALYSIS
740
[CHAP.
11
Solid -absorber Rod. The general solutions (11.72)

and (11.73) are used here to derive the criticality condition for the eccen
tric solid absorber control rod. For this analysis we apply the following
set of boundary conditions (refer to Fig. 11.9):
b. Application
to
(1)
$2
(2)
= 0
at di = 6,
at*
= 0
As before, & denotes the effective rod radius [refer

along with (11.19).
to Eqs. (11.25)], and d\ is the radial distance out from the center of the rod.
It
is clear
that the above conditions
are essentially the set (11.33).

The interface conditions
(11.74)
apply strictly only when the flux dis
tributions around the rod are sym
metrical, and this will not be true for
the case of the eccentric rod or a
ring of rods.
However, it has been
shown12 that to good approximation
for the neutron

expressions
fluxes can be divided into two parts,

the
Fig.
11.9
Eccentric control rod.
one containing
two functions which
vary strongly near the rod but are
symmetric about the rod, and the

other being a function which varies slowly near the rod but is not sym
metric about it.
It will be assumed in the analysis which follows that
the radial derivative at the rod of the slowly varying function can be
neglected by comparison with the derivatives of the two strongly varying
functions.
Hence in evaluating the derivatives with respect to d\, we
shall consider only the strongly varying symmetric functions.
Also,
of
when applying (1)
(11.74), we evaluate the strongly varying functions
at d\ = 6* and the slowly varying functions at the center of the rod.
On the basis of the assumptions and postulates set forth above, we
select the following pair of functions to describe the rapid variation in
the flux distribution in the vicinity of the control rod:
SoFoMi) +
JFoKoGSdi)
(11.75)
But these functions are symmetric about the center of the rod; therefore,
they cannot satisfy the boundary condition (11.19) at p = R in the
1 F.
T. Adler, "Efficiency of Control Rods as a Function of Their Position in a

Cylindrical Pile in the One Group Picture," Chalk River Laboratory,
MT-22'2,
Feb. 12, 1947.
!J. Codd and C. A. Rennie, "Two Group Theory of Control Rods in a Thermal
Reactor," Harwell, A.E.R.E., R/R 818, 1952.
SEC. 11.3]
CONTROL-ROD
THEORY
741
Thus we must add a suitable general solution

neighborhood of the rod.
for fa, in the nature of a correction, and determine the arbitrary con
stants in such a way as to satisfy all four conditions given by (11.19)
and (11.74).
2(p,<p,z)
We select, therefore,
= H(z)
|ai80F0(7rfi)
+ ajffoKoGSdi)
^ ao
[aia7n(7p)
a23Hn/n(i8p)]e<,",J
(11.76)
Note that the terms in ($>nY(yp) and yLK(f}p) have been omitted so as
to avoid the singularity at p = 0 [cf. Eq. (11.73)]. For the fast flux we
take the corresponding form
*i(p,>,c) =
H(z)
{SoFo(7<*i)
+ MToGMi)
[n^n(7p)
n = ao

In working with
+ SJIU/.OJp)]**"*}
(11.77)
it will be convenient to
With the aid of the
use series expansions for the singular functions.
addition theorem for Bessel functions1 we can write (refer to Fig. 11.9)
these expressions for the fluxes
F0(7<*i)
Koifidi)
Jm{ya)
"V"
2
Yêi"'
(11-78)
Im(fia) Km(Mc""*
where d\ = a2 4- p2 2ap cos i(j>,<P,z)
H(z)
fSoJm(7) Fm(7p)
+ So/m(0a) Km(fip)
+ dmJm(yp) +
3irm/m(0p)!e<m*
ae
2(p,<P,z)
H(z)
(11.79)
|oi[6o/(to) ^(7p) + aJm(7p)]
a2[ff0/m(/3a) /Cm03p)
3Tlm/m(/3p)]
je*-'
(11.80)
In order to satisfy the boundary conditions at the outer surface of the
reactor (11.19), we must equate to zero the coefficients of eim* for all m,
and <2 evaluated at p = R. The simultaneous solution of the
resulting expressions yields the following set of relations between the
in both
1 See, for example,

G. N. Watson, "Theory of Bessel Functions,"
bridge University Press, New York, 1944.
chap.
XI,
Cam
REACTOR ANALYSIS
742
[CHAP.
11
coefficients:
am =
-SoJ47a)^ll
Jm{yR)
<n-81)
-S0U(fia) Krm{f
(11.82)
3TCm
which
apply for all
it will
be convenient
In applying the interface conditions (11.74),

m.
to denote the series indicated in (11.76) and (11.77)
by the functions
BO
P(p,f) =
[mJm(7P) + mmIm(0p)]ei""'
(11.83)
(11.84)
-
X
Q(p)>)
[a!ttmym(7p)
a29Tlm/(/3p)]e<-"'
m=
P and Q are slowly varying functions in the vicinity of

accord with the premises established above, we shall neg
lect the derivatives dP/ddi and dQ/ddi at di = 6, and evaluate P at
Then, the application of
d\ = 0 (p = a, ip = 0) instead of at di = &.
Note that both

the rod.
In
condition
(1) from
(11.74) yields [using (11.84) to replace the series in
(11.76)]
8=
-9.()^-QM
(11.85)
OC
where
Q(a,0)
[a^J^ya) +
The application of condition
(2) on the fast flux gives
6,=
wherein
we have neglected
(11.86)
as3ltm/m(/3a)]
-5o
(11.87)
the derivative of
P with
respect to di as
planned.
The criticality condition for this system is obtained by eliminating the

constants
dm, and
ffo,
/o,\ 0K1(m
3Tlm
between
(11.81),
(11.82), (11.85), and

and
The
(11.81)
(11.82) in (11.86).
(11.87).
when
substituted
into
resulting expression for Q(a,0)
(11.85) yields one
For a second equation we use (11.87). If
equation relating 80 and 50.
we divide one by the other we obtain
80,
First we
use the relations
Kt((ib,)
IKda)
V
,
F(7&.)
- ^(7a)
Vt^:
/i(|Sa)
,
>
f=
(0#)
-^(7a)
m 1
(11.88)
CONTROL-ROD
SEC. 11.4]
THEORY
743
wherein we have used the relations

J n
I-n
'
l)"Jn
In
Yn
K_n
l)nYn
gg^
situations, all the / terms in the numerator of

small in comparison to K0(fibt), and the series in
the denominator may be neglected in comparison to the term in J\.
Thus a good approximation to the criticality equation for the bare
For most practical
(11.88) are negligibly
cylindrical reactor with eccentric control rod
rtfrju
Note that
is
rFo(7feJ_/^jp(igx(^)l
further reduction
is possible when b > 0; then,
(n.90)*
Eq. (11.53)
may be applied.

11.4
Ring of Control Rods in Two-group Model
In the present section we develop a general

a. Analytical Model.
method for computing the effectiveness of a ring of equally spaced solid
Specific criticality
absorber rods on the basis of the two-group model.
calculations are carried out for the case of zero fast current at the rod
The analytical approach used in these calculations
surface, as in (1 1.74).
follows that of Codd and Rennie.1
The more general situations of two
rings of control rods, and one or two rings with central rod, have been
These calculations were performed for the
analyzed by Garabedian.2
bare cylindrical reactor; however, Garabedian considers also the problem
In all these analyses,
of the single ring of rods in a reflected reactor.
condition that the fast flux remains finite inside the rods
Eqs.
to
[refer
(11.25)].
The problem of determining the criticality condition and the flux
distribution in the two-group model for a ring of gray rods has been
he applies the
considered
The method used
of the
Nordheim-Scalettar approach and may be applied when either the rods
or the reactor, or both, are multiplying media.
b. Application to Solid -absorber Rod.
The derivation of the criti
by Avery.3
is
a generalization
cality condition for a ring of M solid-absorber control rods in a bare

cylindrical reactor follows the procedure used for the eccentric rod. The
As before, we select
configuration to be analyzed is shown in Fig. 11.10.
a pair of functions for each rod which are singular at the center of the rod;
1 Codd
and Rennie, op. cit.

H. L. Garabedian, "Control Rod Theory for a Cylindrical Reactor," AECD-3666,
Aug. 9, 1950.
R. Avery, "Two-group Diffusion Theory for a Ring of Cylindrical
Rods," ANL5729, June, 1957.
1
REACTOR ANALYSIS
744
[CHAP.
11
thus, we take for rod m [cf. (11.75)]
SoFoMm) +
If
we choose
such pairs for the
(11.91)
VoK0(Pdm)
rods, and add also a general solution
Rod 1
11.10
Fig.
Ring of control rods.
Angular coordinates for ring
11.11
of M rods.
to help meet all the boundary conditions, then the solutions for the fluxes
can be constructed in the form [cf. Eqs. (11.76) and (11.77)]
{8
m-
Y0(ydm)
fj
*i(p,*,) = H(z)
Ko(0dm)
1
m
[./(tp)
/C0(y3dM)
o,3RB/.C8p)]e*-'}
(11.93)
[a,aJ(7p)
a
is
it
a2ff
to use
series representation for each of the
Thus, as in (11.78), we write for rod m (refer to
convenient
functions.
(11.92)
m=
+
singular
Y
1
Y0(ydm)
m=
Again,
3TCn/n(fr>)]e<'}
t(.P,<p,z) = H(z)
|a,So
Y(ydm)
Fig. 11.11)
J.(ya) F.(7p)e-
(11.94)
the azimuthal coordinate measured counterclockwise from

0m
the angular
the radial line passing through the center of rod m.
then
coordinate measured from rod
is
>p
If
- jj (m -
1)
1,
9m
where
is

Fig.
(11.95)
CONTROL-ROD
SEC. 11.4]
THEORY
745
is,
Note that for a ring of M rods the angular interval between rods is 2t/M ;
therefore, <pm = 2r(m \)/M.
the interval between rod m and rod 1
If we use (11.95) in (11.94) and sum over all M rods as indicated in
m=
J.(ya)
Y.{yp)e<
Y0(ydm)
(11.92), then
=. oo
_-..<
u/*
(11.96)
m1
_t
1
1_-
j|f
e-'-"IM
if
g-Unfm-UlM
is
Now
an integer
(11.97)
_____ _
is
__
if
g-T.(m-i)/M
AT
not an integer
Thus all terms in Eq. (11.96) vanish except those in the form m Ym
and e"Mf, and these appear with factor M. Equation (11.96) therefore
reduces to
Yo(y<L)=
Jj..(7o) YM.(yp)eiM'*
(11.98)
KmMW'*
[SoJj_.(7cO F.v>(7p)
ffo/_f.03a)
[^n(Yp)
3Il,/.(|8p)]ei-}
a-a W
= H{z)
{M
',
Similar results can be obtained for the functions Ko(fidm) the substitution
of these, along with (11.98), into the expression for the fast flux yields
n o
(11.99)
is
The corresponding form for 2

obtained by inserting the factor eti in
each term involving
regular Bessel function of the first kind (J) and the
factor a2 in each term with modified Bessel function of the first kind (/).
In order to satisfy the boundary conditions at the outer surface of the
reactor, we must equate to zero the coefficients of like functions e*B* (that
If this
all exponentials with n = Ms).
done for both <fi and fa,
is
is,

the following relations are obtained:

GLm.
3Tl_f. =
All other
9TC
-M&,Ju.(ya) YT"jy
(11.100)
-M^lu.ifia)
(11.101)
JM.{yR)
and ftn are zero.
KTM'^?1
REACTOR ANALYSIS
746
[CHAP.
11
In view of the symmetry of the present system, it will suffice to satisfy

interface condition (11.74) at one rod, say at m = 1.
For this purpose we write the flux equations (11.92) and (11.93) in a
form that separates out the contributions due to rod 1 ; thus,
the rod-reactor
= ff(z)[SoF(7di) + SoKotfdy) + V(P,<p)]

2{p,t,z) = H(z)[a1S0F0(7di) + atfoKotfdO + tr(p,*>)]
Mp,<p,z)
(11.102)
(11.103)
V and U denote all the remaining terms in (11.92)
Here the functions
and (11.93), respectively; these terms are all slowly varying functions in
The appli
the vicinity of rod 1 [cf. P and Q in (11.83) and (11.84)].
cation of condition (1) from (11.74) to (11.103) yields an equation analo
gous to (11.85); this is
\ai/
where
f^>--^
Y0(yb,)
U(a,0)
O!80
Y0(ydlm) +
a2ff
K0(pdlm)
m=2
m=2
+
(11.104)
aiF0(-yo*)
[aiJ(7o) +
a23Tl/n(/Sa)]
(11.105)
n =
and
dim
denotes the distance between the centers of rod
Fig.
In the general
11.12
Distance between
and rod m.
rod centers.
case, the distance between the centers of rod m' and rod m
is given by (refer to
Fig.
dm-m
11.12)
= 2a sin
(m'
m)
(11.106)
The application of condition (2) from (11.74) to (11.102) yields a

relation identical to (11.87), if we neglect the derivative of V with
This equation, together with (11.104), provides two rela
respect to d\.
SEC. 11.4]
CONTROL-ROD
THEORY
747
tions involving
The coefficients d and 3TC in (11.105) may

S0 and ff0.
with the aid of (11.100) and (11.101). The simultaneous
solution of (11.104) and (11.87) then yields the criticality condition for
This is found to be
the system.
be eliminated
m==2
Y(yb.) +
"
Ya(ydlm)
- MJKya)
m=2
-1
- 2M V J2v.(7a) jf^fj
=1
(11.107)

where we have again used the relations (11.89).

As in the case of the eccentric rod [cf. Eq. (11.88)], a good approxima
tion to (11.107) can be obtained by neglecting all the terms in the numer
ator which involve the modified Bessel functions of the first kind, and the
sum over s indicated in the denominator; then, the criticality condition
for the ring of M solid-absorber rods in a bare cylindrical reactor may be
written
R)
m=2
-fe)Sw[K,(*-)+SK'(H)
m =2
This relation will
be adequate for most computations.
<1I108)*
CHAPTER
HYDROGENOUS
12.1
12
SYSTEMS
The Special Treatment Required for Hydrogen
a. General Considerations.
Most of the discussion of reactor calcu
lations in this work has been based on the age-diffusion model of neutron
behavior.
This model is very convenient to use and has yielded valuable
results in innumerable cases.
However, there is an important class of
to
which
this
model
is
If a system includes
poorly applicable.
systems
a large proportion of hydrogen, the approximations involved in the agediffusion method are really not valid. As indicated in the discussion of

Sec.
6.1a, the age-diffusion
method
would
not be expected to hold in
situations where the neutrons make appreciable numbers of collisions

with light nuclei; nevertheless, in a practical sense, this model yields
The case of hydro
acceptable results for any nonhydrogenous material.
is
its
in one elasticsince
the
neutron
can
lose
all
unique
energy
gen
scattering
collision,
and the energy-angle
relationship
(7.134) takes a
very special form.

The practical advantages of the use of water as a moderator and as a
coolant make it necessary to consider methods of improving the agediffusion approximation.
The more accurate treatment required for hydrogenous systems can
be separated into two classifications:
1. A complete set of equations covering the slowing down and diffusion
process, replacing the age-diffusion set, may be found in two ways:
(a) An intuitive approach using physical understanding of the neutronproton elastic scattering process may form the basis for a set of equations.
(b) The original Boltzmann equation may be examined in an attempt to
discover a practical approximation which is an improvement over the
age-diffusion approximation. It is this method which is emphasized in
the present chapter.
2. The essentials of the age-diffusion
method may be retained, but

various parameters of the calculation may be chosen to give better results
for hydrogenous reactors: (a) The group diffusion models discussed in
Chap. 8 may be applied with particular choices of the group constants
These constants are not neces
designed to yield improved accuracy.
sarily calculated in any prescribed manner but are obtained from meas
748
SEC. 12.1]
HYDROGENOUS
SYSTEMS
749
or by trial and error in the calculations,
(b) An empirical
kernel may be used with the usual one-velocity diffusion
equations for the thermal neutrons.
In general, a particular method of the second class will give useful
results only for a limited range of types of reactors; thus, in undertaking
a new design analysis, considerable effort must be applied to find the
appropriate constants.
b. Utility of the Measured Neutron Age in Water.
One very impor
urements

slowing-down
tant quantity involved in all the methods indicated above is the neutron
A simple example
age to thermal (or to indium resonance) in water.
of the one approach (6) would be the application of the criticality condition
(6.80) to a bare water-moderated system, wherein for the fast leakage
losses one uses exp ( B2rtb) with the age to thermal as obtained from
More fundamentally, an experimental determination of the
experiment.
actually
in
water
gives the spatial distribution of neutrons of various
age
Thus an obvious
energies as they slow down from a fission source.
requirement for a proposed calculational method is that when it is applied
to the geometry of the experiment it should yield the observed distribu
In practice this is most conveniently approached by requiring
tion.
that the calculational method give accurately a finite number of the lower
If, in a spherical geometry,
moments of the slowing-down distribution.
the spatial distribution of neutrons reaching thermal energy at a distance
r from a fission source is Kth(r), then the nth moment r" is defined as
4tt
rnKtb(r)r2 dr
(12.1)*
The most important of these is the second, r"-, and in many cases, to
require that the second moment be adequately represented is regarded
In the age-diffusion approximation, the second moment is
as sufficient.
proportional to the
age, and
it was shown previously that

rHu)
6t(w)
In analogy to this result, it is convenient to define

in approximations
lethargy u is given by
use
(6.52)
a generalized age
other than the age diffusion;

Age = lsW(u)
for
thus, the age to
(12.2)*
Then, the simplest and most direct criterion for evaluating a proposed
calculational
method for water-moderated
systems is that it yield
correctly the measured age to thermal in water (i.e., the correct second
moment).
In the sections which follow,
we examine some of the improvements

The discussion
over the age-diffusion approximation presently in use.
from
which
calculation
of
the
the
are developed
equations
of the
age
REACTOR ANALYSIS
750
[CHAP.
12
The primary purpose

is deferred until the last section of this chapter.
of the present chapter is to demonstrate the basic ideas on which the
are based and to indicate the general direction for the
In no case is a complete and
development of a calculational method.
(including various regions, boundary
detailed method of computation
conditions, neutron balance, etc.) given, such as was done previously
improvements
for the age-diffusion
Once the basic ideas are understood, the

detailed system applied to the age-diffusion case may be easily adapted
to the cases presented here.

