Introduction To T 00 Cars

=
ill
atk^mmm/ttiat
FOURIER'S SERIES AND INTEGRALS

PROPERTY OF
DOT OrPHYSICS
.
fi
ASTRONOMY
M.VEKSITY OF
FLORIDA
WNEWILH
FLORIDA
32601
INTRODUCTION TO THE
THEORY OF FOURIER'S
SERIES AND INTEGRALS
BY
H.
S.
CARSLAW,
Sc.D., LL.D., F.R.S.E.
PROFESSOR OF MATHEMATICS IN THE UNIVERSITY OF SYDNEY

FELLOW OF EMMANUEL COLLEGE, CAMBRIDGE
AND FORMERLY LECTURER IN MATHEMATICS IN THE UNIVERSITY OF GLASGOW
THIRD EDITION, REVISED AND ENLARGED
DOVER PUBLICATIONS, INC
COPYRIGHT
First Edition, 1906.
Second Edition, 1921.

Third Edition, 1930.
America
Printed and bound in the United States of
PREFACE TO THE THIRD EDITION

In preparing this new edition the text has been carefully revised,
and considerable new matter has been introduced without
altering
the numbering of the sections or the character and aim of the book.
The use
some
of the
upper and lower limits of indetermination
of the proofs
of the
uniform convergence of
earlier
series are included.
tion
and the Second Theorem
The
sets of
examples on
of
Mean Value
Infinite Series
simplifies
Additional tests for
edition.
Term by term
are treated
integra-
more
fully.
and Integrals have been
enlarged by the insertion of questions drawn from recent Cambridge

Scholarship and Intercollegiate Examinations, as well as from the
The introduction
papers set in the Mathematical Tripos.

tions of
bounded variation extends the
of func-
which
class of functions to
the elementary discussion of Fourier's Series given in the text

applies.
In the chapters dealing particularly with Fourier's Series space
has been found for the Riemann-Lebesgue Theorem and
Theorem under
quences, and for Parse val's
fairly
conse-
its
general
con-
ditions.
For
all
ordinary purposes the discussion of the properties of
Fourier's Series
and Fourier's Constants given
in the text will,
it is
hoped, be found both sufficient and satisfactory.
For the
specialist
who
wishes to go further a treatment of the
Lebesgue Definite Integral
is
given in a
new Appendix, which
takes
the place of the former Appendix containing a detailed bibliography

of Trigonometrical
and Fourier's
tried to show, in as simple a
Integral
is,
differ from,
Integral.
and
in
what
Series.
way
In this Appendix
as possible,
to,
have
what the Lebesgue
respects the rules to which
and are superior
it is
those for the classical
subject
Riemann
PREFACE TO THE THIRD EDITION
VJ
So
many
Trigonometrical Series, and

papers are being written on
and Fourier's Integrals, that
on Fourier's Series, Fourier's Constants,

a mere
list of
And
pages.
their titles, brought
it is
doubtful
if
up
such a
to date,
list is of
would cover many

value to the
much
disposal other works from

In any case he has now at his
It
this kind can be obtained.
which bibliographical information of
at the ends of the
of books and memoirs given
is hoped that the lists
omissLon of the
Appendix II will make up for the
student.
chapters and of
detailed bibliography.
having had
proofs I am fortunate
In the revision of part of the
Walker of the University of Sydney,
the assistance of Mr. George
His criticism and sugat Emmanuel College, Cambridge.
and now
to me.
gestions have been of great service
Emmanuel College,
Cambridge, 2ith September, 1929.
H.
S.
CARSLAW.
PEEFACE TO THE SECOND EDITION

This book forms the
and
Fourier's Series
first
volume
Integrals
of the
and
the
new
Since 1906 so
and
Fourier's Series
much advance
it
book on
now for some time out of

made in the Theory of
has been
Integrals, as well as in the
Heat Conduction, that
cussion of
my
Mathematical Theory of the
Conduction of Heat, published in 1906, and

print.
edition of
mathematical
dis-
has seemed advisable to write
new work, and to issue the same in two volumes. The

volume, which now appears, is concerned with the Theory of
a completely
first
Infinite Series
Series
and
and
Integrals, with special reference to Fourier's
The second volume
Integrals.
will
be devoted to the
Mathematical Theory of the Conduction of Heat.
No
one can properly understand Fourier's Series and Integrals
without a knowledge of what

infinite series
and
integrals.
is
involved in the convergence of
With
these questions
is
bound up the
development of the idea of a limit and a function, and both are

founded upon the modern theory of real numbers. The first three
chapters deal with these matters.
Integral
tion
is
is
In Chapter IV the Definite
treated from Kiemann's point of view, and special atten-
given to the question of the convergence of infinite integrals.
The theory
whose terms are functions
of series
and the theory
of integrals
are discussed in Chapters,
theories are closely related,
The treatment
and can be developed on similar
of Fourier's Series in Chapter
Dirichlet's Integrals.
lines.
VII depends on
There, and elsewhere throughout the book,
the Second Theorem of

of the argument.
of a single variable,
which contain an arbitrary parameter

and VI. It will be seen that the two
Mean Value
will
be found an essential part
In the same chapter the work of Poisson
is
adapted to modern standards, and a prominent place is given to

Fejer's work, both in the proof of the fundamental theorem and in
PREFACE TO THE SECOND EDITION
Vlll
Chapter IX
of Fourier's Series.
the discussion of the convergence
Fourier s
Phenomenon, and the last chapter to
is devoted to Gibb's
Pringsheim, who has greatly
In this chapter the work of
Integrals.
which Fourier's Integral Theorem
extended the class of functions to
applies, has
Two
been used.
appendices are added.
The
first
deals with Practical Har-
Analysis.
monic Analysis and Periodogram
graphy of the subject
is
given.
The functions treated
An
interval (a, 6) for
number
finite
of
monotonic.
is
finite in
in this
which /(x)
open partial
If
number.
functions.
book are " ordinary

is
into a
defined can be broken up
function
in each of which the
intervals,
they are isolated and

infinities occur in the range,
demands of the
functions will satisfy the
Such
Applied Mathematician.
+ i, +
chiefly with the
of integration, associated
Fourier s
into the Theory of
The modern theory
name
In the second a biblio-
i,
of Lebesgue, has introduced
more complicated nature

and Integrals functions of a far
up a
are engaged m building
Various writers, notably W. H. Young,
much more advanced than anytheory of these and applied series
developments are m the mean-
Series
These
thing treated in this book.
specialising
Pure Mathematician
time chiefly interesting to the
has been
Real Variable. My purpose
the Theory of Functions of a
Applied Mathematician.
the difficulties of the
to
remove some
of
much
occupied some time and
The preparation of this book has
Calculus to
course in the Infinitesimal
of it has been given as a final
the
assistance
For
my
students.
To them
it
and
for
revision of the proofs
owes much.
many
valuable suggestions
am much
Mr. IP kk
Mr. R. J. Lyons and
indebted to Mr. E. M. Wellish,
University
the
Mathematics
Thorne of the Department of
H.
Sydney.
Emmanuel College,
Cambridge, Jan.
1921.
S.
CARSLAW.
CONTENTS
.....
Historical Introduction
CHAPTER
Rational Numbers
Numbers
3-5. Irrational
6-7.
8.
9.
10.
Relations of Magnitude for Real
Dedekind's Theorem
Dedekind's Axiom
The Development
System
of the
of
of
II
AND SERIES
an Aggregate-
Limiting Points of an Aggregate
15.
Convergent Sequences
Divergent and Oscillatory Sequences
Monotonic Sequences
---_-._.
--....
The Upper and Lower Limits of Indetermination

Theorem on an Infinite Set of Intervals
-
19-23. Infinite Series
......
of a
42
-
43
46
47
III
FUNCTIONS OF A SINGLE VARIABLE.

CONTINUITY
25. lim/(x)
37
41
CHAPTER
The Idea
-34
35
17. 2-17. 4.
24.
36
16.
Theorem
14. Weierstrass's
18.
29
--.'....
The Upper and Lower Bounds
17. 1.
25
27
28
Real Numbers
INFINITE SEQUENCES
Aggregates
20
21
Numbers
The Linear Continuum.
11. Infinite
13.
--.._.
--...._
CHAPTER
12.
RATIONAL AND IRRATIONAL NUMBERS

lujjuivo
SECTION
1-2.
TAGE
Function
LIMITS
AND
55
56
CONTENTS
PAGE
SECTION
58
---"'""
-
Some Theorems on Limits
26.
27. Iim/(a0
60
the Existence of a
Sufficient Conditions for
28-29?3. Necessary and
Limit
Oscillation of a
The
29. 4.
31.
66
The Heine-Borel Theorem
31. 2.
an
32. Continuity in
66
at a Point
Bounded Function
Continuous Functions of Continuous Functions

1. Properties
30.
61
Discontinuous Functions
34.
Monotonic Functions
35
Inverse Functions
36
71
73
Infinite Interval
33.
67
73
75
76
Merence
a Function as the
The Possibility of Expressing

Increasing Functions
two Positive and Monotonic
Variation
Functions of Bounded
of
78
-
80
36. 2.
Variables
Functions of Several
37.
84
CHAPTER IV
THE DEFINITE INTEGRAL
89
-
Introductory
38.
91
The Sums S and s

40 Darboux's Theorem
a Bounded Function
of
Integral
41 The Definite
Conditions for Integrabihty
Sufficient
42. Necessary and
39.
92
94
43
Integrable Functions
44
45-47.
Some
The
".
49
"
"
97
'
"
Number
(a, b) has an Infinite
b)
any Partial Interval of (a,
of Points
Function integrate in
of Continuity in
48
95
-
The
51-56.
57-58.
100
Integral
Properties of the Definite
First
105
"
Theorem
Definite
Limit
of
Integral
Mean Value
considered as a Function
of
its
Upper
106
Value
The Second Theorem of Mean
Infinite
Integrand.
Bounded
Infinite Integrals.
50 1-50
99
2.
The Mean Value Theorems
59-61. Infinite Integrals.
for Infinite Integrals
"
112
-
123
125
Integrand Infinite
Examples on Chapter IV.
107
-
Interval
134
-
'
CONTENTS
xi
CHAPTER V
THE THEORY OF INFINITE SERIES, WHOSE TERMS ARE
FUNCTIONS OF A SINGLE VARIABLE
SECTION
PAGE
62. Introductory
63.
The Sum
64*
Repeated Limits
of a Series of
may
Continuous Functions
be Discon-
tinuous
65-69.
]QQ
j^2
Uniform Convergence
I44.
Term by Term Integration

Term by Term Differentiation
The Power Series
70. 1-70. 2.
71.
72.
73.
156
-
162
Extensions of Abel's Theorem on the Power Series
74-76. Integration of Series.
Infinite Integrals
Examples on Chapter
161
CHAPTER
.*
168
172
igi
VI
DEFINITE INTEGRALS CONTAINING AN ARBITRARY
PARAMETER
77.
Continuity of the Integral
F(x,y)dx
ja
78. Differentiation of
the Integral
79. Integration of the Integral
80-83. Infinite Integrals
189
F(x, y)dx
F(x, y)dx
191
F(x, y)dx.
Uniform Convergence
192
poo
84. Continuity of the Integral
$00
F(x, y)dx
198
F(x, y)dx
199
/QO
86. Differentiation of the Integral
F(x, y)dx
87. Properties of the Infinite Integral f"

88. Applications of the
89.
The Repeated
F(x, y)dx
200
201
Preceding Theorems
Integral
dx
f
Examples on Chapter VI
f(x, y)dy
202
-
209
212
CONTENTS
CHAPTER
VII
FOURIER'S SERIES
PAGE
SECTION
21 K
-
90. Introductory
(First
91-92. Dirichlet's Integrals.
Form)
93. Dirichlet's Conditions
(Second Form)
94. Dirichlet's Integrals.

95.
Fourier's Series
Proof of the Convergence of
The Cosine
97.
The Sine Series

Other Forms of
98.
241
"'".-"."."
Fourier's Series
Fourier's Series
99-100. Poisson's Discussion of
101. Fejer's
-----
Series
96.
227
250
Theorem
102.
Two Theorems on
103
Feier's
258
the Arithmetic Means
Theorem and Fourier's
Series
265
Examples on Chapter VII
CHAPTER
VIII
SERIES
THE NATURE OF THE CONVERGENCE OF FOURIER'S
CONSTANTS
FOURIER'S
OF
PROPERTIES
AND SOME
"'
104.
105.
106.
The Order
269
Terms
of the
Consequences
The Riemann-Lebesgue Theorem and its
-r,
in
infinite
is
Discussion of an Example in which /()
The Uniform Convergence

Differentiation and Integration
107-108.
109.
tt)
273
110. Parseval's
Theorem on
of Fourier's Series
of Fourier's Series
275
-
282
284
Fourier's Constants
CHAPTER IX
PHENOMENON
THE APPROXIMATION CURVES AND THE GIBBS
IN FOURIER'S SERIES
111-112.
113-114.
The Approximation Curves

2. The Gibbs Phenomenon
115.
x
q
o
The Trigonometrical Sum 2
116.
The Gibbs Phenomenon
v
2,
289
293
-
2r^l
sin (2r- l)s
>
sin( 2r-l)a:
J,
2r-l
for the Series 2
-297
3q 3
CONTENTS
x iii
SECTION
117.
1.
117. 2.
PAGE
The Gibbs Phenomenon for Fourier's Series in generalThe Gibbs Phenomenon for Fourier's Series when summed by
Fejer's Arithmetic Means
305
308
CHAPTER X
FOURIER'S INTEGRALS
118. Introductory
119. Fourier's
Integral
in its Simplest
120-121.
Theorem
Sommerf eld's Discussion
Examples on Chapter
Appendix
I.
II.
Practical
the
Arbitrary
-311
Function f(x)
312
More General Conditions
Appendix
for
Form
122. Fourier's Cosine Integral

123.
for f(x)
315
and Fourier's Sine Integral

of Fourier's Integrals
320
321
322
Harmonic Analysis and Periodogram Analysis
Lebesgue's Theory of the Definite Integral
323
329
Index of Proper Names
363
General Index
365
HISTORICAL INTRODUCTION
A
trigonometrical series
+ (a
said to be a Fourier's Series,
is
+ (a 2
cos x-\-b 1 sin x)
if
cos 2x
+b 2
sin 2x)
the constants a
a l3 b lt
...
satisfy
the equations
"
= 2:~\\f(x)dx,
= If'
=-f
/(#) cos
rac cfo,
~n
""
>
and the Fourier's
Series
many important
In
corresponds tof(x)
there
is
f(x) sin
is
is
rac <&c
said to correspond to the function f(x).*
cases the
sum
of the Fourier's Series
equal to f(x); but
if
the function
is
which
arbitrary,
no a priori reason that the series should converge at all

- x, -), nor, if it does converge at a point, is there
(
in the interval
any a
priori reason that its
sum
for that value of x should hef(x).
Fourier in his Theorie analytique de

first
(
Ghaleur (1822) was the

an arbitrary function, given in the interval
to assert that
7r, 7r),
la
could be expressed in this way.
ously that the expansion
is
He proved
quite rigor-
true for certain simple functions, which
he needed in the problems of the conduction of heat; and, though

he did not develop his proof for the general case with the precision
the importance of the theorem demanded, the substantial accuracy
of his
method must be admitted.
This correspondence
is
That the expansion was possible
sometimes denoted by
00
f(x)~a + 2 (a n cos nx + b n sin nx),

i
the notation being due to Hurwitz, Math. Annalen, 57 (1903), 427.

c.i.
at that
an arbitrary function, as such was understood

the date at which his work
time, was assumed to be true from
have been freely used in
series
these
became known. Since then
of mathematical physics.
equations
the solution of the differential
Series-or the
Fourier's
called
For this reason they are now
function
f(x)and the
the
corresponding to
in the case of
Fourier's
Series
coefficients in the series,
Fourier's Constants, for that

are called Fourier's Coefficients, or
function.
and still is having

The Theory of Fourier's Series has had
of the theory of functions
an immense influence on the development
and importance of these series
of a real variable, and the influence
of the power series m the
field are comparable with those
in this
general theory of functions.
FIRST PERIOD
The question
[1750-1850]
of the possibility of the
expansion of an arbitrary
and cosines
of
series of sines
function of x in a trigonometrical
the eighteenth century
multiples of x arose in the middle of
of strings.
connection with the problem of the vibration
the solution of the
The theory of these vibrations reduces to
differential equation
?t*~
and the
earliest
attempts at
its
ex*'
solution were
made by d'Alembert,*
d'Alembert and Euler obtained

Euler,t and D. Bernoulli.; Both
the solution in the functional form
y = <p(x +at) +ip(x -
at).
between them lay in the fact that

the string to be given
cV\lembert supposed the initial form of
The principal
by a
difference
Euler regarded it as lying

single analytical expression, while
different parts of which
curve,
arbitrary continuous
aiona any
expressions.
might be* given by different analytical
*
Mem.
tie
tfe*. cit.,
VAcademie de Berlin, 3
4 (1748), 69.
(1747), 214.
frc
cit.,
9 (1753), 173.
Bernoulli,
on
when
the other hand, gave the solution,

in the
form
the string starts from rest,
of a trigonometrical series
y = A 1 sin x cos
and he asserted that
at+A 2
sin 2x cos 2at
+ ...
this solution, being perfectly general,
must
The importance
discovery was immediately recognised, and Euler pointed
contain those given by Euler and d'Alembert.

of his
out that
statement of the solution were correct, an arbitrary

function of a single variable must be developable in an infinite
if
this
series of sines of multiples of the variable.
obviously impossible, since a series of sines

odd, and he argued that
these properties
it
This he held to be
is
both periodic and
the arbitrary function had not both of

could not be expanded in such a series.
if
While the debate was at this stage a memoir appeared in 1759*

by Lagrange, then a young and unknown mathematician, in
which the problem was examined from a totally different point of
view. While he accepted Euler's solution as the most general, he
mode
objected to the
a satisfactory solution
number
and he proposed to obtain
of demonstration,
by
first
of particles stretched
considering the case of a finite
on
'a
weightless string.
From
the
solution of this problem he deduced that of a continuous string
making the number

that
when the
of particles infinite.f
initial
|*1
initial velocity
by
by
way he showed
displacement of the string of unit length
given by f(x) and the

at time t is given by
y =2
In this
is
F(x), the displacement
00
Yj ( sm nirx sm n ^ x cos mrat)f{x')dx'

'
ri
H
QsTVJ
oo
y] -
1^
(sin nirx' sin nirx sin nirat)F(x')dx'
This result and the discussion of the problem which Lagrange

gave in this and other memoirs have prompted some mathematicians to deny the importance of Fourier's discoveries, and to
attribute to Lagrange the priority in the proof of the development
of
an arbitrary function in trigonometrical
that in the formula quoted above
it is
series.
the order of summation and integration, and to put

*Cf. Lagrange, (Euvres, 1 (Paris, 1867), 37.
\loc. cit., 37.
It
is
true
only necessary to change

t
= 0,
in order
of the function f(x) in a

shall take the definite
coefficients
the
series of sines, and that
that
we may obtain the development
which we are now familiar. Still Lagrange

Burkhardt remarks,* the fact that
did not take this step, and, as
example of the ease with
he did not do so is a very instructive
obvious conclusion
almost
which an author omits to draw an
been undertaken
has
from his results, when his investigation
purpose was to demonwith another end in view. Lagrange's
solution and to defend its general
strate the truth of Euler's
obtained
against d'Alembert's attacks. When he had
integral forms with
conclusions
proceeded to transform it into the funchis solution, he therefore

Having succeeded in this, he held
tional form given by Euler.
demonstration to be complete.
these series was due
The further development of the theory of
of the reciprocal
expansion
the
of
to the astronomical problem
his
between two planets in a series of cosines of

As early as 1749 and
radii.
multiples of the angle between the
published discussions of this
1754 d'Alembert and Euler had
of the distance
definite integral expressions for

question in which the idea of the
in
Series may be traced, and Clairaut,

the coefficients in Fourier's
which practically contained these
1757 t gave his results in a form
and
Again, Euler, J in a paper written in 1777
coefficients.
the method of multiplying
published in 1793, actually employed
both sides of the equation
+2a 2 cos 2x + ... + 2a n cos nx + ...
f(x) = a +2a x cos x
by term between the
cosnz and integrating the series term
by
limits
and
tt.
In this
way he found
1 f
that
a n = -\ f(x) cos nxdx.

7TJ
papers seem to have had no effect upon

It is curious that these
of the Vibrations of Strings in which,
the discussion of the problem
Euler, Bernoulli, and Lagrange were
as we have seen, d'Alembert,
explanation is probably to
about the same time engaged. The
results were not accepted with
be found in the fact that these
oscillirenden
*Burkhardt, "Entwicklungen nach
32.
Math. Ver., Leipzig, 10, Heft II (1901),
Functionen," Jahresber.
Art. iv. (July 1757).

\Paris, Hist. Acad. Sci. (1754 [59]),
94 (May 1777).
XPetrop. N. Acta, 11 (1793 [98]), p.
d.
confidence,
and that they were only used in determining the

whose existence could be demonstrated
coefficients of expansions
by other means.
It was left to Fourier
to place our knowledge of the theory of
trigonometrical series on a firmer foundation.
The methods he adopted were suggested by the problems he met

Theory of the Conduction of Heat. He discussed the subject in various memoirs, the most important having
been presented to the Paris Academy in 1811, although it was not
printed till 1824-6. These memoirs are practically contained in
his book, Theorie analytique de la Chaleur (1822).
In a number of
in the Mathematical
special cases he verified that a function f(x), given in the interval

(
7r,
x),
can be expressed as the sum of the
a +(a 1
where
cosx+b 1
If
= ^\
77
If
bn
sin 2x)
+ ...
77
a n = -\
f(x)dx,
A7TJ - n
series
+(a 2 cos 2x+b 2
sin z)
f(x) cos nx dx,
7TJ - n
/(#) sin nx dx,
(n^l).
J ir
Some
of the proofs he
gave for the general case of an arbitrary
function are far from rigorous.

Euler.
But
One
is
the same as that given by
in his final discussion of the general case (Cf. 415,
416 and 423), the method he employs
is
perfectly sound,
and not
unlike that which Dirichlet used later in his classical memoir.
However, this discussion is little more than a sketch of a proof,

and it contains no reference to the conditions which the arbitrary
function must satisfy.
Fourier made no claim to the discovery of the values of the
coefficients
9A7TJ
f(%)dx,
-jt
f(x) cos nx dx,
n^l.
f(x) sin nx dx,
7TJ _
We have already seen that they were employed both by Clairaut

and Euler before this time. Still there is an important difference
between Fourier's interpretation of these integrals and that which
was current among the mathematicians of the eighteenth century.
by whom they were employed (with the possible

determination of the
exception of Clairaut) applied them to the
demonstrated by
been
had
coefficients of series whose existence
to the reprethem
apply
other means. Fourier was the first to
earlier writers
The
in the sense in which

sentation of an entirely arbitrary function,
made a distinct advance
that sum was then understood. In this he
asserts that when
upon his predecessors. Indeed Kiemann*
Academy in 1807, stated
Fourier, in his first paper to the Paris
be expressed in such a
that a completely arbitrary function could
Lagrange that he denied the
series, his statement so surprised
It should also be noted that
possibility in the most definite terms.
function might be given
he was the first to allow that the arbitrary
parts of the interval;
different
in
by different analytical expressions
be used for other
could
series
also that he asserted that the sine
other functions
for
series
functions than odd ones, and the cosine
that, when a
see
to
first
than even ones. Further, he. was the
value outside
its
variable,
function is defined for a given range of the
and it follows that no one
that range is in no way determined,
understood the representation ot
before him can have properly
series.
an arbitrary function by a trigonometrical
the Paris Academy
from
received
work
his
The treatment which
his contemporaries viewed

evidence of the doubt with which
ot
His first paper upon the Theory
results.
is
his
arguments and
Heat was presented
The Academy, wishing to enmade the

extend and improve his theory,
in
1807.
courage the author to

the subject of the grand pnx
question of the propagation of heat
submitted his Memoue sur
mathematiques for 1812. Fourier
candidate for
at the end of 1811 as a
la propagation de la Chaleur
Lagrange,
The memoir was referred to Laplace,
the prize
but, while awarding him
Legendre, and the other adjudicators;
^rigour
praise with criticisms of
the prize, they qualified their
at
published
not
was
paper
the
and
of his analyst and methods,t
^7~R^n7"trber die
ZL, GoUiên,
JLt
Werke
eine trigone.
Darstellbarkeit einer Function durch
and Matkemalucke
Abh. Ges. Wiss., 13 (1867), 2,
(2 Aufl., 1892), p. 232.
que
Ouvrage, en observant cependant
la
mamere dont
Auteur parvient
the time in the
Memoir es
de
VAcademie des
Sciences.
always resented the treatment he had received.

his treatise in
change, the
first
1822, he incorporated in
it,
When
Fourier
publishing
practically without
part of this memoir; and two years later, having
become Secretary of the Academy on the death of Delambre, he

caused his original paper, in the form in which it had been communicated in 1811, to be published in these Memoires.% Probably
this step was taken to secure to himself the priority in his discoveries, in consequence of the attention the subject was receiving
at the hands of other mathematicians.
It is also possible that he
wished to show the injustice of the criticisms which had been
passed upon his work.
After the publication of his treatise,
the results of his different memoirs had become known,
recognised that real advance had been

cussion of the subject
made by him
and the substantial accuracy
when
it
was
in the dis-
of his reasoning
was admitted.
equations n'est pas exempte de difficulties, et que son analyse, pour les integrer,
laisse encore quelque chose a desirer, soit relativement a la generalite, soit meme
du cote de la rigueur."
%Memoires de VAcad. des 8c,
It
is
4, p. 185,
and
5, p. 153.
interesting to note the following references to his
work
in the writings of
modern mathematicians:
Kelvin, Coll. Works, Vol. Ill, p. 192 (Article on "Heat," Enc. Brit., 1878).
"Returning to the question of the Conduction of Heat, we have first of all to
say that the theory of it was discovered by Fourier, and given to the world through
Academy in his Theorie analytique de la Chaleur, with solutions of problems naturally arising from it, of which it is difficult to say whether their uniquely
original quality, or their transcendently intense mathematical interest, or their
perennially important instructiveness for physical science, is most to be praised."
the French
Darboux, Introduction, (Euvres de Fourier, 1 (1888), p. v.

"Par l'importance de ses decouvertes, par l'influence decisive qu'il a exercee sur
le developpement de la Physique mathematique, Fourier meritait l'hommage qui
est rendu aujourd'hui a ses travaux et a sa memoire.
Son nom figurera dignement a cote des noms, illustres entre tous, dont la liste, destinee a s'accroitre
avec les annees, constitue des a present un veritable titre d'honneur pour notre
pays.
La Theorie analytique de la Chaleur
que Ton peut placer sans injustice
.
a cote des ecrits scientifiques les plus parfaits de tous les temps, se recommande
par une exposition interessante et originale des principes fondamentaux; il eclaire
de la lumiere la plus vive et la plus penetrante toutes les idees essentielles que nous
devons a Fourier et sur lesquelles doit reposer desormais la Philosophic naturelle;
mais il contient, nous devons le reconnaitre, beaucoup de negligences, des erreurs

de calcul et de detail que Fourier a su eviter dans d'autres ecrits."
Poincare, Theorie analytique de la propagation de la Chaleur (1891), p. 1, 1.
"La theorie de la chaleur de Fourier est un des premiers exemples de l'application de l'analyse a la physique; en partant d'hypotheses simples qui ne sont
He
The next writer upon the Theory of Heat was Poisson.
the
of
employed an altogether different method in his discussion
by a
function
arbitrary
an
of
question of the representation
onwards, which
trigonometrical series in his papers from 1820
e
Mecanique (2 ed.,
are practically contained in his books, Traite de
He began
(1835).
Chaleur
1833) and Theorie mathematique de la
with the equation
7.2
-2 = 1 +2>] hn cos n(x'-x),
tUl
l-2hcos(x'-x)+h
h being numerically less than unity,
and he obtained, by
inte-
gration,
{l-h 2 )f(x')dx'
r*
].
l-2hcos{x'-x)+
n
=\
J
While
it is
f(x')dx'
+22^1"
-n
true that
f(x')cosn(x'-x)dx'.
J -7r
by proceeding
to the limit
we may deduce
that at a point of continuity or ordinary discontinuity

f(x) or
is
l[f(x+0)+f(x-0)]
equal to
liml^M*
/(zW+-i>n f
f(x')cosn(x'-x)dx'\
l,
the value A
are not entitled to assert that this holds for
this
for
converges
series
the
that
unless we have already proved
Fourier's Series,
This is the real difficulty in the theory of
value.
lost sight of in
been
has
and this limitation on Poisson's discussion
we
generalises, Fourier en a deduit une

autre chose que des faits experimentaux
une theorie complete et coherente.
serie de consequences dont l'ensemble constitue
par eux-memes, mais ce qui
interessants
certes
sont
obtenus
a
qu'il
Les resultats
pour
employee
a
qu'il
y parvenir et qui servjra
methode
l'est plus encore est la
cultiver une branche quelconque de
touiours de modele a tons ceux qui voudront
importance
J'ajouterai que le livre de Fourier a une
la physique mathematique.
lui doit peut-etre
pure
l'analyse
que
et
mathematiques
des
capitale dans Fhistoire
plus encore que l'analyse appliquee."
4.
Boussinesq, Theorie analytique de la Chaleur, 1 (1901),
methode (i.e. his method of inte"Les admirables applications qu'il fit de cette
a la fois, assez simples et assez
sont,
Heat)
of
grating the equations of Conduction
geometres de la premiere moitie de ce
generales, pour avoir servi de modele aux
d'autant plus utiles, qu'elles ont pu, avec de legeres
siecle- et elles leur ont ete
de la
etre transportees dans d'autres branches
plus,
modifications tout au
l'Hydrodynamique et dans la Theorie
Physique mathematique, notamment dans
de
l'elasticite."
some presentations
There are, however, other

which Poisson's method has led to most notable
results.
The importance of his work cannot be exaggerated.*
After Poisson, Cauchy attacked the subject in different memoirs
published from 1826 onwards using his method of residues, but
directions in
treatment did not attract so much attention as that given

about the same time by Dirichlet, to which we now turn.
Dirichlet's investigation is contained in two memoirs which
his
appeared in 1829f and 183 7. J

The method which he employed
we have already referred to in speaking of Fourier's work. He
based his proof upon a careful discussion of the limiting values
of the integrals
j/^lnf dx...,b>a>0,
as
increases indefinitely.
ju
sum
By
this
means he showed that the

+0) +f(x-0)) at every
+0) +f(ir - 0)) at x 7r,
of the Fourier's Series ioif{x) is \(f{x
point between
ir
and
ir,
and J(/( -
ir
provided that f(x) has only a finite number of ordinary discontinuities and turning points, and that it does not become infinite
in
7r, 7r).
In a later paper, in which he discussed the expan-
sion in Spherical Harmonics, he
f(x)
must remain
finite is
showed that the
restriction that
not necessary, provided that
f(x)dx
converges absolutely.
SECOND PERIOD
The
principal
names
[1850-1905]
in the First Period are those of Fourier
and
and the position as left by Dirichlet was that, when the

function f(x) is bounded in the interval ( - tt, x), and this interval
can be broken up into a finite number of partial intervals in each
of which f(x) is monotonic, the Fourier's Series converges at every
point within the interval to \[f(x+0) +f(x-0)], and at the endpoints to J[/( - 7r +0) +f(ir - 0)]. These sufficient conditions and
Dirichlet,
*For a
full
treatment of Poisson's method, reference
paper, "Introduction to the Theory of Fourier's Series/'

f Journal
fur Math., 4 (1829).
%Dove's Repertorium der Physik, 1 (1837), 152.
^Journal fiir Math., 17 (1837).
may
Ann.
be made to Bocher's
Math. (2), 7 (1906).
of
10
their extension to the
unbounded function
cover
most
of the
cases that are likely to be required in the applications of Fourier's

Series to the solution of the differential equations of
mathematical
physics.
In the Second Period we pass more definitely into the domain of

the pure mathematician, and the first name we meet is that of
His memoir* TJber die Darstellbarkeit einer Function

durch eine trigonometrische Reihe formed his Habilitationsschrift at
Riemann.
it was not published till 1867, after his

most important developments in mathematical
Gottingen in 1854, but

death.
It led to
analysis, as well as to the discovery of
many striking
properties of
trigonometrical series, in general, and of Fourier's Series, in parHis aim was to find a necessary and sufficient condition
ticular.
which the arbitrary function must satisfy so that, at a point x in the

interval, the corresponding Fourier's Series shall converge to f{x).
had shown that certain conditions were sufficient. The

Riemann set himself to answer has not yet been solved.
But in the consideration
quite probable that it is not solvable.
Dirichlet
question
It
is
of the
problem he realised that the concept
should be widened.
study of Fourier's
Cauchy
in 1823f
And
Riemann
the
of the definite integral
Integral
we owe
to the
Series.
had defined the
much
function as the limit of a sum,
He
in elementary text-books.
into partial intervals
by the
a=a
definite integral of a continuous
in the
way
it is still
treated
divided the interval of integration
points
ax
...
a n _ v a n = b.
The sum S was given by the equation
S = (a 1 -a
)f(x 1 ) +(a 2
-a 1 )f(x 2
+...
+(a n -a n _ 1 )f(x n ),
where x r is any point in (a r _ l5 a r ).

He showed that, when the number of points of section tends to
infinity and the length of the largest partial interval tends to zero,
the sums
S tend
to a limit.
The
definite integral
f(x)dx he
defined to be this limit.

If
the function
*See note on p.
f-Cf.
is
continuous in
(a, b)
except at the point
c,
in
6.
Cauchy, Resume des legons donnees a VEcole
infinitesimal, 1 (Paris, 1823), pp. 81-84,
and (Euvres
roy. Polytechnique sur le calcul

(2), 4, p.
122-29.
the neighbourhood of which
is
taken to be the
sum
may
it
be bounded or not, the integral
of the limits
Cc-h
hm
11
Cb
and
f(x)dx
when these limits exist.

number of points c v c 2
And
cm
...
if
lim
f(x)
f(x)dx,
discontinuous at a
is
the interval
is
each of which contains only one of these points.

parts the preceding definition
is
applied,
when
then the sum of the numbers so obtained

from a to b.
is
finite
divided into parts
To each
of these
this is possible;
and
taken as the integral
In dealing with the bounded function, Riemann did not assume

it was continuous in the interval, or had only
a finite number
of discontinuities therein.
But he used the sum S as before, and
that
the integral
f(x)dx was defined as the limit of these sums S,
provided this limit existed.
He obtained a necessary and sufficient

condition for the existence of the limit, and placed the definite
integral on a wider and purely arithmetical basis.
With Riemann's
definition of the integral
f(x)dx, for a
'
bounded
function given in the text in a slightly modified formfunctions

that were previously without an integral became integrable. A
striking
example* due to him was the sum

,
(2x)
of the series,
(3x)
where (nx) stands for the positive or negative difference between nx

and the nearest integer, unless it lies midway between two consecutive integers, when (nx) is to be taken as zero.
This function is
discontinuous for every rational
is
an odd number, prime to
n;
number of the form p/2n, where p

and there are an infinite number
of points of discontinuity in every interval,
however small.
fundamental theorem proved by Riemann deals with the
Fourier's Constants
f(x)
-J
nx
He showed
dx.
that for any
bounded and integrable function f(x) these constants tend to zero

as n tends to infinity.
And this holds also for the integral
J
*Ct. loc.
cit.
6.
/(*) cos nx dx
12
This theorem shows at once that,

rate in ( - 7T, tt), the convergence of
in
hood
x,
tt)
if
f(x)
bounded and
is
integ-
its Fourier's Series at a point
depends only on the behaviour
of f(x) in the neighbour-
of that point.
also led to
Riemann was
series of the
of trigonometrical
type
a +(a x cos x
when
examine the theory
+6 X
+(a 2 cos 2x +6 2
sin x)
sin 2x)
the coefficients are not Fourier's Constants.
+ ...
He
obtained
The most important question

many
could be represented by
to be answered was whether a function
This reduces
(-tt, tt).
interval
an
more than one such series in
of the properties of
such
series.
to the question whether the
sum
of a trigonometrical series in
vanish can be zero right through

was
The discussion of this and similar problems
the interval.
onwards;
1870
Cantor, from
carried on, chiefly by Heine and G.
Theory of
papers Cantor laid the foundation of the
which the
coefficients
do not
all
these
in
influence
another example of the remarkable
had upon the development
the theory of Fourier's Series has
in this place to state that
sufficient
It will be
of mathematics.
of the trigonocoefficients
the
all
Cantor showed in 1872 that
Sets
Points,
of
metrical series
(
tt, tt),
In'
with
must vanish,
if
its
sum
is
zero at
converges in (-tt,
tt)
points of
nth order.
a trigonometrical
to f(z), where f(x) is integrable the
He also settled the
Series ior f(x).
1875 P. du Bois-Reymond provedf that
series
all
the exception of the points of a set of the

if
series must be the Fourier's

for a continuous function
question as to whether the Fourier's Series
only an example of a
always has /(a?) for its sum; for he gave not
Fourier's Series did not
function, continuous in ( - tt, tt), whose
constructed another,
also
he
but
converge at a particular point,
points of an everythe
at
whose Fourier's Series fails to converge
where dense
set.
Many
years later Fejer gave several
much
simpler examples.J
received
the convergence of Fourier's Series
by Stokes (1847) and
attention, especially after the introduction
The nature
of
*Math. Annalen, 5 (1872), 123.