12.2
case.
Elementary Development of the Goertzel-Selengut Equations
Some improvement over the use of the ordinary age-diffusion equations

for the hydrogenous case may be obtained by starting with the agediffusion model but giving neutron scattering collisions with hydrogen
special attention. It has been shown by Goertzel and Selengut1 that
the equations resulting from this approach may be derived directly in a
In developing these
certain approximation to the Boltzmann equation.
relations in the present formulation, we will use a somewhat intuitive
approach in order to display more clearly the physical basis of the model.
Consider the balance of neutrons in volume dr about r and with
At steady state the neutrons lost from the
lethargies in du about u.
differential volume dr du will be just balanced by the gains, as indicated
In the Goertzel-Selengut approach the leakage term is
in Eq. (6.2).
given by the usual age-diffusion form (6.3), and the absorption losses are
given by the form 2(m) 4>(r,u) dr du. The volume sources will be
terms, however, are
represented by S(t,u) dr du. The slowing-down
given special attention.
q(l,u) =
These we divide into two parts:
number of neutrons slowing down past lethargy w

per unit time per unit volume around r which made
their
last scattering
collision with
a nonhydrogen
nucleus
of neutrons slowing down into lethargy
interval du about u per unit time per unit volume
r;(r,w) du =
about r which made their last scattering collision
number
(12.3)
with hydrogen
The slowing-down
density given here is related to the flux
usual way, viz.,

g(r,)
where
2j0)
Zi>(M)*(r,W)
The new quantity
in the
(12.4)
individual ({2,), for all materials

7j(r,) may be computed from the
is taken as the sum of the
except hydrogen.
1 G.
tf>(r,u)
Goertzel and D. Selengut (unpublished).
sec. 12.2]
HYDROGENOUS
integral relation (4.97)
SYSTEMS
751
thus,
ij(r,M) du =
f* 2(u')
4>{T,u')e'-
du' du
(12.5)
where (H) refers to the hydrogen nuclei, and it is implied that there
are no sources above u = 0.
In order to determine the net gain in
neutron density in dr du from scattering collisions in and out of this
differential
2W)(u)
volume, we require also
(j,u)
number of neutrons being scattered out

dr du = of dr about r and du about u because of
(12.6)
collisions with hydrogen nuclei

and g(r, u + du).
[q(r,u) +
7(r,u)
Then, the net gain may
be
du] dr
2(u) (r,u)
- [q(t,u + du)
written
du] dr
(12.7)
Note that in (12.7) each scattering collision with hydrogen is counted

with Eq. (12.5) in which each neutron lost by
scattering from a given lethargy interval is counted as a gain to some
interval at higher lethargy.
If we use expression (12.7) for the slowing-down term, and the defini
tions for the other loss and gain terms given above, then in the present
case Eq. (6.2) may be written

as a loss to be consistent
- D(u)
V2<Kr,w) du
+ 2(u)
2(u)
4>(r,u)
{t,u) du
du + q(r,u + du)
S(i,u) du + q(r,u) + ij(r,w) du
(12.8)
But, from (12.5)
(r,u)
-v(t,u) +
-D(u)
V24>(t,u)
20(m) (r,w)
2(u)
*(r,u)
- ^-v(r,u)
- |-(r,w)
du
du
with q{i,u) related to (i,u) by (12.4).

In applying these results to a reactor calculation
(12.9)*
+ S(r,u)
(12.10)*
the pair (12.9) and

is
taken for each region containing hydrogen along with, as
(12.10)
needed, ordinary age-diffusion equations for nonhydrogenous regions.
To this system is applied the usual spatial boundary conditions, thermal-
group relations, and coupling equations.

results using this set is given in Sec. 12.4.
A discussion of computational
The derivation of the Goertzel-Selengut equations given above brings

out clearly the fact that, while the distribution of neutrons in energy
is properly handled through exact treatment of the slowing-down in
hydrogen, the treatment of the space-energy correlation during the slow
ing-down process is not improved over that of the age-diffusion equations.
752
RKACTOR
ANALYSIS
A possibility exists for such improvement

diffusion coefficient, but as will be shown
derivation of these equations from the
that D(u) is to be calculated as in Eq.
on the same basis as all other substances
[CHAP.
in the calculation
12
of the
in the next section, a consistent

Boltzmann equation indicates
(7.161), with hydrogen treated
present.
Solution of Boltzmann Equation for Hydrogenous Systems
12.3
of Second Moment for Mixtures of Nuclear Species.

The present analysis is a more general development of the ideas previ
For this purpose we require a statement
ously considered in Sec. 7.6a.
a. Computation
of the Boltzmann equation
for a system containing
a mixture
of
The direct extension of the steady-state form of (7.110)
nuclear species.
is
a-V(T,v,a)
+ St(y)
2^
S(r,v,0)
X /' W) <Kr,y''') ^"'^
dv' dQ'
= 1
The conditional frequency qm is defined according to (7.114), and the

superscript (i) means that the scattering collision occurred with a
nucleus of type i. In the present application we consider the case of
one-dimensional
symmetry so that spatial variation occurs only in the
x direction, and the only angular variable of interest is p, the cosine
of the angle between the x axis and the direction of motion of the neutron.
The appropriate one-dimensional form of (12.11) is [cf. Eq. (7.336)]
Q
/*
q- (x,u,n) + 2,(u) (x,u,n)
= -S(z,u,ai)
X L dU'
=
Si'>(M,) *(x-u'>'1')*W(u''i;M'>Q')
dQ'
(l2-12)
2^
fa-
where we have transformed from the speed

The
are defined according to Eq. (7.127).
to the lethargy
variable.
We obtain the solution to Eq. (12.12) by expanding the flux and the
For the
frequency functions 8<0 in infinite series of Legendre functions.
frequency function we use the general form (7.129), and for the flux
[cf. Eq. (7.74)],
CO
KW)
(^P)
Mx,u)PM
(12.13)
similar
series
is
If
defined also for the source function S(x,u,p).
with
the
addition
used
in
theorem
(12.12),
along
Eq.
are
these expansions

(r,v,a)
sec. 12.3]
SYSTEMS
HYDROGENOUS
753
(7.21), and the orthogonality properties for the Legendre functions are
applied, the result is an infinite set of coupled equations in the coefficients
-2,(m)
>
Jf
2(u') (x,w')
*"+l(x'M) +
Ix
where
and
+ Sn(x,u) +
<fn(x,u)
a,
(> tt)
8'(m;m',Q') du'
*-l(:r'M)
(12-14)
(12.15)
|^
Ai is the mass number of the tth nuclear
species.
Note that the
range of integration for the lethargy variable extends over the usual
collision interval appropriate to the particular species.
It is of interest
to mention that the set (12.14) may also be derived directly from the
general form (7.125) by applying the procedure used in Sec. 7.2f. The
is
obtained by the application of (7.77).

(12.14)
To proceed further we require
specification of the coefficients
For this purpose we introduce the assumption that the scattering col
lisions are isotropic in the center-of-mass system of coordinates.
In that
case we obtain
(^',0) = F"<(r')e""U
a,
(12.16)
[cf.
Eq. (7.134)]
^
f,(M,')
as demonstrated
i[(Ai +
l)e -
(At
in the derivation of Sec. 7.3e.
- l)e-*]
If
with
(12.17)
we use the expression
(12.16) in (12.14) and approximate the infinite set by the first three
harmonics, we obtain, for
system with isotropic sources,
i(x,u) + Z,(u) 0(x,u)
= S0(x,u)
i=
dx
+3dx
Z'fl(u') *(*,')
-f^ du'
(12.18)
-f1^
(12.19)
if
+ 2<M *i(x")
AT

is
\p,
computation requires first that Eq. (7.125) be integrated term by term

over the range of the azimuthal variable
in order to obtain an expres
sion suitable to the one-dimensional
The next step
to
problem.
multiply through by some Pm(n) and integrate over p. These operations
yield an equation in the coefficients
The final reduction to Eq.
2p(vf) *,(*,') Uu,W)
du'
REACTOR ANALYSIS
754
2d
5 dx
3d
*1(X'M)
2(') *,(*,')
./u + lnoci
Z_/
=
It
these
[3f?-(",M,)
2(1
12
+ 2|(u)
*3(X'M)
A'
[CHAP.
11
e-'-
du'
*i)
(12.20)
for the demonstration of some of the properties

equations to work with their Fourier transforms; we introduce
is convenient
*(,) = t^Tj
1
So(s,w)
f"
/_
/"
*(x,u)e*
da;
(S0(a;,)e<
dx
of
(12.21)
The transformation of Eqs. (12.18) through (12.20) yields

isî(s,u) + 2,(w) *0(s,m)
= o(s,u)
A1
+y.r
/
-y
*o(s,m) +
2<(')
*o(,')
du'
(12.22)
du'
(12.23)
e"'-
<*w'
(12.24)
*2(s,u) + 2,(u) *i(s,u)
2>(m')
,(,') r.(,')
1= 1
^ *i(s,u)
^ (s,u)
2,(m) 2(s,u)
.V
y
Z_/
i=
./u+lna,
2i<1(w')
*(,') [3r](/?'"')
A1
x
ar<)
11
If
the medium under consideration and the source distribution are

in x, then the total flux <p0(x,u) is symmetric in x, and conse
Thus, we may represent
quently, 0(s,u) and 0 are symmetric in s.
symmetric
*o and
go
in the following power series in s:

s4
s2
*o(,)
= *oo() + *o2(w)
2j + *o() Ti
So(,)
' - '
(12-25)
Soo(m)
+ Soj(w)
2]
S<(M)
4j +
'
Then, the form of Eqs. (12.22) through (12.24) shows that it is sufficient
to represent the higher-order coefficients in power series beginning with
SEC. 12.3]
HYDROGENOUS
the power corresponding
755
SYSTEMS
to the order of the coefficient; thus,
*,(,u) = *u(m)s + (u)

+
^
*,(,) =
*(u) ^ + ,() +
*,(,) =
*()
+ (u)
|^
|* +
(12.26)
(12.27)
(12.28)
The lethargy-dependent coefficients defined by these equa

determined by substituting these series into Eqs. (12.22)
The result
through (12.24) and equating coefficients of like powers of s.
is a set of relations between the coefficients; for example,
and so on.
tions
are
AT
S,(u) *0o(m) = Soo() +
2'(w') *oo(')
du'
(12.29)
*(')
du'
(12.30)
i=l
| So2(m)
iV
2>(u')
i=\
2'(m')*u(m') f.(,') y6"-^ d'
2,(u) *u(u)
*oo(")
(12.31)
If
the cross sections and the source distribution are known, these equa
thus, (12.29) yields *0o, which
tions may be solved simultaneously;
when applied to (12.31) yields
This in turn gives *<>2 from (12.30),
and so on.
particular coefficients follows from the discussion
In
the generalized definition of the age (12.2).
to
distance
the
mean
squared
age, we require
given by Eq. (7.343), which in t'ie
expression
is
The interest in these

given in Sec. 12.1b on
order to compute the
suitable
lethargy u.

/""
I 2,(u) *02(m)
tn(w) +
present case reduces to the form
S-=
(12.32)
Fourier transforms
beginning with
discussion
to
previously
(12.21).
[refer
Eq. (7.364) and extending through Eq. (7.374)] that these are related
This relation may also
It
was shown
be expressed in terms of the
REACTOR ANALYSIS
756
[CHAP.
to the moments of the function, and in the present case we obtain
Eq. (7.374)]
PTC _
r(3Va')*.(,tt)1
*o(s,u)
[
J._o
..jaw
12
[cf.
(1M8)
*oo()
the last equality following from (12.25).

We find then that Eqs. (12.29) through
(12.31) are sufficient for com

Equations (12.22) through
puting the correct second moment of the flux.
(12.24) show that, by working with the transformed equations, only
the first two terms of the series (12.13) for the flux are required to obtain
the second moment and, therefore, the generalized age accurately.
Pi
The accurate
set of
approximating equations to the Boltzmann

(12.24) by neglecting *2 and
equation is obtained from (12.22) through

higher terms; thus,
isVi(s,u) + 2,(u) *0(s,w)

+
- 0(s,u)
V
-l
Z'(m') (,')
du'
(12.34)
du'
(12.35)
+ 2,(u) *i(s,u)
u+
lna.
2(u') *!(,') f,(M,')
b. Reduction to the Age-diffusion Equations.
As in
a,-
Sec. 7.3g, the basis
2(m)
where 2,(zt)
is
These approximations
isî{s,u)
*o(s,u)]
(12.36)
AT
(12.37)
for
=* 2<0(m) i(s,m)
[2>(u)
1,
*i(s,u')
+ ('
2,
2(u')
=* 2i(w) *o(,)
*0(,')
2>(tt')
reduce Eqs. (12.34) and (12.35) to
*0(s,m) = o(s,m) + 2,(u) *0(s,?i)
- A [J2.(u)
*,(,) + 2,(u) *,(,)
mo2.(m)
*.(8,u)]
*,(s,u)
(12.38)
(12.39)
defined according to Eq. (4.121), and

N
7o2.(u) ^
^2i()
that
u)
of the reduction to the age-diffusion equations is the assumption that the

collision densities are slowly varying with lethargy over the range of
integration required on the right-hand sides of Eqs. (12.34) and (12.35);
is,

IS
= So(s,w)
N
(12.40)
sec. 12.3]
HYDROGENOUS
SYSTEMS
757
If we define
{, and Mo" are related to A( in the usual way.
be
Eq.
to
then
reduced
to the form
may
according
(7.161),
(12.38)
D(u)
The quantities
sW{u) ,(,) + Z.(u) *(,)
= So(,)
- ^ IfS.(u)
*.(,)]
(12.41)
The Fourier inversion of Eq. (12.41) yields

by the elimination of
the familiar age-diffusion equation (7.164).
The treatment
c. Reduction to the Goertzel-Selengut Equations.
above demonstrates that the age-diffusion equations rest on two principal
approximations:
(1) The angular distribution of the flux1 can be repre
sented by the first two harmonics [refer to Eqs. (12.38) and (12.39)].
(2) The collision densities are sufficiently slowly varying over the lethargy
range In 1/a, for each species so that they may be represented by one or
These
two terms in a Taylor expansion [as in (12.36) and (12.37)].
in
Sees.
7.3f
and
The
7.3g.
conditions were established previously
gen
range (u + In an) > (u) is the entire lethargy scale2 from u 0 to u;

How
thus, approximation (2) is inapplicable in the case of hydrogen.
ever, since the approximation is made separately for each nuclear species
over different lethargy ranges, there is no inconsistency involved in using
approximation (2) for some substances and dropping it for others. In
particular,
to
obtain
the
Goertzel-Selengut
equations:
for all substances
(1)
Equation
*(,)
Z(m)
+ 2i'(w) *o(s,m)
J*
*(,)]
[So2,>(m)
= 8o(,)
ti(,tt) + 2,(u) *,(,)
i.
is
1,
1,
is,
except hydrogen; that

(12.36) holds
= N, which we use to denote

=
but not for
.
.
N
for all
retained for all
With these con
hydrogen.
(2) Equation (12.37)
ditions, Eqs. (12.34) and (12.35) become
(12.36) is applied
V\u') o(,u')e"'=
mo2.(w)
*i(s,u)
du'
(12.42)
(12.43)
H. Hurwitz and P. Zweifel, J. Appl. Phys., 26, 923 (1955), have considered the
expansion of the scattering function, rather than the flux, in spherical harmonics; by
dropping all terms beyond the nth, they obtain the so-called Bn approximation, which
has been shown to be more accurate than the corresponding P approximation devel
oped here.
it
is
is
assumed here, as usual,

Actually, the lethargy scale extends from to u, but
taken to correspond to a sufficiently high energy so that
that the zero of lethargy
there are
negligible number of neutrons of higher energy in the system.
1

eral method for improving the age-diffusion equations for application to

hydrogenous systems is achieved by discarding some or all of approxi
mation (2). In the particular case of scatterings from hydrogen, the
ANALYSIS
REACTOR
[CHAP.
Here, the index 0 refers to all substances other than hydrogen, that
12
is,
758
N-l
2(u)
N-l
fcZ* ()
(12.45)
fo2i()
(12.44)
i=i
- ir [o2(w)*o(s,w)]
= So(s,m)
s2D(u) *o(,) + Zi(tt) *o(,)
2(w)
2(m')
If
Eqs. (12.42) and (12.43), we obtain for *0
from
we eliminate
[using (7.161) for D(u)]
?0(s,u)
*0(,u')e-'-" *'
(12.46)
The inversion of this equation yields:

2
d2
-D(u)
= S0(x,u)
+ 2() *0(x,)
<^0(x,m)
- 2(u) <,(*,)
(*,) + (*,)
(12.47)
ij(i,)
JoU
g(x,u) = |02<0)(u)
(x,u)
(12.48)
>o(x,M')eu'-"
2Jff>(u')
dw'
(12.49)
is
Thus we obtain the one-dimensional

and the application of Eq. (12.9).
form of the Goertzel-Selengut
equations
(12.9) and (12.10) given
previously.
d. Reduction to the Consistent Pi Approximation.
Another improve
ment on the age-diffusion equations for use in the hydrogenous media
which has been sufficiently widely used to merit discussion here
the
so-called "consistent Pi approximation."
improvement over the Goertzel-Selengut
This model also represents an

equations.
In the consistent
Pi
is
approximation, neither of the expansions (12.36) or (12.37)

to hydrogen, though both are used for all other nuclear species.
tions (12.34) and (12.35) become, for this case,
iao(8,i0 +
[o210)(m)
2(m)
*o(,m)]
applied
Equa
+ 2i'(M) *(S,U)
Jo
2<*'(w')
*,(8ft*')-'- du'
(12.50)*
du'
(12.51)*
*i(s,m) = M?Zi0)(u) *i(s,u)
2<H(w')
*i(8,u')e*("'
>
- ir
du
= So(s,u)
f"
M*i(*,u) + 2,() *o(,)
^""(w)
where
N-l

with the definitions
f?Z()
(12.52)
HYDROGENOUS
sec. 12.4]
SYSTEMS
759
Some
These equations are, in general, difficult to handle analytically.
for
the
consistent
set
to
the
has
been
Pi
given
treatment
analytical
Biedenharn.2
special case of pure hydrogen by Simon1 and by Welton and
In the present discussion we shall content ourselves with a comparison
of the various models on the basis of numerical calculations.
12.4
Numerical Results
The primary pur

of Methods by Age Calculations.
of the various
the
compare
accuracy
discussion
is
to
pose of the present
models for hydrogenous systems on the basis of their prediction of the
Thus the information given here will supplement
neutron age in water.
the detailed discussion of age computations given in Sec. 7.6 which
a. Comparison

includes a development for water-metal systems based on the FermiMarshak method. The comparison of the methods is drawn on the basis
Besides the relative ease of
of the computed age-to-indium resonance.
measurement (as contrasted with the age to thermal) it has the advantage
of easier interpretation because the thermal motion and molecular effects
are not generally important above the indium resonance (1.44 ev).
According to the general definition of the age, the quantity of interest
is the second moment of the neutron flux at the energy of the indium
resonance.3
It is also possible to calculate the second moment of the
This quantity will be slightly smaller than the

slowing-down density.
second moment of the flux because of the "last flight" of the neutrons
It is clear also
after reaching the specified energy (refer to Sec. 7.6c).
that the second moment of the flux corresponds more closely to the
physically measured quantity. Some calculated results are given in
Table 12.1.
An acceptable experimental value at the present time is (see Table 6.1)
TIn
= 26.7 cm2
The agreement between the more accurately calculated values, as repre

sented by the Monte Carlo and consistent Pi approximations in Table
12.1, and the experimental
result is seen to be quite good.