\Abh.
d.
Bay. Akad., 12 (1875),
p. 117.
JCf Journal fur Math., 137 (1909)

.
28 (1909), 402.
138 (1910), 22.
Rend. Circ. Mat. Palerrm,
13
Seidel (1848) of the concept of uniform convergence.
known
since Dirichlet's
time that the
conditionally convergent,
It
had been
series were, in general,
only
and that their convergence

depended upon the presence of positive and negative terms. It
was not till 1870 that Heine showed* that if f(x) is- bounded
and integrable, and otherwise satisfies Dirichlet's Conditions in
(
- 7r,
(a, b),
at
all;
Fourier's Series converges uniformly in
its
7r),
if
which contains neither inside
any
interval
nor at an end any discon-
it
tinuity of the function.
The importance attached to the question of uniform convergence

was due to the impression that term by term integration would only be permissible, if the series converged uniformly.
It was not till much later that it was found that a Fourier's Series
of the series
could be integrated term by term, even
if
the series
itself
did not
converge.
The
sufficient conditions of Dirichlet
now
conditions,
classical,
and Jordan.
Lipschitz
associated
were succeeded by three

with the names of Dini,
Dinif in 1880 showed that the Fourier's
Series for the integrable function f(x) has lim
\[f(x+h) +f(x-h)]
its sum at any point in

( -tt, it) for which this limit exists, provided that there is a positive S such that
for
is
\f(x+t)+f(x-t) -lim [f(x+h)+f(x-h)]\
a convergent
dt
integral.
special case of Dini's criterion
Lipschitz.J
had been given
in 1864
by
This can be put in the form:
The Fourier's Series for f(x) converges
at
to
limi[f(x+h)+f(x-h)l
A->0
when
this limit exists, if there is

I
A* +*) +f(x -t)~ lim
[f(x +h) +f(x -h)]\< Ct\
0<t^S
when
where
a positive S such that
C and
k are positive numbers.
*Journal fur Math., 71 (1870), 353.

fCf. Dini, "Serie di Fourier e altre rappresentazioni
analitiche delle funzioni di
variabile reale" (Pisa, 1880), p. 102.
una
JCf. Journal fiir Math., 63 (1864), 296.
14
was
Fourier's Series
The treatment of
by the introduction
simplified
by Jordan*
bounded variation and his
of his functions of
the integrate function

criterion states that the Fourier's Series for
fix) converges to
neighbourhood of which f(x)
at every point in the
of bounded
is
variation.
Constants were
During this period the properties of Fourier's
obtained, when
results
important
also examined, and among the
to cite that
sufficient
it is
the Riemann integral was still used,
when
which,
according to
usually called Parseval's Theorem,f
f{x)
and [f(x)f are integrable
in
tt, tt),
00
7Tj
Also,
if
f(x)
-
where a n
bn,
-*
and
*
integrable,
g(x), as well as their squares, are
f( x )g( x ) dx
= 2a oo +
+ h ^n),
and a n p n are the Fourier's Constants
iox f(x)
and
g(x) respectively.
If Fourier's Series for /(a) is
not convergent,
it
may
converge
to
or other of the methods
remarkthe
discovered
divergent series is adopted. Fejer in 1904
summed by the method of
able theorem* that, when the series is
at every point
arithmetical means, its sum is J [/(*+<)) +/(*-<>)]
only condition attached to
in (-tt, tt) at which f(x0) exist, the
shall be integrable in ( - tt, tt), and,
f(x) being that, if bounded, it
of
when one
if
unbounded, that
f(x)dx shall be absolutely convergent.

J'
THIRD PERIOD
The theory
"summation" applied
[1905-
of Fourier's Series, as built
up by
Dirichlet,
Riemann,
of the nineteenth
Cantor, Dini, Jordan and other mathematicians
processes
limiting
the
of
century, with a fuller understanding
*Cf.
Comptes Rendus, 92 (1881), 228, and Jordan, Cours
V Analyse,
(1- ed., 1882),
Ch. V.
Soc.
tCf. de la Vallee Poussin, Ann.
Math. Annalen, 57 (1903), 175.
+Cf.
Math. Annalen, 58 (1904), 51.
sc.
Brux.,
17B
(1893), 18,
and Hurwitz,

15
involved, placed in the hands of applied mathematicians a quite

satisfactory instrument.
But the properties of the series, which
we owe to them, failed in many ways to give a theory with which

the pure mathematician could be fully content. Unity, symmetry
and completeness were still wanting. In this respect the last
twenty-five years have seen a great improvement, due, chiefly, to
the new definition of the definite integral put forward in 1902 by
Lebesgue in his Paris thesis Integrate, longueur, aire* and further

developed in his Legons sur V integration et la recherche desfonctions
primitives (1904).f
Lebesgue's integral
is
founded upon the subtle and rather difficult

In the modern theory of
idea of the measure of a set of points.
functions of a real variable, Lebesgue's integral (or one of the others
associated with
the
Riemann
it)
is
But for practical purposes

The progress which we now
the specialist; and in no department of
indispensable.
integral will suffice.
describe lies in the field of
pure mathematics has greater activity been displayed in recent

years than in the theory of trigonometrical series.! Most important contributions have been made by Lebesgue himself, Fejer,
Hobson, Hardy and Littlewood, de

Young.
The
first
point to notice
according to Riemann's
is
that,
la Vallee
if f(x) is
definition,
Poussin and
W. H.
bounded and integrable
also integrable with

Lebesgue's definition, and the integrals are equal. But a bounded
function may be integrable with Lebesgue's definition, and fail to
be integrable with Riemann's.

function
is
integrable
(L),
is
It is convenient to say that a
when
Lebesgue's definition, and that
it
it
it is
is
integrable according to
integrable (R),
integrable according to Riemann's definition.

(L),
when
it
is
If f(x) is integrable
but not bounded in the interval of integration, the Lebesgue

converges absolutely. Unbounded functions may be
integral
integrable (L), but not integrable (R);
The fundamental theorems
and conversely.
of integration apply to
both integrals,
but one of the advantages for our present purpose of the Lebesgue
integral is that a function integrable (L) need not be continuous
*Annali di Mat.
fA
J
(3),
7 (1902), 231.
and enlarged second edition has appeared in 1928.

account of work in this field is to be found in Hobson's Theory
revised
A full
of a Real Variable, 2 (2nd ed., 1926), Ch. VIII.
of Functions
16
integration, as
"almost everywhere"* in the interval of
if
Also
(fi).t
integrate
fn (x) is integrate
with a function
is
lim
(L),
integrable (L).
exists, finite or infinite, this limit is
fn (x)
with Lebesgue integrals, under

of the relation:
eeneral conditions,! we can make use
more important
fix)
If
the case
lim
still,
then
fn (x),
f(x)dx= lim
and
And
much more
fn (x)dx.
that Riemann s
Returning to Fourier's Series, we remark first
of a bounded
Constants
theorem, according to which the Fourier's
to infinity
tends
n
and integrable function /(z) tend to zero when
or,
more
generally, that
nz
Hm "/(*)
<fe
with the
= 0-applies
or not, integrable
Lebesgue integrable to any function, bounded
usually referred to as the Riemann-Lebesgue
This is now
Theorem or Fundamental Lemma and may
(L).
be
stated
as
follows:
If f(x) is
integrable (L) in
(a, b),
S
lim) f{x)
then
nxdx = 0.
This was proved by Lebesgue in 1903.

and sin nx
Now the sum s n (x) of the terms up to those in cos nx
integrable (L) can be written
of the Fourier's Series for f(x)
//
1 f**
^ xn sin (2n + l)a ,
..(*)
[/(*+2a)+/(-2a)]
=;
^-
<*>
7TJ
and we
this that
from
find
ff
lim *(*) -/(*)
,.
=7T
n _>oo
lim
f*
sin
0(a)
(2n+l)a~ ,
da
K
=/( +2a) +f(x - 2a) - 2/(s).

Hence, by the Riemann-Lebesgue Theorem,
,sin(2n+l)a
,.
f*
^
0(a)
where
lim
for
s n (s)
some positive
*A property
is
=/(x),
if
lim
f Cf
Appendix
Appendix
Unnales
0(a)
<*a
A.
said to hold almost everywhere in
points except those forming a set of measure
+ See
>
n -ôoJO
an
interval,
if it
holds for
all
zero.
II, 10.
II, 15, 18.
Sci. de
VEcole Normale
(3),
20 (1903), 453.
sur les series trigonometriques (Paris, 1906), 61.
Also see Lebesgue, Lemons
17
Also the question of the convergence of the Fourier's Series for

f(x) at a point in the interval ( - tt, tt) depends only on the
behaviour of f(x) in the neighbourhood of that point.

In 1905 Lebesgue gave a new sufficient condition for the convergence of the Fourier's Series for f{x), which included all the
previously
known
conditions.*
Another point to notice

integration
is
that the question of term by term
Fourier's Series does not depend, as used to be
of
thought to be the case, on the uniform convergence of the

Indeed, with the usual notation, we havef
J
iT
f{ x )dx = aQ (x +tt)
+2 ~( a
n sin
nx +6 n (cos
nir
- cos
series.
nx)),
where x is any point in ( - tt, tt), for any function integrable (L),
whether the Fourier's Series converges or not. And the new
series
converges uniformly to
f(x)dx in the interval (-tt,
tt).
Term by term
integration can then be continued indefinitely.

This result can be used as a test in determining whether a trigonometrical series is a Fourier's Series.
If, on integrating the series
term by term,
it fails
to converge in the range
cannot
tt, tt), it
00
be a Fourier's Series.
In this
way
it
^ sin nx
can be seen that Y\
w^2 log
is
not a Fourier's Series, as the integrated series diverges at x =

0.%
Again Parse val's Theorem, that
[/(*)N*=2V+i>B* + &A
l\'
holds for any function f(x), whose square is integrable (L) in

- tt, tt), and a similar remark applies to the relation
(
^)_J(x)9(x)dx = 2a a
+f> a
n n
+bj n
where a n b n and a n (5 n are the Fourier's Constants for the functions

f(x) and g(x), whose squares are integrable (L) in ( - tt, tt).
,
*Cf Math. Annalen, 61 (1905), 82, and Lebesgue, Lemons sur

.
les series
trigonometri-
ques, p. 59.
f Cf.
Lebesgue, Legons sur
{This example
les series trigonometriques, p. 102.
due to Fatou, Comptes Rendus, 142 (1906),

examples are given by Perron, Math. Annalen, 87 (1922), 84.
is
Cf. Lebesgue, Legons sur
les series
Fatou, Acta Math., 30 (1906), 352.

c.i.
trigonometriques, p. 100.
p. 765.
Other
18
As g(x) csgi be put equal to zero in the partial intervals ( - ir, a)

and (ft, 7r), it follows that when/(a?) and g(x) are functions whose
squares are integrable (L) in ( - w, it) and (a, /?) respectively, the
integral
f(x)g(x)dx
Fourier's Series
But one
may
be obtained by substituting for f(x)
its
and applying term by term integration.

most remarkable results which follow from the
of the
use of the Lebesgue integral in the theory of Fourier's Series
converse of Parseval's Theorem,

the Riesz-Fischer
Theorem
known from
its
is
the
discoverers as
:*
oo
Any
trigonometrical series for which V] (a n 2
+6 n 2
converges
is
the Fourier's Series of a function whose square
made
Reference has already been
is
integrable (L) in
to the application of
summa-
by Fejer's arithmetical means to Fourier's Series. This

method is a special case (C, 1) of the general Cesaro sum, usually
denoted by (0. r). A great deal of work has been done in the
tion
investigation
of
sufficient
conditions
that
Fourier's
Series
be
The results obtained by

thrown
light on ordinary
have
fractional,
when
r
is
method,
this
integral.
r
when
is
summation,
Cesaro
and
convergence
Another field in which much progress has been made is the
investigation of the behaviour and properties of Fourier's Constants
when Lebesgue integrals are used. The Parseval and RieszFischer Theorems belong to this class, and extensions of both
have been made, when the condition that f(x) and g(x) shall be
functions whose squares are integrable (L) is replaced by a more
summable
(C, r) at a point in
ir, it).
general condition.
The convergence problem
for Fourier's Series
is still
unsolved.
no property of the arbitrary function /(x), integrable (L)

There
for
in (
7T, tt), which is known to be both necessary and sufficient
sufficient
simple
are
There
Series.
the convergence of Fourier's
is
known not to be necessary, and the necessary

obtained are known not to be sufficient and the same
conditions, which are
conditions
remark applies to summation by an assigned Cesaro mean.

*Cf. F. Riesz, Comptes Rendus, 144 (1907), 615-619, 734-736.
Fischer, Comptes Rendus, 144 (1907), 1022.
Youii, W. H. and Grace Chisholm, Quarterly J. of Math., 44 (1912), 49.
19
REFERENCES.
Burkhardt, "Entwicklungen nach
oscillirenden Functionen," Jahresber

10 (1901).
"Trigonometrische Reihen und Integrale bis etwa 150,"
Enc. d. math
Wiss., Bd. II, Tl. I, p. 819 et seq.
(Leipzig, 1914).
d.
Math.
Ver.,
Gibson,
"On
the History of the Fourier Series," Proc. Edinburgh

Math. Soc,
11 (1893).
Hilb
Enc.
d.
u.
M. Riesz, "Neuere Untersuchungen
math. Wiss., Bd. II, Tl. Ill, p. 1189
iiber trigonometrische
et seq.
Reihen,"
(Leipzig, 1922).
Plancherel, "Le developpement de

dans
le
la theorie des series trigonometriques

dernier quart de siecle," Uens. math., 24
(1924-5).
Plessner, "Trigonometrische Reihen" in Pascal's

Repertarium der hoheren
Mathematik (2 Aufl. Leipzig, 1929), Bd. I, Tl. Ill, Kap.
XXV.
RlEMANN,
he.
Cit.
Sachse, "Versuch einer Geschichte der Darstellung

willkurlicher Functionen
emer Variabeln durch trigonometrische Reihen,"

Schlomilch's
(1880), and Bull, des sciences math. (2), 4 (1830), 43.
Tonelli, Serie trigonometriche (1928, Bologna).
Zeitschrift,
25
[CH.
CHAPTER

THE SYSTEM OF REAL NUMBERS
The question of the convergence of
of satisfactory treatment when
capable
only
Infinite Series
of irrational numbers
conception
the
the difficulties underlying
1.
Rational Numbers.
is
have been overcome.
For
this reason
shall first of all give a
we
short discussion of that subject.
The idea
of
number
is
formed by a
series
of
generalisa-
We begin with the positive integers. The operations

tions.
always
of addition and multiplication upon these numbers are
possible; but if a and b are two positive integers, we cannot
determine positive integers x and
a = b+x
a = by
y,
so
that
are satisfied, unless, in the
the
equations
case, a
is
and
of
b.
multiple
a
is
a
case,
second
the
in
greater than b, and,
are
numbers
negative
and
fractional
To overcome this difficulty
introduced, and the system of rational numbers placed at our
first
disposal.*
The system of rational numbers is ordered, i.e. if we have two

greater
different numbers a and b of this system, one of them is
than the other. Also, if a>b and b>c, then a>c, when a, b
numbers
Further, if two
of the system.
and
c are
we
numbers a and b are given,

number greater than the
rational
another
can always find
different rational
rational
*The reader who wishes an extended treatment of the system of
(Leipzig,
numbers is referred to Stolz und Gmeiner, Theoretische Arithmetik
und Funktionenlehre
1900-1902) and Pringsheim, Vorlesungen uber Zahlen(Leipzig, 1916).
20
1-3]
one and
less
any two
different rational
of rational
2.
than the other.
21
from this that between

numbers there are an infinite number
It follows
numbers.*
The introduction
of fractional and negative rational

numbe justified from two points of view. The
fractional
numbers are necessary for the representation of
the subdivision
of a unit magnitude into several
equal parts, and the negative
numbers form a valuable instrument for the
measurement
bers
may
of
magnitudes which may be counted in opposite

directions.
This
may be taken as the argument of the applied
mathematician
On
the other hand there
is the argument of the

pure mathethe notion of number, positive and
negative
integral and fractional, rests upon
a foundation independent
matician, with
whom
of measurable magnitude, and in

whose eyes analysis is a
scheme which deals with numbers only, and
has no concern
per se with measurable quantity. It is

possible to found mathematical analysis upon the notion of
positive integral number.
Thereafter the successive definitions of the
different kinds of number, of equality and inequality
among these numbers, and of the
four fundamental operations, may be
presented abstractly.!
3. Irrational Numbers.
The extension of the idea of number
from the rational to the irrational is as natural,
if not as easy as
is that from the positive integers
to the fractional and negative
rational numbers.
Let a and b be any two positive integers.
The
equation afi = a
cannot be solved in terms of positive integers
unless a is a perfect
b power.
To make the solution possible in
numbers are introduced.
But
it
will
haS a
^ "^ ^^ amt6ger
SUCh
^
SltTt^ r^
SGt
th-T^T
When we
say that
it
thingS
general the irrational

be seen below that the
^ ^
finite
number
of
nUmb -
me ^bers, we mean
f
has an infinite number of members,

we mean that
* ords ' however Iarge n may be th
^^^V*-
re^^^^r^^:
(or
.
when
its
it
has
-*- -
>
abie)
- tne
In this case there is a one-one correspondence

between the members of the
ne set
and the set of positive integers 1, 2, 3, ...
.
t(.
Hobson
Proc.
London Math.
Soc. (1), 35 (1913), 126 ; also

the
Theory of Functions of a Real Variable,
1 (3rd. ed., 1927), 11.
same author's
22
[<*.
which
not confined to numbers
coefficients
algebraical equations whose
arise as the roots of
of irrational
system
numbers
is
ar
extension from the abstract
much for the desirability of the

extension U also evident
From the concrete the need for the
side
measurement of any quanti y to
We have only to confer the
a
assigned
So
.,.
is
divisibility
which the property of unlimited
Take any segment
indefinitely.
TtSght line L produced
of thi
as origin or
definite point of the line
as unit of length, a
for the positive
directions of right and left
zero point, and the"
corresponds a
To every rational number
senses.
Zl
and negative
.2
-1
FIG.
1.
number is an mteger, the point

of unit segments one
taking the required number
fs ob led by
If it is a fraction pjq
direction.
afterSother in the'proper
into q equal parts
length
dividing the unit of
it is obtained by
definite point
on the
line.
If
the
according as the sign is

these to the right or left
measures of
These numbers are called the
posit ve of negative.
said to be
are
segments
^responding segments, and the
correspondThe points
with the unit of length.
takngTof
and
fZUable
comm
ing to rational
There are,
IS
Lm
as
points.
numbers may be called rational
points on
number
of
however, an infinite
points.
are not rational
please
we
as
nearly
the
Although we may
by choosing more and
line
approach
mo
quite reach them in

line, we can never
atiTnal points on the
of the points coincase
example is the
way The simplest

of which
.diagonal of a square, the sides
cSnlw'th one end of the
line L
the
ong
a
the diagonal lies
unit of length, when
h
are
with any rational point.

and its other end coincides
any other case of incommensurThus without considering
,n poira-s than
ability!'
we
see that the line
is infinitely richer
in numbers.
thpsustem of rational numbers
all
we desire to follow arithmetically
Hence it is clear that if
rational numbers are
the straight line, the
the properties of
this system by the
extend
and it will be necessary to
tumcS.
numbers.
creation of other
3, 4]
23
Returning to the point of view of the pure mathematician,

now describe Dedekind's method of introducing the
irrational number, in its most general form, into analysis.*
Let us suppose that by some method or other we have divided
4.
we
shall
all the rational numbers into two classes, a lower

class A and an
upper class B, such that every number a of the lower class is less
than every number f$ of the upper class.
When this division has been made, if a number a belongs to
the class A, every number less than a does so also;
and if
number
/?
belongs to the class B, every
number
greater than
ft
does so also.
Three different cases can
The lower
(I)
class
arise
can have a greatest number and the upper

no smallest number.
class
This would occur,
every number
upper
class all
(II)
if, for example, we put- the number

5 and
than 5 in the lower class, and if we put in the
the numbers greater than 5.
less
The upper
can have a smallest number and the lower

no greatest number.
This would occur if, for example, we put the number
5 and
class
class
all
the numbers greater than 5 in the upper class, while in

the lower
we put all the numbers less than 5.
class
It
is
impossible that the lower class can have a greatest
number m, and the upper class a smallest number n, in the

same division of the rational numbers; for between the rational
numbers m and n there are rational numbers, so that our
hypothesis that the two classes contain all the
rational numbers is
contradicted.
But a third case can

(III)
arise
The lower
class can have no greatest number and the upper

no smallest number.
For example, let us arrange the positive integers and
their
class
squares in two rows, so that the squares are

underneath the
numbers to which they correspond. Since the square of a
fraction in its low est terms is a fraction
whose numerator and
Dedekind (1831-1916) published his theory in Stetigkeit

und irrationale
Zahlen (Braunschweig, 1872); English translation
in Dedekind's Essays on Number
(Chicago, 1901).
24
[oh. i
there are not rational

denominator are perfect squares,* we see that
...
numbers whose squares are 2, 3, 5, 6, 7, 8, 10,11,
,
12
4-
4 5 6 7 8 9 10 11 12 13 14 15 16 ...
squares are as near
However there are rational numbers whose
instance, the numbers
these numbers as we please. For
1 2 3
1-5, 1-42, 1-415, 1-4143,
...
1-4142,
1, 1-4, 1-41, 1-414,
...
2,
the squares of the terms

form an upper and a lower set in which
the squares of the terms in the
in the lower are less than 2, and
find a number in the upper
upper are greater than 2. We can
their squares differ
and a number in the lower set such that

from 2 by as little as we please.f
contaimng al
Now form a lower class, as described above,
rational
positive
all the
negative rational numbers, zero and
class
upper
an
and
numbers whose squares are less than 2;
are
squares
numbers whose
all the positive rational
set
containing
one
Then every rational number belongs to
class is less than
Also every number in the lower
or the other.
lower class has no greatest
every number in the upper. The
number.
smallest
number and the upper class has no
class
greater than 2.
obtained a division of all the

of this kind, the lower class
rational numbers into two classes
upper class no smallest
having no greatest number and the
5
When by any means we have
by
number, we create a new number defined

and we say that
call it an irrational number,
all
the rational numbers
rational
numbers
of its
of its
upper
lower
class,
is
defined
described above.
by
it is
and
less
greater than
than
all
the
class.
Such divisions are nsually called
number J2
We
this division.
sections.%
The
the section of the rational
Similar sections would
irrational
numbers
define the irrational
is
The system of irrational numbers
numbers , 4/5, etc.
a
into
numbers
divisions of the rational
e iven by all the possible
rational
every
B, such that
lower class A and an upper class
^.^^ "'
needed,
a formal proof of this statement is
Pure Mathematux (5th
translation, p. 14, or Hardy, Course of
~If
fCf.
Hardy,
loc. cU., p. 8.
{French, coupure; German, SchniU.
ed., 1928), 6.
EngBsh
4-6]
number
is
in one class or the other, the
numbers
class being less
lower class
than the numbers of the upper

has no greatest number, and the upper
25
of the lower
class,
class
while the
no smallest
number.
In other words, every irrational number
is defined by its section (A, B).
It may be said to "correspond" to
this section.
The system of rational numbers and
irrational numbers
together make up the system of real
numbers.
The
rational
numbers themselves "correspond" to
divisions of rational
numbers.
For instance, take the rational number

m. In the lower class A put all
the rational numbers less than m,
and
itself.
In the upper class B put all
the rational numbers greater than
m. Then
corresponds to this division
oi the rational numbers.
Extending the meaning of the term section,

as used above in the definition
of the irrational number, to
divisions in which the lower and upper
classes
have greatest or smallest numbers, we
may say that the rational number
corresponds to a rational section (A, B),*
and that the irrational numbers
correspond to irrational sections. When
the rational and irrational numbers
are defined in this way, and
together form the system of real numbers,
the
real number which corresponds
to the rational number
(to save confusion
it is sometimes called the
rational-real number) is conceptually
distinct from
m. However, the relations of magnitude,
and the fundamental operations
for the real numbers, are defined
in such a way that this rational-real
number
nas no properties distinct from those
of m, and it is usually denoted by
J the
same symbol.
6. Relations of Magnitude for Real

Numbers. We have extended our
conception of number. We must now
arrange the system of real numbers
order; i.e. we must say when two
numbers are equal or unequal to, greater
or less than, each other.
In this place we need only deal with cases

where at
is irrational.
An irrational number is never equal to
least
one of the numbers
a rational number.
They
are always
different or unequal.
Next, in 5, we have seen that the irrational

number given by the section
(A, B) is said to be greater than the
rational number m, when
is a member
of the lower class A, and that the rational
number m is said to be greater than
the irrational number given by the section
(A, B), when m is a member of the
upper class B.
a nUmber m COUld corres nd

<*> two sections:
P
the one named in
!
and that in which the lower class A contains
all the rational numbers
less than m and the upper class B,
m and all the rational numbers greater than m
lo save ambiguity, one of these sections only
must be chosen.
rati
^*T
l
the text,
26
Thev are different
"x.tL-
two
sections
when they are given by different
,.
by
h
r.",
ratlnal
rr,-
numbers, one
of
>
them
is
_,
< is
used in aea
If
- - -
greater than the other,
end of
and
5.
nnmbers there
as will
(I)
is an infinite number
of real nnmbers,
system
the
similar property holds for
different rational
numbed A
now be shown:
Between any W o
sr- .-
the
sections, as explained at
Between any two
7
of
:S
different real
are given
r= * -- * ';rr'."
or unequal,
that the notation

It will be observed
sr^
[ch.
different real
numbers
of rational numbers.
rational, the property
a and a are
is
a, a' there are
an
** number
known.
:,t^." r.\zs s -'
'
îrririr=^l r:r - ...
greater than a
than a and
and these numbers are less
the case when a <a.
to
applies
proof
P
A similar P
general
proved can be made more
The"lmlt which has just been
&er of
tkere are an infinUe
two different real numbers
(II) Between any
B' of
a';
irrational numbers.
a<a./
given numbers, and suppose
n' hP the two 8
and p", such that
numbers
Tl any two
atlna !
and
p" there
can show that between
/J
^Tââl
number.
must be an
If this
does not
< ?
<^
we
<f number, the
irrational
lie
between , and ,. by
^H
6-8]
27
For we can
it a suitable rational number we can make it do so.
two rational numbers m, n, such that m<i<n and (n - m) is less than
- ft). The number ft - m + i is irrational, and lies between ft and ft'
adding to
find
{ft'
8.
We
Dedekind's Theorem.
shall
now prove
very im-
portant property of the system of real* numbers, which will be
used frequently in the pages which follow.

If the system of real numbers
in such a
(i)
way
is
divided into two classes
and B,
that
each class contains at least one number,
(ii)
every
number belongs
(iii)
every
number in
to
one class or the other,
the lower class
is less
than every number
in the upper class B;
number a such that

every number less than a belongs to the lower class A, and
every number greater than a belongs to the upper class B.
The separating number a itself may belong to either class.
Consider the rational numbers in A and B.
These form two classes e.g. A' and B' such that every rational
number is in one class or the other, and the numbers in the lower
class A' are all less than the numbers in the upper class B'.
As we have seen in 4, three cases, and only three, can arise,
then there is a
The lower
(i)
class
class
A' can have a
The rational number

it is
number
and
the
upp
the number a of the theorem. F< r

number a less than m belongs to the
a member of this class. Also every real number
is
clear that every real
class A, since
b,
greatest
B' no smallest number.
m is
greater than m, belongs to the class B.
rational, since b
If b is irrational,
This
is
evident
then belongs to the class B', and

a rational number n between
if
b is
B' is part of B.
we can take
Then n belongs to B, and therefore b does so also.

(ii) The upper class B' can have a smallest number
m and b.
and
the
lower class A' no greatest number.

It follows, as above, that the rational
number
is
the
number
a of our theorem.
(iii)
The lower
class
class
A' can have no
greatest
*It will be observed that the system of rational
property.
number and
the
upper
B' no smallest number.

numbers does not possess
this
28
[ch.
be the irrational number denned by this section (A', B').

Every rational number less than m belongs to the class A, and
every rational number greater than m belongs to the class B.
We have yet to show that every irrational number less than m
Let
belongs to the class A, and every irrational
m to the
number
greater than
class B.
But this follows at once from 6. For if m' is an irrational

number less than m, we know that there are rational numbers
between m and m'. These belong to the class A, and therefore
m' does so
also.
m>
m.
argument applies to the case when
a belongs to
number
separating
the
discussion
above
the
In
to the upper
belongs
it
(i);
case
in
rational,
is
and
the lower class,
similar
class,
and
is
again rational, in case
belong to either
class, in case
(ii);
it is irrational,
and may
(iii).
The Linear Continuum. Dedekind's Axiom. We return now

has been
to the straight line L of 3, in which a definite point
length.
of
taken as origin and a definite segment as the unit
We have seen how to effect a correspondence between the
The
rational numbers and the "rational points" of this line.
"rational points" are the ends of segments obtained by marking
9.
Fig.
2.
equal to multiples or sub-multiples

of the
of the unit segment, and the numbers are the measures
off
from
on the
line lengths
corresponding segments.
be a segment incommensurable with the unit segment.

The point A divides the rational points of the line into two classes,
such that all the points of the lower class are to the left of all the
The lower class has no last point, and
points of the upper class.
Let
OA
the upper class no
We
first
the measure of
point.
an
the segment
then say that
is
irrational point of the line,
OA
is
the irrational
and that
number
defined
the rational numbers.
by this section of
Thus to any point of the line L corresponds a
and to different points of the line correspond
numbers.
real
number,
different
real

8-10]
There remains the question

correspond a point of the line
For all rational numbers
affirmative.
amounts to
To
every real
number does
29
there
we can answer
the question in the
When we turn to the irrational numbers, the question
If all the rational points of the line are divided

a lower and an upper, so that the lower class has no
last point and the upper class no first point, is there one, and only
one, point on the line which brings about this separation ?
this:
into two classes,
The existence of such a point on the line cannot be proved.

The assumption that there is one, and only one, for every section
of the rational points is nothing less than an axiom by means of
which we assign its continuity to the line.
This assumption is Dedekind's Axiom of Continuity for the
line.
In adopting it we may now say that to every point P of
the line corresponds a number, rational or irrational, the measure
of the segment OP, and that to every real number corresponds a point
P of the line, such that the measure of OP is that number.
The correspondence between the points of the line L (the linear
continuum) and the system of real numbers (the arithmetical conis now perfect.
The points can be taken as the images
numbers, and the numbers as the signs of the points. In
consequence of this perfect correspondence, we may, in future, use
tinuum)
of the
the terms
10.
number and point
in this connection as identical.
The Development of the System of Real Numbers.
instructive to see
It is
how
the idea of the system of real numbers,

as we have described it, has grown.* The irrational numbers,
belonging as they do in modern arithmetical theory to the realm
of arithmetic, arose
their aid.
from the geometrical problems which required

first as an expression for the ratios of
They appeared
incommensurable pairs of lines. In this sense the Fifth Book of

Euclid, in which the general theory of Ratio is developed, and the
Tenth Book, which deals with Incommensurable Magnitudes, may
be taken as the starting point of the theory. But the irrationalities
which Euclid examines are only definite cases of the ratios of
incommensurable lines, such as may be obtained with the aid of
ruler and compass; that is to say, they depend on square roots
d.
*Cf. Pringsheim, "Irrationzahlen u. Konvergenz unendlicher Prozesse," Enc.

math. Wiss., Bd. I, Tl. I, p. 49 et seq. (Leipzig, 1898).
30
alone.
lines
The idea that the
determined
ratio of
[ch.
any two such incommensurable
a definite (irrational)
number did not occur
to
him, nor to any of the mathematicians of that age.

Although there are traces in the writings of at least one of
idea that
the mathematicians of the sixteenth century of the
number,
rational
every
every irrational number, just as much as
sequence
ordered
possesses a determinate and unique place in the
still considered to
of numbers, these irrational numbers were
is
only from certain cases of evolution, a limitation which
arise
commanding position of Euclid's methods

to the belief that the problem of finding
partly
Geometry, and
th powers of
th root of an integer, which lies between the n
the n
partly due to the
two consecutive
integers,
was the only problem whose solution
could not be obtained in terms of rational numbers.

The introduction of the methods of Coordinate Geometry by
Calculus
Descartes in 1637, and the discovery of the Infinitesimal
Leibnitz and Newton in 1684-7, made mathematicians regard
by
applicability of number
this question in another light, since the
Coordinate
spatial magnitude is a fundamental postulate of
to
"The view now prevailed that number and quantity

objects of mathematical investigation, and that the two
Thus
similar as not to require careful separation.
Geometry.
were
the
were so
number was applied
to quantity without
any
hesitation, and,
to
conversely, where existing numbers were found inadequate
that
ground
sole
measurement, new ones were created on the
every quantity must have a numerical measure."*
was reserved for the mathematicians of the nineteenth

century notably Weierstrass, Cantor, Dedekind and Heineto
Until their writings
establish the theory on a proper basis.
expression for the
an
as
upon
looked
was
number
appeared, a
which was
another
by
line
of
a
result of the measurement
It
regarded as the unit of length. To every segment, or, with the

natural modification, to every point, of a line corresponded a
number, which was either rational or irrational; and by

an
the term irrational number was meant a number defined by
definite
infinite
set of arithmetical operations (e.g. infinite decimals or
continued fractions).
*Cf. Russell, Principles of
The
justification
Mathematics (1903), Ch.
for
regarding such an
XIX,
417.
10]
31
unending sequence of rational numbers as a definite number was

considered to be the fact that this system was obtained as the
equivalent of a given segment
measurement
as those
other segments.
However
this that, conversely,
by the
which gave a
any
it
same methods of
number for
any way follow from
aid of the
definite rational
does not in
arbitrarily given arithmetical represen-
tation of this kind can be regarded in the above sense as an

irrational
number; that
is
to say, that
we can
consider as evident
the existence of a segment which would produce by suitable
measurement the given

has the credit of
first
arithmetical
representation.
Cantor*
pointing out that the assumption that a
segment must correspond to every such sequence is

it admit of proof, but involves an
actual axiom of Geometry.
Almost at the same time Dedekind
showed that the axiom in question (or more exactly one which is
definite
neither self-evident nor does
equivalent to
it)
gave a meaning, which we can comprehend,
any sufficient definition,

had been spoken of as the continuity of the line.
To make the theory of number independent of any geometrical
axiom and to place it upon a basis entirely independent of
measurable magnitude was the object of the arithmetical theories
associated with the names of Weierstrass, Dedekind and Cantor.
The theory of Dedekind has been followed in the previous pages.
Those of Weierstrass and Cantor, which regard irrational
numbers as the limits of convergent sequences, may be deduced
from that of Dedekind. In all these theories irrational numbers
appear as new numbers, to each of which a definite place in the
domain of rational numbers is assigned, and with which we can
operate according to definite rules.
The ordinary operations of
arithmetic for these numbers are defined in such a way as to be in
agreement with the ordinary operations upon the rational numbers.
They can be used for the representation of definite quantities, and
to them can be ascribed definite quantities, according to the axiom
of continuity to which we have already referred.
to that property which, so far without
*Math. Annalen, 5 (1872), 127.
32
REFERENCES.
Bromwich, Theory
Dedekin'd,
of Infinite Series (2nd. ed., 1926,
Stetigkeit
und
translation in Dedekind's Essays on
De la Vallee
duction,
App.
I).
irrationale Zahlen (Braunschweig, 1872); English
Poussin, Cours
Number
(Chicago, 1901).
d? Analyse,
(5e ed., Paris, 1923), Intro-
1.
Dini, Fondamenti per la Teorica delle Funzioni di Variabili Reali (Pisa,

1878), 1-9.
(4 e ed., Paris
Goursat, Cours
a"'
Hobson, Theory
of Functions of a Real Variable,
Analyse,
1923), Ch.
1
I.
(3rd ed., 1927), Ch.
Knopp, Theorie und Anwendung der unendlichen Reihen

1924; English translation, 1928), Kap. I-II.
Pringsheim, Vorlesungen
I, Kap. I-III.
iiber
(2 Aufl.,
Zahlen- und Funktionenlehre,
I.
Berlin,
(Leipzig,
1916), Absch.
Russell, Principles of Mathematics (1903), Ch.