The improvements discussed in this chapter
b. Reactor Calculations.
may be incorporated into the multigroup method of calculating reactors
which has been developed for use on fast computing machines (refer to
Sec. 8.8).
Examples
of such programs are discussed by Alexander
and
'A. Simon, "Neutron Slowing Down by Hydrogen in the Consistent Pi Approxi

mation," Oak Ridge National Laboratory, ORNL-2098, July 19, 1956.
L. C. Biedenharn and T. A. Welton, "Some Remarks on the Slowing Down of
Neutrons in Hydrogen," Oak Ridge National Laboratory, ORNL-2107, Aug. 21, 1956.
' That is, t,v, in the notation of Hurwitz and Zweifel, op. cit.
REACTOR ANALYSIS
760
[CHAP.
12
The "Eyewash" code described by

Given1 and by Kinney et al.*
Alexander and Given was developed for the UNIVAC and incorporates
the Goertzel-Selengut method in a 30 lethargy-group program designed
A similar 30-group code called
for a multiregion spherical reactor.
"Medusa" is also available for the UNIVAC, and this is based entirely
When applied to nonhydrogenous

on the age-diffusion model.
these two methods give essentially identical results.'
Table
12.1
op Models on Basis of Aoe of Fission Neutrons

in Water at Standard Conditions
Comparison
Model for
calculation
Age (cm1) to
indium resonance
Age-diffusion
17.5
Monte Carlo
25.6
Consistent
reactors
Pi
25.3
Reference
Comments
Oxygen mass
assumed infinite
Second moment of
slowing-down
density
R. R. Coveyou,
ORNL-
2081, p. 144.
H. Hurwitz and P.
Appl. Phys.,
Zweifel,
J.

26, 923 (1955)
Goertzel-Selengut
. . .
30.9
H. Hurwitz and P.
Zweifel,
Appl. Phys.,
J.
26, 923 (1955)
As an example of the results which may be obtained by use of the

different approximations consider the following reactor system:
Core: Sphere of radius 12.06 cm containing
5.94 kg of H20
8.88 kg of U02F2 (containing 6.84 kg of 93 per cent enriched
uranium)
Reflector: Spherical
shell of pure water, 10.56 cm thick
For the analysis of this system the lethargy scale was divide into 29 fast
The spatial inte
groups plus a thermal group corresponding to 183F.
which
of
0.754
cm
yielded 16 space
intervals
gration was carried out using
The results of this compu
points in the core and 14 in the reflector.
tation are given in Table 12.2.
J. H. Alexander and N. D. Given, "A Machine Multigroup Calculation: The Eye
Wash Program for UNIVAC," Oak Ridge National Laboratory, ORNL-1925, Sept.
1
1955.
15,
" Multigroup
W. E. Kinney, R. R. Coveyou, J. G. Sullivan, and B. J. Garrick,
Calculations for Multiregion Reactor," Applied Nuclear Physics Division Annual
Report for Period Ending Sept. 10, 1956, ORNL-2081, pp. 133-139.
* Actually there are some slight differences because of different assumptions in the
averaging process for obtaining group constants from cross-section data.
1
sec. 12.4]
HYDROGENOUS
This somewhat extreme

between
assembly
case
SYSTEMS
761
was chosen to exhibit the difference
and the Goertzel-Selengut

models.
The
age-diffusion
is very small, and therefore the neutron leakage plays an
the
important role in determining
the neutron economy.

The effect of the
different approximations in the treatment of the energy distribution of
neutrons after scattering collisions with hydrogen will be magnified
because of the resultant effect on the neutron leakage in the fast range.
An examination
diffusion
Table
of the results given in Table 12.2 shows that the agegives a higher estimate of the multiplication
approximation
12.2
Multigboup
Model
Calculations fob
Wateb-modebated
Fraction of Fraction of
Total
neutrons
absorp
neutrons
absorbed
tions in
escaping
in reactor
core
Reactob
Fraction of Fraction of
neutrons
fissions
due to
escaping
as fast
thermal
neutrons
neutrons

Age-diffusion
(Medusa)
Goertzel-Selengut
(Eye Wash) ....
1.17
0.789
0.763
0.211
0.107
0.375
1.00
0.663
0.770
0.337
0.239
0.430
Evidently this is due to the

constant than does the Goertzel-Selengut.
greater fast leakage in the Goertzel-Selengut case, being over twice as
This is entirely consistent
large as that predicted by the age model.
with the larger computed
(refer to Table 12.1).
age
for water
as obtained from the
G-S model
In contrast to this discrepancy in the multiplication constant, the
This
spatial distribution of the group fluxes does not differ appreciably.
point is illustrated in Fig. 12.1, which shows that the computed power
distributions for the two methods are nearly identical.
The multigroup calculational method based on the Goertzel-Selengut
model has also been tested in the calculation of critical assemblies.
E. G. Silver1 has compared the computed critical masses from the Eye
Wash code with the experimentally determined values for the case of
enriched uranium assemblies in water.2 Unfortunately the evaluation of
the analytical model was not clear cut since it was necessary to apply a
from the actual completely reflected parallelepiped geome
transformation
try to a three-region spherical model. Some aspects of this problem were

discussed previously in Sec. 8.6.
These considerations are discussed in
1 E. G. Silver, "Comparison
of Calculated Critical Mass for BSR with Experi
ment," Applied Nuclear Physics Division Annual Report for Period Ending Sept. 10,
1956, Oak Ridge National Laboratory, ORNL-2081, pp. 47-58.
The ORNL Bulk Shielding Reactor.
REACTOR ANALYSIS
762
12
[CHAP.
2.4
2.0
,-
1.6
a
s
0.8
0.4
10
12
14
16
Radial coordinate (units of 0.754 cm)
Fig.

12.1
Comparison of the power distributions
Goertzel-Selengut models.
obtained from age-diffusion
and
A partial set of the results obtained from the compari

detail by Silver.
son are listed in Table 12.3.
As may be seen, the comparison is entirely
Since the results of Table 12.2 indicate that the agesatisfactory.
diffusion model would have given considerably different results, it may be
concluded that the G-S model represents a worthwhile improvement.
Table
12.3
Comparison
by
with Experiment of Critical

Goertzel-Selengut Method
Masses
Critical mass of Um, kg

Core dimensions, cm
61.3
45
61.3
38.5
31
40.5
Computed
Experimental
3.242
3.246
3.26
3.27
Computed
CHAPTER
13
PERTURBATION THEORY
13.1
The
The Scope and Methods of Perturbation Theory

described
in this chapter
designed to meet the

general class of problems involving small changes which affect the nuclear
features of a reactor system and which require, in general, a recalculation
The methods developed for treat
of the principal reactor parameters.
methods1
are

ing situations of this sort utilize the fact that the changes are sufficiently
small that the characteristics of the perturbed system may be constructed
This approach is widely used in the
from the analysis of the original.
problems of physics and engineering science because it frequently leads
to substantial simplifications in the calculational procedure.
As an illustration of such problems, consider the situation wherein a
at some point within a
critical assembly.
Since the precise nuclear balance of the system has
been upset, it can no longer sustain itself in steady-state operation, and
it is to be expected that the over-all flux level in the reactor would fall in
small lump of absorbing
material
is introduced
time, eventually to zero. In order to restore the steady-state condition,

it would be necessary to add a certain amount of positive reactivity to
If it were
the system (presumably by means of the control system).
possible to calculate the required additional reactivity, the time-depend
ent behavior (at least after long time) of the uncompensated case would
be predicted as an exponential decay of the flux level with a time con
stant given by the ratio of the neutron lifetime in the system to the
required extra multiplication [see, for example, Eq. (9.23)].
(It would
be supposed that the small amount of absorber added would have a negli
Thus the central problem is the

gible effect on the neutron lifetime.)
calculation of the reactivity decrease resulting from the introduction of
It may be seen at once that the result is dependent upon

Clearly, if the
the position in the system at which the absorber is placed.
lump were placed near the outer boundary of the assembly it would absorb
the absorber.
much fewer neutrons than if
it were placed near the center.
1 The
The effect
presentation in this chapter represents a departure from the remainder

of the book in that the mathematical background required is considerably higher.
No attempt will be made to supply the necessary basic information.
The material
presented should be of value as a reference, in any event.
763
REACTOR ANALYSIS
764
[CHAP.
13
of the lump on the reactivity of the system is therefore much smaller

Further, the energy dependence
when positioned near the outer edge.
of the absorption cross section and the neutron-energy spectrum at the
location of the lump also have important influences on its effectiveness.

For example, in a highly thermal reactor it would be expected that a
material with a high thermal absorption cross section would produce a
greater reactivity change than would a material with high absorption
resonances in the epithermal range and low absorption cross section in
the thermal.

The calculational method for approaching such problems as mentioned

here involves the solution of the neutron-balance equations for the sys
By the
tem, with small changes allowed in the original parameters.
application of the well-known mathematical techniques of perturbation
theory, one can obtain the corresponding variations in the multiplication
This general procedure is applied
constant and in the neutron flux.
Before undertaking detailed calculations, it
throughout this chapter.
will be instructive first to examine the over-all problem from a general
viewpoint.
a. General Methods.
In this section we present the formalism of the
perturbation method, leaving the general discussion of the physical inter
pretation to the next section. The presentation given here is in no sense
derivation.
It is intended merely as an outline
can
be
of a procedure which
applied rigorously to many particular cases.
Suppose that the reactor equations for the critical assembly can be
a rigorous mathematical
written in the general form

= 0
O*0(r,w)
(13.1)
Here <o(r,M) denotes the neutron flux as a function of space and lethargy,
and O symbolizes an operator chosen appropriately for the system at
For example, in the one-velocity model the operator O would have
hand.
the form
O =
-V
DV +
(13.2)
vXf
However, much more complicated operators are also encompassed in the

general treatment; e.g., in a later section the operator O will be taken as
a matrix.
It will be convenient for the analysis which follows to define
operators Oi and 02 such that
0=0^-
vOi
(13.3)
Here the intent is to bring the quantity v explicitly into the calculation.
the one-velocity case, for example, we have from (13.2)
In
O! =
-V
DV +
Oi
S0
The average number of neutrons per fission
2/
(13.4)
is very useful for the present
PERTURBATION
SEC. 13.1]
THEORY
765
type of problems, since we may imagine that nonuniform changes in the

neutron parameters of the reactor system are compensated by a uniform
Then, as discussed
change in v, so that the system remains critical.
previously [cf. Eqs. (4.167) and (5.178)], the change in the effective
multiplication constant brought about by the nonuniform changes is
given by
T
If
" "
<13 5>
then, in the reactor system, changes in the

operator O are implied; therefore, the flux in the perturbed system will
By allowing hypothetical changes in
also be different from the original.
we allow
alterations,
the perturbed system may also be critical, so that (denoting quantities

associated with the perturbed system by a prime)
v,
O''0(T,u)
(01
v'
Oi)
4>'0(r,u)
= 0
(13.6)
The form of this equation implies

a recalculation of the entire reactor

of
the
value of v' for criticality being a pri
problem, the determination
mary objective.
The complex reanalysis required here is not always
If the changes introduced into the system are small,

necessary, however.
it is possible to make use of the knowledge already available about the
unperturbed system from the solution of Eq. (13.1).
Suppose for each
of the primed quantities
Q of
Eq. (13.6)
Q'
we take the form
Q + SQ
(13.7)
where the unprimed quantity refers to the unperturbed system, and the
change &Q in passing from the unperturbed to the perturbed system is
If we use the form (13.7) everywhere in (13.6), and
regarded as small.
apply also (13.1) and (13.3), and neglect all terms containing the product
of more than one small quantity, we obtain
SO 0o
+ O
54>o
(Oi
v02)
80 +
(BOi
v 8O2)4>0
&v Oi4>o
= 0
(13.8)
Equation (13.8) is only intended to illustrate the basis of the approach

in this section and will not be of direct further use.
However, it may be
noted that Eq. (13.8) can be formally obtained from (13.1) by appli
cation of a differentiation operation through the equation.
Equation (13.8) is still not directly useful, being a very complicated
differential equation involving the new operators and the new fluxes. In
order to obtain a more tractable form for calculation, we must introduce
another function.
Let ô(r,w) be an as yet unspecified function of space
We multiply Eq.
and lethargy which will later be carefully defined.
(13.1) through by the function \ff0 and integrate over the reactor volume V
If we use Eq. (13.3) and collect terms so as to
and over all lethargies.
REACTOR ANALYSIS
766
solve for
v,
[CHAP.
13
we obtain
Jv
dr
Ju
^0(r,w) Oi^0(r,t<) du
(13.9)
O2*0(r,w) du
jydr Ju^0(r,M)
Changes are now allowed in the operators Ox and 02 and in v to retain

criticality ; there are then resultant changes in
and ypo, in general (the
latter function is not as yet associated with the reactor system, but will
The altered quantities may then be formally

later).
obtained to a first approximation, as shown in Eq. (13.8), by applying a
The result is
differentiation through Eq. (13.9).
be so associated
hv =
jydi Ju [&MOi
vO2)4>0
+ ô(Oi
JY
dr
S02)o + MOi
\f/0 O2<0
- ?02) ] du
du
(13.10)

Because of Eq. (13.1) the first of the three terms under the integral in
the numerator of Eq. (13.10) vanishes.
We consider now the last of the
In order to evaluate this term we introduce two new oper
three terms.
ators
0[ and OJ, which are related to Oi and 02 by the equations
f
Jv
dr
fy
dr
f foOiF
Ju
fu
du =
toOtF du
f
jv
dr
fr
dr
f
Ju
fu
FOtyo du
(a)
FOfyv du
(6)
where F(t,u) denotes some arbitrary function.

last integral in (13.10) becomes
Jydr
JuM<>i
v02)8odu
Jvdr
O'Mr.M) ^ (OJ
With this notation the
0(Ol
We are now prepared to specify the function

it satisfy the equation
- xOô du
ô(r,w)>
- vO\) Mr,u)
(13.11)
(13.12)
and we require that
= 0
(13.13)
With this choice the last integral of Eq. (13.10) vanishes, and the change
in the multiplication constant may be expressed, finally, as
5,
5k
r- = =
K
"
jYdxjuUiOi-v60i)^du
v \dx / Vo02<ôdu
(13.14)
form in which the principal results

be cast.
The fractional change in the multiplication
to
first
constant is calculated,
order, as an integral over the changes in
associ
the reactor operators, with a pair of weighting functions ^0 and
Equation (13.14)
of this chapter will
is of the general
PERTURBATION
SEC. 13.1]
THEORY
767
ated with the unperturbed system.

[The integral in the denominator of
The symbol
Eq. (13.14) may be regarded as a normalization factor.]
denotes the neutron flux in the original system; \f/a is closely related to
the neutron flux and is calculated from Eq. (13.13), which involves only
is,
Note that the changes in the flux

parameters of the original reactor.
It
of course, possible by
are not calculated in this approximation.
an extension of this method to obtain approximations to the perturbed
flux; however, such
quite involved and the

results are seldom of direct interest for reactor work.
b. The Statistical Weight and the Importance Function. Let us turn
is
calculation
generally
of the physical interpretation of the result given in

Eq. (13.14). The situation discussed above in which the neutron absorp
locally increased will serve as an illustrative example.
tion cross section
Suppose we concentrate on a thermal reactor and use the one-velocity
approximation [thus eliminating the energy dependence in Eq. (13.14)]
is
to a consideration
is
that the extra absorber
is
for thermal neutrons.