Stolz u. Gmeiner,
Absch. VII.
Tannery,
Ch.
Theoretische
Arithmetik
XXXIV.
(Leipzig,
Introduction a la Theorie des Fonctions,
1902),
Abth.
II,
(2 e ed., Paris, 1914),
I.
And
Pringsheim,
Enc.
d.
"Irrationalzahlen
math. Wiss., Bd.
I, Tl.
u.
Konvergenz
unendlicher
Prozesse,"
I (Leipzig, 1898).
WM
CHAPTER
II
INFINITE SEQUENCES
AND SERIES
We
are accustomed to speak of the

numbers, the integers which
prime
positive integral numbers, the
These are all examples of infinite sets
are perfect squares, etc.
11. Infinite Aggregates.
which have more than a finite number of terms.

In mathematical language they are termed aggregates, and the
theory of such infinite aggregates forms an important branch of
of
numbers or
sets
modern pure mathematics.*

The terms of an aggregate
may
be
finite or infinite.
are
all
different.
usually called infinite aggregates, but sometimes
them simply
as aggregates.
Their number
In the latter case the aggregates are
we
shall refer to
After the discussion in the previous

if we speak of an aggregate
an aggregate of numbers. The
chapter, there will be no confusion

of points
on
a line instead of
two notions are identical. We associate with each number the

It may happen that, however
point of which it is the abscissa.
we go along the line, there are points of the aggregate further

In this case we say that it extends to infinity. An aggregate
said to be bounded on the right, or bounded above, when there is
far
on.
is
be taken as the founder of this theory, which the Germans call

In a series of papers published from 1870 onward he showed its
importance in the Theory of Functions of a Real Variable, and especially in
the rigorous discussion of the conditions for the development of an arbitrary
* Cantor
may
Mengenlehre.
function in trigonometric series.
Reference may be made to the standard treatise on the subject by W. H. and

Grace Chisholm Young, Theory of Sets of Points (1906), and to the earlier chapters
of Hobson's Theory of Functions of a Real Variable, Vol. I, already cited.
The most recent book on the subject, from the advanced point of view, is
Mengenlehre, by Hausdorff (2 Aufl., Berlin, 1927).
33
AND SERIES
INFINITE SEQUENCES
34
[ch. ii
no point of k to the right of some fixed point. It is said to be

bounded on the left, or bounded below, when there is no point of it to
the left of some fixed point. The aggregate of rational numbers
greater than zero
rational
numbers
The aggregate of
bounded on the left.
than zero is bounded on the right. The
positive numbers less than unity is bounded
is
less
aggregate of real
above and below; in such a case we simply say that

The aggregate of integral numbers is unbounded.
12.
an
it is
bounded.
When
The Upper and Lower Bounds of an Aggregate.

number
M which
aggregate (E)* is bounded on the right, there is a
possesses the following properties:
no number of (E) is greater than M;

however small the positive number e
of (E) greater than
We
can arrange
may
be,
there is a
e.
the real numbers in two classes,
all
relative to the aggregate.
number
number x
and B,
be put in the class
will
if one or more numbers of (E) are greater than x.

in the class B if no number of (E) is greater than
It will
x.
be put
Since the
classes,
bounded on the right, there are members of both

and any number of the class A is smaller than any
number
of the class B.
aggregate
By
is
Dedekind's Theorem
( 8)
the two classes, such that every
the class A, and every
number
there
is
number
greater than
number
than
less
M separating
M belongs to
M to the class B.
We
of our theorem.
number
shall now show
of
(E) greater than M.
number
In the first place, there is no
+h (h> 0). Then the
For suppose there is such a number
+\h, which is also greater than M, would belong to
number
would not separate the two classes A and B.
the class A, and
In the second place, whatever the positive number e may be, the
number M-e belongs to the class A. It follows from the way
in which the class A is defined that there is at least one number
that this is the
of (E) greater
than
This number
It
may
M-e.
is
called the upper
belong to the aggregate.
*This notation
term ensemble.
is
convenient, the letter
bound
of the aggregate (E).
This occurs
being the
when the aggregate
first letter
of the
French
INFINITE SEQUENCES
11-13]
AND SERIES
35
number of terms.
But when the aggregate
number of terms, the upper bound need not
For example, consider the rational numbers whose
contains a finite
contains an infinite
belong to
it.
squares are not greater than
2.
This aggregate is bounded on
upper bound being the irrational number J2, which
does not belong to the aggregate.
On the other hand, the aggregate of real numbers whose squares are not greater than 2 is also
bounded on the right, and has the same upper bound. But
the right,
its
J2
belongs to this aggregate.
upper bound
of the aggregate (E) does not belong to
must be an infinite number of terms of the aggregate
- e, however small the positive number e may
between
and
If the
there
it,
number of such terms, there would

be no term of (E) between the greatest of them and M, which is
contrary to our hypothesis.
be.
It
If
there were only a finite
can be shown in the same way that when an aggregate (E)

the left, there is a number m possessing the following
bounded on
is
properties:
no number of (E)
is
smaller than
m;
however small the positive number

of (E) less than
The number
defined in this
of the aggregate (E).
the aggregate
when
it
As above,
may
be,
there is a
number
way
it
is
called the lower
may, or may
bound
not, belong to
has an infinite number of terms.
the aggregate has only a

to
m+e.
finite
number
of
terms
it
But when
must belong
it.
13. Limiting Points of
an Aggregate.
'
There are an
infinite
111
"<""
2'
number
Consider the aggregate
3'
of points of this aggregate in
however small, extending from the origin to the

Such a point, round which an infinite number of points
interval,
aggregate cluster,
More
is
any
right.
of
an
called a limiting point* of the aggregate.
a will be a limiting point of the aggregate (E) if,

however small the positive number e may be, there is in (E) a point
other than a whose distance from a is less than e.
If there be one
definitely,
French, point
limite;
German, Hdufungspunkt.
INFINITE SEQUENCES
36
AND SERIES
[oh. ii
there will be an infinite

such point within the interval (a e, a -he),,
and a n were the
number, since, if there were only n of them,
point other than a whose
nearest to a, there would not be in (E) a
*.
In that case a would not
\a-a
than
H
a was less
distance from
hypothesis.
be a limiting point, contrary to our
one limiting point. The
than
An aggregate may have more
form an aggregate with
rational numbers between zero and unity
every point of the
an infinite number of limiting points, since
that some of
noticed
It will be
a limiting point.
segment
(0, 1) is
to it, and some, namely

the limiting points of this aggregate belong
its end-points, do not.
the irrational points of the segment and
section,
In the example at the beginning of this
1
1
ij
the lower bound, zero,
3'"' n'"'
'
a limiting point,
and does not belong to
upper bound, unity, belongs to the aggregate,
The
the aggregate.
is
2'
and is not a limiting point.

The set of real numbers from to 1, inclusive,
which is identical with its limiting points.
14. Weierstrass's
and
Theorem.
is
An infinite aggregate,
an aggregate
bounded above
below, has at least one limiting point.
and have
Let the infinite aggregate (E) be bounded,
bounds.
for its upper and lower
We
can arrange
the aggregate (E).
to
the real numbers in two classes relative
A
class
the
to
belong
to
said
be
will
number x
terms of (E) are greater than

contrary case.
be said to belong to the class (B) in the
when an
will
all
M and m
infinite
number
of
x.
It
to the class B, there are

belongs to the class A and
in the class A is less
numbers of both classes. Also any number
than any number in the class B.
two
Theorem, there is a number ^ separating the
Since
By
Dedekind's
p-e
However small the positive number e may be,
interval
the
Thus
B.
class
the
to
e
belongs to the class A, and p +
of the aggregate.
contains an infinite number of terms
classes.
Hence
/u is
a limiting point.
numbers aând h taken
the difference between two real
With this
of (a - b).
modulus
or
value
and to call it the absolute
""^rTr^l^note
positive,
by
notation
\x
\a
- b\,
+ y\^\x\ + \y\,
\xy\
= \x\\y\.
INFINITE SEQUENCES
13-15]
As
will be seen
from the example
AND SERIES
bounds
of 13, the
37
M and m
may be limiting points.

An infinite aggregate, when unbounded,
need not have a limiting

the set of integers, positive or negative.
But if the
aggregate has an infinite number of points in an interval of finite
point;
e.g.
length, then
15.
of
it
must have at
..
7/
when some law
is
speak of an
infinite
sequence
,.
given according to which the general term u

n
be written down.
The sequence
is
We
Convergent Sequences.
numbers
may
least one limiting point.
u^
said to be convergent
u^
and
have the limit A, when, by indefinitely

increasing n, the difference between A and u becomes, and thereafter
n
remains, as small as we please.
This property
is
to
so fundamental that
it is well to put it more

The sequence is said to be convergent and to
A, when, any positive number e having been chosen,
precisely, as follows:
have the limit
as small as we please, there

I
is
A _ u n\ <
a positive integer v such that
e,
provided that
n^v.
For example, the sequence

l
I
2'
'
has the limit zero, since l/n

than 1/e.
The notation that
is
3>"'
is less
employed
than
'"
e
for all values of
in this connection
n greater
is
lim u n = A,
and we say that as n tends
The
to infinity,
u n has
the limit
A*
employed to denote an arbitrarily small

number, as in the above definition of convergence to a
limit as n tends to infinity.
Strictly speaking, the words as small
as we please are unnecessary in the definition, but they are
inserted
letter e is usually
positive
as
making
clearer the property that
is being defined.
very frequently have to employ the form of words
which occurs in this definition, or words analogous to them, and
We
shall
*The phrase " u n tends to the
limit
as
n tends to
infinity"
is
also used.
INFINITE SEQUENCES
38
AND SERIES
[oh.
with them by
advised to make himself familiar
following sequences are convergent
formally testing whether the
the beginner
or not
is
1,11
()1,
(6)i,
5.
2'
W
2
<<*
(C)
1+ + 2*
1 +
1+
2'
2
1.
l
'
'
-1
en.
-| 1--
- 1 '--
J
-
distinct limits
sequence cannot converge to two
and B.
Then there are only a

were possible, let *<\ \A-B\.
the interva
of the sequence outside
finite number of terms
A lnis
value
the
to
converges
(A - e A +.), ^nce the sequence
limit B
the
also
has
sequence
contradicts the statement that the
interval
the
in
terms
of
finite number
for we would only have a
B as centre.
of the same length with
contained in the
test of convergency
the
of
application
If this
The
definition involves the
of the limit
knowledge
it.
frequently be impossible to use
when
the convergence of a sequence,
The
we
Thus
it
wil
requ.red criterion for
are not simply asked to
not the limit is contained

whether a given number is or is
principle of convergence:in the fundamental general
existence of a hmit
A necessary and sufficient condition for *
u lt u 2 u 3 ...
to the sequence
test
is that
\u^-u
as
that
n becomes
positive integer v exists such
integer
p.
n^v, for every positive
small as we please
when
More exactly:
A necessary and
to the
sufficient condition
uv
sequence
u2
for the existence of a
u3
..
7
limti
...
we
has been chosen, as small as
positive number e
such that
be a positive integer v
shall
please, there
n^v, for every positive integer p.
is that,
if
We
any
-u\<6, when
shall first of all
show that the condition
is
necessary;
i.e. it
In the words of
bezeichnet (Allg. Funct.-Theone, pp.,
loc. cit.,
Enc.
d.
math. Wiss. p. 66.
6,
260)-ist der
un damentalen
AND SERIES
INFINITE SEQUENCES
15]
the sequence converges, this condition

if
this condition is satisfied,
words, the condition
The condition
(i)
is satisfied;
secondly that,
the sequence converges;
in other
is sufficient.
is necessary.
Let the sequence converge to the limit A.

Having chosen the arbitrary positive number
We know
39
that there
e,
then take
Je.
is
\A-u n \<ê, when n^v.

But
Therefore
\u n+p
+3,
- u n ) = (u n+v -A)+(A-u n ).
- u n \^\u n+p -A\
<
+\A-u n
n>
if
fc
for every positive integer p,
v,
The condition is sufficient.

must examine two cases; first, when the sequence contains
an infinite number of terms equal to one another; second, when
(ii)
We
it
does not.
(a)
Let there be an
Then,
\
infinite
number
of terms equal to A.
if
u n+P -u n \<e, when n
we may take u n+v = A
for
^ v,
and p
some value
is
any positive
of p,
integer,
and we have
\A-u n \<, when n^v.

Therefore the sequence converges, and has
(b)
Let there be only a
number
finite
of
for its limit.
terms equal to one
another.
Having chosen the arbitrary
We know that there

\
u n+ v ~
number
n^N,
e,
then take \
N such that
for every positive integer p.
we have
\
all
positive
a positive integer
<|f, when
It follows that
Therefore
is
u n~ u N\<2 when
>
n^N.
the terms of the sequence
u n+s ...
Un+i, u n+2
within the interval whose end-points are u N - Je and u +|e.
N
There must be an infinite number of distinct terms in this
sequence.
Otherwise we would have an infinite number of terms
,
lie
equal to one another.
AND SERIES
INFINITE SEQUENCES
40
[oh.
ii
formed by the distinct

Consider the infinite aggregate (E)
u v u 2 u s ...
terms in
limiting point
(Cf.
bounded and must have at least one

above interval.
within, or at an end of, the
aggregate
This
is
14.)
.,
were
point A', for if there

There cannot be another limiting
and the formula
we could choose e equal toftA-A'] say,
\u n+p
-u n \<e, when n ^ v,
shows that
all
the terms of the sequence
u2
uv
w3
...
within an interval of length

except a finite number, would lie
A' are limiting points of the
\\A - A'\. This is impossible if A,
aggregate.
one limiting point A.

Thus the aggregate (E) has one and only
We shall now show that the sequence
u 3 ...
uv u2
,
converges to
We
as
n tends to
qo
A = (u n - u N + (u N - A).
\u n - A\^\u n -u N +\u N -A\
un -
have
Therefore
<
<
|e
ie,
when
when
6j
n^N,
n^N.
for its limit.

Thus the sequence converges, and has A
theorem:
We have therefore proved this
convergence
A necessary and sufficient condition for the
u 2 u z ...
u\,
sequence
,
positive
is that, to the arbitrary
of the
number
e,
there shall correspond
positive integer v such that

Un+j>
_ Un
<e
when
n^vjor
every positive integer p.
It
is
ea^to show
that the above condition
may
be replaced by the
following:
In order
that the sequence
U\,
uv
**3>
that, to the arbitrary positive
nuiy converge, it is necessary and sufficient

integer n such that
there shall correspond a positive
K+p -nl < /or every Positive

It is clear that
if
number
e,
integer p.
the sequence converges, this condition
is satisfied
by n = v.
INFINITE SEQUENCES
15, 16]
Further,
number
to the
AND SERIES
41
and is an arbitrary positive number,

there corresponds a positive integer n such that
if
<=
M n+?> un\
\un+P - u n+p ,\
!
But
< h e fr every
^ \un +v
< 2 + l
positive integer p.
-un +
- n
are any positive integers.
also satisfied, and the sequence con
\
\u n+p .
when p\ p"
>
Therefore the condition in the text
is
verges.
and Oscillatory Sequences.*
16. Divergent
Wj,
u2
u3
does not converge, several different cases
In the
(i)
if
the terms
first place,
When
arise.
may have
any positive number A, however
the sequence
...
the property that
large, is chosen, there is a
u n >A, when nîv.

In this case we say that the sequence
+ oo and we
diverges to
divergent,
is
and that
it
write this
lim u n = + oo
n>-oo
(ii) In the second place, the terms may
have the property that
any negative number - A is chosen, however large A may be,
.
if
there
is
u n <-A, when n^v.

In this case we say that the sequence
diverges to
oo
and we write
lim
u.
divergent,
is
and that
it
this
00
7t>OU
The terms
when n
is
sequence may all be very large in absolute value,

very large, yet the sequence may not diverge to +oo or
of a
- oo
A sufficient
whose general term is
to
illustration of this
is
given by the sequence
l) n n.
After some value of n the terms must all have the same
sign,
the sequence is to diverge to + oo or to - oo the sign
being
positive in the first alternative, and negative in the second.
if
When
(lii)
to
+00
or to
the sequence does not converge,
oo
it
is
and
does not diverge
said to oscillate.
*In the first edition of this book, the term divergent

was used as meaning
merely not convergent. In this edition the term is applied
only to the case of
divergence to +oo or to -oo, and sequences which oscillate
infinitely
among
are placed
the oscillatory sequences.
INFINITE SEQUENCES
42
AND SERIES
[oh. ii
oscUlate finitely, if there is

oscillatory sequence is said to
n; and it
such that \u n \<A,for all values of
positive number
An
when there is no such number.

term is ( - 1)"
For example, the sequence whose general
is (-l)-n
term
general
finitely; the sequence whose
said to oscUlate infinitely
is
oscfilates
oscillates
infinitely.
We may
quences by
i.e.
se-
and divergent
distinguish between convergent
has a finite limit,
saying that a convergent sequence
divergent sequence
a
is a definite number;
lim u n = A, where
has "an 'infinite
limit, i.e.
lim w n = +00 or lim
u -
co
and the terms

co
must be remembered that the symbol
conventional
purely
tend to infinity, have
infinite, infinity and
which the
in
Phrases
infinity;
meanings. There is no number
previously,
us when we have
term is used have only a meaning for
them.
to
meaning
by definition, attached a
are using a short and
When we say that n tends to infinity, we
that assumes an endless
convenient phrase to express the fact
become and remain greater than
series of values which eventually
So far we have supposed n
any arbitrary (large) positive number.
But
it
through integral values only.

in this connection, to advance
removed later.
remark applies to the phrases
restriction will be
similar
divergence to
to our
This
+o or
earlier use of
and oscillating infinitely, as well as

sequence and infinite aggrethe terms an infinite number, infinite
has been attached to the
In each case a definite meaning
gates.
to
co
employed only with that meaning.

by the
our work might be simplified
It is true that much of
the
assuming
by
and
-co,
introduction of new numbers + co,
have
we
which
line
upon the
term, and
it is
existence of corresponding points
these
numbers. But the creation of
nsed as the domain of the
matter
a
be
would
these points,
numbers, and the introduction of
for separate definition.
17. 1.
the sequence
Monotonic Sequences. If the terms of
u lt u 2i u 3 ...
,
relations
satisfy either of the following
or
u x û 2 û z ...
u 1 û 2 û 3 ...
the sequence is said to
be monotonic.
û n
û ni
,
...
...
INFINITE SEQUENCES
16-17. 2]
In the
first case,
AND SERIES
43
the terms never decrease, and the sequence
may
be called monotonic increasing) in the second case, the terms never

increase,
and the sequence may be
Obviously,
when we
called monotonic decreasing*
are concerned with the convergence or
divergence of a sequence, the monotonic property,
need not enter
The
till
tests for
if
such
exist,
after a certain stage.
convergence or divergence are extremely simple in
the case of monotonic sequences.
If the sequence
is
Uv
monotonic increasing, and
u%t
its
WjJj ...
terms are
number B,
the sequence is convergent
P such
un
that
fS
($
tk
B for every positive
than some fixed
all less
and has for
its
limit
a number
integer n.
Consider the aggregate formed by distinct terms of the sequence.
bounded by u x on the left and by B on the right. Thus it

must have an upper bound /? (cf. 12) equal to or less than B,
and, however small the positive number e may be there will be a
It is
term of the sequence greater than j3 - e.

Let this term be u v
Then all the terms after u v _ 1 are to the
If any of them coincide
right of ft e and not to the right of /?.
with p, from that stage on the terms must be equal.
Thus we have shown that
.
\j3-u n \<e, when
n^v,
and therefore the sequence is convergent and has fi for

The following test may be proved in the same way
its limit.
If the sequence
Uv
Wgj
Wjjj ...
monotonic decreasing, and
its terms are all greater than some fixed

number A, then the sequence is convergent and has for its limit a
number a such that u n
for every positive integer n.
It is an immediate consequence of these theorems that a monotonic sequence either tends to a limit or diverges to + oo or to - oo
is
âÂ
17. 2.
The Upper and Lower Limits of Indetermination of a Bounded Sequence.
Let u lf u 2 u 3
,
Let
and
and so
...
be a sequence, bounded above and below.

1 be the upper bound of u ly u 2 u 3 w 4
M
M
be the upper bound of u 2 %, uif
. . .
. . .
on.
*The words
words in
and steadily decreasing are sometimes employed

and when none of the terms of the sequence are equal, the
steadily increasing
in this connection,
the stricter sense are
added.
AND SERIES
INFINITE SEQUENCES
44
mv m
Similarly let
m3
2,
sequences.
M ^M \=M^ ...^m r
Then
Thus
ii
be the lower bounds of the corresponding
...
[oh.
lim
Mn exists
( 17. 1.).
n>qo
Let this limit be A.
To the
But
number
arbitrary positive
such that
A = Mn <A +
v is
^.M v when
<A+
when
e,
i?
n,
Also at least one of the set
v,
when n^v.
the upper bound of u v u v+1 w, + ,,
Therefore u n
Thus u n
there corresponds a positive integer
...
v.
v.
u v u VJrl u v + 2
,
...
(say
u A)
greater than
is
- e,
and thus greater than A - e.

Take v' a positive integer > N.
A^M '^M
Then
And
u v ,) of the
way we have the infinite
at least one (say
In this
UA
all
A-
greater than
u v u v +1
set
',
'
u' +2 ,
...
is
greater than
A-
e.
set
Uyrr, ...
Uy>,
e.
associated with the bounded

thus shown that there is a number
... which has the following properties
We have
sequence u x u 2 u z
,
<A
+ e,for all positive integers greater

If t is an arbitrary positive number, u n
e,for an infinite number of
andu
on
e;
depending
integer
a
n
definite
than
>A-
positive integers.
Similarly for the lower bounds
Denoting this limit by
A,
x,
we have
ra 2
ra3 ,
...
we
see that lim
mn
the corresponding result; u n
exists.
>A-
e,
for
depending on e; andu n <k + e, for

all values of n greater than a definite value
an infinite number of values of n.
indetermination
The numbers A and A are called the upper and lower limits of
write
and
we
sequence*
the
of
A = limwn A = Umi*n
n>oo
n>oo
,
It is clear that
AiS
A.
v v v 2 , v z ... be sequences of positive terms, the first

17. 3. Let u x u 2 u 3 ...
equal to unity.
sequence being bounded above, and in the second lim vn being
,
Then lim (unvn ) =lim u n
sequence u x u 2 u z ... is
It is clear that as u x u 2 u 3 are positive and the
is bounded
bounded above, the sequence of positive terms u x v x u 2 v 2 u zv 3 ...
,
above.
French la
German
:
tion.
e.g.
indeterminaplus grande limite and la plus petite limite, or limites $

Other terms are used
obere and untere U nbestimmtheitsgrenze.
oberes (unteres) Limes; and Limes superior
obere (untere) Hdufungsgrenze
(inferior).
INFINITE SEQUENCES
17.2-17.4]
If possible, let
Take 2e=fx' -
Km
/j.,
=//> /x=!im u n
(un vn )
and
45
the upper bound of u v u 2 u z ... be K.

a positive integer v 1? such that
let
Since lim vn = l, there
AND SERIES
is
n>oo
K~
I<^
when w ="i-
Kv-wn = KIK-l|< ffx^<ie, when n^v v

Therefore
un vn <un + e, when n =; i^.
Thus
But
un =/x, there
since lim
n>x>
is
a positive integer
i'
un </x + Jf, when n ^

w n vw
Therefore
where
But
1/
< /a +
the larger of the integers
is
vt
we know
since lim (u n vn ) =/*',
n-><x>
when
and
w.
such that
a,
y2.
jg
r,
v2.
that un vn
> // -
e,
for
an
infinite
number
of values of n.
Therefore y! cannot be greater than
ft.
can be shown that // is noi less than

Hence /x'=/z and the theorem is proved.
Similarly
it
fx.
If the sequence u lt u 2 , u z ...converges, then lim wn = lim un lim un .

n>-oo
n>qo
n>oo
Since lf w 2 w 3 , ... converges, it must be a bounded sequence, and lim u
n,
w-^ 00
lim un both exist.
17. 4.
(i)
n-*oo
We
have to show that they are equal to lim u

n
n>oo
Let
lira
n*
un =l, lim^n =A, and limun = A..

->oo
If possible, let
Then
there
is
A>
n-w
I,
and take 2e=A-l.
a positive integer
v,
such that
un <l + e, when n^v.

But we are given that un > A - e, for an infinite number of values of
These two inequalities cannot both be true.
Thus A cannot be greater than I.
In a similar way we can show that A cannot be less than I.
AÂ.
But
Therefore
(ii)
n.
we must have
A=X=
l.
Conversely, if the upper and lower limits of indetermination of the bounded

u v u 2 , u 3 , ... are equal, then the sequence converges to their common
sequence
value.
We
are given that
lim un =
n>x>
A = A =lim un
n>oo
Thus, to the arbitrary positive number e, there correspond positive integers*

vx and v 2 such that
u
+ e, when n v
,
<A
i=
un > A - e, when n iS
v2,
INFINITE SEQUENCES
46
\u n
Hence
where
v is
the larger of
v %.
->qo
n>co
n><x>
this notation
and
positive integers v x
Km un = lim un = lim un
With
[ch. ii
-A\<e, when n^v,
the two
Thus
(iii)
AND SERIES
con.
easy to establish the general principle of
it is
the existnamely that a necessary and sufficient condition for

c Ms
number
positive
any
that,
is
...
u
if
3
ence of a limit to the sequence u x u 2
positive integer v, such that
a
be
shall
there
please,
we
as
small
been chosen, as
-u < e, when n^v and p is any positive integer.
It* +
vergence (p. 38),
condition is necessary for the

There Ts no difficulty in showing that this
39.)
(Cf.
sequence.
p.
the
of
convergence
that the condition was
The difficulty in our former proof was to show
sufficient.
But,
it is
clear that
with this condition, the sequence u x u 2 u 3

,
is
. .
bounded.
exist.
upper and lower limits of indetermination therefore
unequal.
be
and
k
A
let
before,
as
notation
the same
Its
With
2=A-A.
Take
There
is
a positive integer
\u n+P
such that
-u n \<, when
n^v
and p = l,
2,
(1)
< A + |c, for an infinite number of values of n

u n > A - |e, for an infinite number of values of n
(2)
un
But
And
Let
v,
v',
v" be the
first
positive integers greater than
which
(3)
satisfy (2)
and
(3) respectively.
Then |tv -uv'\> e, contrary to (1).

Therefore A and A are equal, and by
18. Let
l3
A2 Az
,
...
be
an
(ii)
lim un exists.
each
infinite set of closed intervals,
lying within it and having

lying entirely within the preceding, or
length of
n tend to zero
with it a common end-point; also let the
as
n tends
to infinity.
Then
there is one,
and only
one, point
which
internal point of all, or,

belongs to all the intervals, either as an
end-point
of all.
common
a
and after a definite stage, as
a+
a.
Fig.
3.
Let the representative interval A n be given by

an ^ x ^ bn
.
Then we have
and
â 3 a v
a 1 ^.a 2
...
from
AND SERIES
INFINITE SEQUENCES
17.4-19]
of end-points
Thus the sequence
a2
a lt
has a limit, say
a,
47
and a n
â for
a3
>
(!)
every positive integer n
( 17. 1).
Also the sequence of end-points
has a limit, say
Now
less
it is
than
/?,
and
bn
6 3 ...,
bv
h2
for every positive integer
ft
(2)
under the given conditions,
clear that,
ft
( 17. 1).
cannot be
a.
Therefore, for every value of n,
-a n >f$-a = 0.
lim (b n -a n = 0.
bn
But
Therefore this
a=
It follows that
fi.
common limit of the
sequences
(1)
and
(2) satisfies
the inequalities
an
^.a^b n
and thus belongs to
for every positive integer n,
the intervals.
all
Further, no other point
values of
Since
we would have
be
is
y)
can satisfy a n
^ y ^ bn
for all
same time
at the
lim a n
which
(e.g.
n.
^y
and
lim b n
n>-qo
n><x>
impossible unless y = a.
19.
The Sum of an
an
infinite sequence,
^ y,
Infinite Series.
and
let
2~
sn
the successive
sums
2'
'
= u 1 +u 2 +u 3 +
...
+u n
be formed.
If the sequence
is
convergent
series
and
and has
sv
the limit
s2,
s,
u1
s3,
...
then s is called the
sum
of the infinite
+u 2 +u 3 + ...
this series is said to be convergent.
*This result also follows at once from the fact that,

then lim (a n - b n ) = a - /3. (Cf. 26, Theorem I.)
if
lim a n = a and lim & n = /3
INFINITE SEQUENCES
48
It
must be
carefully noted that
infinite series is
limit,
AND SERIES
what we
the limit of the
sum
call
of
[oh.
sum
the
n terms
ii
of the
of
Thus we have no right to assume without

are necessarily true
proof that familiar properties of finite sums
for sums such as s.
When lims n =+oo or lims n =-oo, we shall say that the
as
n tends to
infinity.
to
infinite series is divergent, or diverges
may
+00 or -
00
as the case
be.
- 00 then it
does not tend to a limit, or to +co or to
of these
definitions
the
to
according
oscillates finitely or infinitely
oscillates
series
the
that
say
In this case we shall
16.
,
If s n
terms in
finitely or infinitely
in 15 for the convergence of a sequence

convergence of the series in
allow us to state the criteria for the
The conditions obtained
either of the following ways:
any positive
The series converges and has s for its sum, if,
is a positive
there
number e having been chosen, as small as we please,
integer v such that
s - s n < e, when n^v.
(i)
(ii)
the series
necessary and
is that, if any
for the convergence of

has been chosen, as small
positive
number
integer v such that

as we please, there shall be a positive
_ Sn
n^v,
e when
1
5n+j)
<
p.^\
converges, lim
It is clear that, if the series
u n = 0.
This
is
con-
a necessary condition for

tained in the second criterion. It is
e.g. the series
convergence, but it is not a sufficient condition;
1+1+4 + is
divergent, though
If
lim u n = 0.
we denote
u n+1
+u n+2 + ..-+u n+T
or
s n+p
-s n by
,
Rr
condition for convergence of

the above necessary and sufficient
the series may be written
R \<e, when n ^ v, for every positive integer p.
I
*Cf. footnote, p. 41.

t
As remarked
positive
number
replaced by: To the arbitrary

in 15, this condition can be
n such that
integer
positive
a
there must correspond
\s
n+P - n \<e for every
positive integer p.
AND SERIES
INFINITE SEQUENCES
19, 20]
+u 2 +u z +...
u1
Again, if the series

converges
and has
converges
and has s-s n for
s for its
For we have
sum,
the series
sum.
its
s n+p
Also keeping n fixed,
49
= s n + p Rn
clear that
it is
lims n+p = s.
Km p R n = s-s n
Therefore
P-+CO
Thus
if
we
write
Rn
for the
sum
of the series
U n+1 + w n+2 +"
we have
The
first criterion for
= s n +R n
convergence can
now be put
in the
form
\R n \<c> when n^v.
Rn
and
is
PRn
usually called the remainder of the series after n terms,

or s n+p - s n a partial remainder.
,
20. Series
whose Terms are
Let
u1
all Positive.
+u 2 +u 3 + ...
be a series whose terms are all positive.

series either tends to
Since the terms are
The sum of n terms of this
+ oo
limit, or it diverges to
all positive,
the successive sums
s1
= u1
s3
= u 1 +u 2 +u 3
form a monotonic increasing sequence, and the theorem stated

above follows from 17. 1.
When
a series whose terms are
all positive is convergent, the series
we obtain when we take the terms in any

convergent and has the same sum.
order
we please
is also
This change of the order of the terms is to be such that there

be a one-one correspondence between the terms of the old
series and the new.
The term in any assigned place in the one
series is to have a definite place in the other,
will
c.i.
AND SERIES
INFINITE SEQUENCES
50
u1
s 2 = u 1 +u 2
s 3 = u 1 +u 2 +u 3
s1
Let
[oh. ii
Then the aggregate
(U),
which corresponds to the sequence

5 3'
5 2>
5 1>
'
sum
of the series.
bounded and its upper bound
the series obtained
for
aggregate
Let (V) be the corresponding
the understanding
on
please,
by taking the terms in any order we
is less than s.
(W)
in
we have explained above. Every number
be a number
must
s, there
In addition, if A is any number less than
a number of (W) greater than
of (U) greater than A, and a fortiori
bounded on the right, and its upper
A. The aggregate U') is thus
therefore the same as the
bound is s. The sum of the new series is
s is
is
the
sum
of the old.
It follows that if the series
ux
whose
must also
following theorems
u1
If the series
convergent and
all its
we obtain by changing
diverge.
may
be proved at once by the use of
the second condition for convergence
is
terms are all positive, diverges, the series
the order of the terms
The
+u 2 +w 8 + ...
( 19):
+u 2 +u 3 + ...
terms are positive, the series we obtain from
this, either
terms,
(I)
by keeping only a part of
or
(2)
by others, either positive

by replacing certain of its terms
or inferior to them,
equal
or zero, which are respectively
or
(3)
its
by changing the signs of some of
its
terms,
are also convergent.
The trigonoand Conditional Convergence.

later, belong
investigate
shall
metrical series, whose properties we
presence of
the
to
due
convergence is
to the class of series whose
would
series
the
that
sense
both positive and negative terms, in the
sign.
taken with the same
if all the terms were
21. Absolute
diverge
series with positive
convergent,
same sign
when
and negative terms
the series in
converges.
which
all the
is
said to be absolutely
terms are taken with the
AND SERIES
INFINITE SEQUENCES
20-22]
51
In other words, the series
ux
is
is
absolutely convergent
+u 2 +u z + ...
when the
series of absolute values
convergent.
It is obvious that
an absolutely convergent series is also convergent in the ordinary sense, since the absolute values of the
partial remainders of the original series cannot be greater than
those of the second series.
There are, however, convergent

which are not absolutely convergent:
e.g.
- 1 +1
...
is
convergent.
+|+}
...
is
divergent.
series
Series in which the convergence depends upon the presence

of both
positive and negative terms are said to be conditionally convergent.
The reason for this name is that, as we shall now prove, an

absolutely convergent series remains convergent, and has the same
sum, even although we alter the order in which its terms are
taken; while a conditionally convergent series may converge for one
arrangement of the terms and diverge for another. Indeed we

we can make a conditionally convergent series have
any sum we please, or be greater than any number we care to
name, by changing the order of its terms. There is nothing very
shall see that
extraordinary in this statement.
The rearrangement
the
s n , as
this function
has a
of the terms
new function of n, say s' n instead of the old* function

sum of the first n terms. There is no a priori reason why
introduces a
limit,
should have a limit as n tends to infinity,

that this should be the same as the limit of s *
n
s' n
or, if it
22. Absolutely Convergent Series.
convergent series remains the
The sum of an absolutely

same when the order of the terms is
changed.
Let
be the given absolutely convergent series; (') the

composed of the positive terms of (S) in the order in which
they appear; (") the series composed of the absolute values of
(S)
series
the negative terms of (S), also in the order in which they appear.
If the number of terms either in (') or
(") is limited, the
theorem requires no proof, since we can change the order of the
*Cf. Osgood, Introduction to Infinite Series
(1897), 44.
INFINITE SEQUENCES
52
sum
[ch. ii
to the
sum, which includes the terms of (S) up
its
altering
number, without
of the class which is limited in
same
are of the
and we have just seen that when the terms
terms in the
last
AND SERIES
finite
sign, as in those
which
in the
follow, the alteration in the order
convergent series does not affect
be the
sn
In this
sum
sum
let n'
sum.
infinite series
Let a be the sum of the

values of the terms of (S).
Let
its
formed by the absolute
n terms of (S).
terms be positive and n" negative.
of the first
Let sn be the sum of these n' terms.

these n" terms, taking
be the sum of the absolute values of
Let
in which they appear in (S).
in each case these terms in the order
>
= s' ~~ s"">
sn
Then
s-<ff,
s n -J-
Now,
as
n increases
VV
,
Thus, as n increases
never diminish.
frm two infinite

in 17. 1, whose
examined
monotonic sequences such as we have
without limit, the successive values of
terms do not exceed the fixed number

fore,
a.
...
These sequences, there-
and s"
lim (s) =*'-*",
tend to fixed limits, say,
Thus
Hence
the
sum
s'
convergent series (S) is equal to

of the "absolutely
series formed one

of the two infinite
in which they appear, and the
with the positive terms in the order
the order
negative terms, also
other with the absolute values of the
the difference between the
sums
in which they appear in (S).