Let the additional
introduced nonuniformly, be denoted by 5S0(r);
absorption, which
so that
further, choose the normalization of ^-o and
so
dr =
02
(13.15)
8Z(rWo(r) *o(r)] dr
(13.16)
As further illustration of this point, suppose the absorption

only at point r0, so that 520(r) = oS(r r0); then,
Wo(r0)
is
jv
Then, from (13.14)
jrMr)
changed
Aa0{rB)
is
just the frac

This shows that ô(ro), with appropriate normalization,
in
the
neutron
multiplication constant per
tional change
absorbed per
unit time.
As Eq. (13.16) shows, changes in the absorption cross section at various
function
points in the reactor are weighted by
W(i)
Mr)
is
If
*o(r)
(13.17)
Another interesting interpretation

a good approximation to the
<0(r)
is
called the statistical weight.
also possible.
of Eq. (13.16)
flux in the perturbed system (as
is
which
is
essentially assumed in this treatment),
it
is
a good estimate of the additional rate of

the product 52a(r) o(i)
in the perturbed system over the
removal of neutrons at the point
original system. Then from (13.16)
may be seen that this extra
is,
in this
removal of neutrons
weighted by the function ô(r), which
For the simple one-velocity case
sense, called the importance function.
is

REACTOR ANALYSIS
768
[chap.
13
later), the function ô(r) is just equal to the unper

<o(r).
(Here we are not concerned with normalization of
either the statistical weight or the importance function, but only with
the relative values of either from space point to space point.) Thus for
the one-velocity case, the importance (to the multiplication constant of
the system) of the introduction of extra neutrons (or the removal of
neutrons) from the system varies as the original neutron flux, that is,
will
be shown
flux
Importance of neutrons at r
,
a n
model)
-x
(in one-velocity
Further, the statistical weight varies

W(t)
Thus the influence
l(i)
...
~fo(r)
, .
= 0o(r)
(13.18)
v
as the square of the flux,
in one-velocity
model
that
is,
(as
turbed
(13.19)
placed in
very strong at the center, where the flux
thermal reactor
high.
In a slab reactor, for example, the statistical weight for absorption would
fall off as cos2 x. (The implications of this behavior are discussed fur
neutron
absorption
is
model
is
ther in subsequent analysis when the one-velocity
considered in
detail.)
Of course, such results
as expressed in general terms by Eq. (13.14)

encompass much more complicated behaviors than the simple one-velocity
Even in the one-velocity case changes in the diffusion coeffi
absorption.
cient bring about somewhat different considerations than those presented

above; the energy dependence of the changes has not been discussed here
is
a
Still
it
broadly applicable conclusion that the importance

tends
to
behave
function
space-wise as the neutron flux, and thus the
statistical weight, in so far as changes in the absorption and fission cross
at all.
sections are concerned, tends to vary spatially as the square of the flux.
c. The Adjoint Function and Operators.
We return now to the formal
outline of the perturbation method as started in Sec. 13.1a.
In the
terminology the operators 0\ and 0\ as defined by

Eqs. (13.11a) and (13.116) are called the "adjoint operators" to dand 02;
which satisfies Eq. (13.13) involving the adjoint operators
similarly
said to be the "adjoint function" to 0.
to those of Eq. (13.1) for 0,
It clear that the emphasis of the succeeding discussion must be placed
on the calculation of the adjoint function \f><>
for the various models con
It also clear, from Eq. (13.13), that the calculation of the
sidered.
very much like the calculation of the neutron flux
adjoint function
With appropriately chosen boundary conditions [not specified in
itself.
order to reach the result (13.14), and which will be selected in the exam
ples below for computational convenience] the machinery used for the
calculation of the flux may be applied directly to the construction of the
adjoint function. The advantage of this procedure
seen from Eq.
and one for
(13.14); the two calculations one for
on the original
is
usual mathematical
is
is
is
is

thermal
is
of extra
PERTURBATION
SEC. 13.1]
THEORY
769
system, allow an easy evaluation of the effects of many types and dis
tributions of small changes from the original.
In fact, in some cases,
as for the one-velocity model mentioned above, only one calculation is
required since the adjoint function is just equal to the flux itself.
This
situation arises when the operators in the reactor equation are self-adjoint,
Before turning to the calculation of
i.e., when Of and O are identical.
adjoint functions for special models, however, we shall examine one
general property of the adjoint function.
is
it
is
r',
Consider the following problem.

A burst of neutrons of lethargy u
is introduced into a just-critical reactor at space point
and
required
that the resultant neutron flux be found as
function of space and
Since the system
lethargy for all subsequent time.
critical, neutron
into the system will
on the average just reproduce itself

rise
to
a
indefinitely, giving
persistent steady-state flux corresponding
to a neutron density of just one neutron in the reactor volume.
Now
the phrase "on the average" here implies an average taken over neutrons
introduced
into the system in the fundamental mode. It

clear that
there will arise a persistent, ultimate flux (in the fundamental mode)
from the introduction of
neutron at any point in the system, but the
magnitude of this ultimate flux will vary with the point of introduction.
would be expected that a neutron introduced near the
outside would give rise to
much smaller ultimate flux than that from
the introduction of
neutron near the center of the system, since the
fraction of the neutron which would "leak out" before the flux settled
a
it
For instance,
=
v
0<f,(i,u,t^',u')
at
>
0
is
The flux in the reactor for all time
down into the fundamental mode would be higher in the former case.
We now attempt to solve this problem in general terms and find the form
and magnitude of the ultimate flux.
taken to satisfy the equation
(r,u,t;r',u>) +
S(t)
h{u')
(13.20)
is
is
r,
at
Here >(r,w,<;r',w')
the neutron flux of lethargy u, at space point
time after the introduction of one neutron of lethargy u' at position r'.
denotes the reactor operator of Eq. (13.1). The approach
The symbol
to the solution of this problem
through expansion of the flux in space
2,
0,
= X,<fc(r,u)
1,
0f(r,)
it
is
more specialized case in Sec. 5.4.

Suppose that
modes as was done for
following equation:
which
satisfy
possible to find functions <v(r,i which
satisfy the
following the
equation:
.
(13.21)
is
is
X,
The eigenvalues
are assumed to be arranged in order of increasing
taken to be zero so that
magnitude with increasing index, and X0
just the reactor flux of Eq. (13.1). The basic assumption here
<o(r,u)
that the eigenvalues of Eq. (13.21) can be arranged in order of increas
The eigenvalues and eigenfunctions of
ing magnitude of the real part.
is

is
introduced
REACTOR ANALYSIS
770
[CHAP.
13
(13.21) may, of course, be complex; indeed, little is known of them

except for the simplest cases, e.g., the one-velocity case discussed later
However, it does seem reasonable to expect, from the
in this chapter.
results obtained below for the time-dependent behavior, that when the
operator O corresponds to a physically operating system the assumption
In the case of complex eigenfunctions, more
used here must hold.
It is
convenient definitions of the adjoint operators might be chosen.
important to point out again that the present discussion is merely an
outline of a procedure which might be carried out in detail for particular
cases.
We now expand the time-dependent flux of Eq. (13.20) in the functions

<fc;
thus,
(T,u,t?',u')
a,;r',u') *(r,u)
(13.22)
Here
<,
a departure from the procedure of Sec. 5.4 must be made because

are not necessarily orthogonal.
However, we
the members of the set
which
of
functions
are
to the set
orthogonal
&(r,u)
construct
a
set
can
dr
fu
fv
in the sense
fc(r,u) 0,(r,) du
(13.23)
5,7
is
is
is
unique (except for normalization, which

Further, the set ^,(r,u)
unimportant here; we are not concerned here even with the dimensional
not
ity of the adjoint function which can be specified; but this also
important). Now multiply Eq. (13.21) through by the function &(r,u)
If
du =
we introduce the adjoint operator
jydt
ju4>j04>idu
0+
dr
du = Xrf
frOti
fy
dr
X,
fv
and integrate over space and lethargy.

(13.24)
of Eq. (13.13),
fydr jui[0^j\du
= X,50
(13.25)
The form of Eq. (13.25), along with the uniqueness of the functions
must satisfy the equation
shows that the
Otyy(r,tt) = X^,(r,)
(13.26)
Thus the orthogonal functions are just the adjoint functions, and ^0(r,w)
just the flux adjoint of Eq. (13.13).
We now introduce the expansion (13.22) into Eq. (13.20), multiply
The resultant
through by ^(r,w), and integrate over space and lethargy.
equations for the coefficients in the expansion are
- d,(')
X,ay(*;r',u') =
is

i(i,u)
5)
fc(r>')
(13.27)
sec. 13.2]
PERTURBATION
THEORY
771
This equation shows the expected result that the coefficients of all the
higher modes decay exponentially in time (t > 0) and that the coefficient
Thus the ulti
of the fundamental mode is a constant after the burst.
mate form of the flux
long after the burst is
x(t,u;i'u')
(13.28)
4>0(r,w)
others) to the calculation

reactor system.
is
it
if,
We find, then, that the relative values of the adjoint function to the
neutron flux from one space point to another gives directly the ratio
of the magnitudes of the ultimate fluxes from bursts of neutrons intro
Again it may be seen that the adjoint function
duced at these points.
measures the importance of the introduction (or removal) of neutrons at
a given point in a reactor assembly.
It should be pointed out that this rather interesting approach to the
adjoint function is not of direct importance for developing useful results
In fact, this line of thinking
which will be obtained in this chapter.
for example,
can be overextended
taken to imply that an extra
source of neutrons can be compensated by an appropriate change in the
multiplication constant.
In the sections which follow we develop with the aid of various reactor
models the application of the techniques discussed above (as well as
of the effects of parameter
changes on the
is
In many cases
possible to use Eqs. (13.11) to find explicit forms
for the adjoint operators so that (13.13) may be written as an explicit
set of equations for \po(T,u) (along with appropriate boundary conditions).
In the general case this particular approach leads to the consideration
of the adjoint to the Boltzmann equation in which all the generality
might be preserved, including angular dependence of the functions.
On
it
the other hand, the procedure represented by (13.20) might correspond

to the time-dependent Boltzmann equation, and the eventual flux result
ing from the introduction of one neutron might be found in the most
general case.
However, such ambitious programs must meet with
is
it
exactly the same difficulties as are encountered in applying the Boltz

mann equation directly to reactor-flux
calculations.
Further,
hardly desirable to know the importance function to better approxima
known, since the basic result, as typified by
tion than the flux itself
is
Eq. (13.14),
uses the two functions
is
symmetrically.
Consequently,
customary and proper to base perturbation theory and the calculation
of the adjoint function on the model used for the calculation of the flux
it

vMr',W)
itself.
13.2
Perturbation in the Infinite Multiplying Medium
The Unperturbed System.

The discussion of the case of the infinite
multiplying medium, in which the disturbances are changes in the energy
a.
REACTOR ANALYSIS
772
[CHAP.
13
dependent cross sections only, is of value as an introduction to methods

The full power of the per
which will be of use in subsequent analyses.
turbation techniques is not needed, since all the important results may be
easily obtained from the explicit form of the criticality condition (4.162) ;
however, this very same fact makes this case an interesting one for the
introduction to the methods since the results obtained by somewhat
methods may be easily checked from considerations of the
criticality condition.
For this treatment we consider the lethargy range 0 < u < uB, where
the two limiting values of the lethargy are chosen so that practically no
neutrons are found in the system with lethargies beyond these limits.
Thus, lethargy zero should correspond to an energy which is a practical
upper limit for the fission spectrum, and lethargy w0 should correspond
All functions are treated as continu
to an energy just below thermal.
ously varying over this lethargy range, and the special conditions in the
region of thermal energy are imagined to be handled by allowing the
quantity (the average logarithmic changes in the energy per collision)
All physically inter
to take on very small values in the thermal range.
esting results can be obtained as limiting cases of the lethargy-dependent
functions obtained in this way.
The results for the unperturbed system are summarized from the deri
vations of Sec. 4.6. The appropriate form of the neutron-balance relation
(4.145) for the case of the normalized fission spectrum j(u) is seen to be
q(u) + P(u) q(u) = rj(tt)
y(u') g(u') du'
where 0(w) and y(u) are defined by Eqs. (4.161).

equation is
fj,
9() =
(13 29)
The solution to this

(13.30)
and for the criticality condition we previously obtained
the form [cf.
(') exp
/"
0(u") du"] du'
j(u') exp
y(u) du
fo"
Jou
Eq. (4.162)]
1
0(u") du"] du'
(13.31)
The Adjoint Equations.

The first step in the derivation of the
function by the technique of adjoint functions
to cast
Eq. (13.29) into the form of Eqs. (13.1) and (13.3) and to multiply
function ^(u), as yet unspecified.
The next step
through by
to
integrate over lethargy and then solve for v.
b.
is
is
importance
du
/*o
9^
dH
^ Q^
rut
rue
Jo
9f(") J(u) du
rut
Jo

elaborate
7(w') g(u') du'
dU
(13.32)
PERTURBATION
sec. 13.2]
THEORY
773
Now allow changes in the cross sections, implied in
and 7, and conse

changes are also imaginable: the
f)
quent changes in q and q*.

(Other
fission spectrum j might be altered by, for example, the introduction
of new species of fissionable nuclei; this is a rather unimportant case, and
presented here may be easily generalized to include it.

might change, if it is imagined that u0 corresponds roughly
energy, if the temperature were changed.
However, as
the treatment
The limit
to
Mo
thermal
chosen, o corresponds to an energy so far below thermal that practically

no neutrons are found there, and the changes consequent upon the
The
temperature change may be regarded as changes in /3 and 7.)
application of (13.29) eliminates

written
rut
g'(u) /3(u) q(u) du
/
hv =
rut
$(w)
gf(M) j(w) du
du /
du
Jo
y(u') q(u') du'
~ "i(M)
S9(M) + ?t(w) ^(M) *9(m)
rut
rut
Q*(u) j(m) du
q*(u') Sy(u') q(u') du'
Tut
?t(M)
Jo
v
may be
rut
fw
Tut
Jo

and the result
one integral,
rw
Jo
y)
~i
*?*U') du'
dH
y(u') q(u') du'
(13.33)
The first term on the right-hand side of (13.33), involving the crosssection changes 5/3 and 67, is in the desired form; the last term on the
right corresponds to the last term on the right of Eq. (13.10) and will
We first look for the operators adjoint to the
be treated analogously.
of
the
differential
The only one that pre
equation (13.29).
operators
sents any difficulty is the differential operator d/du; following the require
ment of Eq. (13.11a), consider
fut
0
qt (M)
=
*9(M) du
But,
C**
0
d
du
Sq^
f"'d
Jo du
t?t(M)
5?(M)1 du
fu
d
*?() du 9f(w) du
/
Jo
(13.34)
du = 9Ûo) 5^Mo) ~ 9^0) 5^0) = 0
(13-35)
t-
The last step follows since the slowing-down
density, originally and

after the change in constants, is zero at both u = 0 and u = Uo; conse
For this case,
quently, no specification of gf at either limit is required.
then, the operator
adjoint to d/du is d/du.
The numerator
of the
REACTOR ANALYSIS
774
[CHAP.
13
g*() + P(u) q*{u)
5q(u)
q1{u)
joy{u']
&q(u,)
q*(u')
vy{u)
du'] du']
~
n{u)
"(tt) Sq{u)
5q{u)
[?t(M) Tu
fa"
last term on the right-hand side of (13.33) may be written

du
du
(13.36)
qfM
P(u) q'(u)
seen to be
n(u)
j(u') ?'(') d'
f"
of 3t(u)
the required specification
is
Thus, by proper choice of the function q*(u), the entire last term on the
From (13.36)
right-hand side of Eq. (13.33) may be made to vanish.
(13.37)*
is
is
take
and
if
we use (13.30) for q(u) and note that

With this requirement we find the solu
which
may
be
compared
Finally, from (13.33)
we
J*
y(u') exp
jj
tion to Eq. (13.37) for the adjoint function to

gt(tt)
(13.38)
f(w) du =
/0"
"
analogous to (13.31),
have exchanged roles.
j is
which
7

is
it
may be seen that besides the sign change

By comparison with (13.29),
on the derivative term the fission spectrum and the fission cross section
have exchanged roles for the calculation of the adjoint slowing-down
The calculation of the adjoint slowing-down density
density.
very
much like a "backward" calculation of the slowing-down density itself;
the adjoint function
started in the thermal region with a high value,
The "adjoint neutrons"
the fission cross section acting as
source.
are then followed up in energy (because of the negative derivative) with
losses occurring by absorption just as in the case of the slowing-down
Finally, at high energies the value of the fission spectrum
density.
multiplied by the adjoint function determines the number of "adjoint
This feature of the
neutrons" started up the scale again from thermal.
lethargy adjoint functions has lead to the use of the expression "forward"
calculation when referring to the calculation of the slowing-down density
or the flux from the slowing-down equation (13.29), as contrasted with the
adjoint or "backward" calculation of Eq. (13.37).
still to be chosen; we
The normalization of the adjoint functions
be
/3(u")
du"]
du'
(13.39)
with (13.30) for the "forward" function.

for the change in the eigenvalue, or
obtain
PERTURBATION THEORY
sec. 13.2]
775
multiplication constant,
Sv
5k
In this result,
fun
/
qi(u) 5/S(w) q(u) du
fun
5y(u) q(u) du
(13.40)*
indeed plays the role of an importance function

In the first
for the introduction of neutrons, as discussed in Sec. 13.1.
integral on the right-hand side of Eq. (13.40), h&{u)q(u) are the extra
neutrons removed per unit volume per unit time by the additional
absorption cross section, assuming that the change in the slowing-down
Then q^(u) is a measure of the (relative) impor
density itself is small.
gf(w)
Note that an importance

tance of removing a neutron at lethargy u.
function is not needed in the second integral; dy(u)q(u) gives the extra
fissions, and all fission neutrons appear with the same distribution in
lethargy regardless of the energy of the neutron causing the fission (at least

by assumption in this model).

By substituting the explicit forms for
c. The Importance Function.
q(u) and q*(u), from (13.30) and (13.39), respectively, into Eq. (13.40),
a complete expression for the fractional change in multiplication con
This is not of particular interest, however, and
stant may be obtained.
is not recorded here.
In fact,
as mentioned
previously, the entire tech

nique used here is unnecessary for this simple case, and the value of the
discussion is the ease with which the basic ideas, to be used later, can be
For example, the result (13.40), with the explicit forms of

exhibited.
the functions used, may be obtained directly by a differential technique
Even the importance
applied to the criticality condition, Eq. (13.31).
function Eq. (13.39) may
be readily constructed from a physical under

of the model; the relative importance of a neutron introduced
at lethargy u is just proportional to the number of fissions it would
eventually cause in slowing down from u toward wo- Equation (13.39)
standing
is the mathematical
statement of this fact.
In
the very simple case in

which fissions are caused only by thermal neutrons, it is evident that the
relative importance of a neutron introduced at lethargy u is just the
probability that the neutron escape absorption in slowing down to ther

mal; thus in this case, wherein fissions are caused only by thermal neu
trons, both physical considerations and Eq. (13.39) lead to the result
,
v.
,,
,.v .. ,..,,
(13.41)*
The behavior of the slowing-down density and its adjoint are presented
in Fig. 13.1. The thermal range is represented as a small interval in
lethargy just below u = w0; throughout this interval the functions are
rapidly varying. Changes in the absorption cross section, as seen from
Eq. (13.40), are weighted by the function q*(u)q(u); from Fig. 13.1 this is
seen to be nearly a constant over the lethargy interval below the fission
REACTOR ANALYSIS
776
[CHAP.
13
Lethargy
Thermal range
-|
In fact this may be easily seen to be true

spectrum and above thermal.
in
which
fissions occur only at thermal and the fission
in the limiting case
-u0
delta function at the zero of lethargy; then,

represented as
lethargy range
a
is
spectrum
>
q(u)
in the intermediate
(13.42)
(13.43)
and
qy(u) q(u)
= constant
in lethargy
"statistical weight"
(13.44)*
is
it
[Actually, this result can be somewhat misleading unless

noted
that the statistical weight of Eq. (13.44) really applies to changes in
= 2/2(, as seen from Eq. (13.40).]
/3
13.3
Perturbation in the One-velocity Model of the Bare Reactor
is
a. Eigenfunctions of the Bare Reactor.

The method of adjoint func
tions introduced in Sec. 13.1, and applied to the multiplying medium in
Sec. 13.2, may also be applied directly to the one-velocity diffusion equa
tion.
The general development for these systems
given in Sec. 13.4.
In the particular case of the bare reactor, we discard the method of
adjoint functions in favor of an alternate approach which yields more
information.
Consider
bare reactor in the one-velocity approximation in which the
flux satisfies the equation
a

Fig. 13.1 The slowing-down density q(u) and the importance function g'(u) for the
infinite multiplying medium.
-V DV
2tf>
vXfpfi
(13.45)
sec. 13.3]
PERTURBATION
THEORY
777
In this equation the parameters D, 20, 2/; and pfi are, in general, func
tions of position; the parameter p/( is taken to represent the losses in the
fast range, so that in the application of any of the results of this section
P/i may be replaced by whatever estimate of the fast losses, leakage and
In the undis
absorption, is regarded as satisfactory in a particular case.
turbed homogeneous system, the unperturbed flux o(i) satisfies the
equation
-Z)oVJ0o(r) +
S>*,(r)
v^pîT)
The solution to
of position.
where D0, 2^', and 2}0> are independent

this equation may be written in the form
(13.46)
*0(r) = cFo(r)
(13.47)
where, in the usual way, F0(r) satisfies

= 0
V*F0(r) + BlFa(T)
(13.48)
and Bo is determined from the requirement

F(ft)
= 0
(13.49)
The vector ft denotes the outer extrapolated

The critical equation for this system is
BS(ft) =
It
will
be of interest
gl
in connection
P/'
Do
boundary
of the reactor.
(13.50)
with the perturbations to
be dis
cussed here to consider the complete set of eigenf unctions FH(t) of the
bare reactor, of which F0(r) is the lowest mode. These functions are
solutions to the equation [cf. Eq. (8.260)]
V!F(r) + BlFn(t)
It
is convenient to construct
thus [cf. Eq. (8.249)]
v
= 0
with F(ft)
= 0
(13.51)
this set so that the F(r) are orthonormal;
F(r) Fm(r)
dx =
Smn
(13.52)
where V denotes the volume enclosed within the surface represented by

ft.
The set Fn{i) may be ordered to correspond to the eigenvalues Bn
such that B* < B\+l. It should be noted that in the present formulation
this set of functions must be constructed in complete generality in order
perturbations.1
involving nonsymmetric
Appropriate functions for the four elementary reactor geometries are
listed in Table 13.1.
to
meet
1 For
physical
situations
example, the block-reactor eigenfunctions must include the sine functions as

well as the cosine, and in the spherical and cylindrical cases, the angular harmonics
are required.
ANALYSIS
REACTOR
778
[chap.
13
b. Change in Multiplication Constant.

Now suppose that in the bare
is
made in the cross sections; thus,
system under consideration a change
2(r) =
2,(r) =
X'f + 2(r)
2}>
+ SS^r)
(13.53)
(Problems involving changes in the diffusion coefficient are deferred until

The flux in this perturbed system may be defined,
the next section.)
then,
(r)
<fo(r)
+ *(r)
In order to maintain steady-state operation,
(13.54)
we must allow a
hypothetical
change in v.
f-+
The substitution of
hv
(13.55.)
forms into Eq. (13.45), along with the appli
these
cation of (13.46), yields

-D0V2S<Kr)
+ ^TH{r)
vvpp,M{r) + p2,(t) P/I0(r) + v2P//*o(r)
20(r)
>(r)
(13.56)
where we have neglected products of the various differential quantities.

be solved for the fractional change in k by expanding

This equation may

Table
1.
Orthonormal
13.1
Set ok Eioenfunctions
Four Elementary Reactor Geometries
for the
Infinite slab reactor
(A.
*".(*)
cos Bx
sin Bx
n even
n odd
A.
n = 0, 1, 2,
2. Rectangular block reactor

Fi,k{x,y,z) = Aijtf(liX)g(?iijy)h(ntz)
BU
with i,
j,
= <? +
m) +
k = 0, 1, 2, .
1)
n*
Aijk
"I
,
and
my
2c^
(j +
1>
= (0JC)"i
and
... .
Jf(Ux) 3
kx
'
ifcos
.
[sin kx
1)
.
,
g(m<y)
"v
"
fcos m<y
cos ntz
(sin ntjy
sin
i,j,k
even
odd
sec. 13.3]
PERTURBATION
THEORY
779
Table 13.1 Orthonormal Set of Eioen functions for the

Four Elementary Reactor Geometries (Continued)
Cylindrical
3.
with i,
li =
= 0, 1, 2
7
2a
(i +
and
A.,
t i
For

4.
reactor
and k
1, 2, 3,
. .
= fcth zero of Bessel function
,*
1)
|^[-^V^,(Û;+1(vyi)]|-i
t i
/Bi
f coa j6
j there
each value of
are
lain j9
two functions Fiit,
Jj
i. i even
/cos i<2
?, j odd
(sin /;2
one with cos j$, the other with sin j$.
,
Spherical reactor
FijkirM)
- Aiiltfik(r)
Bl
g<i(9)
Ay(*)
()"
with pa = Hh zero of the spherical Bessel function"

. . . , and Jb
1, 2, . . . .
Atik
~{'[-\
?o() =
Pj
(cos 9)
fcos
(sin
jV>
(*
j,-,
and the indices i,
= 0,
1,
+*!)/->)!'}
spherical Bessel function

associated Legendre function
for each value of there are two functions
Fiji, one with cos jrp, the other with sin jip
P. M. Morse and H. Feshbach," Methods of Theoretical

the perturbed
Physics," McGraw-Hill
flux (13.54) in a series using the complete set of eigenTherefore set
functions given by (13.51).

*(r) = ^
n=0
= cF(r) +
(13-57)
n=l
Here the contribution from the original flux has been separated out so
Note
that the last sum on the right of Eq. (13.57) represents 5<(r).
=
also that we have taken a0
c for convenience since the magnitude of
REACTOR ANALYSIS
780
[chap.
13
is not of interest here; only the spatial variation of &

relative to
The analysis of the per
will be considered in the present calculation.
turbed flux, taking into account its actual magnitude and buildup in time,
is an entirely different problem and is not examined here.
The substitution of the expression (13.57) into (13.56), followed by
multiplication through by 0o(r) and integration over the volume enclosed
by the extrapolated outer surface of the reactor, gives the following result :
&
- =^
Sk
5k
fv
*I
*o(r)[vp/,8S,(r)
- Z.(r)]*,(r)
di
(13.58)*
-.

In this expression the change in the multiplication constant resulting
from a change in the parameters is calculated

which it is evident that for this case
Importance function
Statistical weight
~
~
in the usual form, from
4>o(r)
4>l{r)
The same result will be obtained in Sec. 13.4, somewhat more generally,
and will be discussed in more detail there.
For the present section, we shall concentrate on the extra information
If Eq. (13.57) is multi
to be obtained by the eigenfunction expansion.
plied through by/rm(r), where m ^ 0, and integrated over the volume V, an
expression may be obtained for the coefficients an of the expansion for
With the application of Eqs. (13.50) and (13.52) the result may
6>(r).
PERTURBATION
SEC. 13.4]
written
be
a =
-f
THEORY
- 2.(r)]*0(r)
DM - be
F,(r)[irp2,(r)
781
dx
(13-6)
Since the F(r) are known, the specification of the perturbing functions
52/ and 52a will allow explicit evaluation of the coefficients a from the
and, therefore, also an infinite series representation for
flux <(r) may be conveniently given the same
The
perturbed
50(r).
normalization as <o(r) by requiring that
above relation
f
By
*(r) di
= 0
(13.61)
(i)
may be directly compared with o(x).

Figure 13.2
shows an example of the results of a calculation using Eqs. (13.57) and
(13.60) for the special case of a bare slab with locally added absorber.
so doing

13.4
Perturbation in the One-velocity Model of
General System
a. General Formulation.
The treatment of the general reactor prob
lem in the one-velocity model using the method of adjoint functions is
based on the neutron balance relation (13.45).
It is understood that in
the present application all the quantities appearing in that equation are,
in general, functions of the spatial coordinates so that this equation may
be used to describe any and all regions of the reactor.
For convenience
in writing, we combine the two factors 2/ and p/( into a single function
2/i, where
2/((r)
- 2/(r) p(r)
(13.62)
Then (13.45) becomes
-V
D(r) V*(r) + 2(r) *(r)
= *2/((r)
*(r)
(13.63)
In order to determine the required modification
in v to compensate
for a general change in any of the nuclear parameters, we first multiply
through Eq. (13.63) by some, as yet unspecified, function >f(r) and
integrate over the volume V enclosed within the outer extrapolated sur
face of the reactor.
The resulting equation when solved for v yields
L*f(r)[-V.D(r)
= }
V + 20(r)] *(r) dx
-.
Jv
*t(r) 2/J(r) *(r)
dx
(13.64)
As before, we next allow variations in the cross sections, which in turn

cause variations in 4> and >f.
Formally, this is accomplished by apply
ing a differential operator through Eq. (13.64).
Again the original differ
ential equation (13.63) may be used to eliminate some terms. The
&v
&D(r) V + 52n(r)
Jr *t(r)
+
**(r)[-V
/"
13
'
resulting expression is
L
^
[chap.
ANALYSIS
REACTOR
782
Z(r)] *(r) dr
2(r) *(r) dr
D(r) V + 2(r)
jv
- r2,,(r)]a+(r) dr
(13.65)
*t(r) 2/l(r) *(r) dr
The first term on the right side of this equation is of the desired form.
The immediate problem is to choose ^(r) according to the method out
lined in Sec. 13.1 so that the second term vanishes.
b. The Adjoint Function.
The first step in the analysis is the determi
nation of the operators adjoint to those within the brackets in the inte
The only one of these which pre
grand of the last term of Eq. (13.65).
In order to treat this term
sents any difficulty is the operator V D(r) V.
we introduce the following identities:
DV8dr
D
Jv 6V
dr =
Jv
- Jy (V^)
(5 D V^) dr - jy
(W
jyV
V5tf>)
dr
(V</)
(D V) dr
(13.66)
(D V*f) dr
(13.67)
If the functions whose divergence is required in the first integral on the

right-hand side of each of these equations are nonsingular and continu
ous, Gauss' theorem may be applied; then,
jv
(W VS) dr
(8D
Vtf) dr
(4>W V80)
Jg
(8 D V^)
dS
(13.68)
dS
(13.69)
S,
where S denotes the extrapolated surface of the reactor and dS a differ

ential area on that surface.
Since both the perturbed and unperturbed
fluxes (and thus 8<p) must vanish on this surface, the integrals of (13.69)
If it now be required that
vanish on the surface
vanish.
the
D(r)
V5<Kr) dr =
80(r)
Jy
tf{r)
Jv
is
Thus the first integrals on the rightintegrals of (13.68) also vanish.

hand side of Eqs. (13.66) and (13.67) vanish, and inasmuch as the order
of scalar multiplication
immaterial, the second integrals are equal,
and we conclude that
D(r) V^(r) dr
(13.70)
f(r)
is
the function #f(r)
is
second term in (13.65) may be made to vanish
if
ft
V,
0,
and Z)(r)V<^t(r) are continuous and nonsingular throughout the

where
denotes the surface S.
We find,
volume
and <^(ft) =
Z)V
self-adjoint.
then, that under these conditions, the operator
Since 2(r) and f2/((r) are obviously self-adjoint, the numerator of the
if

Jv
sec. 13.4]
PERTURBATION
THEORY
783
-V
But this
D(r) V*t(r)
selected so as to satisfy the equation

2o(r) *t(r) = *2(r)
^(r)
(13.71)*
it
is
Since
precisely the equation satisfied by the flux (13.63).
the boundary conditions specified above for
are identical to those for
must be concluded that
**(r)
*(r)
- 82(r)
'
6S/'(f)
U-
*C(r)V]^(r) dr
(13.72)*
Sk
'ir
- 5, =
from (13.65) as
lv
be calculated
With the above result the change in the multiplication constant may
rF2y,(r)*(r)dr
Importance function
Statistical weight
~
~
(r)
(13.73)
4>\r)
(13.74)
is
it
In Eq. (13.74) we have directly the result mentioned in the discussion

following Eq. (13.19). In conjunction with this
important to point
is
out the great effectiveness of changes in fission and absorption cross

sections at the center of the reactor where the neutron flux
For
high.
in
cubic
reactor,
bare
the effective
example,
homogeneous one-velocity
a
in changing the multiplication constant of adding a given amount of

absorber material at the center of the cube
eight times as great as
the same amount of absorber were spread uniformly over the reactor.
c. Changes Involving the Diffusion Coefficient.
Equation (13.72)
if
is
ness
information not previously considered regarding the effect of

In order to gain insight into the
changes in the diffusion coefficient.
contains
sk
it
is
nature of the weighting functions for changes in the diffusion coefficient,

a continuous
consider, for the moment, cases in which 8D(t)V4>(t)
function. Following the technique of Eqs. (13.66) through (13.69)
can be shown that the portion of (13.72) involving the diffusion coefficient
may be written in the form
5D(r)[V<Kr)]2 dr
/v

is
In so far as the changes in the absorption and fission cross sections are
seen to be formally identical to that of Eq.
concerned, this result
(13.58); however, Eq. (13.72) applies more generally to a multiregion
reactor.
It must be concluded, then, that (for absorption and fission
changes) in the one-velocity diffusion approximation [cf. Eqs. (13.59)],
We find, then, that for the
Z(r) *2(r) dr
case of changes in the diffusion
Statistical weight
~ (V)2
(13.75)*
coefficient,
(13.76)*
784
REACTOR ANALYSIS
[chap.
13
In the case of a bare slab reactor, the statistical weight for absorption
and fission cross-section changes would vary across the slab as cos1 Bx
[from Eq. (13.74)], whereas the statistical weight for changes in the
These forms
diffusion coefficient would go as sin2 Bx [from Eq. (13.76)].
in
that
the
changes
This figure reveals
absorption
are shown in Fig. 13.3.

and fission cross sections are most important if made near the center of
the reactor; changes made near the edge are not as important since extra
neutrons there are likely to leak out and therefore unlikely to contribute
Very nearly the opposite behavior is observed
to the next generation.
for changes in the diffusion coefficient; changes near the center are
unimportant while changes near the edge contribute heavily.
We may
approach the understanding of this result by considering the introduction
of a small void at various points in the reactor (neglecting the effects
of the displaced absorbing and fissionable material).
The presence of
the void increases the effective diffusion coefficient in the nearby region,
and this, according to (13.75), will result in a reduction in the reactivity
of the system.
This is seen to depend on the fact that neutrons are
streaming from regions of higher importance to regions of lower impor
In
tance, and an enhanced diffusion coefficient will aid this streaming.
regions where the gradient of the flux is large, the increased diffusion
coefficient will be most effective.
At the center, where the gradient
vanishes, no finite increase in D will make the net current anything but
zero, and thus will not affect the reactivity.
This increased net current
due to an increased diffusion coefficient accounts for one factor of V<
in Eq. (13.75); the other arises from the fact that the change in impor
tance function due to streaming is proportional to the change in the flux
itself since the flux is self-adjoint in this case.
PERTURBATION
sec. 13.5]
THEORY
Perturbation in the Two-group Model
13.5
For the purposes of the present treatment

equations in the form [cf. Eqs. (8.147)]
-V
-V
785
V*,(r) +
D,(r) V*,(r) +
Dx(r)
Zg>fr)
2i2>(r)
*i(r)
*,(r)
=
=
we write the two-group
rZ,(r) +.(r)
pc(r) 2^(r) ^(r)
"
where the various symbols and subscripts have the previous definitions.
2, DiVfa, and DtVfa
In this analysis we require that the functions
i,
be continuous and nonsingular throughout the volume V (as previously

specified) and that i(ft) = #2(ft) = 0, where ft denotes the outer sur
face of V. In the present formulation all parameters appearing in (13.77)
are taken generally as functions of spatial coordinates so as to allow for
all variations, including those involving regional differences in the
system.
It is convenient
notation.
Thus
and customary in analyzing this system to use matrix

matrix and the oper
we define the fluxes as a column
ators as square matrices.
In this construction,
Eqs. (13.77) may
be

written
If
the square matrices are represented by the operator symbols Oi and

O2 [cf. Eqs. (13.2) and (13.3)] and the column matrix of the fluxes by 4,
then (13.78) reduces to the expression
O,*
vO&
(13.79)
This equation is in the form of the general relation (13.1), and may there
fore be treated by the method outlined in Sec. 13.1.
The analysis begins by introducing a yet unspecified row matrix 4f :
441
(13.80)
As in the earlier calculations, this quantity is multiplied into Eqs. (13.79),

the result integrated over volume V and solved for v.
Jy *t02* dr
Next,
we allow general changes in the parameters
(13.81)
of the system which

cause changes in <P and <b*.
In order to maintain the system at steady
state, we introduce a change also in v.
This operation is performed by
differentiating through Eq. (13.81); the differentiated quantities are
REACTOR ANALYSIS
786
preceded by the symbol
jv 8&(Oi
13
[CHAP.
8.
- f02)*
dx
Jy
*+(50i
+
Sv =
*+02*
02)4 dt
[ *f(Oi
v02)*dr
(13.82)
dx
The first integral in the numerator vanishes by virtue of (13.79). The

second integral is of the desired form.
As previously, we begin with
some manipulations of the third integral.
The procedure is the same as
outlined in Sec. 13.1, and used in subsequent sections.
New operators
0\ and
fy *HOi
Thus
that
02 are defined such
we require
that
v02)*
*f
dx =
**(Olt
- vO\)*l dt
(13.83)
satisfy the equation
0{*l

Jv
vOfrl
(13.84)
where the subscript denotes the transposed row or column matrix.