(S) does not
Now any alteration in the order of the terms of
in the case
that
seen
have
we
change the values of s' and s"; since
not alter
do
we
positive
all
convergent series whose terms are
of a
the
sum by
rearranging the terms.
It follows that (S)
remains
changed
and has the same sum when the order
exists
correspondence
a one-one
in any way we please, provided that
and the new.
between the terms of the old series
of its
convergent
We
series
add some other results with regard to

which admit of simple demonstration:
Any
series
value to the
terms
absolutely convergent
whose terms are either equal or inferior
corresponding terms of an
also absolutely convergent.
is
absolute
absolutely convergent series is
An
AND SERIES
INFINITE SEQUENCES
22, 23]
53
absolutely convergent series remains absolutely convergent tvhen
we suppress a
certain
number
of
its
terms.
u1
+u 2
+...
v1
+v 2
+...
are two absolutely convergent series whose
+v x Jr{u 2 + v 2
(u 1 -v 1 )+(u 2 -v 2
series
(u
and
are also absolutely convergent
and
sums
+...
+...
their
sums
are
and V,
are equal to
the
UV
re
The sum of a conditionally
23. Conditionally Convergent Series.
convergent series depends essentially on the order of
Let
(S)
be such a
must both be
its
terms.
The positive and negative terms

number, since otherwise the series would
series.
infinite in
converge absolutely.
in
Further, the series formed by the positive terms in the order

which they occur in (S), and the series formed in the same way
by the negative terms, must both be divergent.

Both could not converge, since in that case our series would be
equal to the difference of two absolutely convergent series, some
of whose terms might be zero, and therefore would be absolutely
convergent (22). Also (S) could not converge,
converged and the other diverged.
if
one of these
series
We
can therefore take sufficient terms from the positive terms

their sum exceed any positive number we care to name.
In the same way we can take sufficient terms from the negative
terms to make the sum of their absolute values exceed any number
make
to
we
care to name.
Let a be any positive number.
First take positive
numbers from
appear, stopping whenever the
negative terms from
(S), in
sum
(S) in the order in
greater than
which they
Then take
the order in which they appear, stopping
is
a,
whenever the combined sum is less than a. Then add on -.as many
from the remaining positive terms as will make the sum exceed a,
stopping
when the sum
first
exceeds
a;
and then proceed to the
negative terms; and so on.
way we form
a new series (S') composed of the same

which the sum of n terms is sometimes greater
than a and sometimes less than a.
In this
terms as
(S), in
INFINITE SEQUENCES
54
the series (S) converges. Let

Then, with the usual notation,
Now
and
terms be u v u 2 u 3
,
...
A v (Fig. 4) correspond to the sums obtained

when
in (S'), as described above,

v
its
when n^v.
| n |<e,
Let the points
AND SERIES
groups of positive terms and
taken.
groups of negative terms have been
a
ElG. 4.
Then
it is
clear that
(a-A v and {B -a)
are each less than
contains at least one term of (S), and

since each of these groups
absolute value of the last
{B v -a) are at most equal to the
(a v
),
term in each group.

all v' terms.
Let these 2v groups contain in
e
is less in absolute value than
The term t*V in ('), when n'^v',
interval
the
within
the sums s' n lie
Thus, if we proceed from A v
.
(a-e, a+e), when ri^v'.

'
s n - a\
In other words,
<
similar
when
n'
^ /.
limsV = a.
Therefore
e,
argument holds
the only difference

terms of the series.
for the case of a negative
being that now we begin with
number,
the negative
theorem:
have thus established the following
arrange
series is given, we can so
If a conditionally convergent
converge
series
new
the
sum
of
terms as to make the
We
the order of the
any value we care
to
to
name.
REFERENCES.
Bromwich, loc.
De la Vallee
Ch. I-IV.
cit., 1
(5e ed., 1923), Introduction 2, Ch.
XI,
(4* ed., 1923), Ch. I and VIII.

VIII.
Mathematics (5th ed., 1928), Ch. IV and
Pure
Hardy, Cmê of
VIII-X.
Ill,
Ch.
Knopf,' he. cit, English translation,
Kap. Ill, V, Absch. II, Kap. I-III.
Pringsheim, Joe <*, Bd. I, Absch. I,
II, Absch. IX.
Stolz u. Gmeiner, he. cit, Abth.
Goursat, he.
^ngsheim,
d.
cit.,
Potjssin, he.
cit., 1
unendlicher Prozesse," Bne.

"Irrationalzahlen u. Konvergenz
math. Wiss., Bd.
I, Tl. I
(Leipzig, 1898).
CHAPTER
III

LIMITS AND CONTINUITY
24.
The Idea of a Function.
we speak
of a function of x,
In Elementary Mathematics, when
we usually mean a
obtained by certain operations,
some
cases,
variable x
range
from the nature

indicated.
is
unlimited;
is
in the last
\x\
e.g.
x2
Jx,
real expression
log x,
sin" 1^.
In
of the operations, the range of the
In the
first of
in the second,
the above examples, the
xÔ;
in the third
x>0; and
^1.
In Higher Mathematics the term "function of x" has a much

more general meaning. Let a and b be any two real numbers, where
b>a. If to every value of x in the interval a^x^b there corresponds
a (real) number y, then we say that y is a function of x in the interval
(a, b), and we write y=f(x).
Sometimes the end-points of the interval are excluded from the
domain of x, which is then given by a<x<b. In this case the
interval
is
included
(i.e.
said to be open at both ends;
^ x ^ b)
it is
said to be closed.
when both ends
are
An interval may be
a<x^b).
open at one end and closed at the other (e.g.

Unless otherwise stated, when we speak of an interval in the
rest of this work, we shall refer to an interval closed at both ends.
And when we
say that x lies in the interval (a, b), we mean that

but when x is to lie between a and 6, and not to coincide
with either, we shall say that x lies in the open interval (a, &).*
Consider the aggregate formed by the values of a function
a^x^b,
/(#),
*In Ch.
either,
II, when a point x lies between a and b, and does not

coincide with
we have referred to it as within the interval (a, b). This form of words is
convenient, and not likely to give rise to confusion.
55
56
[ch.
hi
aggregate is bounded (of 11),

given in an interval (a, b). If this
lhe
in the interval,
bounded
we say that the function /(*) is
of the aggregate
and m, the upper and lower bounds
numbers
function in toe
the
of
bounds
and lower
(cf 812) are called the upper
and no lower
bound
And a function can have an upper
interval.'
.
'
bound, and
The
vice versa.
difference
(M-m)
is
function
called the oscillation of the
the interval*
may
be
a function
It should be noticed that
interval.
the
in
interval, and yet not bounded
/(0)=0,
E.g. let
and /(z) =
determinate in an
when x>0.
x in the interval Sx ga,

Then /(*) has a definite value for every
But/(z) is not bounded in
where a is any given positive number.
to
make/(z) exceed any number we care
this interval, for we can
near to zero.
name, by letting x approach sufficiently
not attain its upper and
Further a bounded function need
and need not be members of
in other words,
lower bounds;
values of /(*) in the interval.

the aggregate formed by the
Eg
/(0)
let
= 0,
and f(x) = l
-x when O^spSl.
interval (0,
This function, given in the
unity.
zero, but not its upper bound
25. lim f(x).
limit
In the previous chapter
when nô
of a sequence
attains its lower
1),
we have
u v u 3
...
bound
dealt with the
In other words,
where n is a positive
we have been dealing with a function f(n),
->>
the limit of this function as
integer, and we have considered
limit
the
and
x
the function of the real variable
We
now to
when xâ.
pass
Differential
The idea is familiar enough. The
the
But for our purpose we must put
Calculus rests upon it.
what
and a definition of
on a precise arithmetical footing,
of fix)
matter
exactly
definite
meant by the limit

value, must be given.
is
of a function of x, as
x tends to a
when, any positive

have the limit b as x tends to a,
there is a positive
please,
as we
number e having been chosen, as small
x
for which
all values of
number r, such that f(x) - b\ <e, for
0<\x-a\ ^t].
fix) is said to
'
*Hobson,
loc.
the
cit'l (3rd ed., 1927), 280, uses
term fiuctwdion.
AND CONTINUITY
LIMITS
24, 25]
In other words \f(x)-b\ must be less than

rj, a+rj) except the point a.
57
e
for all points in
the interval (a -
When
this
Mm f(x) = b,
x>a
condition
we employ the notation
satisfied,
is
for the phrase the limit off(x), as x tends to a, is
we say that/(x) converges to 6 as # tends to a.

One advantage of this notation, as opposed
to \\mf(x)
b,
and
= b,
is
x=a
that
we say nothing about what happens

In the definition it will be observed that a
about the behaviour of f(x) for all values of x
brings out the fact that
it
when x
is
statement
equal to
is
a.
made
such that 0<\x-a\^rj.
The
first of
these inequalities
is
inserted
expressly to exclude x = a.
Sometimes x tends to a from the right-hand only (i.e. x>a), or

from the left-hand only (i.e. x<a).
In these cases, instead oi 0<\x-a\ ^rj, we have 0<(x-a) ^rj
(right-hand) and 0<(a - x) ^ rj (left-hand), in the definition.
The notation adopted
for these right-hand
Hm/(z)
is
and
x-*a0
assertion that lim f(x)

x>a
lim f(x)
x->a+0
It is convenient to use f(a
and similarly /(a -0)
for
limits
lim f(x).
xâ + O
The
and left-hand
=b
=
thus includes
lim f(x)
= b.
x-^-aO
+ 0) for
lim f(x)
x-+a+0
when
lim f(x)
when this limit exists,
this limit exists.
x->a
When/(z) has not a limit as x-â it may happen that it diverges

+ oo or to - oo in the sense in which these terms were used
in 16.
Or, more precisely, it may happen that if any positive
number A, however large, is chosen, there corresponds to it a positive
number rj such that
}
to
f(x)>A, when 0<\x-a\^rj.

In
this case
we say
that limf(x)
= + oo
x-^-a
Again,
it
may happen that if any negative number - A is chosen,

A may be, there corresponds to it a positive number rj
however large
such thatf(x)< - A, when
In
this case
we say
that
0<\x -a\^r\.
lim f(x) = -
oo
ZH
The modifications when /(a 0)= 00
are obvious.
58
When
limf(x) does not exist, and
oTto -oo,
to +00
as x->a,
when
[ch.
hi
f(x) does not diverge
said to oscillate as x-+a.
it is
It
that
in' some neighbourhood of
oscillates finitely &f(z) is bounded
a in
of
neighbourhood
oscillates infinitely if there is no
It
point.*
which f(x) is bounded. (Cf. 16.)

when x->a only
The modifications to be made in these definitions
obvious.
from the right, or only from the left, are
Sum.
I. The Limit of a
Limits.
26. Some General Theorems on
= and lim g(x) = p, then lim [f(x) +g(x)]=a + ftf
a
If lim f(x)
as small as we please.
Let the positive number e be chosen,
numbers rj v r) 2 such that
Then to \e there correspond the positive
Thus,
if
77
\f(x)-a\<ie, when
0<|x-a|^^ 1
\g(x)-P\<he, wiien
0<\x-a\^r] 2
not greater than
is
rj 1
or
rj
<
<
2,
\f{x)+g(x)-a- P\^\f(x)-a\+\g(x)|e
P\,
when 0<|z-a|^,
when 0<|x-a| ^*7-
|e,
e>
Urn [/() +g(x)]=a + P-
Therefore
the sum of any number of

This result can be extended to
the Limits.
The Limit of a Sum is equal to the Sum of
functions.
II
The Limit
of a Product.
If \imf(x)
=a
and \img(x)
= P,
x-^-a
x-+a
then lim[f(x)g(x)]=ap.
f(x)=a + </>(x)
Let
\im<p(x)
Then
xâ
From Theorem
wh^
and
g(x)
= P +y>(x).
]imip(x)=0.
x -+a
f(x)g(x)=ap+P<p( x ) + a
Also
*f(x)
and
^W +0(
our result follows
if
a Ov( a O-
hm[</>(x)y>(x)]
= 0.
in the neighbourhood of
satisfy a certain condition
the condition is satisfied
a positive "number A such that
isîTto"
there
"
x a
neighbourhood* meant to include the point =
denned by \x - a\ fr h.
îm'eti^the
this case it is
x-a
when
itself.
In
n-o
functions of the positive integer n, as
tThe corresponding theorem for
of
in the argument of certain seetxons
useful
is
and
way,
same
provedTthe
,
is
the previous chapter.
AND CONTINUITY
LIMITS
25, 26]
Since
tends to zero as x-â and
<j>(x)
\p(x)
a proof of this might appear unnecessary.

is
required,
it
could run as follows
59
tends to zero as #-a,
But
if
Given the arbitrary position number
e,
we have,
x
when 0<\x-a\ ^t} 2
\f( )\< Je,
if
rj
is
not greater than
lim
(I),
r\
or
rj
when 0<\x-a\^r).
\<p(x)xp{x)\<,
Therefore
as in
when 0<\x-a\^rj p
\<P(x)\< Je,
Thus,
a formal proof
[(p(x)y)(x)]
= 0.
XHt
This result can be extended to any number of functions.

Limit of a Product is equal to the Product of the Limits.
III.
The Limit of a Quotient.
lf\imf(x) = a^0, then lim
(i)
=-.
This follows easily on putting f(x)

expression
<p(x)
// limf(x) = a, and \img(x)
(ii)
The
*-*a
= <p(x) +a
+a
= p^0,
then lim
x-+a
This follows from
(II)
and
and examining the
x -+a
(III
[-^1=
L^(^)J
fi
(i)).
This result can obviously be generalised as above.

IV. The Limit of a Function of a Function,
lim f [<(*)].
XHI
Let
lim
<f>(x)
= b and
x-+a
Then
We
lim f(u) =f(b).

m->-6
lim/[<(z)]
XHl
=/[ lim
XHl
</>(x)].
'
are given that lim f(u) =f(b).
Therefore to the arbitrary positive
number
number
there corresponds a positive
such that
rf 1
!/[#*)] -/(6) <e, when \<l>(x) -b\^ Vl

are given that lim </>(#) =b.
XHl
Therefore to this positive number rj there corresponds a positive
x
I
Also
(1)
we
yjsuchthat
Combining
a positive
when 0<\x-a\^ v
the arbitrary positive number
\<f>(x)-b\<r]l
and (2), to
number rj such that
(1)
\fMx)]-f(b)\<e, when 0<|*-a| ^rj.

Thus
lim
XHl
f[<f>(x)]
=f(b) =f[lim
<(*)].
number
(2)
there corresponds
60
[ch.
hi
or
+ oo
EXAMPLES.
1.
If
is
n
a positive integer, lim x =0.
2.
If
is
a negative integer,
according as n is odd or even.

n
n
rif n =0, then x = 1 and lim x
+a 1 xn - 1 +
lim (a x n
3.
*=+>; and
lim
lim
.> = - oo
x-^v
x-^ +
= 1.]
...+a n _ 1 x + a n )=a n
x-^0
/ an"
5.
+ a, w
-1
+-+flm-i + fl )-g
unless bw =0.
lim
:r
=a n
if
any positive or negative integer
is
x-â
6.
P(x)
If
then
= a x m + a x x m - x + ...+ a m _ x x + a rrv
\im P(x) = P(a).
x-^-a
7.
1
P(x) = a x m + a^- + ...+ a m x x + a m
- 1 ...+b _ x + b
n
n
+
x
b1
n
Q(x)=b x
n 1
__
Let
and
P(z)
,.
If
Hm
/"(*)
exists, it is the
P(a)
same
.,
lim
and
x ^
lim
/(as) =ri,
then
f(x)=l^0, then lim
The converse does not

12.
Let
/(a;)
+ a).
flf()
= ft
=P-
|/(a:)|=0,
10. If lim/(a;)=0, then lim
11. If lim
a-h<x<a + h,
f{x)<g{x) for
-[f
as lim f(x
xâ
9
ni
|/(a:)|
and conversely
|l|.
hold.
be denned as follows:
/(a;)=a;sinl/aj,
when
x^0\
'"
/(0)=
Then lim
f(x) =f(a) for
all
values of
a.
x->a
21. lim f(x).
"the limit oif(x)
precise definition of the
when x tends
to +co (or to
meaning
of the
oc )" is also
+co
any
term
needed.
positive
if,
the limit b as x tends to
f(x) is said to have
positive
a
is
there
number e having been chosen, as small as we please,
number
such that
\f{x)-b\<e, when
x^X.
AND CONTINUITY
LIMITS
26-28]
When
we
A similar notation,
61
write
lim f(x)
= b.
lim f(x)
= b,
xy oo
used when /(a) has the limit b as x tends to - oo and the precise
term can be obtained by substituting "a negative
is
definition of the
number - X" and "x ^ - X"
in the corresponding places in the
above.
When
it is
lim f(x)
is
From
clear that only positive values of
used instead of
lim
x are in question, the notation
f(x).
the definition of the limit of f(x) as x tends to oo
it
follows that
f(x)=b
lim
> + 00
carries
with
/(-]= b.
lim
it
\x '
x-> +
And, conversely,
lim f(x)=b,
if
x-^+0
then
lim
Similarly
we have
f(-
lim f(x)
-> oo
The modifications
in the
/ \X
'
above definitions when
lim
(i)
lim
x>0
+ oo
or
oo
f(x)=+ao
or
oc
f(x)
X> + oo
and
lim
(ii)
X
will
00
be obvious, on referring to
And
25.
oscillation, finite or infinite, as
x tends to +
oo
or to
oo
is
treated
as before.
28.
necessary and sufficient condition for the existence of a

x tends to a. The general principle of convergence.*
limit to f(x) as
A
to
necessary
and
f(x) as x tends to a
chosen, as small as
that \f{x")
we
is that,
for the existence of a limit

positive number e has been
when any
please, there shall be
~f(x')\<efor
all values of x'
a positive number
r\
such
x" for which
0<\x"-a\<\x'-a\^r}.
(i)
The condition
Let lim f(x)
See
29. 2
is necessary.
= b.
footnote, p. 38.
and 29. 3.
Another treatment
of this question
is
given below in
62
hi
[oh.
please.
be a positive number, as small as we
r\ such that
number
positive
a
Then to Je there corresponds
Let
Now
let x',
\f(x)-b\<ie, when 0<\x-a\^rj.

x" be any two values of x satisfying
0<\x" -a\<\x' -a\ ^rj.
\f(x)-f(x')\^\f(x")-b\
Then
<
(ii)
The condition
ie
+ \f(x')-b\
+
h
is sufficient.
v e 2> e 3'
numbers such that
e
Let
be a sequence of positive
cn
+i< n
lim
w*ao
and
e*
'
V*
V2'
such that
numbers
be corresponding positive
-a\<\x' -a\^r) n \
\f(x")-f(x')\<e n when Q<\x"
(n=l, 2, 3, ...).J
Vv
Let
assume that
Then, since 6 n+1 <e, we can obviously
Now take e x and the corresponding r) v
x" = x.
In the inequalities (1) put x' = a +rj 1 and
r] n
(1)
^r) n+v
Then we have
0<\f(x)-f(a^rll )\<e v when 0<|a;-a|<ihTherefore
0<|s-a|<?i. -(2)
f(a+r]l )-e i <f(x)<f(a+rjl )+e 1 when
,
In Fig. 5 f{x) lies within the interval

centre at/(a +>?i),
when 0<\x-a\<rj 1
Ax
with
of length 2e x ,
A,
A
\
/
Now take
We
ft)
7(t)
}(a+%) -<,
Fig.
e2
7&5Z
5.
and the corresponding ry 2 remembering that

,
rJ2
have, as above,
f(a+ V2 )-e 2 <f(x)<f(a+ri 2 )+e 2
when 0<|a?-a|<i ?1
^rjl
(3)
cannot extend beyond
in (3)
Since !fcîh,.thfi interval for z
is in the open interval
/(a
and
)
+^
in
*
2
(2),
the interval for
Av
AND CONTINUITY
LIMITS
28, 29.1]
63
Therefore, in Fig. 6,f(x) now lies within the interval A

2 which
entirely within
v or lies within it and has with it a common
end-point. An overlapping part of
,
lies
f( a
{/(+*72)-*2>
could be cut
off,
+V2)+ 2}
in virtue of (2).
A,
A-
/
f(a+-n
f(a+r, 2)-e 2
)-^
~\
,
~ ~~~
f(a+vJf(a+V,)f(a+ V ^+e 2
f(a+r,\)+e,
Fig. 6.
In this
way we
obtain a series of intervals
Û ^2> ^3> *
each lying entirely within the preceding, or lying within it and
having with it a common end-point; and, since the length of
>
A n ^2e n we
,
Km A n = 0,
have
we
for
are given that lim
n-*x)
If
en
= 0.
n-*oo
we denote the end-points
and p i9
^3
f} 2 ,
...
of these intervals by a
v a 2 a 3 ...
where P n >a n then we know from 18 that
lima n = lim f} n
,
n*x>
Denote
We
We
where
n+oo
common limit by a.
now show that a is the
this
shall
can choose
en
any given
limit of f(x) as
in the sequence e 1}
e2,
xâ.
so that 2e n <e,
e 3 , ...
number.
Then we have, as above in (2) and (3),
e is
positive
an</(z)<j#n> when 0<\x-a\<rj n
B^
an
^p n
\f(x)-a\<^ n -a n
Therefore
<2e
<
It follows that
when 0<\x-a\<r) n
e,
lini/(a?)
= a.
XHl
As a matter
of fact,
we have not obtained

|/(a:)
in the above, but
-a|<e, when
when 0<\x-a\
However, we need only take
<r) n
r)
0<\x-a\^r1n
smaller than this
ln ,
and we obtain the
inequalities used in our definition of a limit.
29. 1. In the previous section
we have supposed
that x tends to
64
a from both
hi
for
slight modification in the condition
be
easily
can
only
side
tends to a from one
The
sides.
convergence when
[ch.
it
made.
and
Similarly, a necessary
+ 00
to
of a limit tof(x) as x tends
as
small
as
chosen,
has been
number
for the existence
is that, if the positive
number
we please, there shall be a positive
X such that
x">x'^X.
\f(x")-f(x')\<e, when
we have,
lim f(x),
In the case of
same way, the condition
x" <x'
when
\f(x")-f(x!)\<e,
The conditions
in the
^ -X.
x tends
for the existence of a limit to f(x) as
for the
can, of course, be deduced from those
0.
existence of a limit as x tends to +0 or to argument given in the preceding section is simpler
to +00
or to
00
Actually the
when we
+ 00
deal with
or
and the case when the variable

from these
00 ,*
deduced
tends to zero from the right or left can be
two, by substituting
x=a
tends to
it
a,
we must
substitute
1
.
of the Bounded
The Upper and Lower Limits of ^determination
Function
in
when
*;
29. 2.
As
x=
when xâ.
f(x),
there
17. 2,
called the upper
considered.
and
They
is
some advantage
to be obtained
by using what are
at the point
lower limits of iwletermination of the function
in much the same way as in that section.
are defined
Take a sequence
?h>
where
>/
>
^,
>/
,;
?/ 3 ,
2,
and
...
7 /"=
n>-ao
when <\x-a\
lim
'
îh and
be the upper bound of f(x),
and so on.
>/ 2
bound, when <\x - a\
m 2 ... be the corresponding lower bounds.
Similarly let
lf
m and this monotonic sequence
...
Let
Then
~?L
Therefore lim
~>x
its
upper
M ^M
below.
is
bounded
x,
M n exists, and we denote
it
by A.
1)
It is clear that
where
will give
any other sequence
>h>
W> Vz -
',
r/
and
',
...
^ ^=
the same limit A.
1 (4 Aufl., Leipzig, 1923), 33.

*Cf Osgood Lehrbuch der Funktionentheorie,
also be established in this way.
general principle of convergence of 15 can
The
LIMITS
29.1-29.3]
To the
number
arbitrary positive
such that
AND CONTINUITY
e,
65
there corresponds a positive integer
a ^Mn < A + when n ^ r.

< A + when 0< \x -a\ ^
number a < 7/, and let
= a.
v,
e,
Thus
f(x)
Now
take a positive
~=kM v
e,
k^M^M
Then
> Jf
>/.
rj.v
v.
one value xs in < \x - a\ "5- 77 v

Therefore /(a:) > A - e for at least one value x x in < \x - a\ ^k a.
If we now take j3 < \x A -a\ and proceed in the same way, we see that
f(x) > A - e for at least one value x x in
< x - a t:- fi.
And
/(#)
.v
for at least
>
It is thus clear that
points at which f(x)
This number
and we write
x^-a,
when
>A-
is
<
x-a
^ a, there are an infinite number of
e.
called the upper limit of indetermination of f(x),
x-+a
If
an
e is
when
A = lim f(x).
<\x - a\
arbitrary positive number, there is a neighbourhood
'
r).
such that for every point of this neighbourhood f(x) <

+ e; and in every neighbourhood of a, however small, there is a point (other thin a) at which f(x) > A - e.
The
If
lower limit of indetermination
similar way,
and we write A = lim
an
is
when
of f(x),
x-+a,
is
obtained in a
f(x).
<
arbitrary positive number, there is a neighbourhood
\x
-a\ =rj,
such that for every point of this neighbourhood f(x) > A - e; and in every neighbourhood of a, however small, there is a point (other than a) at which f(x) < A + c.
It is clear that
lim f(x)
^ lim
f(x).
X>a
also be extended to the case when we approach a from
x-+a
These definitions
may
the right-hand or the left-hand.
In this
way we have
lim f(x) and lim f(x),
x-^-a
which are conveniently written f(a + 0) and f(a -
and
lim
f(x),
0).
zhi+0
+O
Similarly for
lim f(x)
x-â-0
which are written f(a - 0) and f(a - 0).
x>a
29. 3.
(i)
The following theorems
//lim
x-+a
(ii)
(iii)
f(x) exists, then lim

x-+a
Conversely, if"lim f(x)

x>a
common
are obtained at once
=lim
(cf. 17. 4).
/(x)=lim /(a:)=lim
Z>a
J(x).
x->a
f(x), then lim f(x) exists
x>a
and
is
equal to their
xrti
value.
The general
when
principle of convergence for lim f(x),
x->a:
and sufficient condition for the existence of a limit tof(x) as x tends

to a is that, when any positive number has been chosen,, as small as we please,
there shall be a positive number rj such that
f(x") -f(x') < for all values of
x', x" for which
< \x" - a\ < \x' - a\ rj.
When the upper and lower limits of indetermination are used the proof that
this is a sufficient condition for the convergence oif(x) as x^-a is much shorter
and simpler than that given in 28. See also 17. 4 (iii).
necessary
66
The
29. 4.
Bounded Function
Oscillation of a
an
oscillation of a function in
have denned the
a point.
Let a be a point in the interval in which f(x)
In
at a Point.
We now
interval.
hi
[oh.
24 we
define the
oscillation at
sequence, where
Mn
>
r;
>
and lim
r; 3 ...
?/
is
given,
and
rj
?/ 3?
x , ?/ 2?
For any positive integer n
= 0.
let
n>jo
be the upper and lower bounds of f(x) in the neighbourhood of
defined by \a-x\-~7)n the point a itself now being a point of the
and
x=a,
>; x
mn
neighbourhood.
Then
as in
we have
29. 2,
M
m
Also lim
Hi*
7 y2' 7
M n and lim m
n exist
?~
M ^M
2
^m
-'-_
mv
~ Mv
...^
3
.
and are independent
of the particular sequence
chosen.
M and m
respectively, then
// these limits are

of f(x) at the point a.*
It is clear that the oscillation at
in the interval a -
?/
^x~a+
Also the oscillation at a
and lim /(a;) and
is
is called the oscillation
the limit of the oscillation of the function
as y -> 0.
rj,
is
(M -m)
the difference between
(i)
the greater of f(a)
the smaller of /(a) and lim f(x).
(ii)
x-+a
x-xi
At a point where f(x)
is
continuous, the oscillation
is
zero,
and at any other
from zero.
If the oscillation at x = a is k, then in every neighbourhood a the oscillation of f(x) is greater than or equal to k.
point
it is
different
30. Continuous
Functions.
continuous when x = x
if
The function
f(x)
is
r;
|?
â+
to
said
f(x) has a limit as x tends to x
and each of these limits is equal to/(^ ).

Thus/(x) is continuous when x = x if, to the arbitrary
number e, there corresponds a positive number r\ such that
7;,
be
from
either side,
positive
\f(x)~f(x )\<c, when \x-x Q \^r}.
When f(x)
is
defined in an interval
is
continuous in the interval
of x between a
tof(a),
and
andf(b-0)
In such cases
(a<x<b), and
and is equal
convenient to
we
shall say that
it
continuous for every value
if f(a+0) exists
exists
it is
(a, b),
(a, b), if it is
and
is
equal
tof(b).
make a
definition of continuity at a point,
tinuous at the end-points a and b
slight
change in our
and to say that
when
f(x)
is
con-
these conditions are
satisfied.
It follows
*It
may
from the definition of continuity that the sum or
be noted that Hobson
(loc. cit. I
(3rd ed., 1927), 300) uses the terra saltus.
AND CONTINUITY
LIMITS
29.4, 31.1]
67
product of any number of functions, which are continuous at a

is also continuous at that point.
The same holds for the
point,
quotient of two functions, continuous at a point, unless the denominator vanishes at that point (cf. 26). A continuous function
of a continuous function
is
also a continuous function
(cf.
26
(IV)).
The polynomial
P(x)
is
continuous for
all
= aQx n +a 1 x n - 1 +
+ a n _ 1x+a n
...
values of x.
The rational function

R(x) = P(x)/Q(x)
continuous in any interval which does not include values of x
making the polynomial Q(x) zero.
is
The functions
sin x, cos x, tan x,
functions sin _1rc, cos~%, tan _1#,
and the corresponding
etc.
are continuous except, in
'etc.
certain cases, at particular points.

e
is
continuous everywhere; log x
is
continuous for the interval
x>0.
31.
1.
Properties of Continuous Functions.*
several important theorems

reference will frequently be
these proofs
we
rely only
We shall now prove
on continuous functions, to which
made
on the
later.
It will
be seen that in
definition of continuity
and the
results obtained in the previous pages.
Theorem
Let f(x) be continuous in the interval
and
Then
the interval (a, b) can always be broken up into a finite number of
partial intervals, such that \f(x') -f(x")\< when x' and x" are any
two points in% the same partial interval.
Let us suppose that this is not true. Then let c = \(a-\-b).
At least one of the intervals (a, c), (c, b) must be such that it is
impossible to break it up into a finite number of partial intervals
which satisfy the condition named in the theorem. Denote by
(a l9 6 X ) this new interval, which is half of (a, b).
Operating on
let
I.
the positive
number
e be chosen,
as small as
we
(a, 6)f,
please.
This
section follows closely the treatment given
by Goursat,
loc. cit., 1 (4 e ed.,
1923), 8.
fin these theorems the continuity of f(x)

x rg/6), as explained in 30.
interval (a
X i.e.
in or at t/he ends of the partial interval.
is
supposed given in the closed*
68
same way
(a v b ) in the
we have done with
as
[oh.
(a, b),
hi
and then
such
proceeding as before, we obtain an infinite set of intervals
endof
sequence
The
18.
of
as we have met in the theorem

converges, and the sequence of end-points
of each being the same, say a.
b, 6 l5 b 2 ... also converges, the limit
ascribed
Also'each of the intervals (o n b n ) has the property we have
It is impossible to break it up into
to the original interval (a, b).
the condition
a finite number of partial intervals which satisfy
points a, a lt a 2
...
named
in the theorem.
the
Let us suppose that a does not coincide with a or 6. Since
is a
there
that
know
=
we
x
a,
when
function f(x) is continuous
|e when \x-a\^r).
positive number r\ such that f{x) -f(a)
than r\. Then the
less
is
a
n)
Let us choose n so large that (b n
a+rj), for we
(a-rj,
within
interval (a n b n ) is contained entirely
|
<
know
that a n
g a < bn
and x" are any two points

follows from the above that
Therefore,
in the interval (a n) b n ), it
and
\fix')-f{a)\<h
f
But
\f(x
-f(x")mf(x')
Thus we have
and
x'
if
\f(x")-f(a)\<\e.
-f(a)\ +\f{x") -f(a)\.
f(x') -f{x")
<
e,
our hypothesis leads to a contradiction.
either
There remains the possibility that a might coincide with
argument
above
the
in
required
The slight modification
or 6.
is
obvious.
Hence the assumption that the theorem is untrue leads

case to a contradiction, and its truth is established.
Corollary
of
(a,
b)
I.
into
Let
a,
partial
xv x2
...
intervals
x n _ v b be a
satisfying
mode
the
in every
of subdivision
conditions
Theorem I.
Then
\f(x)mf(a)\+\f(x)-f(a)\
<\f(a)\+
e,
whenO<(z-a) ^{x 1 -a).
Therefore
\f{x 1 )\<\f(a)\
In the same
way
\f(x)\^\f(x 1 )\^\f(x)-f(x 1 )\
< 1/0*1)1+
<|'/(o)|+
e>
2e,
wh.en0<(x-x 1 )^(x 2 -x 1
when0<(^-^ 1 )^(^ 2 -
a;
i)-
of
LIMITS
31.1]
AND CONTINUITY
69
Therefore
Proceeding in the same way for each

we obtain from the n th interval
successive partial interval
\f(x)\<\f(a)\+ne, when 0<(a?-a? n _ 1 )
Thus we
whole interval
see that in the
^ (b-x n _ t
).
(a, b)
\M\<[M\+ne,
It follows that
is
a function which
bounded in that
is
continuous in a given interval
interval.
Corollary
II.
Let us suppose the interval (a, b) divided up
n partial intervals (a, xj, (x v x 2 ), ... (x n _ 1 b), such that
\f( x ') ~f( x ")\< f r an y ^ wo points in the same partial interval.
Let ?7 be a positive number smaller than the least of the numbers
into
-x-i),
(x 1 -a), (x 2
...
points belong to the
Now
(b-Xn^).
x" in the interval (a,
such that
b),
same
take any two points

\x'
-x"\ ^rj.
If
x'
these
and
two
we have
partial interval,
\f(x')-f(x")\<le.
On
the other hand,
interval, they
must
if
lie
they do not belong to the same partial

in
two consecutive
this case it is clear that \f(x')
-f(%")\<\e
partial intervals.
In
+ Je = e.
Hence, the positive number
e having been chosen, as small as we

number r\ such that \f(x') -f(x")\<e,
when x', x" are any two values of x in the interval (a, b) for which
\x'-x"\^ V
please,
there is a positive
We
(a, b).
started with the assumption that f(x) was continuous in

It follows from this assumption that if x is any point in
this interval,
a positive
and
number
any arbitrary
rj
such that
\f(x')-f(x)\<e,
positive
number, then there
is
when \x'-x\^yj.
To begin with, we have no justification for supposing, that the

same rj could do for all values of x in the interval.
But the
theorem proved in this corollary establishes that this is the case.
This result
is
usually expressed
We
by saying that
f(x)
is
uniformly
(a, b).
have thus shown that a function which
is
continuous in an
interval is also uniformly continuous in the interval.
70
Theoeem
hi
is continuous
Iff(a) andf(b) are unequal and f(x)
takes at least once
b), as x passes from a to b,f{x)
II.
in the interval
[oh.
{a,
every value between f {a) andf(b).
suppose that f(a) and f(b) have different signs,

We shall show that for at least one
and f(b)>0.
First, let us
f(a)<0
e.g.
value of x between a and b,f(x)
From
= 0.
we
the continuity of f(x),
see that
it is
negative in the
Conneighbourhood of a and positive in the neighbourhood of b.

makes
which
f(x)
sider the set of values of x between a and b
Then
Let X be the lower bound of this aggregate.
positive.
a< X<b. From the definition of the lower bound f(x) is negative
But lim f{x) exists and is equal to /(A).
or zero in a^x<X.
z-A-0
Theref ore/( X)
But/( X) cannot be negative.

being a positive number, then there is a
also negative or zero.
is
f(X)= -m, m
positive number r\ such that
For
if
\f{x)-f(*)\< m wnen
\^-M=v>
>
The function /(z) would then

+rj and X
be negative for the values of x in (a,' b) between X and X
must
would not be the lower bound of the above aggregate. We
since f(x)
when x=
continuous
is
X.
= 0.
therefore have /( X)
be any number between f{a) and /(&), which may

The continuous function
be of the same or different signs.
x = a and x = b. By the
when
cf>(x)=f(x)-N has opposite signs
least one value
case we have just discussed, cp(x) vanishes for at
Now
of
let
x between a and 6, i.e. in the open

Thus our theorem is established.
Again,
if
f(x)
above that
we show
is
bounded
it is
that
it
continuous in
interval
(a, b).
we know from
(a, b),
in that interval.
Corollary I
In the next theorem
attains these bounds.
Theorem III. If f(%) is continuous in the interval {a, b), and

M, m are its upper and lower bounds, then f{x) takes the value
and
the value
We
shall
Let
least
m at
show
least
= l{a+b)]
once in the interval.
that/(^)
one of the intervals
interval
Theorem
we
I,
=M
at least once in the interval.

the upper bound oi f(x) is equal to M, for at
first
(a, c),
(c, b).
and proceed as
we obtain an infinite set
bisect
it,
Replacing
before.
(a, b)
by
this
In this way, as
of intervals (a,
6),
(a lt b ),
AND CONTINUITY
LIMITS
31. J, 31.2]
71
tending to zero in the limit, each lying entirely within

the preceding, or lying within it and having with it a common end(a 2
b 2 ),
...
bound oif(x)
point, the upper
common
Let X be the
in each being
M.
limit of the sequences a, a lt a 2
show that f(X) = M.