The change in the multiplication constant is finally calculated as
5,
"
_Sk_
/F
-,*0,)*dr
*'(<>!
Jv *+02*
dx
In order to determine the adjoint operators it is necessary to examine

in detail the operator terms of Oi and 02 of Eq. (13.78); these are all
multiplicative, and thus self-adjoint except for V D{V and V Z)2V,
which may be treated exactly as in Sec. 13.4.
The result is that all the
individual operators
which
make up the matrix operators
may be regarded as self-adjoint if we require that 4>\, 4>\,

and continuous and that 4>I(ft) =
Z)jV<^ be nonsingular
0\ and 02
D{V\,
and
4>2(ft)
= 0.
Then the adjoint operators 0[ and 0\ are the transposes of the operators
Oi and 02, respectively (note that all quantities are real here), so that
Eq. (13.84) becomes
-V-
D,V +
0
2^'
-v.
d2v + 2i2,j
For direct comparison with the

expanded to
-V
-V
ZMr) V*I(r) +
>2(r) V^(r) +
u\\
- "is,
oj
uj
(i3-8b)
"forward" equations (13.77) this may

Zg>(r)
2<2,(r)
[{x)
*l(r)
= pc(x)
2^>(r)
= v2,(r)
*Kr)
t(x)
be
'
PERTURBATION
SEC. 13.6]
Finally
THEORY
the change in the multiplication
787
constant
is calculated
from
(13.85):
fv
*J*[Pc2,]*i dl
- fv &D2Vt
- fy <^Z02 dr +
Jy
V$s dr
dr]
4\SZf^
(13.88)*
It
It follows from this development that the method employed here

directly applicable to any multigroup model.
Perturbation in the Age-diffusion Model
13.6
a. Adjoint Equations for the Age-diffusion Theory.

The methods of
the preceding sections may be combined to determine the equations
adjoint to the age-diffusion relations (6.54). The solution to the adjoint
equations yields the functions adjoint to the slowing-down density and
the thermal flux; these will act as importance functions in the treatment
of small changes in the parameters of the system.
Since all the essential
is
features of the method have already been discussed in considerable detail,

only an outline of the procedure
given here, along with
summary of
the results.1
-V
where
"i(u)
PM q(r,u)
q(r,u)
[y(r,u') q(r,u') + Z}"*(r) 5('

=
Ah(r) V*(r) + 2?(r) *(r)
4>(r)
[|frÛ
J~
D(r,u)
f~
-V
For the present application we use the following form for the agediffusion equations:
5(u'
du'
(13.89)
u*) q(r,u') du'
(13.90)
utb)]
denotes the thermal flux, and
*">-
US
All remaining symbols and notation, including the "th," have the usual
definition.
Since all quantities
(with the exception of the fission
spec-
This treatment follows that of M.

Nielsen given in "The Adjoint Equations
and Perturbation Theory for
Reflected Reactor," Oak Ridge National Laboratory,
ORNL Y-F 10-31, Jan.
8,
J.

is
must be remembered that Eqs. (13.87) apply throughout all regions

of the reactor and that the integrals of (13.88) extend over the "extrapo
lated volume."
1951.
REACTOR ANALYSIS
788
[CHAP.
13
trum j) have been specified as general functions of the spatial coordinates,

Eqs. (13.89) and (13.90) apply throughout a multiregion system.
The analysis proceeds by putting the age-diffusion equations into
operational form:
Oix = "0*K
the matrix operators
wherein we have introduced
-VDvir+zS
*2'
- J~ ( - Of-
Oi =
(13.92)
- I* (V
du
Z),hV
+ 2*){-
du
- *)J
(13.95)
du signifies that the quantity operated upon
to be put inside the braces and integrated over lethargy, with g{u) as
(13
Jo
#()
is
The notation
weighting function.
Next we multiply through Eq. (13.92) by an as yet unspecified matrix
and over the entire lethargy
integrate over the extrapolated volume
range, and solve for v.
du
(13.96)
dl
lv
jo"
_ Io"du fv*?0*dt
{K*(01
"
dl
(13 97)
now defined by
LxWdu
taken as the row matrix

JCf
is
xf
If
'
du
- ^(r.t*)
(13.98)
fO0x*
sk
vOid
--- - -
/o
Olst
&v
as usual.
du
fv
lo
J-L"
Then,
xt(Cl ~ "0j)5x dl
where
xf
is
du
lv
Now perform the general variation of parameters and functions in Eq.

(13.96) by applying the delta operator, as used previously in this chapter;
and, again, choose operators adjoint to Oi and O2 such that
/o"

Um
d|
q{t,u)
2)"(r){-
û)
/Q"
du
T(r,t)(0
J()
/"
(13.93)*
Jv Kt0*
,k)]
df
(13.99)*
(13.100)
THEORY
PERTURBATION
sec. 13.6]
789
then the equations for the adjoint functions gf and

be [using conditions (13.97) and (13.98)]
-V
D(r>M)
5(u
^(r,u) + 0(r,)
gt(r|)
- uj^r) + ry(r,)
AhV^(r)
Zf*t(r)
rZjNr)
_A
u ?t(r )
jT"
may be shown to
j(') gt(r,u') du'
(13.101)*
^"
" WlÂ
2<(r,u)
#f
j(u') gtfr.u') d'
(13.102)*
DVg+, and Du.Vtf)* must be everywhere nonsingular and con

With the functions
and
calcu
and gf(&) = <*(&) =
Eq.
of
coordinates
and
lethargy,
lated as functions
the spatial
(13.99)
may be used to calculate the effect of any (small) change in parameters.
Note that the pair (13.101) and (13.102) are "backward" equations
q1
<f
4>f,
0.
and gf,
tinuous,
integral
Jo
in just the sense discussed for the infinite medium Eq. (13.37).
The
general procedure would be to choose a normalized space form for the
j(w') qf(T,u') du'
and
proceed
to
obtain
from Eq.
>f(r)
is
is
"up"
is
(in energy) to u =
[note that as in Eq. (13.37) there
in sign of the lethargy derivative in (13.101) as compared
a change
with the
"forward" Eq. (13.89)].
Application to Multigroup Method. The entire scheme described

readily adapted to the multigroup technique discussed in Sec. 8.8.
The adjoint set (13.101) and (13.102)
very like the forward set (13.89)
and (13.90).
In similar manner to that used in Sec. 8.8, the adjoint
equations may be converted into multigroup equations convenient for
numerical integration.
Further, once the forward calculation
set up,
the adjoint equations are easily solved since the constants involved are
b.
is
is
above
is
gf
the same, and the same network in lethargy and space variables may be
used.
Thus the calculation of the importance functions
and tf, in
this case, add very little to the time and effort already invested in obtain
In fact, the second calcu
ing the slowing-down density and the flux.
as
lation may serve
the same eigenvalue

This "adjoint thermal flux" </f(r) acts as

source for begin
(13.102).
ning the calculation of the "adjoint slowing-down density"; the appro
u = w,h, and the integration
carried
priate starting point in lethargy
numerical check on the forward calculation since

should be obtained in both cases.

Adler, F. T., 681, 740

Alexander, J. H., 760
Anthony, G. W., 698
Antonov, A. V., 557
Arnette, M. R., 407
Avery, R., 743
Copson, E. T., 189

Corinaldesi, E., 260
R, 110, 135, 137, 141, 760
Coveyou,
Cowling, T. G., 127
Cramer, H., 34
Creutz, E., 639, 678
Bakshi, P., 672

Bartels, W. J. C, 252
Bate, R. R., 110, 135
Behrens, D. J., 699
Bethe, H. A., 678
Biedenharn, L. C, 759
Biot, M. A., 124
Bitter, F., 39
Bj0rnerud, E., 646
Dancoff, S. M., 83, 639, 666, 688, 691

Dardel, G. von, 557
Davison, B., 487, 724
E., 642
Dayton,
Dikarev, V. S., 642
Dirac, P. A. M., 375
Dismuke, N. M., 407
Dresner, L., 320, 636-637, 640, 670, 672,
Blatt,
Dwork,
J. M.,
2, 6, 13, 70, 147, 317
I.
678, 683, 687
J.,
253
Edlund, M. C, 690
Eggler, C, 637
Egiazarov, V. B., 642
Ehrlich, R., 521
Ergen, W. K., 601
Feinberg, S. M., 628, 704

399
Fermi, Enrico,
Feshbach, H., 163, 317
Feynman, R. P., 486-487, 512
Fleck,
A., 592, 601
Frankel, S., 180
Frenkel, J.,
Case,
K. M.,
Campbell, E. C, 557
Carlson, B., 441, 446
Carlvik,
689
Carslaw, H. S., 141
Carter, R. S., 697
J.
2,
I.,
Blosser, T. V., 285

Bohr, N., 2, 5
Bonilla, C. F., 3, 7, 9
Born, M., 681
646
Bornstein,
Bo the, W., 260
Breit, G., 83
Brown, L. J., 675
Burgov, N. A., 642
I.,

NAME INDEX
249, 367
Chapman, S., 127

Charpie, R. A., 566
Chernick, J., 633, 638, 692, 699
Churchill, R. V., 203
Clancy, E. F., 253
Codd, J., 740
Condon, E. U., 83
Galanin, A. D., 628, 704, 705

Garabedian, H., 322, 722, 743
Carrick, B. J., 760
Gast, P. F., 738
Ginsburg, M., 639, 666, 688, 691
Given, N. D., 760
791
REACTOR ANALYSIS
792
Monk, A. T., 681

Morse, P. M., 163
Muanker, W., 681
Murphy, G. M., 119, 199
Murray, R. L., 735
I.,
Glasstone, 8., 690

Glauber, R. J., 145
Goertzel, G., 317, 750
Goldstein, M., 446
Goldstein, S., 593
Guggenheim, E. A., 659
642
Gurevich, I.
J.
724
Ornstein, L. S., 399

Osborn, R. K., 110, 135
Pershagen, B., 689

Pettus, W. G., 642, 643
Placzek, G., 249, 367, 678
Pomerance, H. S., 636
Y., 642
Pomeranchouck,
Prandtl, L., 593
Prohammer, F. G., 471, 609
Pruett,
R., 675
Pryce, M. H. L, 659
699
Kaplan,
Karman, T. von, 124
Keepin, G. R., 566, 568
King, L. D. P., 620
Kinney, W. E., 141, 760
Klema, E., 260
Kushneriuk, S., 724
Lamb, \V., 145

Leak, P., 681
Lessig, R. H., 638
Lewis, W. B., 620
Littler, D. J., 659
Maoklin, R. L., 636

Madecv, V. G., 642
Margenau, H., 119, 199
Marshak, R. E., 399, 406
E., 125, 130
Mayer,
Mayer, M. G., 125, 130
Meghreblian, R. V., 150
Messiah, A., 145
Mitchell, A. C. G., 675
Raffle, J. F., 659

Ramanna, R., 557
Rennie, C. A., 740
Richtmyer, R. D., 684
Ritchie, R. H., 260
Roe, G. M., 678, 681
Rose, M. E., 681
Rosenthal, L., 681
Rothcnstein, W., 686
Royston, R. J., 252
Sacks, R. G., 145

B., 633, 692
Sampson,
E., 144, 315, 633
Sanders,
Scalettar, R., 724
Scott, F. R, 557
Seitz, F., 628
Selengut, D., 750
Silver, E. G., 761
J. J.
109
I.,
Kahn, H.,
J.
I.
Jaeger, J. C, 141
Johnson, M., 646
Jupnik, H., 639, 678
J.

5,
Halliday, D.,
8-9, 13
Halperin, J., 620
Harvey, J. A., 143-144, 315
Hellstrand, E., 642
Hendrikson, J. K., 681
Hildebrand, F. B., 76, 184
Hinman, G. W., 681
Hoffmann, F. de, 249, 367
Holmes, D. K., 150
Hughes, D. J., 143, 317, 633, 637, 697
Hurwitz, H., Jr., 253, 521, 757
Nelson, E., 180

Nering, E. D., 675
Newmarch, D. A., 656-658
Nielsen, M. J., 787
Niestlie,
W., 735
Noderer, L. C., 349, 638
Nordheim, G., 407
Nordheim, L. W., 407, 646, 665, 681,
NAME INDEX
Simon, A., 759
Siostrand, N. O., 557
Sneddon, I. N., 140
Snyder, T., 639, 678
SokolnikofT, E. S., 65
Sokolnikoff, I. S., 65
Sommerfeld, A., 200, 337
Soodak, H., 407
Spinney, K. T., 686
Stein, J. M., 546
Stelson, P. H., 557
Stoughton, R. W., 620
Sullivan, J. G., 760
Sweeney, D., 446
J. H., 657
Teller, E., 145, 704
Thie, J. A., 688
Thomson, D. B., 557
Tietjens, O. G., 593
Tittle, C. W., 260
Trubey, D. K., 285, 638
Tsien, H. S., 609
Tait,
Uhlenbeck, G. E., 399
793
Vernon, R., 633, 638
Wade, J. W., 285

Ward, A. G., 615
Watson, G. N., 280, 708
Weinberg, A. M., 349, 601, 638, 686, 690,
692
Weisskopf, V. F., 2, 6,
13, 70, 147,
317
Welton, T. A., 486-487, 512, 601, 759

Westcott, C. H., 144
Wheeler, J. A., 681
Whit-taker, E. T., 708
Wigner, E. P., 628, 633, 638-639, 643,
669, 678, 686, 690-692
Wilson, A. H., 382, 386, 441
Wimett, T. F., 568
Wright, W., 557
Yamauchi,
H., 317
Zemach, A. C, 145
Zimmerman, E. L., 726
Zweifel, P., 757


SUBJECT INDEX
Absorption cross section, values for
moderators, table, 165
Absorption probability, 101
Activation foil, 229
Adjoint function, 768
for age-diffusion model, 789
conditions on, 782, 786
for one-velocity model, 782
orthogonality of, 770
to slowing-down density, 774
Adjoint operator, 768, 770
for age-diffusion model, 787
for Boltzmann equation, 771
self-adjoint, 769, 782, 786
for two-group model, 786
Age-diffusion theory (see Slowing-down
diffusion theory)
Albedo, definition, 193-195
determination of, 195
general expression for, 196
parameters involved in, 196
reflector properties in terms of, 418420
in semi-infinite medium, 196, 420

of spherical shell, 427
used in determining neutron crossings
of interface, 197-198
Angular delta function, 368
"As good as new" hypothesis, 33-35
Atomic mass unit, 3
Average behavior, 31
Average chord length, 376
Average path length, 249
Avogadro's number, 47
Barn, 41
millibarn, 46
Barrier penetration, 6
Base temperature, 312
Beryllium and beryllium oxide, thermal
neutron properties of, table, 165
for, in case of
small argument, 735
modified, 184, 280, 708
regular, 184, 207, 708
zeroes of, 207
Beta rays, 9
Binding energy, average per nucleon, 3, 7
chemical, 78
total, 2-3, 7
Boltzmann equation, 22, 25, 28, 269, 400
application to hydrogenous systems,
Bessel function, expression
752
assumptions and limitations of, 352

first-order approximation to, 360-364
Fourier transform of, 402
isotropic scattering assumption in,
358
neutron balance, 353

Pi approximation, 756
reduction in consistent Pi approxima
tion, 758
transport, 354
Boltzmann 's constant, 111
Bound nuclei, hydrogen in water, 146147 (Fig.)
(See also Molecules)
Boundary conditions, for clad fuel ele
ment surrounded by gap, 654-660
comparison of extrapolated and noreturn current condition, 191
continuity-of-flux limitations, 440
general discussion, 174
at interface between two media, 175,
350, 390
in reflected reactor, 423

in slowing-down diffusion model, 421
time-dependent, 547
for transport theory, 349-350, 383, 390
at vacuum interface, 175, 189, 349
Boundary-value problem, 27
Breeding, 71
Breeding ratio, 24
REACTOR
796
Breit-Wigner resonance, 141, 319, 636

with Doppler broadening, 678-681
evaluation of resonance integrals with,
673
formula, 673
Brookhaven graphite reactor, 699
Buckling, approximation for, in twogroup theory, 480-481
in completely reflected reactor, 479483
for Feynman-Welton method, 517

in one-velocity reflected reactor, 437
in pulsed neutron-beam experiment,
558

Burnout poison, 610,
621
Cavities and gas passages, 698

Chain-reacting system, 14, 29, 53
neutron cycle in, 293
self-sustaining, 50, 55
Chemical bond, effect of, on scattering
cross section, 147
natural frequencies of, 146
Christy-Wheeler kernel, 411, 413
Cladding, fuel element, effect of, on dis
advantage factor, 660
on thermal utilization, 655-660
Coefficient of thermal expansion , 3 1 1
Collision, average number per unit path
traveled, 37
billard-ball type, 73
expected number, 32
scattering, 15
Collision density, in heterogeneous cell,
665-668
in NR approximation, 667
in NRIA approximation, 667
total, 103, 633
for hydrogen, 51, 66, 86, 98-100
in two-component mixture, 664-668
Completely reflected reactor, one-group
model,
476
two-group model, 479