For suppose/(A) = M-A, where h>0. Since f(x)
at x = X, there is a positive number rj such that
bv b2
b,
...
We
...
and
shall
is
continuous
/(*)-/(*) <i*, when |a?-A|îy.

I
Thus/(a?)<M-JA, when |i-A|=^..
Now
take n so large that
(b n
-a n
will
be
than
less
The
rj.
interval (,&) will be contained wholly within (X-r), X+rj).
The upper bound of f(x) in the interval (a n b n would then be

from M, contrary to our hypothesis.
Combining this theorem with the preceding we obtain the
)
different
following additional result:
Theorem
IV.
Iff(x)
is
upper and lower bounds, then it takes at

interval the values M, m, and every value between
are
its
(a, b),
least
and M,
once in this
M and m.
Also, since the oscillation of a function in an interval

as the difference
between
its
was defined
upper and lower bounds
(cf.
24),
and since the function attains its bounds at least once in the
interval, we can state Theorem I afresh as follows:
Iff(x) is continuous in the interval (a, 6), then we can divide (a, b)
into a finite number of partial intervals
(a,
x x ),
(x v
x 2 ),
(x n _ v 6),
...
in each of which the oscillation off(x) is less than
any given
positive
number*
And
a similar change can be
property
31. 2.
set
known
made
in the statement of the
as uniform continuity.
The Heine-Borel Theorem.
of intervals, all in (a,
b),
Let an interval
interior point of at least one of the intervals of A.
regarded as interior
and b
Then a set
interval,
The
to
an
interval of the set,
when a
b)
(a,
be given such that every point
and an
infinite
ofa~x~bisan
(The ends a and b being

is the left-hand
end of another interval.)

consisting of a finite number of the intervals of
end of an
the right-hand
argument
of
has
the
same
Theorem I, adapted to this case, leads to the theorem // the

a=-_x~b is less than a giv m number k, then the interval can
:
oscillation at every point of
be divided
is less
up
than
k.
into
a finite number of partial intervals in each of which
the oscillation
72
property
namely, every point of the closed interval
hi
an interior point of
same convention as to the
(a, b) is
of the intervals of this finite set (with the
at least one
ends a and
[oh.
b).
Let us suppose that this is not true. Then let c \{a + b).
all its
At least one of the closed intervals (a, c) and (c, 6) must be such that
intervals
of
set
finite
a
of
interval
points are not interior points of at least one
of the set A.
Denote by
(a v
this
\)
new
interval,
which
is
half of {a,
b).
as
Operating on {a v bj as we have done on (a, 6), and then proceeding
(a, b), (av b x ), (a 2 6 2 ) ...
intervals
closed
of
set
infinite
an
obtain
we
before,
sequence
such that their ends a, a v a 2 ... form a bounded monotonic ascending
b, b v b 2 ,...
ends
their
and
identical),
all
are
n
value
of
(or from and after some
value
form a bounded monotonic descending sequence (or from and after some
,
of
n are
all identical).
Thus lim a n and lim

n>oo
b n exist,
and they are
equal, since b n - a n =
(6
- a).
11>oo
Suppose that lim a n =a, different from a and
b.
nyco
by
Then a is an interior point of one of the intervals of A, say (a', V). And
Then all the
taking n large enough, we can bring a n and b n inside (a', b').
intervals of A,
points of this interval {an , b n ) are interior points of one of the
contrary to our hypothesis.
argument applies to the case when a coincides with a or b.

Thus our theorem is proved.
of 31. 1 and
Special cases of the Heine-Borel Theorem are the first theorem
the theorem stated in the footnote on p. 71.
domain instead of
It can be at once extended to the case of a rectangular
similar
equally useful in dealing with the properties of

Indeed the proof is independent of the
functions of two variables (Cf. 37).
place also in the general theory of sets
finds
a
it
And
dimensions.
of
number
the linear interval
and
is
of points.*
so generally used by English writers that

German and French mathematicians
But
it
now refer to it as Borel's Theorem, and there is no doubt that this is the better
The theorem (for the case of a countably infinite set of intervals) was
The
title
Heine-Borel Theorem
is
has been adopted in the text.
name.
Annates Sci. de
enunciated and proved by Borel. [These, Paris, 1894
(Paris,
des
fonctions
theorie
la
sur
Lecons
51
Vticole Normale (3), 12 (1895),
of which
means
by
that
and
proof
Borel's
of
similarity
the
of
Because
1898).]
established the uniformity of the continuity of a function, given as
;
first
Heine t
Heine-Borel
continuous in a closed interval, it became customary to call it the
But it may well be the case that the theorem is contained imAnd, as
to Heine.
plicitly in similar demonstrations by authors previous
remarks, the theorem is not one of those of which the demonstration
Theorem.
Lebesgue
* Cf.
t
Hobson,
loc. cit. 1
(3rd ed., 1927),
Journal fiir Math., 74 (1872), 188.
73.
The merit lay
offers great difficulties.*
divining
AND CONTINUITY
LIMITS
31.2, 33]
not' in
its interest,
Theorem, and regards the other
Borel's
an
32. Continuity in
some
is
He
it.
enunciating
it,
refers to it
it,
and
always as
as unsuitable.
title
Some of the results of the last

when f(x) is continuous in x ~ a, where
Infinite Interval.
section can be extended to the case
in perceiving
demonstrating
73
number, and lim f(x)
definite positive
exists.
X><x>
Let u=afx.
When
sponding points
^ a, we have 00
We thus obtain a function of u, which is continuous in the closed interval (0, 1

Therefore
it
it is
bounded
in this interval,
and
takes at least once every value between
interval
bounds M, m.
attains its
and m,
).
Also
u passes over the
as
(0, 1).
Thus we may say that /(re) is bounded in the rangef given by x ^ a and the
new "point" x = cc at which f(x) is given the value lim f(x).
,
x--*x>
Msof(x) takes at
least
once in this range
its
upper and lower bounds, and
every value between these bounds.
For example, the function
bound
attain its upper
x = oo
x2
-%
continuous in
is
unitywhen
xÔ,
but
it
(0, oo
It does not
).
takes this value
when
as defined above.
When
denned for
\^h), and
f(x o +0)=f(xo -0)=f(xo ), then f(x) is continuous at x
On the other hand, when/(#) is denned for the neighbourhood
of x
and it may be also for x while f(x) is not continuous at
x it is natural to say that f(x) is discontinuous at x and to call xQ
33. Discontinuous
and
Functions.
neighbourhood
the
of
is
f(x)
0<\x-x
(e.g.
a point of discontinuity oif(x).

Points of discontinuity may be classified as follows:
I. /(%+0) and f(x -0) may exist and be
common value is different from/(:r or if f(x) is
),
equal.
If
their
not denned for x
then we have a point of discontinuity there.

Ex.
f(x)
= (x- x
sin Iftx
Heref(x + 0)=f(xQ -0)=0, and

or
if
we
leave f(x
In a review of
undefined, x
W.
II.
and G.
if
we
-x
),
when x ~^~ x
give/(a;)
any value other than
zero,
is
a point of discontinuity oif(x).
C.
Young's " Theory of Sets of Points " in the
Bull, des sciences math. (2) 31 (1907), 134.
fit
/(oo
is
convenient to speak of this range as the interval
for lira f(x).
(a, oo
),
and to write
"
74
II.
f(x
Then x
+ 0) and f(x - 0) may exist and be unequal.

whether f(x
point of discontinuity of f(x),
denned or
is
Ex.
f(x)
=l
_J
-*# when x < X-
Here
/(*
+ <>)=<>
and f(x -0) = l.
{lt
in
[ch.
is
not.
In both these cases f(x) is said to have an ordinary or 5im^Ze

And the same term is applied when the
discontinuity at x
point x is an end-point of the interval in which fix) is given, and
.
f(x +0), oif{x -0), exists and

defined for x
from f(x
different
is
),
if
is
f(x)
may have the limit + oo

and it may oscillate on one side
III. f(x)
side,
section the cases in which there
oo
as
x^x
on either
Take in this
or the other.
no
is
or
oscillation.
may
These
be
arranged as follows:
(i)
f(x +0)=f(x
Ex.
f(x)
(ii)/(z
+0)=+oo
Ex.
(iii)
-0)= +cq
= l/(* - x
(or
f(x)
f(x +0)
f(x -
= + oo
oo
Ex.
and f(x - 0) = -
= ll{x-x
(or
when x < xo-
)\
or
/()
= 1 l(x
f(x)=x-x
),
oo
(or
= + oo
+ 0)
exists
when
when
a;
>
+ oo).
f(x - 0)
f(x
-x
x<x
when
),
oo )}
0) exists
-co).
(or
(or
oo
j
J
x~xj
In these cases we say that the point x is an infinity of /(&),

and the same term is used when x is an end-point of the interval
in which f(x)
given,
is
and/(x +0),
or/(a; -0), is
+oo or -
oo
of the
say that f(x) becomes infinite at a point x

and that/(z ) = + oo (or - oo ). But this must be
(i),
in
given
kind
a short way of expressing the fact that f(x)
simply
regarded as
It is usual to
It will
tan
\ir is
) as x->x
x has an infinity at
tan
be noticed that
not denned. On the other hand,
diverges to
+ oo
(or to
oo
tan(7r-0)= +co
IV.
When/(z)
oscillates at
and
tan
(Jtt
|tt,
+0)= -
oo
but that
x on one side or the other,
tz
is
said
The oscillation
to be a point of oscillatory discontinuity.
it is infinite
of x
neighbourhood
some
in
bounded
is
whenf(x)
is finite
when
(cf.
there
25).
is
no neighbourhood
of
in
which /(as)
is
bounded
LIMITS
33, 34]
Ex.
(i)
f(x)
(ii)
f(x)
AND CONTINUITY
= sin 1 ftx - x
= l,'(x - x Q sin
when x % x
),
lftx
In both these examples x
75
-x
when x ^ x
),
a point of oscillatory discontinuity.
is
The first oscillates finitely at x the second oscillates infinitely.

The same remark would apply if the function had been given only
,
for
one side of x
The
infinities defined in (III)
infinitely are said to
Monotonic Functions.
34.
oscillates
infinite discontinuity.
The function f(x), given in
the interval
said to be monotonic in that interval if
(a, b), is
In the
and
and the points at which/(ic)
be points of
it is
either
(i)
f(x') ^f(x"),
or
(ii)
f(x') ^f(x"),
first case,
^ x' <x" ^ 6;
a ^ x' <x" ^ b.
when a
when
the function never decreases as x increases
said to be monotonic increasing;
never increases as x increases, and
it is
in the second case, it
said to be monotonic de-
creasing.*
The monotonic character of the function may fail at the endand in this case it is said to be monotonic
points of the interval,

in the open interval.
The properties
of monotonic functions are very similar to those

monotonic sequences, treated in 17.1, and they may be estab-
of
lished in precisely the

(i)
// f(x)
is
same way:
monotonic increasing when
than some fixed number

less than or equal to A.
when
x^X,
x^X,
and
f(x) is less
then lim f(x) exists
and
is
a^+oo
(ii) Iff(x) is monotonic increasing when x^X, andf(x) is greater

than some fixed number A when x^X, then lim f(x) exists and is
->->
greater than or equal to A.

(iii)
If f{x)
and fix)
is
monotonic increasing in an open interval
is greater
then f(a +0) exists
than some fixed number
and
is greater
than or equal
(a, b),
in that open interval,

to
A.
If fix) is monotonic increasing in an open interval (a, b),

andf(x) is less than some fixed number A in that open interval, then
(iv)
f(b
- 0)
exists
*The footnote,
and
is less
than or equal
p. 43, also applies here.
to
A.
76
[ch. hi
These results can be readily adapted to the case of monotonic

decreasing functions, and it follows at once from (iii) and (iv) that
it can only have
iff(x) is bounded and monotonic in an open interval,
ordinary discontinuities in that interval, or at its ends.
It may be worth observing that if f(x) is monotonic in a closed
interval, the same result follows, but that if we are only given that
monotonic in an open interval, and not told that it is bounded,

may have an infinity at either end.
=
\jx is monotonic in the open interval (0, 1), but not
E.g. f(x)
it is
the function
bounded.
At first one might be inclined to think that a function which

a finite
is bounded and monotonic in an interval can have only
interval.
number of points of discontinuity in that
The following example shows that this is not the case:
Let/(z) =
let
when|<^l;
l,
^<x^\\
f(x)
= i,
/(a)
= 2*, when ^+i<
when
and, in general,
let
(n being
Also
any positive
let /(0)
Then
f(x)
is
^,
= 0.
monotonic in the interval
namely at x = ^
(n being
any
number
any
finite
jump would be
fixed positive
(0, 1).
of points of discontinuity,
positive integer).
Obviously there can only be a

continuity at which the
is
=g
integer).
This function has an infinite
where k
number,
if
number
of points of dis-
greater than or equal to
the function
is
k,
monotonic
(and bounded) in an interval.*

Let the function f(x), defined in the interval
35. Inverse Functions.
(a, b),
be continuous and monotonic in the stricter sensed in {a, b).

For example, let y=f(x) be continuous and continually increase from
to
as x passes from a to
Then
to every value of y in {A, B) there corresponds one
value of x in
{a, b).
*It follows that

set of points
is
if
[31.
tCf. footnote, p. 75.
1 ,
Theorem
infinite.
and only one
II.]
there are an infinite
countably
b.
number
of points
rj
discontinuity, this
AND CONTINUITY
LIMITS
34, 35]
This value of #
say
a function of y
is
<$>(y)
which
77
continually
is itself
increasing in the interval {A, B).
The function
We
which
For
let
called the inverse of the function f{x).
is
that
is
</>(?/)
a continuous function of y in the interval in
defined.
it is
value of
<f>(y)
now show
shall
+ eliein
be any number between

Also
x.
let e
and B, and x the corresponding

e and
be an arbitrary positive number such that xQ
(a, b) (Fig. 7).
Let yQ -r] 1 and y + r) 2 be the corresponding values of y.

Then, if the positive number -q is less than the smaller of
x _ Xq
clear that
\
Therefore
Thus
4>{y) is
\<f>(y)
<
<f>(yQ )\
continuous at y
>
<
wne n
e,
\y
when
\y
r]
and
17 2 ,
it is
r).
77.
similar proof applies to the end- points
and B, and
the same argument applies to a function which
is
it is
obvious that
monotonic in the
stricter
and decreasing.
The functions sin - J:z, cos _1 a:, etc., thus arise as the inverses of the functions
sin x and cos x, where
= x ==i \k, and so on.
In the first place these appear as functions of y, namely sin -1 ?/, where
0~/~l, cos -1 */, where 0^2/= 1, etc. The symbol y is then replaced by
the usual symbol x for the independent variable.
In the same way log x appears as the inverse of the function e x
There is a simple rule for obtaining the graph of the inverse function f~\x)
when the graph of f(x) is known. f~\x) is the image off(x) in the line y=x.
The proof of this may be left to the reader.
The following theorem may be compared with that of 26 (IV):
Letf(u) be a continuous function, monotonic in the stricter sense, and let
sense
lim /[((*)] =/(&).
78
Then lim
cf>(x)
exists
and
is
equal to
[CH. Ill
b.
A strict proof of this result may be obtained, relying on the property proved
above, that the inverse of the function/(w) is a continuous function.
The theorem is almost intuitive, if we are permitted to use the graph oif(u).
values,
is familiar with its application in finding certain limiting
where logarithms are taken. In these cases it is shown that lim log u=log b,
and it is inferred that lim u = b.*
The reader
36. l.f Let the bounded function f(x), given in the interval {a, b), be such
that this interval can be divided up into a finite number of open partial
intervals, in each of which
monotonic;
is
f(x)
or,
in
accordance with the more

usual but less exact expression, let the function
have
number of
maxima and minima in the
a
only
finite
interval.
Suppose that the points

x lf x 2 ... xn _ 1 divide this interval into the open intervals
(a, x x ), (x x x 2 ), ...{x n _ x ,b),
,
in
Fig.
number
is
monotonic.
Then we know that f{x) can
x,
tinuities,
which f(x)
only have ordinary discon-
which can occur at the points
a,
x x x2 ,
of points within the partial intervals.
...
xn _ x
b,
and
also at
any
( 34.)
b, and
Let us take first the case where f(x) is continuous at a,
clearer,
matters
make
To
decreasing.
and
increasing
monotonic
alternately
we shall assume that there are only three of these points of section, namely
first interval (a, xx ), decreasing
*n ^2> */(*) Dein g monotonic increasing in the
(Fig.
on
so
and
8).
x
second
(x
the
in
lt
2 ),
as closed,
It is obvious that the intervals may in this case be regarded
I.
the monotonic character of f(x) extending to the ends of each.

Consider the functions F(x), G(x) given by the following scheme:
G(x)
F(x)
=x^kxx
a*i)
f(*i) -/(*)
t/l
fM-f(*%)+/&)
IM -/(*)
X 2 r^ X z= X 3
/(*)-/(?>)+/(*.)
f{x x )-Ax % )+f{x z )-f{x)
xz
**/
Z^-- *'%
^x'=b
Hobson, Plane Trigonometry (7th eel., 1928), p. 130.

fFor an alternative treatment of the subject matter of this section, see
*Cf.
36. 2.
It is clear that F(x)
and that f(x)
[a, b),
AND CONTINUITY
79
and G(x) are monotonic increasing
in the closed interval
LIMITS
35, 36.1]
= F(x)
If f(x) is decreasing in
the
as before,
i.e.
we
and
G(x),
is
start with
a^x^x v
begin with the second line of the above diagram,
bounded
is
in (a,
we can make both
by adding some number
b),
these functions positive
if,
in the
one or both were negative.
original discussion,
It
we
when
G(x) =f(a) -f(x),
Also, since the function f(x)
to both F(x)
(a, b).
first partial interval,
F(x) =/(),
and proceed
- G(x) in
clear that the process outlined
above applies equally well to n partial
intervals.
We have thus shown that when the bounded function f(x),
given in the interval
such that this interval can be divided up into a finite number of partial
intervals, f(x) being alternately monotonic increasing and monotonic decreasing
(a, b), is
in these intervals,
and continuous
at their ends, then
in the interval
II.
we can express f(x) as
the
which are positive and monotonic increasing
difference of two (bounded) functions,

(a, b).
There remains the case when some or
are points of discontinuity off(x),
all of the points a, x x x 2 ... x _

n 1 b
and the proviso that the function is alter,
nately monotonic increasing and decreasing
is dropped.
can obtain homf(x) a new function <f>(x), with the same monotonic
properties as f(x) in the open partial intervals (a, x x ), (x v x 2 ), ... (xn _ 1 b), but
We
continuous at their ends.
The process
is
the curve fixed,
obvious from Fig.
move
the second
We
9.
up
or
need only keep the
down
till its
first
part of
end-point (x l9 f(%!
+ 0))
^
Fig.
9.
coincides with (x lf f(x 1 -0)), then proceed to the third curve
or
down
If
to the
new
position of the second,
the values oif(x) at
a,
f(a + 0),
we must
xx x 2
,
...
and so
and move
on.
xn _ 1 b are not the same as
/K + 0)
treat these points separately.
or
/K-0),
etc.,
it
up
80
function
In this way, but arithmetically,* we obtain the
[ch.
hi
defined as
<j>{x)
follows:
a tLx
In
At
<x v
x1
xx
In
<x <x 2
x2
At
And
<f>(x)=f(x), supposing for clearness f(a)
4>(x)=f(x)+a v
4>(x)=f(x)+a x +a 2
<f>{x)=f(x)+a 1 + a 2 + a 3
=f(a + 0).
so on,
We
can
etc.
being definite numbers depending on f(x 1 0), f{x x ),
<j>{x) and
function
the
to
above
proved
theorem
the
now apply
<j>(x)=$(x)-V{x)m(a,b),
write
in this interval.
$(x) and ^(x) being positive and monotonic increasing
It follows that:
In
a^-.x<x l
At
xx
In
x,<x<x
And
2,
f(x)=$(x)-V(x).
f(x)=<S>(x)-V(x)-a v
f(x) = 4>(x)-V(x)- ai -a 2
so on.
<P(x): the
terms a v a 2 , ... are negative we put them with
finally we obtain, as before, that
Thus
V(x).
with
leave
we
terms
positive
any
If
of the
f(x)
= F(x)-G(x)m(a,
G(x) are positive
b),
and monotonic increasing
where F(x) and

We have thus established the important theorem:
in this interval.
b), is such that this

bounded- function f(x), given in the interval (a,
partial intervals, %n each
open
number
of
a
into
finite
up
divided
interval can be
can expressf(x) as the difference of two (bounded)
of'which f{x) is monotonic, then we
//
the
and monotonic increasing in the interval (a, b).

discontinuities of
be seen from the above discussion that the
can
discontinuities,
ordinary
be
F(x) and G(x), which can, of course, only
discontinuous.
is
-where
f(x)
occur only at the points
while the monotonic properties
It should, perhaps, also be added that,
discontinuities, the number
ordinary
ascribed to f(x) allow it to have only
infinite
(34).
be
may
discontinuity
of points of
functions, which are positive
Also it will
36. 2. Functions of
are
a*
The functions discussed in 36. 1

known as functions of bounded variaby Jordan,? and used by him in the treatment
Bounded Variation.!
special case of the class of functions
tion introduced into Analysis
in this section,
which
The
may
principal properties of such functions are obtained
replace 36.
1.
curves
*It will be noticed that in the proof the
and diagrams are used simply
as illustrations.
treatment of functions of bounded variation (fonctions

t There is a very complete
Lecons sur Vlntegration, (2" ed., Paris, 1928),
Lebesgue's
in
bornee)
variation
a
Cours a"'Analyse 2 (4 e ed., 1922), Ch. II,
Poussin's
Vallee
la
Ch. IV. See also de
2.
e
Paris, 1893), 54.
+Cf. Jordan, Cours d' Analyse, 1 (2 ed.,
AND CONTINUITY
LIMITS
36.1, 36.2]
Let f(x) be bounded in {a, b) and

a mode of division of this interval. Let y y x
Definition.
be
let
,
...
=x
81
xx x 2
,
y n _ x y n be
,
...
xn _ x , x n = b,
the values off(x)
at these points.
Then
XiVr+i ~ Vr) =/(&) -/()
=V ~ n,
where
sum of
is the
the positive differences,
and - n
the
sum of
the negative
differences.
The sum ^\y r+x - y r
is
denoted by
and we have
t,
= 2\y r -n-yr\=P + n.
To every mode of division of (a, b)

sums t, p and n.
When the sums t, corresponding to
bounded and
in
upper bound
their
and thatf(x)
(a, b),
into such partial intervals, there correspond
all possible
modes of division of
T, we say that
is
is the total
(a, b),
are
variation off(x)
of bounded variation in this interval.
is
2p = t+f(b) -f(a),
Since
2n = t-f(b)+f(a),
and
it is
clear that,
if
f(x)
is
bounded
of
variation, the
sums p and n are
also
bounded.
If their
upper bounds are
P and
N, we have
2P = T+f(b)-f(a),
2N = T-f(b)+f(a),
and
and - N are sometimes
called the positive variation
and the
negative varia-
tion in the given interval.
Again
in (a,
where
/S)
of
if f(x) is
variation,
bounded monotonic function is of bounded variation.

bounded variation in (a, b), it is also of bounded variation
clear that a
it is
And that,
if its
aâ<fi = b.
value
is
Further f(x) does not cease to be of

altered at a finite number of points.
bounded
These facts follow at once from the definition just given, and we proceed
to establish further properties of such functions.
now
I. Iff(x) is of bounded variation in

bounded variation in (a, b).
and
(a, c)
(c, b),
where
a<c<b,
it
is
of
of division of (a, b), say, a = x x x

v 2 ... xn _ x x n ~b.
one of these points coincides with c, then its sum t satisfies t = t x + t
2
where t x and t % are the sums for this mode of division of (a, c) and (c, b) respec-
Take any mode
If
tively.
If c lies
between two points x r and x r+1 since

,
\yr+i-yA ^
it is
clear that
But
t
^T
if
x
Tx
and
and
t
x +
T2
^T
2,
\y r+ i
-f(c)\
+ \f(c)-y r
\,.
with the same notation as before.
are the total variations in
(a, c)
and
(c, b)
2.
Therefore
Thus we have shown that f(x)
is
=i
T + T2
of
bounded variation
in (a, b).
respectively,
82
II. Let c
Tt
T,
foe,
T2
and
a point between
being
its total
and f(x)
a, b
variations in {a,
T = TX + T2
Then
be of bounded variation in (a, b),
shows that
T^T
and
b), {a, c)
(c, b).
We have seen that if f(x) is of bounded variation in

variation in {a, c) and (c, 6), and the concluding
Theorem
[ch. iti
bounded
argument of
(a, b), it is also of
line of the
+ T2
(i)
Now take the usual arbitrary positive number e.

which the
There is a mode of division of {a, c) and also of (c,b), for
t,
and
satisfy
T 1 -h<t 1 T 2 -e<t 2
T + T 2 -e<t 1 + 2
,
Thus
But these modes
and
of division of (a, c)
(c,
sums
fe)
form a mode
of division of
(a, b).
t\
Therefore
T.
T 1 + T 2 -e^T
Thus
It follows
from the inequalities
(i)
and
(ii)
that
T = T + T2
X
III. Iff(x) is
bounded in
and
(a, b),
(ii)
this interval
can be broken up into a
(open or closed) in which the function

finite number of partial intervals
in (a, b).
variation
bounded
is
of
tonic, thenf(x)
This follows from Theorem
since in all these intervals f{x)
I,
is
of
is
mono-
bounded
variation.*
infinite number of turning points
be of bounded variation in that interval. But the
following example shows that thi3 is not the case.
It
might be thought that a function with an
in a finite interval could not
Let f(x)
= xz
sin
when x
> 0,
and /(0) =0.
x5
It is easy to
show that there
is
one turning point and only one in

1
,((/*
where n
is
any positive
+ 1)tt) 4
(7T)
integer.
turning points is
the interval extends to the origin, the number of
in the interval
minimum
or
maximum
the
of
value
But the absolute
infinite.
If
/
h
\((rc
-\
+ l)7r)*
any
It follows that the total variation in
-
tt
On
is less
than
r,.
(nir)*
(nrrfj
interval, (0, a),
n*
where a
> 0, is less than
2-5"
the other hand the functions

f(x)
are not of
= sin l- when
X
/(0)=o
bounded variation
>
and
'
>
in
f(x)
=x
/(0)=o
such an interval.
sin
when x >
j
J
LIMITS
36. 2]
AND CONTINUITY
83
IV- Iffi x) is of bounded variation in {a, b), it is the difference of two positive,
monotonia increasing functions; and the difference of two bounded monotonic
increasing functions is a function of bounded variation.
(i)
First, let P(x)
and - N(x) be the
< x < b.
by the definition of P and
interval
(a, x),
Then,
and negative variations
positive
for the
where a
iV,,we have
n*)-m=p{x)-N{x).
Also P(x) and N(x) are positive and monotonic increasing functions of
Thus
f(x)
which establishes the

(ii)
Next
= [P(x) +/() + |/(a)|]
- [N(x) +
part of the theorem.
first
let
= F(x) - G (*),
f(x)
where F(x) and 0(x) are bounded, monotonic increasing functions

Take any mode of division of the interval
a=x
Then
x.
\f(a)\],
xv x 2
in {a, b).
zn _ l9 xn = b.
...
f(xr ^) -f(xr ) =[*(*r+i) ->(*r)] - [#(*r+i) " #(*r)l,
2 |/(W -/(*r)| ^ 2 b(%i)
and
- F(x r )\ + 2[G(xr+1 ) - G(xr )]
^[F(b)-F(a)] + [G(b)-G(a)].
Therefore f(x)
is
bounded variation
of
in (a, b) .
v Iffi(x ) andf2 (x) are two functions of bounded variation, so aref1 (x)f2 (x)
and fx (x) f2 (x).
Also, iff(x) is of bounded variation, and f(x) < some definite positive number
-
in the interval, then llf(x) is of bounded variation.
f1 {x)=F1 {x)-G1 {x)
(i)Let
where
F {x), F2 (x), O {x)

x
and
and f2 (x) = F2 (x)-G2 (x),
G2 (x)
are
positive
monotonic increasing
functions.
ft {x) +f2 (x) = [F^x) + F2 (x)] - [G^x) + G2 (x)],
Then
and the sum
of the given functions is of
Similarly for the difference

(ii)
With the usual
bounded variation by Theorem IV.

and product.
notation,
lyr+i
Vr\
\ttr+i\\Vr\
I*
\f(x)\>fi>0.
Thus the sum
for l/f(x) is less
than Tffx\ where
^
T is
the total variation
of f(x) in the interval.
VI. A function of bounded variation has only ordinary discontinuities.

This follows at once from Theorem IV, for the discontinuities of
f(x) must
also be discontinuities of F(x) and G(x), and these are ordinary
discontinuities
(34.)
84
[oh. hi
number, they are countably
are infinite in
If the discontinuities of f(x)
infinite.
For,
if
we have a sequence rj v
r) 2 , rj z ,
...
when
and
^/i>^2>^3---
li
we know that
there
is
only a
finite
m ^n =0
number
of points at
F(x + 0)-F(x-0)\>rj n
where n
which
positive integer.
any
is
'
n>*
'
c of tUt
b) and continuous at a point
of bounded variation in (a,
ate.
continuous
also
are
N(x)
and
interval, then T(x), P(x)
There is a mode of division of (a, c),
Take the arbitrary positive number
t for it satisfies
sum
say a, x lt x 2 ... xn . v c, such that the
VII. Iff(x)
is
T{c)-le<t^T{c).
Also there
is
a neighbourhood
of c
such that
|/(*)-/(c)l<K when 0<c-x^>q.

It is clear that
outside
it,
we can take x n ^ within
this neighbourhood, for
by adding a point of section within
Also
=1
2/rti
" Vt\ +
it,
if it
were
we do not diminish the sum t
-/(*-i)
^T(c-0) + h>
since T(x)
is
a monotonic increasing function.
T(c)-| <T(c-0) +
Thus
T(c)<T(c-0) + e.
T(c)^T(c-0).
and
It follows that
But, since T{x)
is
monotonic increasing, T{c)

T(c) = T(c - 0).
<fc
T(c
0).
Therefore we must have
Taking a neighbourhood to the right
Thus T(x)
is
continuous at
of a,
we
T(c) = T(c + 0).

x = c, and the same
find in the
same way that
holds for P(x) and N(x).
So far we have dealt only

If to every value of z in the
with functions of a single variable.
a number */, then we have said
interval a < x ^ b there corresponds
37
that y
pi
Functions of Several Variables.
is
a function of z in the interval
(a, b),
and we have written
The extension to functions of two variables

that
To every pair of values of x and y, such
a^xâ',
let
there correspond
x and y in
this
a number
z.
is
immediate
b^y^b',
Then
z is said to be
domain, and we write z=f(x,
y).
a function of
AND CONTINUITY
LIMITS
36.2, 37]
85
we
If
consider x and y as the coordinates of a point in a plane,

and y there corresponds a point in the
and the region denned by a^xâ',
will be
to every pair of values of x
b^yÛ
plane,
a rectangle.
In the case of the single variable, it is necessary to distinguish

between the open interval (a< x< b) and the closed interval
(a^x^b).
in the case of
So,
distinguish between open
boundary
it is
of the region is
and
two dimensions,
it
is
well to
closed domains.
In the former the

not included in the domain; in the latter
included.
In the above definition we have taken a rectangle for the domain
of the variables.
the same
or again,
function of two variables may be denned in

domain of which the boundary is a curve C:
the domain may have a curve C for its external boundary,
way
for a
and other curves,
C"
etc., for its
internal boundary.
function of three variables, or any number of variables, will

be denned as above. For three variables, we can still draw
upon
the language of geometry, and refer to the domain as
contained
within a surface S, etc.
We
two
shall
now
some properties
refer briefly to
of functions of
variables.
function
defined,
if
is
said to be bounded in the
the set of values of
domain
in
which
forms a bounded aggregate. The upper and lower bounds,

m, and the oscillation, are defined as in 24.
M and
said to have the limit
is
f(x,
it is
for all the points of this domain,
z,
I as (x,
y)
y) tends to (x y ), when,
any positive number e having been chosen, as small as we
please,
there is a positive number rj such that
\f(x, y)-l\< for all value's
,
of
(x, y)
for which
|z-*ol=^ \y-Vo\^V
In other words,
\f(x, y)
the square, centre at {x

ordinate axes and
from the domain.
A necessary and
to f(x, y)
number
as
(x, y)
there
1\
),
<d
0<\x-xQ l+.\y-y
\.
must be less than e for all points in

whose sides are parallel to the co-
of length 2rj, the centre itself being excluded
tends to (x
for the existence of a limit
is that, to the arbitrary positive
correspond a positive number rj such that

\M>y')-f(x",y")\<e, where (x',y') and (x",y") are any two
e,
shall
86
[ch.
hi
at (x
y ), whose sides
points other than (x y ) in the square, centre
length 2rj.
are parallel to the coordinate axes and of
in exactly the same
obtained
be
can
theorem
proof of this
,
The
taking the place of the

as in the one-dimensional case, squares
proof.
intervals in the preceding
way
when x = x and
function f(x, y) is said to be continuous
to (xQl y ).
tends
{x,
y)
y=y0i iif(x, y) has the limit f(x y ) as
Thus,/(#, y)
number
trary positive
that \f(x, y)
-f(x
x=x
continuous when
is
e,
there corresponds
y )\<efor
all
In other words,
\f(x,
y)-f(x
a positive
values of (x,
and
\x-x \^r)
and y = y
if, to
the arbi-
number
rj
such
y) for which
\y-y \^V-
must be less than e for all

whose sides are parallel to
)\
points in the square, centre at (x y ),

the coordinate axes and of length 2rj*
,
speak of a function as continuous at a point

= x and y = y' A1S When a functi n
(x Vo) instead of when x
as denned above, for every
of two variables is continuous at (x y),
continuous function of
point of a domain, we shall say that it is a
It is convenient to
>
(x, y)
in the domain.
It is easy to see that
we can
substitute for the square, with
with the same centre.f

centre at (x y ), referred to above, a circle
follows:
as
read
then
The definition of a limit would
to
{x,
y) tend to (x y ), if,
is said to have the limit I as
,
f(x, y)
the arbitrary positive
v such
that
f{x y)
y
number e, there corresponds a positive number

- l\<e for all values of (x, y) for which
0<J{(x-x f+(y-y
2
)
}^ri-
to this definition
a function converges at (xp, yQ ) according
former definithe
to
(based on the circle), it converges according
conversely. And the limits id
tion (based on the square); and
both cases are the same.
defined as follows:
Also continuity at (x y ) would now be
If
f(x, y) is
There
continuous at (x
), if,
to the arbitrary positive
number e,
statements when we are dealing

are obvious changes to be made in these
of the domain in which the function
boundaries
the
of
one
on
)
, y
with a point (x
is
defined.
fWe may
also use
a rectangle, centre at (x
),
and
sides 2y, 2 V ' say.
AND CONTINUITY
LIMITS
37]
there corresponds
a positive number
r\
such that
\
for
all values of(x, y) for
-f(x
f(x, y)
)\
<
which
2
J{(x-x
Every function, which
Hy-yo) }^v-
continuous at (x
is
continuous at (x
finition, is
87
y ) under this deunder the former definition, and

,
conversely.
It is important to notice that
respect to the
two
variables, as
if
a function of x and y is continuous with

it is also continuous when con-
denned above,
sidered as a function of x alone, or of y alone.

For example, let f(x, y) be defined as follows:
f( x> V)
2xu
T 2 wnen a t
x -t-y
1/(0, 0)=0.
>
<z
least
one of the variables
i3
not zero,
Then f(x, y) is a continuous function of x, for all values of x, when y has

any fixed value, zero or not; and it is a continuous function of y, for all values
of y, when x has any fixed value, zero or not.
But it is not a continuous function of (x, y) in any domain which includes
the origin, since f(x, y) is not continuous when x=0 and y=0.
For, if we put x=r cos 0, y =r sin 0, we have f(x, y) =sin 20, which
pendent of r, and varies from - 1 to +1.
is
inde-
However, it is a continuous function of (x, y) in any domain which does

not include the origin.
On the other hand, the function defined by
I
f( x > V)
= -JTZ&TtJ*
when at
V x "Try
least
one of the variables
is
not zero,
1/(0, 0)=0,
a continuous function of (x, y) in any domain which includes the origin.

The theorems as to the continuity of the sum, product and, in certain
cases, quotient of two or more continuous functions, can be readily extended
to the case of functions of two or more variables. A continuous function of
one or more continuous functions is also continuous.
is
In particular we have the theorem:
u = <f>(x,
Let
Let z=f(u,
Then
y), v=\js(x, y) be
v) be
continuous at (x
continuous in (m, v) at (m
z=f[<f>(x, y),
\jy{x t
),
and
let
= <f>(x
),
).
y)] is continuous in (x, y) at (x
yQ ).
Further, the general theorems on continuous functions, proved in

hold, with only verbal changes, for functions of two or more variables.
31,
For example:
If a function of two
it is
variables is continuous at every point of a closed domain,

uniformly continuous in the domain.
In
when
other words,
please, there is
(x', y')
and
number c has been clwsen, as small as we

number ?; such that \f(x', y') -f{x", y")\<e, tvken
are any two points in the domain for which
the positive
positive
{x"', y")
V{(z'-xy + (y'-y")*^ 7h
88
REFERENCES.
De la Vallee
Goursat,
Poussin, he.
he. cit,
cit., 1
(5e ed., 1923), Introduction, 3.
(4e ed., 1923), Ch. I.

\
(5th ed., 1928), Ch. V.

Aufl., Leipzig, 1923), Kap.
Osgood, Lehrbueh der Funktianentheorie, 1 (4
Variables, 1 (Boston, 1905), Ch.
Pierpont, Theory of Functions of Real
Hardy,
he.
cit.
I.
VI
and VII.
Funktionenlehre," Enc.
Pringsheim, "Grundlagen der allgemeinen
1899).
Wiss., Bd. II, Tl. I (Leipzig,
d.
math.
CHAPTER IV
38. In the usual elementary treatment of the Definite Integral,
defined as the limit of a sum,
it is assumed that the function of x

be represented by a curve. The limit is the
area between the curve, the axis of x and the two bounding ordi-
considered
may
nates.
For long
this demonstration was accepted as sufficient.

To-day,
however, analysis is founded on a more solid basis. No appeal
made to the intuitions of geometry. Further, even among the

continuous functions of analysis, there are many which cannot
be represented graphically.
is
E.g. let
and
Then f(x)
is
f(x)
= x sin -, when
/(0)
= 0.
continuous for every value of x but

}
differential coefficient
It is continuous at
and
it is
\f(*Y-A0>l =
\A*)m\*\;
when
0<\x\^ V)
when x ^ 0,
since
it is
if
v <e.
the product of two continuous
30].
It has not a differential coefficient at
sin IJh has not a limit as
x=0, because
~ sm h
and
has not a
x = 0, because
continuous
functions [cf
it
when x=0.
l/(*)-/(0)|<,
Also
x < 0,
hÔ.
It has a differential coefficient at
every point where
points
/ (z):=sin
xxx
cos -.
xÔ, and
at such
90
More curious
[ch. iv
is
Weierstrass discovered a function, which
differential
a
not
has
it
value of x, while
still,
continuous for every

coefficient anywhere.*
This function
the infinite series
is
defined
by the sum
of
<*>
2a
cos bn irx,
a positive number
a being a positive odd integer and 6
less
than
inequality ab>l +fir-t

unity, connected with a by the
functions have been
extraordinary
Other examples of such
for this reason alone it would
given since Weierstrass's time. And
exact arithmetical treatment
an
have been necessary to substitute
Integral.
Definite
the
for the traditional discussion of
rigorous arithmetical
a
give such
RiemannJ was the first to

this chapter is due to him
treatment. The definition adopted in
/(x) will be indicated
The limitations imposed upon the integrand
as
we
proceed.
Integral has now been superIn Higher Analysis the Riemann

one of the others allied to it
seded by the Lebesgue Integral, or by
first memoir, which appeared
This advance dates from Lebesgue's
been done in this field, but the
in 1902 Much has since then
though it is especially
are far from elementary; and,
ideas involved
Theory of Fourier's Series that the

in the rigorous treatment of the
integral are to be found,
advantages of the new definition of the
In an
proper to introduce it into this work.
does not seem

denned and some of its
Appendix|| the Lebesgue Integral is
also it is shown in what way
distinctive properties are obtained;
the more elementary
introduction simplifies and completes
it
its
treatment of the text.

date to this discovery. Weterstrs
seen* impossible to assign an exact
habit,
it privately, .> w* to
conununicated
bnt
himself did not at once poblish it,
1874.
Du Bois-Reymond quotes it a paper published
It
to his pnpils
and friends.
be replaced
tHardyhas shown that this relation can

Tram. Anur. Math. Soc., 17 (1916)].
J^TZ
WeTeitWs
function
is
earlier
work
of
m
6 > 1
and
interestmg dtfcussion of
Derivates," Quart. J. of
"Innnito
paper,
to be found in a
An
Chisholm Young.
Math., 47 (1916), 127, by Grace
DarstdlbarkeU einer
{In his classical paper, Vber die
6.
above,
p.
See
metrische Reihe.
But the
byO<o<l,
Fusion
Cauchy and Dirichlet must not be
durch e,ne trôm-
forgotten.
Mat. (3), 7 (1902), 230.

5-Integrale, Longueur, Aire," Anmdi di
subject will
to books and memoirs on this
See Appendix II, where references
||
be found.
38, 39]
39.
The Sums S and
in the interval
91
Let f(x) be a bounded function, given
s.*
(a, b).
broken up into n partial intervals

x x ), (x v x 2 ), ... (a? n _x, 6),
where
tKx 1 <x 2 ... <xn _ 1 <b.
Let M, m be the upper and lower bounds of f(x) in the whole
interval, and
r mr those in the closed interval (xr _ 1} x r ), writing
a=x and b = x n
Let
S = 1 (x 1 -a)+M 2 (x 2 -x 1 )+...+Mn (b-x n _ 1 ),}
and
s = m 1 (x 1 -a)+m 2 (x 2 -x 1 )+... +m n (6-x n _ )./
x
Suppose
this interval
(a,
M
M
To every mode
of subdivision of (a, b) into such partial intervals,
sum S and a sum s such that 5 ^ S.

The sums $ have a lower bound, since they are all greater than
m(b - a), and the sums s have an upper bound, since they are all
less than
(b - a).
there corresponds a
Let the lower bound of the sums

of the
sums
be
be J, and the upper bound
/.
We shall now show that I^J.

x l3 x 2 ... x n _ ly b
be the set of points to which a certain S and s correspond.
Suppose some or all of the intervals (a, xj, (x 1} x 2 ), ... (x n _ x
to be divided into smaller intervals, and let
Let
a,
>
y*
Vi,
yk -i> %i> y*> y*+i,
be the set of points thus obtained.
The second mode
when
...
yi-i>
Let
M' lt m\
M' 2 m\
,
it is
The part
of
lt
...
of division will be called consecutive to the
2 which come from
be the upper and lower bounds of f(x) in
y 2 ) and so on.
2 which comes from
in (y l9
b)
obtained from
it in this way.
Let 2, <r be the sums for the new division.
Compare, for example, the parts of S and
he interval (a, x ).
first,
x2
(a, y^),
^'ifoi - i)
(a,
xx)
is
+^2^2 - yi). +itf
then
{i - yjc-iY
But the numbers M\, M' 2 ... cannot exceed
v
Thus the part of 2 which we are considering is at most equal
to M^x-l - a).
,
The argument
Coura
which follows
is
d?Analyse, 1 (4e ed., 1923),
taken, with slight modifications, from Goursat's
pp. 171
et seq.
92
2 which comes from
Similarly the part of
equal to
2 (x 2
- x x ), and
Adding these
Similarly
Consider
results
[ch. iv
(x v
x 2)
is
at
most
so on.
we have 2 ^ S.
we obtain
now any two modes
a-
i= s.
of division of (a, b).
Denote them by
a,
and
On
a2
x lt
...
xm _ lt
with sums
b,
(1)
s,
and s'
y 2 ,...y n -i> *>, with sums S*
superposing these two, we obtain a third mode of division
a,
yv
consecutive to
both
(I)
and
S^ 2
Also, since (3)
is
(2)
(3),
(2).
Let the sums for (3) be 2 and <t.

Then, since (3) is consecutive to
(1),
and
consecutive to
But
m2
2^
Therefore
S^s'
S'
<r
s.
(2),
and
a-
^ s'.
<r.
Thus the sum

not
8 and
less
^S
'^s.
and
from any mode of division of (a, b) is

arising from the same, or any other, mode
arising
than the sum
of division.
It follows at
once that I^_J.
For we can find a sum s as near / as we please, and a sum S

we
(not necessarily from the same mode of division) as near J as
an
and
s
an
of
existence
If I>J, this would involve the
please.
for
which s>S.
The argument
of this section will offer less difficulty,
if
the reader follow
curve, when the sums

it for an ordinary function represented by a
will refer to certain rectangles associated with the curve.
40. Darboux's
Theorem.
The sums S and
S and
s tend respectively
multiplied indefinitely,
to J and I, when the points of division are
in such a way that all the partial intervals tend to zero.
Stated more precisely, the theorem reads as follows:
is
If the positive number e

a positive number rj such
is chosen,
that,
for
we please, there
modes of division in which
as small as
all
r\, the sum S is greater

than I by less than e.
smaller
is
s
sum
than J by less than e, and the
please.
we
as
small
Let e be any positive number as
all the
partial intervals are less than or equal to
39, 40]
sums S and
Since the
respectively, there
it
exceeds
Let
this
have J and / for lower and upper bounds

of division such that the sum S for
mode
is
J by less than Je.

mode of division be
a
a,
a 29 ...av _ l9
with sums
6,
S x and
sx
(1)
S X <J +y.
Then
Let
93
rj
be a positive number such that
of (1) are greater
than
all
the partial intervals
rj.
Let
a=x
be any
xv
mode
-xr _ 1 )^rj,
(xr
The mode
e.g. a,
b=x n
x 2 ,...x n __ v
with sums
2 and
s 2 , ....(2)
of division such that
when r=l, 2, ... n.

by superposing
of division obtained
x v x 2 a v x3 a 2
,
x^,
...
xn _ lt
and (2).
we have S 1 i= S 3
with sums
6,
;
(1)
S3 and
s3
and
(2),
(3)
consecutive to (1)
is
Then, by
39,
But
S^J+fa
Therefore
S3 <J+e.
Further,
S 2 -S 3 = I [M(xr _
I
xr )(xr - x^) - M(xr _ l9 a k)(a k - xr _ x )
-M(ak ,xr )(xr -ak )],

M(x\
x") denoting the upper
bound of f(x) in the interval (x', x"),

and the symbol 2 standing as usual for a summation, extending
in this case to all the intervals (x _ x ) of (2) which have one of
r v
r
the points a v a 2 ... a _ x as an internal point, and not an endp
,
point.
From
the fact that each of the partial intervals of (1)

greater than r], and that each of those of (2) does not exceed
we
of
see that
no two
of the a's
can
Lie
between two consecutive
is
rj,
x-s
(2).
There are at most (p-l) terms in the summation denoted by

Let \f{x)\ have A for its upper bound in (a, b).
2.
We
S 2 - S 3 above in the form

S 2 -S 3 = X [{M(xr _ l9 xr - M(xr _ l9 ak )}(ak - xr _ x
can rewrite
+{%-i.
*v)
" M(a k xr )}(xr ,
k )].
But{M(xr _ l9 xr )-M(xr _ l9 ak )} and {M(xr_ v xr )-M(a x )}

k
r
,
are both positive or zero,
and they cannot exceed 2 A.
94
S 2 -S 3 ^2Al(xr -xr ^),
Therefore
summation having at most (p-1) terms, and

at most equal to rj.
S 2 -S3 ^2fr-l)Arj.
Thus
S 2 <J +h + 2 (P " l ) Arl>
Therefore
the
since
[ch. iv
we have
(#r
- V-i) being
S 3 <J -f |e.
seen that
number rj
So far the only restriction placed upon the positive
greater
each
are
has been that the partial intervals of (1)
than
rj.
We
can thus choose
so that
rj
r)<
I(^riA'
S>J +e.
With such a choice of rj,
Thus we have shown that for any mode
of division such that the

equal to a certain
greatest of the partial intervals is less than or
exceeds J by less
positive number rj, dependent on e, the sum S
than
e.
and it is obvious that we can make the

to
both S and 5, by taking the smaller of the two
Similarly for 5
and
same rj satisfy
which we are led in
41.
The
come to the
I;
this
argument.
Definite Integral of
a Bounded Function.
definition of the definite integral of a
f(x), given in an interval
We now
bounded function
(a, b).
said to be
bounded function f(x), given in the interval (a, b), is
sums S
the
J
bound
lower
of
the
integrable in that interval, when
and the upper bound I of the sums s of 39 are equal.

The common value of these bounds I and J is called
written
integral off(x) between the limits a and b, and is
the definite
J\a
It follows
from the
than the sum
f(x)dx*
f(x)dx cannot be greater
definition 4hat
Ja
or less than the
sum
corresponding to any
These form approximations by
mode
excess
and
defect to the integral.

*The bound J of the sums S
bound I of the sums s the lower
is
usually called the upper integral of f(x)
integral.
and the
40-42]
We
95
can replace the sums
S = M 1 (x1 -a)+M2 (x2 -x1 )+...+M n (b-xn _ 1 ),

s=m1 (x1 -a) + m2 (x2 -x1 )+... + w&n(&-#n-i)>
by more general
'
expressions, as follows:
Let g l9 2 ... f
(a, x x ), (x v x 2 ), ...
,
any values
n be
...
x r ),
(xr _ x ,
x in the partial intervals
of
(x n _ v b) respectively.
...
The sum
obviously
-a)+f($ i )(x i -x 1 )+...+f(S n )(b-xn _ 1 )

between the sums S and s for this mode of
1 )(x 1
lies
(1)
division,
^M
partial intervals.
we have m r ^f(g r
r for each of the
when the number of points of division (x r ) increases indefinitely in such a way that all the partial intervals tend to zero,
since
But,
Cb
the sums
S and
common
have a
namely
limit,
I
#
f(x) dx.
Ja
Therefore the
sum
(1)
has the same limit.
Thus we have shown
sum
an integrable function
that, for
/(IJ^-a) +f(h)(x 2 -x 1
+... +/()(&
Cb
has the definite integral
f(x), the
-*n _i)
f(x)dx for
its
limit,
the
number
way
that all
when
Ja
of points of division (x r ) increases indefinitely in such a

the partial intervals tend to zero, l9 2 ,
in these partial
In particular, we
may take
for the values of l9 2
42. Necessary
and
...
a,
n being any values of x
x l9 x 2
...
xn _ l9 or x x x 2
,
...
x n _ l9 b
Any
Sufficient Conditions for Integrability.
a necessary and sufficient condition for the
one of the following
is
integrability of the
bounded function
(a,
...
intervals .*
f(x) given in the interval
ft):
I.
When any
positive
mode of
has been chosen, as small as we

r\
such that
S s<efor
every
division of (a, b) in which all the partial intervals are less
than or equal
We
number
a positive number
to
r\.
S - s<e,
have
S^J
But
*We may
as stated above.
and
sÎ.
substitute in the above, for /(^), /(2 )>
intermediate between
(Mt
in the partial intervals.
rrij),
(M2 m2
,
),
etc.,
f(n)>
anv values
ft
v n2
...
nn
the upper and lower bounds of /(*)
96
J - I<e.
Therefore
And
J must be equal
Thus the condition
is sufficient.
to
I.
I=J, the condition is satisfied.

given e, by Darboux's Theorem, there
Further,
For,
[oh. iv
if
7]
such that
in
which
all
S-J<^
and 7-5<Je
is
a positive
every mode
for
the partial intervals are less than or equal to
S-s = (S-J)+(r-s),
But
number
of division
r\.
I=J.
since
S-s<e.
Therefore
number e has been chosen, as small as we

please, there shall be a mode of division of (a, b) such that S-s<e.
Also
It has been proved in (I) that this condition is sufficient.
and
integrable,
is
as/(z)
I=J,
given
are
For we
it is necessary.
modes
of
number
any
we have shown that in this case there are
II.
When any
such that
of division,
III. Let
positive
o),
o-
S-s<e.
any pair of
be
There shall
positive numbers.
sum of the lengths of the

or equal to w shall be less
division of (a, b) such that the

the oscillation is greater than
This condition
number
e,
is
sufficient.
a mode of
than o\*
For, having chosen the arbitrary positive
take
2{M-m)
where M,
be.
partial intervals in which
and
2(b-aY
m are the upper and lower bounds respectively
Then there
is
mode
sum
of division such that the
of f{x) in (a, b).
of the lengths of the
which the oscillation is greater than or equal to w shall

Let the intervals {x r _ x x r ) in which the oscillation is greater
than or equal to w be denoted by D r and those in which it is less than w by d r
and let the oscillation (M r - m r ) in {x r _ x x r be denoted by w r
partial intervals in
be
less
than
<r.
Then we have, for this mode of division,

S - s = *Z<D r D r + ^Lo) rd r
<
and, by
(II), f(x) is integrable in (a, b).

Also the condition is necessary. For, by (II),
there is a mode of division such that S - s < wo\
if
f(x)
is
Using
integrable in
Dr
S -s = ^(D r D r + ^LiOfdr
^ i^D
(a, b),
d r as above,
r.
*Cf. Pierpont, Theory of Functions of Real Variables, 1 (1905), 498.
42, 43]
wo- > <o^D r

^D f <
Therefore
and
97
cr.
43. Integrable Functions.

I.
Iff(x)
In the
since
is
continuous in
first place,
it is
continuous in
know
Next, we
(a, b), it is integrable
we know
tha,tf(x)
than
less
that, to the arbitrary positive
(a, b).
in the interval,
number
e,
there
such that the oscillation oif(x)

partial intervals less than or equal to r\
in all
in
bounded
(a, b) [cf. 31. 1].
corresponds a positive number

is
is
rj
31.1].
Now we wish to show that, given the arbitrary positive number
[cf.
there
a mode of division such that
is
with the given
number
we know that
e,
S -s<e
for ej(b-a) there
such that the oscillation oif(x)

partial intervals less than or equal to rj.
If
is less
rj
we take a mode
than
is
e/(6
e,
Starting
[42, II].
a positive
- a)
in all
which the partial intervals

then for it we have
of division in
are less than or equal to this
rj,
S-s<(b-a)j-^ = .
'b-a
Therefore f(x)
II.
is
Iff(x)
In the
first
is
integrable in
monotonic in
place,
closed interval (a,
b),
(a, 6).
(a, b), it is integrable
we note that the
in
(a, &).*
function, being given in the
and being monotonic, is also bounded. We

monotonic increasing function, so that
shall take the case of a
we have
for the
j (a) g/^) s/(x 2)
mode
of division given
a,
&f(*-i) Sf(b)
by
x2
#]_,
. . .
x n _ 1}
o.
Thus we have
1
^=/K)K-)+/(^2)(^2-î)--- +/(&)(& -s-i)>
s=f(a) (x x - a) f(x 1 )(x 2 - %)... +f(x n _ 1 )(b - x^J.I
Therefore,
if all
the partial intervals are less than or equal to
rj,
S-s^r][f(b)-f(a)l
since
/( Xl ) _/ (a); f (x2 -f {Xl)>
are none of
them
Since a function
monotonic functions,
are integrable.
. . .
/(&)
-Jiz^j
negative.
of
it
bounded variation (cf. 36. 2) is the difference of two

follows from 45, III that functions of bounded variation
98
If
it
we take
^ </(&)-/()'
S-s<.
follows that
Thusf(x)
[oh. iv
integrable in
is
(a, b).
applies to a monotonic decreasing function.

have seen that a monotonic function, given in (a, b), can
only have ordinary discontinuities, but these need not be finite in
The same proof
We
number
(cf.
34).
We
are thus led to consider other cases in
which a bounded function
is
integrable,
function occur in the given interval.

bility is contained in the following
when
discontinuities of the
simple test of integra-
theorem:
bounded function is integrable in (a, b), when all its points

(a, b) can be enclosed in a finite number* of intervals
sum of which is less than any arbitrary positive number.
III.
of discontinuity in
the
be any positive number, as small as we please, and let

the upper bound of \f(x)\ in (a, b) be A.
By our hypothesis we can enclose all the points of discontinuity
Let
oif(x) in a finite
number
of intervals, the
sum
which
of
is less
than
ejiA.
The part
of
S - s coming from
these intervals
is,
at most, 2
multiplied by their sum.
On
the other hand, f(x)
is
continuous in
all
the remaining
(closed) intervals.
We
can, therefore, break
number
of
up
this part of (a, b) into a finite
of partial intervals such that the corresponding portion
S-s<ie[d.
(I)].
Thus the combined mode

such that for it S - s<e.
Hence f(x)
is
integrable in
of division for the
The
b), is
finite
number of
integrable in this interval.
discontinuities referred to in this
Theorem
be ordinary discontinuities, but, as the function

cannot be
of (a, b) is
(a, b).
In particular, a bounded function, with only a

discontinuities in (a,
whole
is
III need not
bounded, they
infinite discontinuities.
*It will be shown in Appendix II, 10 that a bounded function is also integrable
according to Riemann's definition of the integral, when the points of discontinuity
can be enclosed in an infinite number of intervals, if the sum of the lengths
of these intervals can be made as small as we please, and, in particular, when its
points of discontinuity form a countably infinite set.
43, 44]
IV. If a bounded function

intervals {a,
a x ), (a 1} a 2 ),
...
99
integrable in each of the partial
is
(a J,_ 1
b),
is
integrable in the whole
interval (a, b).
Since the function
mode
there
is
for it
is less
Then
interval
than
S-s
is
where
e[p,
e is
combined mode
than e.
is
the above results
such that the interval
intervals,
of division of the
-s
(a, b)
it is
integrable.
clear that if a bounded function is
can be broken up into a finite number
integrable in (a,
V. If the bounded function f(x)

is also integrable
in
ioif(x) for the
is integrable
in
monotonic
then \f(x)\
(a, b),
S-s
same mode
for \f(x)\ is not greater
42
not
is
than
of division.
It may be remarked that the converse does not hold.

E.g. let/(:r)=:l for rational values of x in (0, 1),
and
f(x) = - 1 for irrational values of x in (0, 1).
Then \f(x)\ is integrable, hut f{x) is not integrable, for
is
b).
(a, b).
This follows at once, since
the condition (II) of
whole
it is
of open, partial intervals, in each of which the function

or continuous, then
S-s
ar _ l3 a r ), such that
any given positive number.
for the
(a, b) is less
Therefore the function
From
integrable in each of these
of division for each (e.g.
satisfied, as
the oscillation
is
obvious that
2 in any interval,
it is
however small.
44. If the
bounded function
f(x) is integrable in (a, b), there are
an
infinite
number of points in any partial interval of (a, b) at which f(x) is continuous.*

Let w 1 >w 2 > o) 3 ... be an infinite sequence of positive numbers, such that
lim g>=0.
>oo
Let (a, ft) be any interval contained in {a, b) such that aâ<p<b.
Then, by 42, III, there is a mode of division of (a, b) such that the sum
of the partial intervals in which the oscillation of f(x) is greater
than or equal
to Wj is less than (/? - a).
If
we remove from
at least part of
(a lt Pj)
than
io
(a, b)
(a, P).
such that
these partial intervals, the remainder
We
can thus choose within
(^ -a x <%{p-a) and
)
(a,
/?)
must cover
new
interval
the oscillation in (a lt p )
x
less
is
x.
Proceeding in the same way, we obtain within (a

v &) a new interval (a 2 p 2 )
such that (P 2 -a 2 )<^(P 1 -a 1 ) and the oscillation in (a
2 /? 2 ) is less than o> 2
,
And
so on.
*Cf. Pierpont, loc.

10.
cit.,
508.
more general theorem
is
given in Appendix
If
r
100
Thus we
find
A Xt A
infinite set of intervals
an
2,
...
[ch. iv
each contained entirely
w->oo, and the

within the preceding, while the length of A n tends to zero as
zero.
to
tends
also
A
in
of
oscillation
f(x)
n
c) which
the theorem of 18, the set of intervals defines a point {e.g.
By
within
lies
all
the intervals.
be any positive number, as small as we please.

than .
Then we can choose in the sequence c^, w 2 ... a number w r less
smaller
number
positive
a
and
(a
interval
&),
corresponding
v
r
Let A r be the
than (c - a r ) and (/? r - c).
Let
Then
and
therefore
fix) -/(c)
we have shown
<
when \x-c\^ V
e,
that /(a)
is
continuous at
c.
interval (a, /?) contains

Since this proof applies to any interval in (a, 6), the
for any part of (a, 0),
continuous,
is
/(x)
which
at
points
an infinite number of
however small, contains a point
Some
45.
Properties of the Definite Integral.
establish
some
bounded
in
I.
(a,
Iff( x )
ft)
of continuity.
of the properties of
We
shall
now
f(x)dx, the integrand being
J a
(a, 6)
and
integrable.
is integrable
contained in
in
(a, b), it is also
integrable in
any
interval
(a, b).
to the arbitrary positive number e

difference
number
positive
n such that the
there corresponds a
partial
the
all
which
in
b)
(a,
of
S-s<e for every mode of division
From
42, I
we know that
than or equal to r\.

can choose a mode of division
intervals are less
We
ends of partial intervals.

Let 2, <r be the sums for the
mode
of this
kind with
of division of (a,
/5)
(a, /?)
as
included
in the above.
0^2-o-^/S- s<e.
Then we have
Thus/(z)
is
integrable in
(a, p) [
42, II].
function f(x) is altered at a finite

If the value of the integrable
obtained is integrable
number of points of (a, b), the function cj>(x) thus
II.
in
(a, 6),
and
its
integral is the
same as
that off{x).
We can enclose the points to which reference is made in a

sum of which is less than ejiA,
finite number of intervals, the
is the upper bound
where e is any given positive number, and A
of |0(x)| in (a, b).
intervals,
The part of S-s for <p(x), arising from these

than
most 2A multiplied by their sum, i.e. it is less
Je.
is
at
44, 45]
On
the other hand, f(x) and
in the parts of (a, b)
which are
101
which is identical with f(x)

are integrable in each of these
</>(#),
left,
parts.
Thus we can obtain a mode of division for the whole of them

which will contribute less than Je to S-s, and, finally, we have
a mode of division of (a, b) for which S-s<e.
Therefore
c/>(x) is
integrable in
Further,
<p(x)
For we have seen in
(a, b).
dx =
41 that
P f(x) dx.
the limit of
(f>(x)dx is
Ja
<P( 1 )(x 1
when the
-a)+ p( 2 )(x 2 -x 1 )+... + cp(Z n )(b-x n 1

(
intervals
xx)
(a,
(x v
x 2 ),
...
(xn _ 1} b)
tend to zero, and
any values of x in these intervals.

put /(fj, /(f a ), .../() for 0(^), 0(f 2 ), ...
0({ w ) in
this sum, since in each interval there are points at which <j>(x) and
v
n are
$2> "
We may
f(x) are equal.
way we
In this
which
is
obtain a
identical with
sum
of the
form lim 2/(r )(a;r -ir _i),
f(x) dx.
Ja
III. It follows
that iff(x)
immediately from the definition of the integral,
is integrable
in
(a, b),
so also is Cf(x), where
is
any
constant.
Again, if
f (x) and f2 (x)
are integrable in (a,
b),
their
sum
is
also integrable.
For, let (S,
the same
Then
s),
mode
(S', s')
and
(2,
of division for
it is clear
a-)
be the sums corresponding to
/1 (^),/2 (^) and/^z) +f2 (x).
that
2-o-^(S-s)+(S'-s'),
and the
Also
result follows.
it is
easy to show that

b
P
Ja
Cf(x)ix = c\ f(x)dx,
Ja
b
and
\fi(x)+f2 (x)}dx=\ A(x)dx
+ \ f2 (x)dx.
IV. The product of two integrable functions
f (x), f2 (x)
is integ-
rable.
To begin
with, let the functions

frfx),'ft (p) be positive in
(a, b).
102
M m
M m
[ch. iv
be the upper and lower bounds of

in the partial interval {x r _ v x r ).
fi(x),f2 ( x )
sums for a
and
(2, a) be the corresponding
Let (S, s), (S', s')
interval.
partial
a
is
x
(x
which
_
in
)
r
certain mode of division
r v
Let
M' r m' r
and/1 (x)/2 (a;)
T,
Then
r;
r,
clear that
it is
M - m ^ M M' - m m' = M (M' - m' + m' {M - m

M -m ^M(M' -m +M\M -m
A fortiori,
r
r)
r ).
where
M,
bounds of fx {x) f2 (x) in (a, b).

inequality by (x r -x r _ x ) and adding the
Multiplying this
sponding
results,
It follows that
integrable in
is
r ),
r)
are the upper
corre-
we have
X-<t^M(S'-s')+M'(S-s).
E-o- tends
and the product oifx (x), f2 (x)
to zero,
(a, b).
throughout the
the two functions are not both positive
that/^z) + c lt
so
c
and
c
2
interval, we can always add constants
If
+c 2 remain
f
The product
(x)
2
is
positive in
(/l(*) +Ci)(/2
then integrable.
But
(a, b).
W +cJ=A(*)/l(*) +C1/2W +C/lW +
c^z) +c f (x) +C& is

2
It follows that/xfaO/afc)
On combining
is
C 1C 2
integrable.
integrable.
these results,
we
see that
iff1 (x),f2 (%) ~'fn( x )
are integrable functions, every polynomial in

fi(x),
is afoo
an
/(*)-/..(*)
integrable function.*
46. Properties of the Definite Integral (continued).
|V(a;)efc=-fV(a;)efc.
I.
In the definition of the sums and
f(x)dx,
we assumed
s,
and
that a was less than
of the definite integral

b.
This restriction
is,
J a
however, unnecessary, and will now be removed.

points
If a>b, we take as before the set of
x v x 2 ... x n _ v b,
a,
,
and we deal with the sums
S=
s
M (x
1
-a)+M 2 (x 2 -x )+...+M n (b-xn _ 1 )A

1
(1)
= m 1 (x 1 -a)+m 2 (x 2 -x t + ... + mJJ>-xn -i)'>

)
function of the n functions

*This result can be extended to any continuous
ed., 1927), 337 (6)].
Hobson, Theory of Functions of a Real Variable, 1 (3rd
[cf
45, 46]
The new sum S

to the
sum
equal in absolute value, but opposite in sign,
is
obtained from
%n-V Xn-1-> ** X l> "

bounds of S and s in (1)
ft,
The existence
of the
definite integral is defined as the
when they have
common
It is thus clear that,
common
and the
value.
Ca
f(x)dx= >'a
f(x)dx*
being any points of an interval in which f(x)
a, b
follows,
value of these bounds,
with this extension of the definition of
Cb
we have
41,
103
is
bounded and
integrable.
II.
and
Let c be any point of an interval
in which f(x) is bounded
(a, b)
integrable.
P f(x) dx = P f(x) dx + P f(x) dx.
Then
Ja
J a
'
mode of division of (a, b) which has not c for a point

If we now introduce c as an additional point of section,
Consider a
of section.
the
sum S
is
certainly not increased.
But the sums S
and
for (a, c)
division, are respectively not less
(c, 6),
than
f(x) dx
of division of (a, b) gives a
f(x)dx
Ja
It follows that
Jfa
we
mode of
If
f(x) dx
sum
(a, b)
s,
gives a
dx
f(x) dx.
J
sum S not
less
than
f(x)dx.
.'c
in the
f(x)dz
(*
of
a f f{x) dx + [* f(x) dx.
Ja
It follows that
mode
Ja
consider the
division of
+\
and
>>
Thus every mode
given by this
sum
+\
J
=s
same way we
s
find that every
not greater than
f(x)dx.
c
r fix) dx
r /(x) &.
Thus we must have
(*/(*)&= r /(*)& +P/w&f.

J
The results proved in 42-45 are also applicable, in some cases with slight
verbal alterations, to the Definite Integral thus generalised.
104
[ch. iv
produced in either direction, it is easy to

show, as above, that this result remains true, provided that f(x)
other.
is integrable in (a, c) in the one case, and (c, b) in the
If c lies
on
(a, b)
and both functions are integrable
47. If f(x)^g(x),
in
(a, b),
f(x)dx:^[ g(x)dx.
then
J
^(x)=f(x)-g(x)^0.
(a, b), and obviously, from the sum
Let"
Then
<f>(x) is
integrable in
dx
<p(x)
s,
^ 0.
Ja
Corollary
g(x)dx^0.
f(x)dx-\
Therefore
is integrable
Iff(x)
I.
\\
We
is
have seen in
(a, b),
then
if
\f(x)\dx.
is
f(x)
integrable in
(a, b),
so also
|/(x)|.
And
The
result follows
Corollary
Iff(x)
We
have
is
f(x)dxU\
43 that
in
is
-\f(x)W(x)^\f(x)\.
from the above theorem.
Letf(x) be integrable and never negative in
II.
continuous at c in
f(x)dx>0.
andf(c)>0, then
(a, b)
(a, b).
have seen in
44 that
if
f(x)
is
integrable in
What
points of continuity in the interval.
is
must
assumed here
(a, b), it
that at one of these points of continuity /(x) is positive.

Let this point c be an internal point of the interval (a,
and
where
b),
not an end-point. Then there is an interval (c', c"),

a<c'<c<c"<b, such that f(x)>k for every point of (c',
k being some
positive
c"),
number.
Thus, since/(z)i=0 in
(a, c'),
\
f(x)dx^0.
Ja
C&'
And, since /(x)

Also,
> k in
since/(z)^0
in
(c',c")A
(c", 6),
[
f{x)dx^
k(c"
-c')>0.
f(x)dx^0.
J c"
Adding these
results,
we have
f(x) dx
> 0.
Ja
The changes
are slight.
in the
argument when
c is
an end-point
of (a, b)
46-48]
Corollary
At a point
Let f(x)
III.
c in (a, b),
Then
f( c )>g( c )-
let
= #(#),
and
105
both be integrable in (a, b).
f(x) and g(x) both be continuous, and
f(x)dx>\ g(x)dx.
Ja
Ja
This follows at once from Corollary II
by writing
<f>(x)=f(x)-g(x).
By
the aid of the theorem proved in
44, the following simpler result
may
be obtained:
Iff(x) >g(x), and both are integrable in
then
(a, b),
Cb
Cb
f(x)dx>\ g(x)dx.
J
.'a
f{x) and g(x) are integrable in (a, b), we know that f{x) - g(x)
integrable and has an infinite number of points of continuity in (a, b).
For,
if
At any one
of these points f(x)
- g(x)
is
positive,
and the
result follows
is
from
Corollary II.
48. The First Theorem of Mean Value.

Let <p(x), \^(x) be two
bounded functions, integrable in (a, b), and let \f/{x) keep the same
sign in this interval;
Also
let
M,
Then we have,
e.g. let \js(x)
in (a,
and multiplying by the factor

m\fr(x)
from
m\
is
^ M,
</t(x)
\js(x),
which
is
not negative,
<p(x)\ls(x)dx^M\
Ja
\f,(x)dx,
Ja
also integrable in (a, b).
<])(x)\]s(x)dx
= [A
Ja
[j,
is
^ <j>(x)\f,(x) ^ M\^(x).
\j,(x)dx^\
Therefore
where
(a, b).
47 that
Ja
since <p(x)\/s(x)
in (a, b).
b),
m^
It follows
^0
be the upper and lower bounds of ^(x) in
\J/(x)d^
Ja
m^
some number satisfying the relation

fjt^ M.
argument applies also to the case when
It is clear that the
\jr(x)
If
fji
(j)(x) is
continuous in
some value
for
of
(a, b),
in
{a, b).
we know
x in the interval
that
it
takes the value
(cf. 31).
We
have thus established the important theorem:

^(x) are two bounded functions, integrable in
If
<j>(x),
(a, b), <p(x)
<
106
being continuous
and
keeping the same sign in the interval,
<p{x)\!s(x)dx
then
where
some
is
Further, if
by
\]s(x)
where
[x,
definite value of
= <p()\
a^x^b.
x in
satisfies the relation
fi
\[,(x)dx,
not continuous in (a,
is
<p(x)
particular case,
<p(x)dx
when
b),
we replace <f>()
are the
m,
m^p^M, and
bounds of <p(x) in (a, b).