Compound nucleus, 2-3, 7, 12
excited, 69-70
first excited state, 70
ground state, 70
Consistent Pt approximation, 758
Constant power distribution, 264
Continuous slowing-down model, 86,
272-273, 278
Control effectiveness, 22-23
ANALYSIS
Control rods, black, 722
effect on net neutron leakage, 727
effective radius of, 724
Feinberg-Galanin method for, 705
gray, 722
shim, 25
in two-group theory, boundary condi
tions in, 728-729, 732-733, 740
central black rod, 728-732
eccentric black rod, 738-743
flux distribution for, 730, 731 (Fig.),
736, 737 (Fig.)
general equations for, 727
hollow central rod, 732-737
ring of black rods, 743-747
worth of, 726
xenon-override, 614
Converter plate, 244
Converting, 71
Coordinate system, center-of-mass, 72,
359
laboratory, 13, 72
Core, 14
Coulomb interaction, 8
Coupling equation in age theory, 287,
364-365, 421
CP-5 Reactor, 321-323
Critical experiments, 23-24, 557
Critical fuel concentration, 58
in diffusion theory, 212
in Feynman-Welton method, 491, 519
in reflected reactor, 419
Critical mass, 22, 213
calculations, 299, 321
Critical size, 299
comparison of, from Serber- Wilson and
diffusion theory, 452 (Fig. 8.18)
Criticality condition, in age theory, 292
in diffusion theory, 209
table, 217
in infinite homogeneous medium, 55,
121
in one-velocity cylindrical reactor,

428-432
in one-velocity spherical reactor, 424
in reactor, with black control rod, 725
with central hole, 726
with gray control rod, 723
in transport theory in Pa approxima
tion, 396-397
Criticality determinant, evaluation of, 469
sample computation of, 472-474
Cross section, absorption, 44, 165
SUBJECT INDEX
Cross section, for compounds, 48
correction to, for non-l/v variation,
144-145 (Fig. )
temperature, 137
effective (see Effective cross section)
escape, 222
fission,
42
inelastic, 41
macroscopic, 36, 46-48
measured,
133, 135
measurement of, 44
microscopic, 40
radiative capture, 42
reaction, 696
scattering, 42
slowing-down, 508
thermal neutron values for, table, 46
total, 44, 72
transport, 164, 697
Crow-flight distance (see Root-meansquare radial distance)
797
Diffusion theory solutions, by kernel

method, 185
for line source in infinite medium, 183
for multiplying block, 213
for multiplying finite cylinder, 215
for multiplying infinite cylinder, 207
for multiplying slab, 198
for multiplying sphere, 205
for plane source in infinite medium, 180
for point source in infinite medium, 182
Dirac delta function, 123, 187, 286, 359
Directed flux, 331, 367, 374
first-order approximation for, 360-361
harmonics of, 331, 339, 357-358, 391
at interface between two media, 393
(Figs.)
in multiplying medium, 378
in purely absorbing medium, 371
at slab surface, 384 (Fig.)
in zero cross section medium, 369
zeroth-order harmonic, 345-347, 363-

364
Dancoff-Ginsburg correction, 688

Decay channel, 70
Decay constant, 37
Decay scheme, 38
Delayed-neutron groups, 566
damping effect in reactor kinetics, 588589
spectrum of, 569

fraction from fission of various nuclear
fuels, 566-568
tables, 567, 568
Delayed-neutron precursors, 566, 569
Density reduction factor, 58
Deuterium, 26
Difference equation, 35, 124
Differential solid angle, 170
Diffusion of gases, Kick's law of, 163
Diffusion coefficient, 164
comparison of transport- and diffusiontheory expressions for, 173
for gases, 164
transport-theory result for, 342
units of, 164
values of, table, 165
Diffusion length, 181
(See also Thermal diffusion length)
Diffusion theory solutions, for elementary
sources in finite geometry, 193
table, 194
energy
Disadvantage factor, 630-631, 646

for clad fuel rod in gas channel, 656657
from diffusion theory, 648

error in, from one-velocity model (dif
fusion theory), 646
as function of cell size, 649 (Fig.)
for thin plates and pins, 651-653
Distribution function, 36
for elastic scattering, 80
for first collision, 33
for mixture of nuclear species, 48
for radioactive decay, 37
Dj() (see Heavy water)
Doppler effect, 309-310, 316-317
(See also Effective resonance integral,
effect of nuclear motion on)
Effective cross section, absorption, 244,

443, 447, 631, 635
for lumped fuel, 639

removal, 379, 443, 497
with thermal motion of nuclei, 318
Effective neutron temperature, 135-137,
316
sample calculations, 232

Effective resonance integral, 634, 639, 688
dependence of, on square root of fuel
surface-mass ratio, 642-653, 676678
REACTOR ANALYSIS
798
Effective resonance integral, effect of

nuclear motion on, 678, 682
geometric dependence of, 676 (Fig.)
for heterogeneous systems, 641-643
for homogeneous mixtures, 637 (Fig.),

668
for infinitely dilute mixtures, 675

l/t> contribution to, 687
temperature dependence of, for U"8
and Th 688-689 (Fig.)
table, 688
for thorium lumps, 642-643
for thorium mixture, 637
for UO, lumps, 642-643
for U3Os lumps, 642
for uranium lumps, 641, 643
for uranium mixtures, 637
using Breit-Wigner formula, 675
Eigenfunctions, 560
for bare reactor, 217, 524, 776-779
table, 778
expansion of integral-theory kernels in,
493-494
for one-velocity integral-theory, 492496
for slab reactor, 200

Eigenvalues, 560
for one-velocity integral-theory, 492496

for spherical reactor, 207-208
Einstein mass-energy equation, 4
Electron volt, 2
Electrostatic repulsion energy, 4
Elementary sources, in diffusion theory,
180-185
in slowing-down diffusion theory, 275281
End
228-229, 305
Energy groups, 421
(See also Multigroup method)
Energy loss, 69
from stationary nuclei, 74
Energy space, 30
Energy spectrum, neutron, 19, 69-71
Equivalence of homogeneous and hetero
geneous systems, 640, 671
accuracy of, 672
based, on flat-flux assumption, 673
on rational approximation to escape
effects,
probability,
672
Equivalence of homogeneous and hetero

geneous systems, in narrow resonance
(NR) approximation, 672
in narrow resonance-infinite absorber
(NRIA) approximation, 672
practical rules resulting from, 673
Equivalent bare reactor dimension, 420,
471, 721
Error function,
304
Escape probability, application of, to

calculation of fast effect, 694
to unit-cell analysis, 666, 669-670
table, 670
from arbitrarily shaped volumes, 372377
average,
373
by chord method, 375-377

in elementary geometries, 671 (Fig.)
Wigner's rational approximation for
669-670, 672, 689
Eta (>;), definition of, 314
temperature coefficient of, 314
Even-odd term, 6
Excitation energy, 2
Exponential attenuation of flux in nonmultiplying media, 228
Exponential experiment, 27, 557
Exponential integrals, 249, 386
recursion relation for, 716
Extrapolated boundary condition, in age
theory, 288-289
compared with no-return current con
dition, 191
use at corners, 214
Extrapolated outer boundary, notation
for, 180
use at corners, 214
Extrapolation distance (length), 179-180

in age theory, 289
transport theory result for, 180
used in control-rod analysis, 724, 730,
733
"Eyewash"
code, 760
Fast effect, 153, 644

in age theory, 292-293, 693
effect of inelastic scattering on, 644,
693
fast-neutron cross sections for, 694,

696-697
table, 697
by multigroup method, 693
SUBJECT INDEX
Fast effect, produced by cascade, 693696
significance of, 692

by two-group method, 693
Fast-fission factor (see Fast effect)
Fast fissions, 152, 293
Fast-neutron cross sections for fast-effect
calculation, 694
Fast-nonleakage probability, 293, 297,
301-302, 559, 565
effect of temperature on, 313-314
Fast-removal cross section, definition of,
457-459
Feinberg-Galanin method, 632, 644
advantages of, over unit-cell method,
629
application to reactor with small num

ber of fuel lumps, 715-718
assumptions in, 705
comparison, with two-group method,

720
with unit-cell result, 714

criticality condition for, 709-710, 716
Fermi age model, use in, 707
neutron balance for, 705, 707
proof of homogenization method, 710714
Fermi age, 240, 275-281

and Christy- Wheeler method, 411
effect on, of cavities and fuel lumps, 701
of inelastic collisions, 702
by equivalent mean-free-path method,
407
for fission neutrons, 416

of fission neutrons in various moder
ators, table, 285
influence on reactor size, 285-286
for metal-water mixtures, 399, 407-408
physical significance of, 283
relation to slowing-down length, 284,
749
temperature dependence of, 313-314

to thermal, 286
values for beryllium and graphite,
table, 415
Fermi age model, 26, 141, 275, 365-366
application of, to bare reactor, 286300
to problems with elementary sources,

275
derivation of general equation, 273

limitations of, 366
numerical integration of, 521
799
Fermi age model, use of, in kinetics

problems, 547, 551
in reactor calculations, 299
with fission sources, 295
with monoenergetic sources, 286
Fermi- Marshak method, 331, 399-407
Feynman-Welton method, application to
thermal reactor with moderating re
flector, 511, 518
criticality condition for two-group
model, 490
lower approximation in, 489, 506, 516
summary of, 487-491
two-group model with moderating re
flector, 511
upper approximation in, 505-506
Fick's law, 163
First-flight losses, 151, 647
First-flight transport model, 247-248
application to thin plates and pins, 651
for treating thin absorber lumps, 646
First-scattering sources, 237-243
Fission, energy release per, 236
fast, 693
heating, 260
process,
sources,
119,
199,
218
spontaneous, 6
in Th", 693
in U"8, 693
Fission energy, average, 19
distribution for U"s, 22
table, 9
Fission fragments, 3, 8
beta decay of, 566
secondary, 620
Fission products, buildup, 614
in constant-flux reactor, 617
in constant-power reactor, 618
decay of, 610
following reactor shutdown, 614
I1",
611, 613
Ndl4, 612
Sml",
611, 613
steady-state concentrations of, 613

Xe1", 611, 613
xenon override, 614
Fission spectrum, distribution of neutron
energies,
286, 295
example (4.7), 158

histogram, table, 416
normalized in lethargy, 119, 122 (Fig.)
UJ", 120 (Fig.)
REACTOR ANALYSIS
800
Flat-flux assumption, 665-666, 689-690

Flux, angular dependence of, 330
average for parallelepiped, 236
depression,
253
from elementary sources, line, 185

plane, 182
point, 183
two-point sources, 185
expansion in Taylor series, in lethargy
variable, 103, 362
in spatial coordinates, 170
harmonics of, 331
l/E variation of, 98, 136, 144
peak-to-average ratio, 218
per unit lethargy, 27
units of, 67
Flux correction factor, 244
Flux depression factor, near absorber foil,

260
definition of, 255-256

for resonance neutrons, 644, 660-661
(Fig.)
for thermal neutrons, 645
Flux slowing-down density relation, in
absorbing medium, 104
in scattering medium, 98
(See also Coupling equation)
Flux transients, decay constant for, 559,
562 (Fig.)
in nonmultiplying medium, 563 (Fig.),
576 (Fig.)
in subcritical multiplying medium, 559,
564-565 (Figs.)
Foil, end effects in, 260
Four-factor formula, 154, 627
Fourier transform, 138, 402, 707
of Boltzmann equation, 402
convolution theorem for, 139
of flux, 413, 707
inverse of, 139, 405
moments of, 405
three-dimensional, 138
Frequency function, 36
for absorption, 225
for azimuthal scattering from station
ary nuclei, 355
conditional, 34, 355, 359
for elastic scattering, 76, 335
for first collision, 33
first moment of, 36
isotropic, in laboratory coordinate sys
tem, 378
in mixture, 48
Frequency function, for radioactive

decay,
37
for scattering, in hydrogen, 99

in hydrogenous mixtures, 404
Fuel, burnup, 22, 24, 620
critical concentration of, 299-300
Fuel enrichment, 18, 718-719
Fundamental mode, for one-velocity in
tegral theory, 500
for slab reactor in one-velocity model,
198, 205
for spherical reactor, 206
Gamma ray, 2
internal conversion process, 39
prompt, 9
Gaps and gas channels, effect of, on age,
701
on thermal diffusion length, 698701
tables,
700, 701
on thermal utilization,
654-660
Gauss' theorem, 782

Gaussian, 130-131
sources,
647, 707
Gaussian kernel, 413

Goertzel-Selengut equations, 751, 760762
from Boltzmann equation, 757

Graphite, 19
thermal neutron properties of, table,
165
Green's function, method of,
181
Half-life, 38
Hankel function of first kind, 184, 189
Harmonics (see Spherical harmonics)
Heat-conduction equation, 24-25, 141,
273-274
Heavy water (DaO), 19
165
Helmholtz equation, 25-26, 381

Heterogeneous fuel-moderator systems,
equivalent homogeneous system,
640, 671-673
merits of, 626-628
Monte Carlo analysis of, 684
of natural uranium, and graphite,
626-627, 646
and heavy water, 646
SUBJECT INDEX
Heterogenous method, advantages of,
704
for reactors with small number of fuel

rods, 704
(See also Feinberg-Galanin method)
Heterogeneous reactors, 17-21
characteristic parameters of, 629

with irregular or multiply periodic
lattice arrays, 629
methods of analysis, 628-629

H,0
(see Water)
Homogeneous fuel-moderator mixtures,
equivalent to heterogeneous, 640
infinitely dilute, 634, 637-638
table, 638
of natural uranium and graphite,
626-627 (Fig.)
resonance integrals for, 637 (Fig.)
table, 638
Homogenized system, 62
Hydrogen, scattering from, 146-147
scattering cross section in water, 147
Hydrogenous systems, 123
age calculation in, as check on theory,
749, 759
methods for treating, 748

reactor calculations in, 759
special treatment for hydrogen, 748
Importance function, 767

for infinite multiplying medium, 775
one-velocity, 767-768, 783
for one-velocity bare reactor, 780
for one-velocity slab reactor, 768
Infinite-medium model, 26, 71
criticality equation for, 121, 151
multiplication constant for, 55
multiplying, with fission sources, 119
with monoenergetic sources, 119
in one velocity, 51
Infinite-slab geometry, flux distribution,
from plane source, 189
isotropic, from collimated beam, 240
Infinite-slab reactor, asymptotic flux in,
202
in one-velocity diffusion theory, 198

Inhour, 571
Inhour equation, for circulating fuel
reactor, 591, 598
for stationary fuel reactor, 571
Integral transport theory, directed flux,
381
801
Integral transport theory, elementary

solution for infinite multiplying
medium, 380
general solution for infinite medium,
380
International Conference on Peaceful

Uses of Atomic Energy, Geneva,
1955, 683
Isotropic scattering, in center-of-mass

system, 77-78, 80, 84, 110, 270, 358360, 404
in laboratory system, 168, 378
Jacobian, 65, 80, 358
Kernel, Christy-Wheeler, 411, 413

for one-velocity integral theory, 491496
for plane source in nonmultiplying slab,

241
transport, 413, 697

(See also Feynman-Welton method)
Kernel method, 186, 491
application, to one-velocity model,
with line source, 188
with plane source, 186, 303
to slowing-down diffusion model,
with distributed sources, 281
with line source, 282
used with first-scattering source, 237
(See also Green's function)
Kinetic theory, 25, 125
Kinetics of circulating-fuel reactor,
asymptotic solution for, 606
effect of circulation velocity on reac
tivity, 601
effective temperature of, 606
excess reactivity of, 600
flux-temperature coupling without de
layed neutrons, 601
fraction of precursors in core, 599 (Fig.)
for low-power systems, 591
open-loop circuit, 602
slug-flow assumption, 591
transit time of circulation, 592, 595
Kronecker delta, 338, 493
Laplace transform, 276-281, 305

inversion of, 277, 281
Laplace's equation, 261
REACTOR ANALYSIS
802
Laplacian operator, 141, 167

Leakage probability, 222
Legendre functions, associated, 337
orthogonality of, 753
recursion relation for, 340
Legendre polynomials, 76, 337
Lethargy variable, 81
average gain per collision, 81-83, 123
Level spacing, 70
Liquid-drop model, 5
Lowest mode (see Fundamental mode)
Lumped fuel, over-all effects of, 628, 643
resonance behavior of, 638
self-shielding effect of, 643
Manhattan Project, 646

Marshak boundary condition, 349, 383384
Matrix, column, to represent
fluxes, 785
row, to represent adjoint fluxes, 785,

788
transposed, 786
Matrix representation of operators, in
age-diffusion model, 788
in two-group model, 785
Maxwell-Boltzmann
distribution, 110,
130-135
speed for, 142
most probable speed for, 150
Maxwell-Boltzmann
flux, 136
Mean free path, 36
for removal reaction, 459
Mean square radial distance, first-flight
corrected, 239
for Gaussian slowing-down source, 240
relation of, to neutron slowing down,
average
749
to
401
"Medusa"
code, 760
Method of images, 305

Migration area (length), in anisotropic
reactor, 703
application of, 467
effect of cavities on, 698, 703
in heterogeneous reactors, 645, 698
for large reactors, 300, 440
Minimum critical volume, 217
Minimum fuel mass, 217, 299-300
Miss probability, diffusion-theory esti
mate of, 657
for fuel lumps in cavity, 655-658
Miss probability, for hollow rod in

cavity, 658
Moderating effectiveness, 117
Moderating power, 1 16
Moderating ratio, 116
table, 117
Moderator, 15
materials, 17
Moderator temperature, 111, 136
Molecules, rotational and vibrational
states,
146
scattering from, 145

Moments, 405
Monoenergetic beam, 67, 134
Monte Carlo method, 109, 135
analysis of heterogeneous systems, 684
application to collision-density calcu
lation, 109-112
procedure, 109
resonance escape probability by, 109
thermal neutron spectrum by, 135
MTR-type fuel elements, 231-232, 321,
470, 483, 733
Multigroup method, 421, 520

adapted for numerical computation,
528
boundary conditions in, 534-536

flux distribution by, 542 (Fig.)
group constants for, 522, 532, 537539 (Figs.)
integration over lethargy intervals, 520
iteration for, 528, 530-533
multiplication constant by, 532-533
removal cross section for, 523
simultaneous equations for, 523-526
spatial integration of group equations,
533-536
using characteristic functions, 527-528
using harmonics, 526-527
Multiplication constant, 56
in age theory, 293-295
for bare reactor in diffusion theory, 211
effect of temperature on, 310
effective,
211
in heterogeneous reactors, 644

infinite homogeneous medium for, 122
in infinite one-velocity model, 21 1
multigroup method for, 531-532
Serber- Wilson method for, 454-456
Multivelocity reactor, with energydependent cross sections, 501-505
with energy-independent cross sections,
496-501
SUBJECT INDEX
Narrow resonance approximation (NR),
633-635, 664, 684-687
system, col
use of, in heterogeneous
lision density, 667
effective resonance integral, 670,
675
in homogeneous system, collision

density, 667, 690
675, 682
Narrow resonance-infinite absorber ap

proximation (NRIA), 633, 635, 684system, col
lision density, 667

677
homogeneous
550
in infinite medium, 222, 550

Neutron net current, definition, 163
divergence of, 167
expressions for, 173
Fick's law approximation for, 162,
transport-theory result for, 343
vector current, 166-167, 344-345
Neutron-nucleus interaction function,
132-134
system, collision
density, 667
effective resonance
relation to measured cross section,

133
integral, 669,
677, 682

Neutron diffusion, analogy to gaseous

diffusion, 164
equation, 26
linearity of, 186
Neutron flux, 72, 161
Neutron life time, in finite medium, 222,
270-271
687, 690, 692

use of, in heterogeneous
in
803
Net neutron leakage, 162, 166-167

effect of temperature on, 310
in heterogeneous reactors, 645
from outer boundary, 178, 365-366
Neumann series, 378
Neutrino, 8-9
Neutron, balance equation, 161-162,
167, 271, 331, 353, 456, 488, 569
cross sections, 72
delayed, 9, 566
tables, 567, 568
density function, 22-24, 28, 53, 64, 72,
161
energy spectrum, 19, 69-71

in unit cell, 660, 691
fast, 52, 134
generations of, 56
kinetic energy of, 64, 73
life history, 14
moderation of, 69-72
photoelectric liberation of, 2
production cycle, 154, 293
prompt, 9, 549
reduced wavelength of, 674, 680
thermal, 74, 126
total track length of, 64, 72, 248
yield, per absorption, 57
per collision, 377, 445
per fission, 11, 13
as adjustable parameter, 210
table, 11
per removal, 503
Neutron partial current, angular dis

tribution of, 258, 260
application to unit-cell calculation,
652, 655
definition, 167-168
in diffusion approximation, 174
Neutron sink, 23, 610
Neutron temperature (see Effective neu
tron temperature)
Neutron wave phenomena, 279, 351
No-return current condition, 176
application of, to slab, 190, 383
to sphere,
386
transport-theory equivalent, 383

Nonabsorption probability, 100
(See also Resonance escape probability)
Nonleakage probability, 58, 218
in age theory (see Fast-nonleakage
probability)
in one-velocity model, 222
Nonproductive capture, 12, 17
Nordheim-Scalettar method, 724, 743
Nuclear accidents, 22
Nuclear densities in mixtures, 47
Nuclear forces, 29
Nuclear fuel, 6
mass calculation, 28
Nuclear mass formula, 6
Nuclear reactions, effect on neutron
economy, 11
Nuclear reactors, applications, 20
classification, 17
(See also Reactor)
Nucleons,
3, 7
804
REACTOR
Nucleus, billard-ball model, 40

effective target area, 41
recoil of, 72
thermal motion, 309
Perturbation
One-velocity model, 26, 160-161

application to flux calculation around
244, 253
boundary conditions to be used with,

174-180
complete set of equations, 174
integral theory, 491-496
time-dependent, 198-205, 273-281,
One-velocity transport equation, 335
application to one-dimensional prob
lems, 345-349
asymptotic solutions to, 442, 445-446
boundary conditions to be used with,

349
diffusion approximation for, 341

time-dependent, 350-352
Optical thickness, 370
Optimum proportions of bare reactor,
216-217
ORACLE, 484
ORNL Research Reactor, 483-485
ORNL Swimming Pool Reactor, 480,
485-486
ORNL
X-10 Graphite Reactor,

Orthonormal set, 336, 493
completeness of, 493
18
approximation, application to hydro

genous
for infinite multiplying medium, ad

joint equations, 772-774
adjoint slowing-down density, 774
for multigroup method, 789
operators, for age-diffusion model, 788
for one-velocity model, 764
variation in eigenvalue (multiplica
tion), for age-diffusion model, 788
for infinite multiplying medium,
774-775
for one-velocity model, 766, 778, 780,
783
352
Pi
theory, general methods,
764
Nyquist's criterion, 609
absorbers,
ANALYSIS
systems,
756
comparison with P3 approximation,

392-399
P,
approximation, 350, 387-388

Partial current, relation to directed flux,
347-349
(See also Neutron partial current)
Penetrability, 669, 691

relation to transmission coefficient,
669
Pershagen-Carlvik correction, 688

Perturbation from diffusion coefficient
changes, 783-784
effect on net leakage, 784

Plutonium, 7, 21, 143
Poisons, buildup after shutdown, 614
burnout, 610, 621
steady-state concentrations, 613
Poisson's summation formula, 711
Position vector, 161
Power production, 20
Probability function, 41
conditional, 34
independent, 40
joint, 35, 40
Prompt critical, definition, 575
Pulsed neutron beam, 223
diffusion coefficient from, 564
experiment, 556
reactor parameters, 556, 561
Radiation, beta, 9
electromagnetic, 12
Radiation damage, 244
Radiative capture, 12-13
Radioactive decay, constant, 37
by emission, of alpha particle, 38
of beta particle, 38
of e~, internal conversion of gamma
ray, 39
of gamma ray, 38
of K, orbital electron capture, 38
half-life, 38
scheme,
38
Radioisotope, table, 20
Reaction, absorption (capture),
deflection, 13
deuteron-deuteron, 557
fission, 31
14, 31
SUBJECT INDEX
Reaction, between neutrons and station
ary nuclei, 134
proton-triton, 557
scattering, 31
Reaction rate, 134, 137, 678
for 1 /v absorber, 140
Reactivity, 310, 571
maximum excursion of, 632
Reactor, breeder, 21, 692
central station power, 20
circulating fuel, 24
CP-1, 18
CP-5, 321
epithermal, 19
fast, 15, 19
Hanford, 21
heterogeneous,
homogeneous,
18, 27
18
hydrogenous, 399, 748

intermediate, 19
multivelocity, 154, 161
package power, 20, 470, 519

poisons,
22
reflected
(see
research,
20
Reflected reactors)
stability, 22
temperature effects on, 308
start-up, 325
thermal, 19, 57, 70, 155
(See also specific reactors)
Reactor heat capacity, 578-579
effect on temperature overshoot, 588589
Reactor kinetics, in infinite delay-time

model, 576 (Fig.), 577, 581
in one-velocity bare reactor, 548,
569
in one-velocity reflected reactor, 555

problems of, 546
stable system, 583
with temperature dependence, for con
stant power removal, 484
with delayed neutrons, 585
without delayed neutrons, 582
linearized solution, 584
critically damped system, 589
maximum temperature rise in, 583,
609-610
overdamped system, 590
power production in, 580
power removal function for, 578
for stationary fuel reactor, 577
805
Reactor kinetics, with temperature

dependence,
system energy balance,
578
using one group of delayed neutrons,

573
Reactor period, 449, 453-456, 551

stable, 573-575
Reciprocity theorem, 375
Reflected reactors, one-velocity model,
cylindrical, 428
general equations, 422
rectangular, 433
spherical, 423
Reflectivity of medium, 195
Reflector, 14
advantages of, 418
extrapolated thickness of, 424
for fast reactors, 419
function of, 418
moderator materials in, 98
movable, 24
neutron moderation in, 419
thick, 426-427
thin, 424-426
Reflector savings, 418, 721
to describe merit of reflector, 419-420
for spherical reactor in one-velocity
model,
426
by two-group method, 322

Relaxation length, 224
Removal density, 488, 498
Resonance, absorption, 107, 317, 634
broad, 106-107
Doppler broadening of, 316, 320, 678,
685-687
peak, 105
practical width, 685
scattering, 107, 317, 634
sharp (narrow), 106
unresolved, 687-688
width, 106, 674
Resonance disadvantage factor, 641, 690
Resonance escape probability, 102, 105,
297, 632-643
in absorbing medium, 115
average,
153
from Breit-Wigner line shape, 673

geometric effect on, 644, 662
in heterogeneous system, 638
in hydrogen, 101
in mixtures, 108
for natural uranium systems, 639
in slowing-down diffusion theory, 278
REACTOR ANALYSIS
806
Resonance integral, 157, 634

approximate formula for, 668
evaluation of, 683
1/v contribution to, 319-320
standard formula for, 639-641, 683,
689
surface-absorption
term for, 639-641,
Scattering integral, 356

Secondary fission fragments, composite
nucleus representation, 620
Self-shielding, over-all, 246
Self-shielding factor, 246
Separation of variables, 199, 289, 550,
560, 579
Serber-Wilson condition, 177, 441, 443-
683
temperature dependence of, 318-321

volume-absorption term for, 639-641,
683
integral)
Root-mean-square radial distance, 224
for fast neutrons, 283-286
(See also Effective resonance
447
results from, compared with continuity-of-flux condition, 452

(Fig.)
Serber-Wilson method, 379, 386
application of, to multigroup model,
441
to multiregion systems, 441
Safety storage calculation, 60

Scattering, azimuthally symmetric, 355-

356
from bound nuclei, 145

deflection, 13
elastic, 13, 15, 19, 31, 69-70, 488, 497
energy loss from, average, 81
average logarithmic, 81-83
forward, 165, 270
in laboratory coordinate system, 78
inelastic, 12-13, 15, 19, 31, 69, 270,
702
interference, 673, 681

isotropic (ee Isotropic scattering)
from multiple centers, 146
p, 79, 157
potential, 13, 317

reactions, 69
resonance,
13, 317
of thermal neutrons from nuclei, 126

Scattering angle, average value of cosine,
80
in lethargy, 84
Scattering frequency function, in energy,
521
assumptions of, 268-269

numerical-integration method for,
time-dependent, 547
from transport equation, 361-364,
756
Slowing-down length, 283-284

Slowing-down time, 554-555
Slowing down track, 87, 270
Slurry, 18
Sn method, 646
Source, continuous distribution,
discrete distribution, 186
fission,
79
general,
125
in hydrogen, collision density for, 98

Slowing-down density, 364-365
with absorption, 102
asymptotic solution for, 95, 103
definition, 88, 271
in infinite medium, 97, 278
in mixtures, of absorbers, 114
of scatterers, 1 13
in pure scatterer, 85
Slowing-down diffusion theory, 269, 273,
521
78, 342
in center-of-mass coordinate system,

75-76, 359
frequency function for, 76, 79
Scattering collision density, asymptotic
solutions for, 95, 98
by collision intervals, 90
Scattering distribution function, in
energy,
to noncritical systems, 453

criticality condition, 449
discontinuity of flux from, 451-453
(Fig.)
multiplication constant, 454-456
Shield, 4
Slowing down, by average jumps, 123-
333, 354
for hydrogen, 99
in lethargy, 84
119
function, 162
monoenergetic, 119
thermal neutron, 219-221
186
SUBJECT INDEX
Spatial modes, leakage by, 496
Spatial trajectory, 30
Spectrum, thermal neutron, hardening of,
132-133
Spherical harmonics, 331, 338

complete set of, 338
complex conjugate, 338
expansion in, 336, 338, 345, 356-358
orthogonality property of, 338, 357,
403
Spherical-harmonics method, 331

application of, to interface problem,
387
to slab reactor, 395
to unit-cell calculation, 646
Spontaneous disintegration, 4
Standard formula for resonance integral,
639-641, 683, 689
Stationary wave equation, 290
(See also Helmholtz equation)
Statistical weight, 767
for changes in diffusion coefficient, 783Generated for wjivans (University of Florida) on 2015-09-22 17:24 GMT / http://hdl.handle.net/2027/uc1.b4118385
784
for infinite multiplying
medium,
776
for one-velocity bare reactor, 780

in one-velocity model, 767-768, 783
Stein model, 546
Surface-absorption term, 639-641, 683,
691
Surface-tension energy,
Thermal energy, 2, 83
Thermal equilibrium between neutrons
and nuclei, 127
Thermal fissions,
Thermal fuel, 8
131
Speed-space,
807
Telegraphist's equation, 25, 351-352

Temperature coefficient, of multiplica
tion constant, 580
parameter, 586
of reactor, 578
Temperature effects on reactor stability,
308
Theorem of detailed balance, 2

Thermal base, 309
Thermal conductivity, 261
Thermal constant, 704-705, 709
by diffusion theory, 709
Thermal diffusion length, effect of
cavities on, 698-701
table, 701
experiment, 226, 302
in heterogeneous system, 632, 698
values for, table, 165
152, 293
Thermal group, 121, 149, 311

cross section, 132
for non-1 /v absorber, 143
for l/v absorber, 142
Thermal motion of neutrons and nuclei
in equilibrium, 125
Thermal-neutron column, 302
Thermal-neutron sources, from age
theory, 438-440
in unit cell, 646-647
Thermal-neutron spectrum, 131
Thermal nonleakagc probability, 222
effect of temperature on, 312-313
Thermal utilization, 57, 631, 654
effect on, of temperature, 311-312
of voids and cladding, 654
in heterogeneous system, 630-631
in homogeneous system, 631
Thorium, 7, 21, 684-685
mixture with moderator, 637
Threshold energy, 5, 70
fast fission, 693, 696
Time absorption, 379
Total track length,
369
comparison in Pi and Pi approxima

tions, 398 (Fig.)
in multiplying medium, 378
of neutron, 64, 72, 248
in purely absorbing medium, 372
in zero cross section medium, 370
Translational motion, average energy of,
125
Transmission coefficient, definition, 245,

254
to describe fuel cladding, 655, 659

expression for, 259
diffusion-theory, 652
relation of, to albedo, 245, 254-255
to penetrability, 669
for sphere, 250 (Fig.)
for thin fuel plates, 653 (Fig.)
use in unit-cell calculation, 652-653
values of, 252 (Fig.)
Transmittance of slab, 192
Transport kernel, for one-velocity plane
source,
413
for one-velocity point source, 697
808
REACTOR
Transport mean free path, 164, 724-725

(Fig.)
Two-group theory, 421-422
comparison with one-group method,
465 (Fig.)
in Feynman-Welton model, 507-518
flux distributions by, 474-476 (Fig.)
general solutions of, 460-467
group constants for, 458-459
neutron balance equations for, 457
reflector savings by, 322
Unit-cell model
cell model)
UNIVAC,
(see
Wigner-Seitz unit-
760
Universal xenon curve, 615

Uranium, natural, 21, 626, 633, 637, 646
U">, 143

U', 143
U", 684-685
ANALYSIS
Velocity space, 132-133
Volume-absorption term, 639-641, 683
Volume fraction, 59, 155
Water (H20), 19
165
Wigner-Seitz unit-cell model, 628, 640,

644-648
approximations in, 712-714
cell radius, 650
comparison with Feinberg-Galanin
method, 629
equivalent homogeneous cell, 631
limitations of, 712-714
multiregional, 660
"Window-shade" method, 722
Yield curve,
10
(Fig.)





THIS
BOOK IS DUE ON THE LAST

STAMPED BELOW
DATE
AN INITIAL FINE OF 25 CENTS

WILL BE ASSESSED FOR FAILURE TO RETURN THIS BOOK
ppn
ON T"F
f WILL INCREASE TO
AND TO J1.00 ON THE
2m-10,'64(E9345s2)
"JOW
22
1954
M '65
RET.
OCT 7
JAN
2 6 1968
JUN
14
1966

JUN 17 IS??
Book Slip-20m-3,'60(A9205s4)458

Call Number:
Meghreblian, R.V.
Reactor analysis.
20U850
PHYSICAL
SCIENCES
WARY
204850
TK9202

Reactor Analysis (Holmes and Meghreblian)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reactor Analysis (Holmes and Meghreblian)

Uploaded by

Copyright:

Available Formats

Reactor analysis [by] Robert V. Meghreblian [and] David K. Holmes.

Meghreblian, Robert V. (Robert Vartan)

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Public Domain, Google-digitized

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

McGRAW-HILL SERIES IN NUCLEAR ENGINEERING

Jerome D. Ltjntz, Associate Consulting Editor

Benedict and Pigford

Nuclear Chemical Engineering

Meghreblian and Holmes

Nuclear Radiation Detection

Control of Nuclear Reactors and Power Plants

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Introduction to Nuclear Engineering

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Oak Ridge National Laboratory

McGRAW-HILL BOOK COMPANY, INC.

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

The purpose of the present work is to provide a complete and consistent

models and emphasis

aid of elementary mathematical

The formal level of the presentation

cussion of the principal physical concepts to be developed.

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

places much more emphasis on the physical aspects themselves of these

the use of analytical methods in the solution of engineering problems.

Emphasis is placed upon detailed presentations of each method in order

In nearly every instance the mathematical treatment

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

several analytical methods are presented in order to

Chap. 5 on diffusion theory; Chap. 6 on the Fermi age model;

applications of these general topics.

Concepts and Nuclear Cross Sections

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Neutron- Nucleus Interactions

Constant and Neutron Flux

Process in the Infinite Medium

Mechanism of Energy Loss by Scattering Collisions

Diffusion Theory: The Homogeneous One -velocity Reactor

One-velocity Diffusion Equation

Applications of the Kernel Method

Fermi Age Theory

The Homogeneous Multivelocity

The Combined Slowing-down and Diffusion Equation

Time-dependent Behavior of the Neutron Flux with Delayed Neutrons

Introduction and Survey

Central Rod in One-velocity Model

The Special Treatment Required for Hydrogen

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

The Scope and Methods of Perturbation

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

Nuclear Chain Reactions

reactors is achieved through a study of the neutron population which

Generated for wjivans (University of Florida) on 2015-09-22 17:17 GMT / http://hdl.handle.net/2027/uc1.b4118385

The first section of this chapter

is a brief discussion of those aspects

and scope of the subject.

In introducing the general subject of chain reac