This is usually called the First Theorem of
As a
[ch. iv
<p(x) is
Mean
Value.
continuous,
= (b-a)</>(),
where
a^^b.
be seen from the corollaries to the theorem in 47 that in certain

b.*
replace at^^^b by a <
However, for most applications of the theorem, the more general statement
It will
cases
we can
in the text
is sufficient.
The Integral considered as a Function of

Let /(a) be bounded and integrable in (a, b), and
49.
its
Upper Limit.
let
F(z) = \*f(x)dx,
where x
Then
is
any point in
+h) is also
(x
if
(a, b).
in the interval,
F(x+h)-F(x) = \
X+h
f(x)dx.
F{x+h)-F(x) = {j,h,
where m^jix^M, the numbers M, m being
Thus
bounds
the upper and lower
x+h).
of f(x) in (x,
It follows that F(x) is a continuous function of
Further,
if
f(x)
continuous in
is
F(x+h)-F(x) = hf{),
When
r
hm
Therefore
Thus when f(x)

in
{a, b),
(a, 6),
x%%x+h.
where
is
and has a
-I
continuous in
=/(*).
F(x+h)-F(x)
v
hô
(a, b),
f(x)dx
cit.,
is.
continuous
differential coefficient for every value of
this differential coefficient being equal tof(x).
*Cf. Pierpont, loc.
(a, b).
(a, b),
h tends to zero,/() has the limit f(x).
mi
x in
pp. 367-8.
x in
48-50.
1]
107
is one of the most important theorems of the Calculus.

shows that every continuous function is the differential coeffi-
This
It
cient of a continuous function, usually called its primitive, or

indefinite integral.
It also gives a
means
of evaluating definite integrals of con-
For iif{x)
tinuous functions.
continuous in
is
(a, b)
and
F(x)=\'f(x)dx,
we know that
we have obtained
other,
We
Suppose that, by some means or
F(x)=f(x).
,-
a continuous function
must then have F(x)~<p(x) +C,
since
-.
dX
<p(x)
such that
(F(x) - 0(ic))=O in
(a, 6).*
To determine the constant
we
C,
use the fact that F(x) vanishes
at x = a.
f(x) dx = (p(x)
Thus we have
<f>(a).
Ja
The Second Theorem of Mean Value.
50. 1.
theorem regarding the integral
</)(x)\js(x)dx of
We now
come
to a
which frequent use
Ja
will
be made, especially in the more symmetrical form given in
The proof
(I),
where
I.
Let
is
<p(x) is
(p(x) be
and
bounded, monotonia decreasing, and never negative in
be bounded and
more than a finite number of times in
(a, b)
let \ls(x)
Then
integrable,
Since
definite
some
such that
\Js(x)
definite value of
Hardy,
x in av=Lx^b.
\js(x)
(a, b),
a,
a2
does not change sign more than a
we can take
...
a n _ ly a n = o,
keeps the same sign in the partial intervals

(a, i),
*Cf.
loc. cit.
sign
Ja
we are given that

number of times in
a = aQ
its
^/(x)dx,
(p(x)\J/(x)dx=(/>(a)
is
and not change
(a, 6).f
Ja
where
(III).
when we begin with the special case taken in

mono tonic decreasing and never negative in (a, b).
simpler,
K,
a 2 )>
K-i,
).
(5th ed., 1928), 228.
|This limitation will be removed in the proof of
50. 2.
108
Then
<f>(x)M x) dx=1,\
Ja
Now, by the
Theorem
First
of
Ca r
-i
Mean
Value,
far
cf>(x) \js(x)
Ja _
r
dx = ju r
where
0(z) ^(x) dx =
Therefore
\ls(x)
Ja _
r
dx,
= j^r = <M ar)-
</>(#r-i)
<f>(x)\j,(x)dx.
Ja
[ch. iv
^ [^(a
r)
- F(a r ^)]
"r-l
where we have written F(x) =\
\j/(x)dx.
.'a
Thus we have
</>(x)^(x)dx=^
Ja
Since F(a) = 0,
fir
[F{a r ) - F(ar _ 1 )\
(1)
we may add on the term
F(a)^(a), and
we
rewrite
(1)
in the form
P 0(z) i/r(x) dx =
()
(a)
fi,)
F(a)
+ j}
(^ - ^.J +
#.)
/*
#(6).
(2)
But none of these multipliers of F(a), F(a x ), ... F(b) are negative.
may, therefore, replace the right-hand side of (2) by
We
M[(4>(a)-fh)Mth-fh)+'''+(/An-l-Pn)+Pnl
where M is some definite number between the greatest and
F(a),
F{a
Since
...
x ),
F(b), or coinciding with
F(x)=
\js(x)dx,
3)
least of
one or other.
we know that F(x)
is
continuous in
{a, b).
Ja
Therefore there
is
number
satisfying
VL=F(S) [cf.31.
It follows
from
(2)
and
(3),
a^x^b,
such that
1].
that
\ls(x)dx,
<j>(x)\fs(x)dx=(p(a)
is some definite value of x in a ^ x l b.

The corresponding theorem for the case when <p(x) is monotonic
increasing and never negative in (a, b) is' stated in (II). It can be
proved in exactly the same way as (I), or deduced from it by the
substitution y = b - x in the integral
where |
(})(x)
\]s(x)
dx.
...
II.
(a, b)
Let
;
<p(x)
and
be bounded, monotonic increasing,
let \j/(x) satisfy the
and never
same conditions as in
(I).
negative in
50. 1]
<p(x)\j/(x)dx
where
some
is
We now
more general case where

the same sign in (a, b).
to the
but not necessarily of
<j>(x) be bounded, and monotonic in

same conditions as in (I).
III. Let
satisfy the
Then
<p(x)\js(x)dx=<j)(a)\
Ja
where
Let
is
some
\Js(x)
dx
(a, b) ;
+ <p(b)\
Ja
definite value
<j>(x) is
Then/(x)
(I)
monotonic,
and
lef\fs(x)
\p-(x)dx,
ofxina^x^b.
be monotonic decreasing, and /(#) = 0(#) -
<p(x)
Using
\Js(x)dx,
ofxina^x^b.
definite value
come
= (p(b)
109
<p(b).
monotonic decreasing, and never negative

we have
is
in (a, b).
\[r(x)dx,
rf(x)\f,(x)dx=f(a)
Ja
where
is
some
definite value of
^x^
x in a
b.
It follows that
<p(x)\[s(x)dx=(<p(a)-<p(b))
Thus
c/>(x)\fs(x)dx=(p(a)
Ja
where
If
use
is
<p(x)
Ja
is
\fs(x)
dx + <f>(b)
Ja
some
definite value of
\^(x)dx.
\fs(x)dx+<p(b)
Ja
Ja
\js(x)dx,
x in a =| x ^
b.
monotonic increasing, we put f(x) = <p(x)-
0(a),
and
(II).
The form
of the
Second Theorem of Mean Value given in
(III) is
the most useful and easily rememberedj
Other modifications
may
be mention 0d
+0) and 0(6 - 0) exist. Also

we may give cf>{x) these values &tx = a and x b respectively, without
changing the monotonic character of <p{x), or the value of the
Since
(p(x) is
monotonic in
integral
(a, b), cp(a
<p(x)\Js(x)dx.
Ja
We
thus obtain the theorem
IV. Let
(f>(x)
bounded and
and monotonic in (a, b) ; and let \[s(x) be

and not change its sign more than a finite
be bounded
integrable,
number of times in
(a, b).
HO
Then
where
<f>(x)\[,(x)dx=<p(a+0)\ \J,{x)dx
some
is
definite value of
x in a
[ch. iv
+^(6-0)
^(x)dx,
^xz.b*
Also it is clear that we can in the same way replace <p(a +0)
and 0(6-0), respectively, by any numbers A and B, provided
that A^<p(a+0) and B^</)(b-0) in the case of the monotonic
increasing function; and A^<f>(a+0), B^<p(b-0) in the case of
the monotonic decreasing function.
We
thus obtain, with the same limitation on
<p{x)
and
\fs(x)
as
before,
V.
0(a) V'C*)
dx=A\
yfr(x)
dx+B\
where A^<f>(a + 0) and B^</>(b-0),if(f>{x)
and A^(p(a+0), B^<p(b-0),
if <p(x)
| being some definite value of x in
a^x^b.
^(x) dx,
is
monotonic increasing,
monotonic decreasing,
is
of f in (I)-(V) need not, of course, be the same, and

depend on the values chosen for A and B.
Theorems (I) and (II) are the earliest form of the Second Theorem
of Mean Value, and are due to Bonnet,t by whom they were
employed in the discussion of the Theory of Fourier's Series.
Theorem (III) was given by Weierstrass in his lectures and du
The value
in (V) it will
Bois-Reymond,*! independently of Bonnet.

Theorem of Mean Value given in 50. 1,
assumed that the second function yfr(x) does not change sign more than a
50. 2. In the proof of the Second

it is
number of times in the interval (a, b).

In this section, we show that this restriction is unnecessary.
It will be sufficient to prove (I), as the other results (II) - (V) follow directly
finite
from
Let
(I).
<l>(x)
let \f/(x)
be
be bounded, monotonic decreasing,
bounded and integrable in
Then
and never negative in
<fi(x)f(x)dx
= <f>(a)\
results hold for (I)
takes the place of

f
Mem.
cour.
(j>(x)\J/(x)dx
and
we
please.
(II)
= <f>(a+0)
\j/{x)dx,
aZk^~b,
(I).
Acad.
roy. Bruxelles,
23 (1850),
8;
also Journal de Math., 14 (1849),
249.
* Journal filr
and
f(x)dx,
where is some definite value of x in a^xZLb.

Let the positive number e be chosen, as small as
e.g.\
-a
)a
Corresponding
(a, b)
(a, b).
Math., 69 (1869), 81; and 79 (1875), 42.
50.
50. 2]
1,
There
a mode of division of
is
such that for
sum 8
the
(a, b),
111
say
it
for
<
cji(x) \f(x)
<j>(x)
\}y(x)dx
+ e,
(i)
rb
$(x)ty(x)dx
andihesumsioT(f)(x)\j/(x)>
also the
sum S
for
rb
<
yj/(x)
-J
\fs(x)dx
+
?(5)'
)a
(2)
rb
6
and the sum

Let
<r
- -1>\ yfr(x)dx
s for \fr(x)
= 2 ^(^r) is( xr)
x r+i ~ x r)
n-1
= ^j>{x r )c r
J'
where
cr
= yr(a: r )(av +1 -av),
n-1
=d
<f>(x
+ 2<t>(xr )(dr -dr _1 ), where d r =c +Ci + ... + c r
=d
[<(*)
c^(x1 )]
+ d1 [$(Xj) - <j>(x2 )] + ...

+ d n _ 2 [<(z n _ 2 - <(*n-i) +dn_1 <f>(x n_1 )].
)
d n _x are negative.
None
Let d v and dQ be the smallest and largest of dQ , d v ... d n_ v
of the multipliers of d
dv
...
Then we have
" +(*r+l)] + </>(*-i))
[^ t<K*r) ~ ^r+i)l + <M*n-i)j = = Jst*)
J
I o
J
I
~ = dq
(3)
dP
is some number satisfying dptk/x dq
Thus
=/x <(a), where
<r
<**
i.e.
<t>(a)
Now
dj,
= 2c r = 2 i*{x r
sum
Therefore the
(a) for (x
And
<r
<f>(a)
/x
cr
x r+1 ~ x r )'
for
yjs(x)
for (x
a^,
...
x v+1 )
~d v ^ the
sum
for
xv ...x P+1 ).
between these numbers
^/{x)dx lies
Also (S -
s)
for
\j/(x)
for (z
*lf
. . .
^ +1
and
^k(S-s)
(xot
for
xv ...x n
#.
i/r(#)
for
)<~ hj(2).
)
d Pn
Therefore
>\
ylr(x)dx
YK
and
dg
similarly
by
+1
Therefore,
Xp
(3)
and
<
'
- -p^-,
\}r(x)dx
(4)
(4),
f(x)dx-2e<d v 4,(a)^ (T ^dq (a)<[
Xq
cli
<l>(a)[
+ 1 ^(x)dx 2e
+
(5)
Ja
Ja
But the sum

<f)(x)\]s(x)
<()
)a
for
s for
(:r
xv
...
xn)
^ ~ the sum # for

<r
cfy(x)\}/(x)
for (x
xx
...
x n ).
112
Therefore by
(1
[a
),
4>(x)yfr(x)dx
<
[ch. iv
e,
rb
<r-e<
and
<f>(x)ijf(x)dx<cr
+ t.
)a
Therefore by
(5),
P+1
<f>(a)\
y(x)dx-3e<<r-e<
<fi(x)\p(x)dx<(r
(/>(a)f^
M,
Let
Then
+1
xp(x)dx
m be the largest and smallest values of

</>(a)(^
)a
+1
f(x)dx
^ m^{a)
and
+ <
+&
(6)
yj/(x)dx in (a, b).
Xq +
cf>(a)\
f{x)dx^
M4>{a).
.'a
Thus we have from
(6),
[b
mcj>(a)-3i<\
And
it
^(x)\f/(x)dx<Mcj)(a)
+ 3e.
follows that
b
m<f>{a)
Hence
where
^{x)f{x)dx^
%-_
\
= <f>(a)\
<f>(x)^(x)dx
is
some
definite value of x in a
M<j>{a).
\js(x)dx,
~ x ^ b.
INTEGRAND BOUNDED.
INTERVAL INFINITE.
INFINITE INTEGRALS.
f(x) dx,
51. In the definition of the ordinary integral
and
in
Ja
the preceding sections of this chapter,
integrand
is
bounded
we have supposed that the
in the interval of integration
from one given point a to another given point

extend this
b.
which extends
We
proceed to
definition so as to include cases in which
the interval increases without limit,
(i)
the integrand has a finite
(ii)
number
of infinite discontinuities.*
/00
I.
Integrals to
oo
f(x) dx.
Let f(x) be bounded and integrable in the interval {a, b), where
We define the integral
a is fixed and b is any number greater than a,
Cx
foo
f(x) dx as lim
For
f(x) dx,
.....
this limit exists.j
the definition of the term "infinite discontinuities," see 33.

more convenient to use this notation, but, if the presence of the variable x
fit
in the integrand offers difficulty,
is
when
we may
f(t)dt
and
replace these integrals
lim
f(t)dt.
by
50. 2, 51]
We
speak of
f(x)dx in this case as an
113
and say that
infinite integral,
it
converges.
On
when
the other hand,
f(x)dx diverges to
infinite integral
Ja
divergence to
f(x)dx tends to
oo
oo
as x-+cc,
oo
and there
we say
that the
a similar definition of
is
*x>
f(x)dx.
of
a
Ve-*dx = \;
Ex.1.
e~*dx=
For
'
e~x dx= lim (1 -e~*)
lim
For
dx=
lim !"<*?= lim
\e x dx = lim
lim
cte
-
=
Vx
log -
Similarly
f* cfo?
lim
Z->oo
dx = -
(e
= oo
l)
3J>QO
00
AW
2(l-
r^=oo.
x>zc
= l.
x>QC
r*fe = oo;
Ex.2.
2.
x>oo J
["*!
And
^=
oo
lim 2{y/x-\)
= tc.
X-+CC
-
oo
These integrals diverge to oo or - oo as the case may be.

Finally, when none of these alternatives occur, we say that the
,
f(x)dx
integral
j
Ex
oscillates finitely or infinitely, as in 16
and
infinite
25.
sin
x dx
oscillates finitely,
x sin x dx
II.
- oo
Integrals to
oscillates infinitely.
f (x) dx.
J -00
When f(x)
b is fixed
is
and a
f(x)dx as lim
J
J -00
We
any number
is
f(x) dx,
less
when
than
the interval (a, b),

b,
we
where
define the integral
this limit exists.
X-^-QoJ X
speak of
J
f(x)dx as an
The cases in which
infinite integral,
and say .that
f(x)dx is said to diverge to

J
oo
finitely or infinitely, are treated as before.
C'*:> LfcVJ-
or to -
oo
it
converges.
or to oscillate
114
[oh. iv
ro
Ex.
tr x dx diverges to
2.
ro
co
sinh x dx diverges to -
sin
x dx
oo
oscillates finitely.
x sin x dx oscillates infinitely.
from -
III. Integrals
If the infinite integrals
oo
^ -
we say
f(x) dx
f (x) dx.
f(x) dx are both convergent,
and
J a
Too
f(x)dx
that the infinite integral]
to their
is
convergent
and
is
equal
sum.
Since
it
to oo
f*
a<a<x,
f(x)dx= \f(x)dx + \ f(x)dx,
follows that,
if
one of the two integrals
f(x)dx
f(x)dx or
J^
converges, the other does.
f(x)dx=\ f(x)dx +
\" f(x)dx=
Similarly,
J X
and,
if
+f
f(x)dx
"
f(x)dx.
\
J a.
J a
x<a<a,
f(x)dx,
* a
f(x) dx converges,
f(x) dx or
one of the two integrals
the other does.

a
P f(x)dx=\J oo f(x)dx + \ a f(x)dx.
Also
J -oo
Thus
/(a?) cfo
foo
**
+ ^f(x) dx =
/(x)d:z
is
J(x) dx + J
J*
f(x) dx,
independent of the point a used in
the definition.
f t^:2 = ^
Ex.
52.
("
e-^&f=2J"e-*
ifa
necessary and sufficient condition for the convergence
of ff(x)dx.
Let
F(x) =
\
J a
f{x)dx.
The conditions under which F(x)
shall
have a limit as z->co
ft
51-53]
have been discussed in
we
integral
27 and 29.
115
In the case of the
1.
infinite
are thus able to say that:

poO
The
I.
f(x) dx
integral
convergent
is
and has
the value I,
when,
Ja
any
is a
positive
positive
number
number
having been chosen, as small as we please^ there
X such that
\I-\ f( x )dx <e, provided

Jo
x^X.
that
And
II.
further:
and
necessary
for the convergence of
poo
the integral
f(x)dx
chosen, as small as
when any
is that,
we
positive
number
positive
number
has been
such that
<e
f(x)dx
\J x'
for
all values
We
of x'
x" for which x" >x'
have seen in
51 that
^ X.
00
f(x)dx converges, then
if
Ja
r
Co.
f(x)dx=\ f(x)dx +
Ja
Ja
f(x)dx,
a<a.
Ja
/bo
It follows
from
(I)
that,
f(x)dx converges, to the arbitrary
if
Ja
number
positive
that

.
f* x
f(x)dx\<, when
Also,
if
These
number
such
x^X.
this condition is satisfied, the integral converges.
results,
and the others given
immediately to the
in 53-58,
can be extended
infinite integral
f(x)dx.
J -00
poo
f(x)dx.
53.
Integrand
Positive.
the
If
positive
when x>a,
it is
clear that
poo
creasing function of
diverge to
I.
x.
Thus
Ja
I f(x)dx
is
a monotonic in-
f(x)dx must either converge or
<x>
It will converge if there is
a positive number
such that
poo
f(x)dx<A when x>a, and
is
y a
Cx
integrand f(x)
in this case
f(x)dxÂ.
=
116
if there is
It will diverge to oo
[oh. iv
no such number.
of
These statements follow from the properties

functions
monotonic
( 34).
Further, there
is
an important "comparison
vergence of integrals
when
test " for the con-
the integrand is positive.
bounded and
Let f{x), g(x) be two functions which are positive,
II.
Also
integrable in the arbitrary interval (a, b).
xâ.
Then,
f f(x)dx
if
is convergent, it
let
g(x)
^f(x) when
j(x)dx
follows that
is
Ja
Ei0O
g(x)dx^
I
f(x)dx.
Ja
For from
47
we know that
P g(x)dx^
f(x)dx,
when x>a.
Ja
Ja
f g (x) dx<\Jz f{x) dx.
Therefore
Ja
Then, from
III.
Ifg(x)
(I),
[ g(x)dx^ \f(x)dx.
Ja
Ja
^ f(x),
and
P f(x)dx
diverges, so also does
g{x)dx*
Ja
g(x)dx^J(x)dx.
This follows at once, since

J
One
of the
most useful
integrals for comparison
is
J^
^, wnere
a>0
=!-i- {x
P%
1-n
xn
We have
1 -*
- a 1 "*},
when n+\,
ja
and
Ja
x = log x - log a,
when n
1.
and
Since the relative behaviour of the positive integrands /(*)
in terms of limits:
only as x-*oo these conditions may be expressed
g(x) matters
if
xôo j"y (x) dx
converges,
When
g(x)lf(x) has a limit as
When
or diverges, as
g(x)lf(x) has a limit, not zero,
f(x) dx diverges.
if
s-oo
\j(z) dx converges.
\jW dx
Merges,
53,54]
Thus, when n
> 1,
lim
x>oo J a
And, when w ^
lim
1,
a1 -"
n-1
I
]
L sy<rbj
converges > since
^Lv^T)^
61868
fsin
a:
g-
o.
..
diverges.
v<ryg) < h-
converges,
sin2a;
since
^-
The
54. Absolute Convergence.
^_
1
-g,
when *=><>
when * = 2
^VfrÎjH?
cte
Xn
rdx
t.e.
xn
X-+CD J h
a
Ex
1 -*
dx
=a =
ft L
X
rdx
Ja
117
when x ^ a > 0.
,
f(x) dx is said
integral
to
Ja
when f(x)
be absolutely convergent
bounded and integrable in the
is
poo
arbitrary interval (a,
and
6),
f(x) dx is convergent.
\
Ja
Since
follows
it
f(x)dx
]Jx
from
'
\f(x)
dx, for
x"
>x'â
(cf.
52, II
that
if
f(x)
47, Cor. I),
dx converges, so also does
Ja
/oo
I
J *'
I
f(x)dx.
Ja
But the converse is not true.

converge, and yet not converge
may
An
infinite integral of this type
absolutely.
For example, consider the integral

x
Jo
The Second Theorem
of
Mean Value
50. 1)
shows that
this
integral converges.
For we have
f^sinx
I
If.sin x dx +-
dx,
,
f
I
sin
x dx,
< x' ^ ^ x".
where
But
I
sin
x dx
and
sin
a;
da;
are each less than or equal to 2.
118
te^fe*?)
sin
[ch. vi
Therefore
J
f* sin
ic
Therefore
Jo
value
^^
x
we
converges, and
cfe
31
i*
integral
^
x
Jo
To prove
^ X,
shall find in
is
any positive
f*
But
dx diverges.
|an
integer.
L -^" L<fo =
have
A dx=\
sin
a?|
^~^-
S
77
sin y
zr-^-Zy dy>
(r-l)Tr
on putting z = (r-
1)tt +?/.
Isind
J^
(r-l)T
1 f
77
r7r J0
>2
2^1
Îsinicl^
Thus
But the
88 that
only necessary to consider the integral
this, it is
"
We
x'
is \ir.
But the
where n
when #">
e,
4
A > -
provided that
its
<
L^**
Thus
Jo
series
7T
a;
on the right hand

n7r
,.
lim
Therefore
n-*<x>J
diverges to
Isind
-1
(w;
= oo
<x>
But when x>n7r,
PIS^l (fc >rJSLl'

&
Jo
..
Therefore
hm
Jo
Msinx|-,
J
L<fe
= oo.
<
fa.
as w->a>
54, 55]
119
When infinite integrals of this type converge, but do not converge absolutely, the convergence must be due to changes of sign
in the integrand as x-^co
.
55.
The
/j-Test for the
Convergence of
f(x) dx.
J a
I.
(a, b)
Let f(x) be bounded and integrable in the arbitrary interval

where a>0.
If there is a number ju greater than 1 such that
x*f(x) is bounded
when xâ, then
f(x) dx converges absolutely.

Ja
Here
\x>
f(x)\<A, where
is
some
definite positive
number and
xâ.
T
l/WK xi
30
But we know that
f
Ja
It follows that
dx
converges.
X
\f(x)\
dx converges.
/oo
Therefore
II.
(a, b),
f(x) dx converges,
and the convergence
is
absolute.
Let f(x) be bounded and integrable in the arbitrary interval

where a>0.
If there is a number ju less than or equal to 1
such that x"f(x) has a positive lower bound when x

diverges to oo
^ a,
f(x) dx
then
Ja
Here we have, as
before,
xrf(x)Â>0, when xâ.
< f(x).
It follows that
But
f
Ja
dx
diverges to
X
III. Let f(x) be
where a>0.
when ju^l.
f(x) dx diverges to oo
It follows that
(a, b),
oo
bounded and integrable in the arbitrary interval

If there is a number /a less than or equal to 1
/
such that xf(x) has a negative upper bound when

diverges to
oo
f(x)dx
Ja
This follows from
must have a
xâ, then
(II), for in this
positive lower
case
bound when x ^
a.
12
But,
xâ;
lim (xi(x)) exists,
if
follows that z*f(x)
it
by properly choosing the
also,
[ch. iv
positive
is
bounded
number X,
in
xf(x)
when this limit is positive,

will either have a positive lower bound,
limit is negative provided
or a negative upper -bound, when this
that
x^X.
Thus, from
(I) -(III),
the following theorem can be immediately
deduced:
arbitrary interval
Letf(x) be bounded and integrable in the
(a, b),
where a>0.
If there
(00
is
a number a greater than
such that lim (xy(x))
f(x) dx converges.
If there
is
a number u
than or equal
less
to 1
such that lim (xrf(x))

X>ao
/oo
exists
exists,
is not zero, then
and
f(x)dx diverges; and the same
is true
J a
if x*f(x) diverges to
We
+cc
or to
ao
as z->oo
use of this test, and refer to it

clear that we are simply comparing the
make very frequent
shall
It is
as the "//-test."
/O0
integral \* f(x)dx with the integral

J a
Ja
and deducing the con-
J,
of the latter.
vergence or divergence of the former from that
Ex.
1.
J^f dx
^
[ "^^
2.
converges, since
since
diverges,
Hm (* x j^f^ji) = L
Km
(x x
^-^) =
\.
It should be noticed that the

x
,
integral
Jo
theorems of this section do not apply to the
x
ax.
sin
/CO
56. Further Tests for the
I.
If
<p(x) is
interval (a, b),
f(x)dx.
Convergence of
bounded when
Ja
xâ, and
and P ^(x) dx converges
integrable in the arbitrary
absolutely, then
4>(x)yf,(x)
dx
is absolutely convergent.
For we have \tp{x)\<A, where

number and x ^ a.
is
some
definite
positive
55, 56]
Cx"
Also
121
fx"
\<p{x)\
\is(x)\dx<A\
J x'
\\fr(x)\dx t
J x'
when x">x'>a.
/oo
we
Since
given that
are
\\fr(x)\dx
converges,
the result
Ja
follows.
-j^dx,
Ex.1.
a
-j^dx converge
when n and a
absolutely,
are
positive.
e~ ax cos bx
2.
o.
cos
mx
2
dx converges absolutely, when a
is
positive.
cte
II. Ze 0(z) 6e monotonic and bounded when xâ.

Let \fs(x) be
bounded and integrable in the arbitrary interval (a, b), and not change
sign more than a finite

\[s(x)
number of times in
the interval.
Also
let
dx converge.
/OO
Then
</)(x)\^(x)dx converges.
Ja
This follows from the Second Theorem of

<p(x)\fr(x)dx=4>(x')\
J x'
\lf(x)dx+<j>(x")\
Jx'
where a<x'
Value, since
\ls(x)dx t
Jf
g f ^ #".
But \<f>(x')\ and |^>(a?")| are each

number A.
Also we can choose X so that
\ls(x)dx
are each less than e/2A,

positive
Mean
less
than some definite positive
and
\}s(x)dx
when x">x'^X, and
number, as small as we
e is
please.
It follows that
If*"
I
<p(x)\js(x)dx
<e, when
x">x'^X
J x'
and the given integral converges.

-
M
1
iX. 1.
..sinx,
f
|
2.
Ja
- e~ x )
ax converges.
dx converges when a > 0.
any given
122
monotonia and bounded when
III. Let <p{x) be
lim
[ch. iv
xâ, and
= 0.
(p(x)
bounded and integrable in the arbitrary interval (a, 6),

and not change sign more than a finite number of times in the interval.
Let
Also
be
\js(x)
let
\Js(x)dx be
bounded when x>a.
Ja
Then
dx
(p(x)\js(x)
is convergent.
Ja
As above,
we know that
in (II),
<t>(x)yj/(x)dx
= <t>(x')\
where a<x'
x.'
^x".
\f/(x)dx\<A
Ja
positive
\l,(x)dx+<p(x")\ \fr(x)dx,
x'
when x>a, where
^(x)dx
is
some
definite
number.
And
\lr(x)dx
y/,(x)dx
\+
|l*'"
\Ja
Jx'
<2A.
"
Cx
Similarly
I
^{x)dx
lim<(z)
Also
<2A.
J s
= 0.
a;->co
any positive number, as small as we

be a positive number X such that
Therefore,
there will
e is
if
\<t>(%)\<Yj,
when x ^X.
It follows that
iCx" <p(x)\)/(x)dx
<<-,
when x">x'^X,
J x'
Joo
<p(x)\l/(x)dx converges.
Ex.1
[ m
XnJ
30
3.
f
I
Jo
rfa:,
T
J
sin c
}a
1+x*
cos ax
sin x
^ <fa converge, when n and a are positive.

X
dx converges.
- cos foe ,
dx converges.
x
please,
56-58]
The Mean Value Theorems

57.
The
First
Let
<f>(x)
be
Let
\j/(x)
keep the same sign in
bounded when xHâ, and integrable in
and,
xâ,
${x)\\,(x)dx
J
m^fi^M,the
We
for the Infinite Integral.
Theorem of Mean Value.
Then
where
123
and
the arbitrary interval (a, 6).
\f/(x)dx converge.
= ix
\Js{x)dx,
inxâ
upper and lower bounds of <f)(x)
have
M and m.
being
when xâ,
m^<f)(x)^M,
^ 0,
if ip(x)
m^{x)^(f)(x)f(x)^M^{x).
Therefore
by
But,
whenâ.
\ls(x)dx^[ $(x)if(x)dx^ lA* iP(x)dx,
56,
I,
<f)(x)xf(x)dx
we
and
converges,
Ja
p*
given
are
that
\fr(x)dx converges.
J
Thus we have from these
inequalities
m\^{x)dx^\ 6(x)f(x)dx^M^
In other words,
<f>(x)f {x)dx=fx
Ja
58.
The Second Theorem of Mean Value.

C
Lemma.
Then F(x)
it
f{x)dx,
Ja
m^/xîf.
where
\fs(x)dx.
Let
poo
f(x)dx be a convergent
continuous when
is
xâ, and
integral,
and F(x) =
bounded in
/(x)dx(xâ).
Also
).
upper and lower
the interval (a, oo
takes at least once in that interval every value between
its
bounds, these being included.
The continuity
of F(x) follows
from the equation

X+h
F(x + h)- F(x) = Further, lim F(x) exists and
is
X>-x>
It follows
from
32 that F(x)
(a, oo
Let
(f)(x)
Let
\Js(x)
be
f(x)dx.
zero.
is
bounded
in the interval (a, a>
if
and
xâ.
integrable in the arbitrary interval (a, b),
change sign more than a finite number of times in the interval.
A Iso
converge.
Ja
where a
as defined
M,
bounded and monotonic when
be bounded
Then
),
m are its upper and lower bounds, it takes at least

the values M and m and every value between M, m.
in that section, and,
once in
<KxMz)dx = 4>{a + 0)\

Ja
^ ^ oo .*
*Cf. Pierpont, loc.
cit.,
654.
ip(x)dx + <f>(oo)\
Jf
^{x)dx,
let
and not
\}r(x)dx
(
'"
124
[oh. iv
Suppose (f>(x) to be monotonic increasing.

arbitrary interval
apply the Second Theorem of Mean Value to the
We
(a, b).
Then we have
b
[
Jo
where
<Hx)f(x)dx=<l>(a+Q)\* f(z)dx+<l>(b-0)\** Mx)dx,

J"
a^^=b.
TOO
Add
B = #oo
to both sides
observing that
JO
<(oo
exists, since
<f>(x) is
does not exceed some definite number

Also Urn
Then
B=0 and
B+
f(x)dx,
<f>(x)xP(x)dx
monotonic increasing in x
^ a and
( 34).
converges
[ 56, II].
b
[
<f>(x)f{x)dx
Ja
^
f(x)dx + ^(b-0)\^ 4s(x)dx + 4>(cc)\ b xfix)dx
= <f>(a + 0)\*
+ cf>{co))
f{x)dx
ib
= <f>(a + 0)\
(1)
xP(x)dx+U + V,
where
U = {<f>(b - 0) - <f>(a + 0)}\~~

j* f(x)dx
and
F = (4>(oo - <Hb-0)}\ b
f(x)dx.
^(*)<** Abounded
(00
Let
M,
in (a, oo
).
m be its upper and lower bounds.

m^"f(x)dx^M,
Then
m^\~+(x)dx^M.
and
Therefore
{<f>(b
- 0) -
<f>(a
+ 0)}m ^ U ^ {<f>(b - 0) - 4>(a + 0)}M,
<(&
-0)}m^V^ {$(<*> )-<f>(b -0)}M.
{cftoo)
Adding
these,
we
see that
- </>(a + 0))m^ U + vm<H*> -<t>(a + 0)}M.

m^/x^Jf.
F
7+ = /*{<( >)- #* + <>)}, where
Therefore
to the limit when &-<*>
proceed
and
in
V
+
(1),
U
for
value
Insert this
+
r4>(x)is(x)dx = cf>(a + 0)\f(x)dx + ix\ct>(cc)-<t>(a 0)},
)
{floo
Then
Ja
where //=lim
This lirniTmust
6->oo
/a.
exist,
since the other terms in (1)
have
limits
when
58, 59]
it
Also, since
mîfA^kM,
follows that
m^
But
/x'rSs
125
takes the value // at least once in the interval
\f/(x)dx
(a,~jx>
).
TOO
Thus we may put // =

Therefore
we have
f
-'a
where a
It
is
of the
==i
''
yjr{x)dx,
^ ' ^
where a
Ja
oo
finally
#r)f(a;)d:r = <(a + 0)P t(x)dx +
;=i
oo
d>(oo
)\ \blx)dx,
'
Jf
clear that we might have used the other forms (III) and (V), 50.
Second Theorem of Mean Value and obtained corresponding results.
INFINITE INTEGRALS.
INTEGRAND INFINITE.
In the preceding sections we have dealt with
f(x)dx.
59.
1,
00
.
the infinite integrals
Ca
f(x)dx and
&
bounded in any arbitrary
Ja
the integrand f(x)
is
f(x)dx,
oo
f(x)dx,
interval,
when
however
large.
further extension of the definition of the integral is required

so as to include the case in which f(x) has a finite number of infinite
discontinuities
First
33) in the interval of integration.
(cf.
we take
when a
the case
discontinuity in
(a, b).
and
in
integrable
is
the only point of infinite
The integrand f(x)
the
arbitrary
is
interval
supposed bounded
(a
+ ,
b),
where
a<a + g<b.
On
this understanding, if the integral
H+Oy we
define the infinite integral
f(x) dx has a limit as
f(x)dx as lim
f(x)dx.
when the point b is the only point of infinite disconand f(x) is bounded and integrable in the arbitrary
&-), where a<b-<b, we define the infinite integral
Similarly,
tinuity in (a, b),

interval (a,
(b
rb-S
f(x)dx as lim
Ja
f(x)dx, when this limit
exists.
f->0 J a
Again, when a and b are both points of infinite discontinuity, we

f(x) dx as the
sum
of the infinite integrals
Ja
f(x)dx and
c being
f(x)dx, when these integrals
a point between a and
b.
exist,
as defined above,
126
This definition
b,
since
is
[ch. iv
independent of the position of
between a and
we have
\f(x) dx
where a<c'<c
Finally,
let
= [fix) dx + f' f(x) dx,
51, III).
(cf.
there be
finite
tinuity in the interval (a, b).
a^x,<x 9
...
<x n ^b.
number of points of infinite disconLet these points be x v x 2 ... x n where

,
We
f(x)dx by
the equation
X2
Xl
P fix) dx = \
when
on
the integrals
f{x) dx
+\
f(x) dx
+\
. . .
f(x) dx,
the right-hand exist, according to the definitions
just given.
It should
be noticed that with this definition there are only to
number of points
be bounded in any partial
be a
to
and f(x) is
which has not one
of infinite discontinuity,
finite
interval of
(a, b),
an interval point or an end-point.
of these points as
Harnack to
This definition was extended by du Bois-Reymond, Dini and
of infinite
points
of
number
infinite
an
has
certain cases in which the integrand
our purpose.
discontinuity, but the case given in the text is amply sufficient for
The modern treatment

Riemann's
of the integral has rendered further generalisation of
discussion chiefly of historical interest.
It is convenient to
speak of the
infinite integrals of this
and
the succeeding section as convergent, as we did when

terms divergent
of the limits of integration was infinite, and the
one or other
and
oscillatory are
Some
before.
writers use the term proper integral for the ordinary integral
when/(a;)
is
for the case
term
employed as
bounded and integrable in the
when
it
interval (a,
b),
f(x)dx,
j
and improper
has points of infinite discontinuity in
(a, b),
integral
reserving the
infinite integral for
\f{x)dx,
\
J-o
)a
f(x)dx or I"
f(x)dx..
J- 00
French mathematicians refer to both as integrates genkralisees\

both as uneigentliche Integrate, to distinguish them from
refer to
Germans
eigentliche
Integrate or ordinary integrals.
ff(x)dx.
60.
Let/(:z)
in
any
have
interval,
f(x)dx.
J -co
Ja
f(x)dx.
J -co
infinite discontinuities at a finite
however
large.
number
of points
59-61]
127
For example, let there be infinite discontinuities only at x

v
... x n in xâ,f(x) being bounded in any interval (c, b), where
x2
c>x n
a^x <x
Let
Then we have,
as
above
\f(x) dx =
f(x) dx
( 59),
X2
Xl
Ja
<x n <b.
2 , ...
Ja
+\
f(x) dx
+f
J xi
f(x) dx
+\
f(x) dx,
Jc
J xn
where x n <c<b, provided that the integrals on the right-hand

side exist.
It will be noticed that the last integral
integral, /(a) being
f(x) dx also converges,
If the integral
J
Too
f(x) dx
we
in
is
an ordinary
(c, b).
f(x) dx by the equation:

Ja
Xl
\*f(x) dx =
f(x) dx
\
Ja
Ja
It
since
is
Xt
+\
f(x) dx
. . .
Jzi
+f
f(x) dx
clear that this definition
+ \f(x) dx.
Jc
Jx
independent of the position of
is
c,
we have
C
f(x) dx
+f
J %n
f(x) dx = f f(x) dx
Jc
+ {"fix) dx,
Jx
c'
where x n <c<c'.
Also
we may
write the above in the form

X2
f(x) dx =
JJ a
\f(x) dx +
Ja
The verbal
f(x) dx
+ ...+ {"fix) dx.
J xi
J xn
alterations required in the definition of
f(x)dx
J -00
/00
are obvious,
and we
define
f(x)dx, as before, as the
sum
of
J -CO
Ca
the integrals
r=o
f(x) dx
and
J -oo
It is easy to
f(x) dx.
Ja
show that
this definition is
independent of the
position of the point a.

61. Tests for Convergence of
f(x)dx.
It is clear that
we
Ja
need only discuss the case when there
is
a point of infinite discon-
tinuity at an end of the interval of integration.
128
If
x=a
f(x)dx,
this limit exists.
from the
It follows at once,
The
I.
we have
the only point of infinite discontinuity,
is
f(x)dx = \im
when
[ch. iv
f(x)dx
integral
definition, that:
convergent
is
and has
the value I
when,
Ja
any
number e having been chosen, as small as we

a positive number r\ such that
please,
positive
there is
<e, provided
f(x) dx
that
0< ^
??.
Ja+
And
II.
further:
and
necessary
the interval
f(x)dx
condition for the convergence of
sufficient
is that, if
any
positive
number
has been chosen,
as small as we please, there shall be a positive number

a+
*
if
such that
f(x)dx\<e, when 0<"<l'=s*7-
\\
Also,
r\
f(x)dx converges, we have
this infinite integral

J a
P f(x) dx = \
It follows
from
(I)
that,
f(x) dx
+ P f(x) dx, a<x<b.
f(x)dx converges, to the arbitrary
if
Ja
positive
number
e,
number y\ such that
f(x)dx\<e, when 0<(x-a)^r).
The
Absolute Convergence.
to be
infinite
integral
absolutely convergent, if f(x) is bounded
arbitrary interval (a
+ f,
b),
where
and
0<<b-a,
f(x)dx
is
said
integrable in the
and
j
\f(x)\dx
converges.
It follows
from
(II)
that absolute convergence carries with it

But the converse is not true. An infinite
ordinary convergence.
converge absolutely,*
integral of this kind may converge, but not
as the following
example shows.
*Cf. 43, V; 47, Cor. I;
and
54.
61]
An example
such an integral
of
It is clear that
is
129
suggested at once by
.'o
54.
dx
converges, but not absolutely, for this integral
reduced to
is
r*^dx
by substituting
\x for x.
ft
-.
Ja
\X
dx
r~ converges,
(l)
if
0<n<l.
For we have
"*
dx
r +t(x-a)
{(ft-a) 1 -"-! 1 -"}.
1-
1 ""
(6 -a)
^J a+ t(x-a) ^\-n
dx
t
r
Therefore
lim
mr.
-.
t^
n
n^l.
Also the integral diverges when
From
we
this
obtain
when 0<-n<l.
results
which
correspond
to
those
of 55.
III. Let f(x) be
+f
(a
where
b),
0< <b- a.
If there is
the arbitrary interval
a number
ju
and
between
fb
such that (x-aYf(x)
is
a<x^b,
bounded when
then
f(x)dx
Again,
IV. Let f(x) be bounded and integrable in the arbitrary interval
(a
+ ,
0<<b-a. If there is a number fi greater than

such that (x - aYf(x) has a positive lower bound when
where
b),
or equal to
a< x^b, or a negative upper bound, then

and
the first case,
And
to
f(x) dx diverges
to
+ <x>
in
in the second case.
oo
finally,
V. Let f(x) be bounded and integrable in the arbitrary interval

(a
+ ,
b),
where
If there
0<<b-a.
a number
is
/i
between
and
such that lim (x - ayf{x)
f(x) dx converges absolutely.

a
If there
lim (x
x-*a+0
the
is
- ay
same
a number
f(x) exists
ju
and
greater than or equal to
such that
Cb
is not zero,
then
f(x) dx diverges; and

Ja
is
true if (x
- ayf(x) tends
to
+ oo
or to
- oo
as x->a
-f 0,
130
We
shall
[ch. iv
infinite integral
speak of this test as the ^-test for the
ff(x)dx, when x = a
is
It is
a point of infinite discontinuity.
clear that in applying this test
we
are simply asking ourselves
the order of the infinity that occurs in the integrand.
The
results
limit b
is
Also,
it is
VI. //
[
can be readily adapted to the case when the upper
easy to show that
and
bounded
is
cp(x)
convergent.
<p(x)yjs(x)dx
then
\Js(x)dx converges absolutely,
(a,
is
and
b),
absolutely
(Cf. 56, I.)
most
tests given in (III)-(YI) will cover
The
in
integrable
would not be
of the cases
which
develop in detail
difficult to
But it
shall meet.
the results which correspond to the other tests obtained for the
we
infinite integral
convergence of the
No
discussion
special
f(x) dx.
f(x)dx,
required for the integral
is
when
a certain
in (a,
b),
number
or for
of points of infinite discontinuity occur
f(x)dx,
f(x)dx, as defined
f(x)dx, and
These integrals all reduce to the sum of integrals of the

types for which we have already obtained the required criteria.
We add some examples illustrating the points to which we have
in 60.
referred.
Ex.
1.
Prove that
i* pr.'(,(1
(i)
d*
converges and that
+ X)y/X
h)
x\
"'"*;
fW= (l+lwx
Let
hm
Then
y/x f{x)
= \.
a-->0
1
The
(ii)
/x-test
thus establishes the convergence ot
f
j
dx
iY+v)77x
fw=x(r^y
Let
limxf(x) = h
Then
x->0
Therefore the integral diverges by the same

Ex.
2.
Prove that
"
Jo
The
integral
is
test.
dx converges, when
^|
X rn
l'
an ordinary
finite integral if
<n <
wÔ.
diverges.
61]
lims*fe!) = l.
Also
Therefore the integral converges

Ex.
131
3.
Prove that
dx
v{a
(1
<n<
when
It diverges
when n^L\.
_ a)) converges.
infinities at x0 and = 1.
have thus to examine the convergence of the two
The integrand has
We
dx
y/{x{l-x)}'
where a
The
some number between
is
f
J
dx
y/{x{\-X))'
infinite integrals
dx
converges,
v\^(l "-#)/
since lim {# 2 /( #)}

z->0
dx
converges,
yp
1 -
since lim ((1 -x)*f(x))
V VM
Show
that
ft*
= !,
i
= l,
x->l~0
#J/
where we have written f(x)
4.
and
each case.
ft-test is sufficient in
Ex.
y/{x{\-x)}
log sin x dx converges
and
is
equal to -
\tt log 2.
Jo
The only
from the
x~0, and the convergence
infinity is at
of the integral follows
//,-test.
Further,
fin-
dx
log sin x
=2
log sin
'
2x dx
= 7r log 2 + 2
log sin x
dx + 2
Jo
= 7r log 2 + 4
log cos x dx
Jo
log sin x dx.

Jo
log sin x dx = 2
But
Jo
Therefore
From
log sin x dx.

Jo
log sin
this result it is easy to
xdx
-\tt log
2.
show that the convergent
log (1 - cos x)dx

Jo
and
log (1
integrals
+cos x)dx
Jo
are equal to - ir log
2.
(4""
Ex.
5.
Show that
cos 2nx log sin

Jo
is
tj.
to
- ~- when
^n
,
a positive integer.
The only
the
xdx converges and is equal
fA
infinity is at
test (or
from the
a?
=0 and
the convergence of the integral follows from
last example).
132
Further, on integrating by parts,
we
see that
[*"",.,
x dx = Ux
log sin
cos
[ch. iv
.'o
2nx cos x
[** sin (2n
An
+ l)x =
1+2
sm a;
3 in(2m
f** sin
j
dx
+ l)x + sin
sin
(2n.
1 )x
a;
| cos2m
x
It follows that
cos 2ru; log sin x
dx= -
From
this
we
-r-.
4w
obtain at once
cos 2wa: log cos x
dx= -
7- cos
4w
IT
and
7J-
cos
nx
log 2(1
- cos x)dx = - -,
cos
wx
log 2(1
+ cos x)dx= - n
Jo
Ex.
6.
nir.
cos
Discuss the convergence or divergence of the

rao
er xx n
integral
W7r.
Gamma
Function
- x dx.
-'0
Let n^-1.
(i)
Then the integrand
< x ^ a,
bounded in
is
where a
is
arbitrary,
and we
roc
need only consider the convergence of
e~ xx n
dx.
-'a
The
jix-test
of 55 establishes that this integral converges, since the order of
any given power

Or we might proceed as follows:
e x is
of x.
greater than
Since
ex
x2
= l+x + ^ + ...
ex
when x > 0,
xr
>
e- x* n_1
and
But whatever n may

It follows that,
= any
(r
positive integer),
<^+i-
be,
we can choose r
whatever n
may
so that r -
n+1
>1
be,
00
e -x x r
-1
dx converges.
a
(ii)
Let0<n<l.
In this case e~ xx n
-x
has an infinity at x=0.
shows that
The
/x-test
e -xx
n-i dx converges.
er xx n
~x
dx converges, and we have just shown that
1;
rao
er xx n
Therefore
-'0
- 1 dx converges.
61]
nÔ.
Let
(iii)
133
In this case e~ xx n - x has an infinity at x=0, and the

e -xx n~i
-'
dx diverges to +
/x-test
shows that
oo
Ex.
7.
Discuss the integral

(x r
Since lim
loga;)=:0,
a;"
-1 log
x dx*
when r>0, the
integral
an ordinary
is
integral,
z-K)
when
n> 1.
Also
we know
that
\ogxdx = x(\ogx-l)
\
L_
<x
log x dx = lim x{(\
It follows that
xÔ
=lim
-'o
Again,
lim (x* x a;"- 1 log
< n < 1,
And when
a:)
{x*
=x(l -logz)-l.
- log
+n ~ l
we can choose a
x)
log
1}
a:)
=0,
1.
if
/x
number
positive
>
//,
less
than
which
satisfies this condition.
x n ~ x log x dx converges, when
Therefore
J
Finally,
< n ^ 1.
we have
lim (a: x a:"- 1
>0
a;"- 1 log
Therefore
log
a:
= lim x n
log x
= oo
when
n.
^ 0.
a;M)
a;
dx diverges, when n
^ 0.
.'o
REFERENCES.
De
Ch.
la Vallee Poussin,
loc. cit.,
I (5 ed., 1923), Ch. VI;
2 (4e ed., 1922),
I.
Dini, Lezioni di Analisi Infinitesimale, 2 (Pisa, 1909),
la
Parte, Cap. I
and
VII.
Goursat, loc. cit., 1 (4e ed., 1923), Ch. IV and V.

Hobson, loc. cit, 1 (3rd ed., 1927), Ch. VI.
Kowalewski, Grundzuge der Differential- u. Integralrechnung (Leipzig,
1909), Kap. XIV.
Lebesgue, Lecons sur V Integration, (2e ed., Paris, 1928), Ch. I and II.
Osgood, loc. cit., 1 (4 Aufl., 1923), Kap. III.
Pierpont, loc. cit., 1 (1905), Ch. XII-XV.
Stolz, Grundzuge der Differential- u. Integralrechnung, 1 (Leipzig, 1893),
Absch. X.
And
Brunel, "Bestimmte
Integrale," Enc. d. math. Wiss., Bd. II, Tl. I (Leipzig,
1899).
Montel-Rosenthal, "Neuere Untersuchungen iiber Funktionen reeller VerC 9 6, Integration und Differentiation," Enc. d. math. Wiss.,
anderlichen: II
Bd.
II, Tl. Ill,
2 (Leipzig, 1923).
This integral can be reduced to the Gamma Function integral, and

vergence or divergence follows from Ex. 6. Also see below, Ex. 5, p. 134.
its
con-
134
EXAMPLES ON CHAPTER
1.
Show
l+x +
00
e -a'xf
cosh 6 dx,
l+z
a;
CX
~dx,
)qX+1
**-.
,,
oo
Show that the

b
s in
Jo
log a?
6
i- --,2 dx.
(to,
-,
Jol-^
'
~a
f<
^& where0<c<l,
[f
---.-da,
r <fcc,
'
-l
)qX-\
~a
rr+1
sin w
cos*0 d0.
following integrals are absolutely convergent:
dx
re_U X
V*
Jo
cos
dX}
&a;
[
J
and
-2^ a?m sin
^^
m > 0),
L.95T)**
where P(z)
n th degree,
Q(x)=Q.
is
?i
a polynomial of the roth degree, and Q(a;) a polynomial of the

of
ro + 2, and a is a number greater than the largest root
< x ^ 1 as follows:
l<x^\, f(x)=-3, \<x^%,
Let f(x) be defined in the interval

/(a?)
=2,
i<x^h
f(x)=4,
and so
that the infinite integral
\<x^h
f(x)=-6,
on, the values being alternately positive
Show
5.
+ sina;
Discuss the convergence or divergence of the following integrals:
)a(x-a)y/{b-x)'
4.
a:
loe
8
,
3.
IV.
that the following integrals converge:
ol+cosz + e*
2.
[ch.
and negative.
f(x)dx converges, but not absolutely.
u
Using the substitution x = e- show that
'
x m l {\og x) n dx
,
converges, provided that ro
And by means
and n > -
>
of a similar substitution,
1.
show that
00
a;
w-1 (log
x) n dx
<
and n > -
\^
converges
when
converges, provided that ro

6.
Show
when
fi
that
^ 0,
x>0
and that
the lower limit a of the integral being some
it
diverges
number greater than
unity.
Deduce that
if
there
is
a number
/x
> 0, such that lim {z(log
a:) 1
+<*/()} exists,
X->00
of
then (*/()&! converges, and give a corresponding test for the divergence
this integral, f(x) being
(a, 6),
where
> a.
in
any arbitrary
interval
iv]
Show
that
J2
On
135
J"-""^ dx converges,
00
7.
fa
(x
+ sin 2 a?)
cos x log
integrating
a;
log
a;
by
da;
parts,
we obtain
j =
cos x llog x dx
sin
log x -
a;
x sin
x j
da;.
Ji
roo
Deduce that
cos x log
cos
aj
da; oscillates infinitely.
a;
log x dx converges,
and
îin 'V
equal to -
is
dx.
Jo
-o
fx" cos x 2 dx by
.
we obtain
parts,
x'
97
cos a,J da-
1
= rr-j-.
sin
- tV1
,/
a;
2x
-.
2x
sin
a:
,
2
+^
x " sin
2)x'
a;
2
7
dx,
where x">x'^>0.
roo
Deduce the convergence
cos x 2 dx.
of
.'
9.
Let f(x) and
number of points
two functions.
Prove that
g(x)
be bounded and integrable in
f(x)g(x)dx converges,
is
series
f(a) +f(a
+f(a + 2) +
2^ + l)-2<-i
11.
From
(i)
the relation
l(ir
f(x)dx converges or diverges.
show that
Jo
lim
n>oo
(ii)
By
integration
+ -L... + -l<2^-l.
sin
by
a;
sin ^
:
.'0
+ 1) and
mJ^ z=2
sin
Deduce that
. .
î+3 ^ +
converges to a value between
parts,
J^...
\ir.
i cos (2r-l)x,
j
dx=2 2
x
\2r -
-.
dx -^..
"
show that
w
f*
lim
wôoJo
dx converge.
values of the positive integer n,
Also show that
lim/(a;)=0.
x>co
all
g(x)
xâ, and
convergent or divergent according as
Prove that for
Ja
Ja
Let /(a;) be monotonic when
Then the
and
\f(x) dx
if
Ja
10.
except at a certain
(a, b),
of infinite discontinuity, these points being different for the
sin 2na:
/ 1
1\
\sina;
xj
da;
=0.
136
-oo
ni)
12
From
(i)
sin
X _TT
the above, prove that
Prove that -if u n =
sin
2nx cot x dx
Jo
vn =
and
u n =^Tr,
then
(ii)
By
integration
by
f*
and
parts,
w sin 2na;
dx,
-i da:.
v = limv n =
n->oo
Jo
show that lim
(v n
- w n ) 0.
n->oo
(iii)
From
the above, prove that

Jo
dx=~=.
A
CHAPTER V
THE THEORY OF INFINITE SERIES, WHOSE TERMS ARE
62.
We
shall
now
consider
We
some
of the
properties of series
of x.
denote such a series by
u x (x) +u 2 (x) +u z (x) + ...

and the terms of the series are supposed to be given
x in some interval, e.g. (a, b)*
for values
of
When we
speak of the
sum
of the infinite series
u i( x ) + u 2( x ) + u z( x )
it is
to be understood:")"
(i)
that
we
settle for
what value
of x
we wish the sum
of the
series;
(ii)
that
we then
insert this value of x in the different terms
of the series;
(iii)
that
we then
find the
sum
s'n (x)
of the
first
n terms;
and
(iv)
that
On
we then
all
find the limit of this
sum
as n->oo
keeping
the time at the value settled upon.
this understanding, the series
u x (x) +u 2 (x) + u 3 (x) + ...

is
said to be convergent for the value x, and to have f(x) for
its
sum,
in 24, when we say that x lies in the interval (a, b) we mean

In some of the results of this chapter the ends of the interval
When this is so, the fact that we are dealing
are excluded from the range of x.
x < 6) will be stated.
with the open interval (a
*As mentioned
that
a^xÊzb.
<
tCf. Baker, Nature, 59 (1899), 319.
137
<
this value of
if,
WHOSE TERMS
INFINITE SERIES
138
x having been
and any
the series,
as we please, there
first inserted
number
positive
in the different terms of
having been chosen, as small
a positive integer
is
[oh.
such that
\f(x)-s n (x)\<, when n^v.

Further,
necessary and sufficient condition for convergence
positive
number
any
has been chosen, as small as we please, there shall
a positive integer
be
is that, if
such that
\Sn+ v (x)-Sn(x)\<e,
when n =
>
A similar convention exists when we are dealing with other limiting processes.
In the definition of the differential coefficient of f(x) it is understood that we
y
first agree for what value of x we wish to know/ (z); that we then calculate
f(x) and f(x + h) for this value of x; then obtain the value of
f(x, a) dx,
Again, in the case of the definite integral
insert in f(x, a) the particular value of a for
before
We
we proceed
to the
hÔ.
finally take the limit of this fraction as
we
fix)
+ h)
-,
and
f(x
summation and
it is
understood that
which we wish the integral
limit involved in the integration.
shall write, as before ( 19),
f(x)-s n (x) = R n {x),

where f(x)
is
the
sum
terms.
remainder after
As we have seen
of the series,
and we
R n (x)
sum
in 19,
is
u n+1 (x) +u n+2 (x)

Also
we
call this
With
R n (x)
the
of the series
+u n+s (x)+...
shall write
p
and
the
shall call
R n (x) = s n+p (x)-s n (x),
a partial remainder.
this notation, the
two conditions
for convergence are
\R n (x)\<e, when n^v\
(i)
(ii)
|
R n (x)
e,
when n^v,
for every positive integer p.*
series
and not
When
s n (x),
may
converge for every value of x in the open interval
for the end-points
there
is
a or
no ambiguity
and write s n
R n (x), pB n (x)
a<x<b
b.
it will
Rn
sometimes be convenient to omit the x
and P R n
in
ARE FUNCTIONS OF A SINGLE VARIABLE
62, 63]
l+x + x + ...
E.g. the series
converges and has
When
1, it
for its
diverges to
The Sum of a
63.
of x
may
139
Series
sum, when oo
<x<1
when x = -
1, it oscillates finitely.
whose Terms are Continuous Functions
be discontinuous.
periodic function of x given
Until Abel* pointed out that the
by the
series
3x-
2 (since- J sin2^+J- sin
...),
which represents x in the interval -7r<x<7r, is discontinuous

at the points x = (2r + 1) 7r, r being any integer, it was supposed that
a function defined by a convergent series of functions, continuous
must itself be continuous in that interval.

Indeed Cauchyf distinctly stated that this was the case, and later
writers on Fourier's Series have sometimes tried to escape the
in a given interval,
difficulty
by
asserting that the
sums
of these trigonometrical series,
at the critical values of x, passed continuously

just before
from the values
those at the points of discontinuity to those just
after. {
sum
This mistaken view of the
The
different errors.
increases,
first
of
such series was due to two
consisted in the assumption that, as n
the curves y = s n (x) must approach more and more
nearly to the curve y=f(x), when the sum of the series

an ordinary function capable of graphical representation.
is
f(x)
These
curves y = s n (x)
we shall call the approximation curves for the

but we shall see that cases may arise where the approximation curves, even for large values of n, differ very considerably
from the curve y=f(x).
series,
It is true that, in a certain sense, the curves

(i)
y = s n (x)
and
(ii)
y=f(x)
approach towards coincidence; but the sense is that, if we choose

any particular value of x in the interval, and the arbitrary small
positive
number
<?,
there will be a positive integer
such that, for
this value of x, the absolute value of the difference of the ordinates
of the curves
(i)
and
(ii)
*Abel, Journal fur Math.,
will
be
less
than
,e
Partie, p. 131.
(Ser. 2), T. Ill, p. 120.

cit.;
when n g
v.
1 (1826), 316.
fCauchy, Cours d 'Analyse (1821),

%0i. Sachse, loc.
Donkin, Acoustics (1870), 53.
Also CEuvres de Cauchy,
INFINITE SERIES
140
WHOSE TERMS
[CH.
not the same thing as saying that the curves coincide

They do not, in fact, lie near to each other in the
geometrically.
neighbourhood of a point of discontinuity of f(x); and they may
Still this is
not do
so,
even where f(x)
is
continuous.
The following examples and diagrams

Ex.
1.
illustrate these points:
Consider the series
z+
(z+l)(2x+l)
1
ujx)
Here
(n-l)x+l
nx+l
and
(*)
Thus, when
x>0,
= !- nx+l
limsn (aj) = l:
91-CO
when x 0,
lims(a;)=0, since
s n (0)=0.
for
The curve y=f(x), when x ^ 0, consists of the part of the line y = 1
=0.
X
at
discontinuous
is
series
x > 0, and the origin. The sum of the
Now
which
examine the approximation curves
This equation
may
be written
more and
As n increases, this rectangular hyperbola (cf. Fig. 10) approaches
of the
shape
the
from
reasoned
=
we
more closely to the lines y l, x=0. If
Fig. 10.
of the axis of y for

approximate curves, we should expect to find that part
when x 0.
curve
=f(x)
the
of
y
portion
a
which < y < 1 appearing as
63]
HI
As sn (x) is certainly continuous, when the terms of the series are continuous,
the approximation curves will always differ very materially from the curve
y=f(x), when the sum of the series is discontinuous.
Ex.
2.
Consider the series
u 1 (x) + u 2 (x) + u 3 (x) +

uJx) =
where
...
x = 0,
nx
(n-l)x
l+n?x2
l+(n-l)W
In this case
and
lim sn (x)
n-x
Thus the sum
=0
for all values of x.
of this series is continuous for all values of
see that the approximation curves differ very materially
y=f{x)
in the
neighbourhood of the
x but we shall
from the curve
t
origin.
y=sn {x).
The curve
has a
maximum
at (1/n, )
l+n2 x2
and a minimum
points on the axis of x just below the
at
1/rc,
- )
(cf.
Fig. 11).
The
maximum and minimum move in towards
n increases. And if we reasoned from the shape of the curves

should expect to find the part of the axis of y from - to \
appearing as a portion of the curve y=f(x).
the origin as
y=sn (x), we
Ex.
3.
Consider the series

,()
+ u 2 (x) + u 3 {x) +
where
= 0,
l+(n-l)W
l+n3x2
Here
and
. .
(n-l) 2x
sn (x)
lim
n2x
'
l+n3x2
sn (x) =0 for all
values of x.
INFINITE SERIES
142
The sum
(cf.
of
~W
[CH.
but the approximation curves

n ) and a minimum at
which have a maximum at (l/V^3
from the curve y=f( x in the neighn )> din?er ver y g rea
the origin. Indeed they would suggest that the whole of the
of the series is again continuous,
Fig. 12),
(-l/y'n 3
bourhood
WHOSE TERMS
axis of y should appear as part of y=f{x).
These remarks dispose of the assump64. Repeated Limits.

tion referred to at the beginning of the previous section that the
approximation curves y = s n (x), when n is large, must approach

closely to the curve y=f{x), where f(x) is the sum of the series.
The second error alluded to above arose from neglect of the
convention implied in the definition of the sum of an infinite
whose terms are functions of

finding the sum has been set out in
series
to
whom
reference
x.
62,
The proper method
of
but the mathematicians
now made proceeded

the sum for a value of
is
in quite a different
at which
x, say x
In finding
a discontinuity occurs, they replaced x by a function of n, which
converges to x as n increases. Then they took the limit when
manner.
In this method x and n approach

value of this limit may quite
the
and
their limits concurrently,
Indeed, by choosing
well differ from the actual sum for x = x
the function of n suitably, it can be made to take any value between
this
f(x +Q) and/(z -0), while in some cases it goes outside
n-âo of
s n (x) in its
new form.
interval
(cf.
Ch. IX, p. 293).
63, 64]
For instance, in the
series of 63,
Ex.
143
1,
"'
z+l + (z+l)(2x+l) +
>n0,
-
'
we have seen that x = is a point of discontinuity.

If we put x=p/n where p is positive, in the expression for s n (x),
and then let w->oo p remaining fixed, we can make lim s n (pjn) take
,
ft=>-co
any value between

have
s n (pln)
and
= -^~,
according to our choice of p.
1,
which
is
lim
independent of
to 1 as
It will be seen that the
and
= -P--,
s n (pln)
P ~r 1
n->3
which passes from
n,
For we
increases
from
to oo
matter at issue was partly a question
words and the misunderstanding of a definition. The confusion

can also be traced, in some cases, to ignorance of the care which
of
must be exercised in any operation involving repeated

we are really dealing here with two limiting processes.
If the series is convergent and its sum is/(#), then
limits, for
f(x) = lims n (x),

n>co
and the
a limit,
limit oif(x) as x tends to x

is
lim /(#):= lim
we may
If
dinate
assuming that there
is
such
given by
of
the
[lim s n (x)]
(1)
use the curve as an illustration, this
we move
point towards which
as
is
the or-
we proceed
along the curve y=f(x), the abscissa getting nearer and nearer
but not quite reaching x
According as x approaches x
to x
from the right or
~Nowf(x
),
the
left,
sum
the limit given in
of the series for
lim[s n (x
n>-x
and
since
we
are
tinuous functions,
now
x=x
is,
by
dealing with a definite
s n (x) is
be/(z +0) or
definition,
)],
number
of con-
a continuous function of x in the interval
with which we are concerned.
Thus
(1) will
s n (xQ )
lim
x->x
s n (x)
WHOSE TERMS
INFINITE SERIES
144
Therefore
jbhe
sum
x=x
of the series for
lim
[lim s n (x)]
n^<x>
x~>jcq
may
[oh.
be written
(2)
The two- expressions in (1) and (2) need not be the same.
are so only when/(x) is continuous at x
They
Uniform Convergence.* When the question of changing the

conorder of two limiting processes arises, the principle of uniform
series
infinite
of
case
the
for
explain
vergence, which we shall now
65.
fundamental. What is involved

the series
in this principle will be seen most clearly by returning to

x
+
i
z+l
(x+l)(2x+ir -'
i
s n {x)
In this series
'It
"
>
><J2
R "W = ^\> x>0
Also
some
n,
=1
x>0
lims n (z) = l when x>0.
and
If
of x, is
positive x
only
is
taken,
number
and # (0)=0
>
e is
chosen, less than unity, and
clear that l/(na>
it is
+l)<e
for a positive
if
n> x
E.g. let
If
= j^q-
3=0-1, 001, 0001,
10- p respectively, l/(* +

w>10 4 105 10 6 ... 10P+ 3
...
And when c =
j^^
and
a;
= 10"*, n must
1)
<
only
when
be greater than 10/'+
fl
if
l/(nar+l)<.
and more terms of the

As we approach the origin we have to take more
the series by less than
of
sum
the
from
differ
series to make the sum of n terms
a given number.
1
When
* = 10- 8 the
,
first
million terms do not contribute
per cent, of the sum.
1-1
rl> "7"
The inequality
shows that when n
is
any given positive number
simple treatment of uniform convergence

Butt. Amer. Math. Soc., 3 (1896).
*A
will
less
than unity,
be found in a paper by Osgood,
64, 65]
and x approaches nearer and nearer to

integer which will
without
all less
...
than
increases
limit.
There
is no positive integer v which

than this e in x ^ 0, the same
all less
zero, the smallest positive
R n (x), R n+1 (x),
make
145
make
will
v
v (x),
(sc),
...
serving for all values of x
in this range.
On
the other hand there
this condition,
definite positive
Such a value
Our
is
a positive integer
the range of x
if
is
given by x
which will satisfy

where a is some
^ a,
number.
of v
would be the integer next above (- - l)
series is said to converge
inxâ, but
uniformly
it
a.
does not
converge uniformly in x =^ 0.
We
turn
now
to the series
u x (x) +u 2 (x) +u 3 (x) + ...

and
define uniform convergence* in
u x (x)
Let the series
an interval as
+u 2 (x) +u
2>
(x)
follows:
+ ...
converge for all values of x in the interval
a^x^b
and
its
sum
be
It is said to
converge uniformly in that interval, if, any

positive number e having been chosen, as small as we please, there
is a positive integer v such that, for all values of x in the interval,
f(x).
\f(x)-s n (x)\<e,
It is true that,
(a, b)
if
when n^v.-f
the series converges, \R n (x)\< for each x in
when n ^ v.
The additional point in the definition of uniform convergence

that, any positive number e having been chosen, as small as
we please, the same value of v is to serve for all the values of x in the
is
interval.
For
this integer v
we must have
\R {x% [A+if*)T,
v
all less
than
e,
no matter where x
lies
...
in (a,
b).
*The property
of uniform convergence was discovered independently by Stokes

Trans. Phil. Soc. Camb., 8 (1847), 533) and Seidel (cf. Abh. Ale. Wiss. Milnchen,
See also Hardy, Proc. Phil. Soc. Camb., 19 (1920), 148.
5 (1848), 381).
(cf.
fWe can also have uniform convergence in the open interval a x

6, or the
half-open intervals a
x
b,
b; but, when the terms are continuous
in the closed interval, uniform convergence in the open interval
carries with it
uniform convergence in the closed interval (cf. 68).
< ^
a^x <
< <
WHOSE TERMS
INFINITE SERIES
146
The series does not converge uniformly

for some positive number (say e ) there
which
will
than
all less
^^ i^^
make
e
for every x in
in (a, b)
is
if
[ch.
we know that
no positive integer
...
(a, b).
It will be seen that the series
x
x~+l
+ (z+l)(2z+l)
converges uniformly in any interval
a^x^b,
where
a, b
are
any
given positive numbers.

zero,
It may be said to converge infinitely slowly as x tends to
origin,
to
the
nearer
still
and
nearer
get
we
as
that,
sense
in the
we cannot fix a limit

to make \R n (x)\<e.
to the
number
It is this
of
terms which we must take
property of
infinitely slow con-
vergence at a point (e.g. x ) which prevents a series converging

uniformly in an interval (x -S, x +S) including that point.
Further, the above series converges uniformly in the infinite
interval xâ, where a is any given positive number.
uniform conIt is sometimes necessary to distinguish between
vergence in an
which
interval,
infinite interval
may
and uniform convergence

we please.
in a fixed
be as large as
series is convergent for all values of x, but

in the infinite interval x !== 0.
uniformly
does not converge
n
is greater than x jn\, when x
(x)
For in this series R n
The exponential
it
is
positive.
n
the series were uniformly convergent in xÔ, x \n\

would need to be less than e when n ^ v, the same v serving for
Thus,
all
if
values of x in the interval.
But
it is
make
n(x)
greater than
we need only take x

than e for n equal to v.
clear that
greater
{y\e)
to
However, the exponential series is uniformly convergent in the

large as we
interval (0, 6), where b is fixed, but may be fixed as
please.
For take
for x
c greater
b.
We know
that the series converges
c.
R n (c)<e, when n^v.

R n (x)<R n (c), when m x ^ b<c.
Therefore
But
Therefore
of
than
x in
R n (x)<e, when n^v,
(0, 6).
the same
serving for
all
values
65, 66]
From
interval
147
the uniform convergence of the exponential series in the

(0, 6), it follows that the series also converges uniformly
in the interval
- b,
b),
where in both cases b
is
fixed,
but
may
be fixed as large as we please.

Ex.
Prove that the
1.
series
l+x + x2 + ...
Ex.
l/( 1
Prove that the
2.
- x)
^x^x <
in
series
(l-x)+x(l-x)+x2 (l -x) +
Ex.
Prove that the
3.
in
^ x = x
...
<1
series
-x) 2 + x(l -x) 2 + x 2 (l-x) 2 +

converges uniformly to (1 - x) in
x
1.
(1
...
^ ^
Ex.
Prove that the
4.
series
A
1+x
L_ + _J_
2 + x2
^3 + x 2
converges uniformly in the infinite interval x

Ex.
Prove that the
5.
"'
^ 0.
series
1.2
273
3T^ +
""
converges uniformly in the interval

as large as
interval
66.
we
please,
and that
it
(0, b), where b is fixed, but may be fixed

does not converge uniformly in the infinite
xÔ.
necessary and sufficient
condition for Uniform Conthe sumf(x) is known, the above definition often
gives a convenient means of deciding whether the convergence is
vergence.
When
uniform or not.
When the sum is not known, the following test, corresponding

to the general principle of convergence (15), is more suitable.
Let
u (x) +u (x) +u (x) +...
x
be
an
necessary
infinite series,
and
whose terms are given in the interval
oufficient condition for the
the series in this interval is that,

if
any
uniform convergence of
number e has been

a positive integer v such
positive
chosen, as small as
there shall be
that, for
interval,
we please,
aU values of x in the
(a, b).
|,2J n (a;)|<e,
when n^v, for
every positive integer p.

(i)
The condition
is necessary.
Let the positive number

Then take |e.
be chosen, as small as we please.
WHOSE TERMS
INFINITE SERIES
148
Since the series

integer
is
uniformly convergent, there
[oh.
a positive
is
such that
v,
when
\f(x)-s n (x)\<ie,
the same
serving
values of x in
for all
i^,
ft
(a, b),
f(x) being the
sum
of the series.
Let
ft",
Then
ft'
'
|
be any two positive integers such that ft">ft'
v(0 - M*)
WV
-/(*)|
'(*)
" *'(*)
/(*)
v.
<e.
Thus \8 n+9 (x) - S n (x)\<, Wnen U = V f r eVer ^ P 0sitlve

the same v serving for all values of x in {a, b).
(ii) The condition is sufficient.
>
We know
that the series converges,
when
inte S eI
this condition is
satisfied.
Let
its
sum be f(x).
Then
the arbitrary positive number e be chosen.
that
there is a positive integer v such
(x)-s (x)\<i, when n^v, for every positive integer p,
\s
Again
let
n+p
the same
serving for
all
values of x in
{a, b).
sv (x)-ie<s v+p (x)<s v (x) +|e.
Thus
\ims u+p (x)=f(x).
Also
p>Q0
Therefore
- \e =g/(a?)
(a?)
^ s
(a?)
+ Je
+\8v {x)-'f{x)\.
But
|* B (aO-/(aON(*)-M*)l
equal to the value
It follows that, when n is greater than or
specified above,
K(x)-f(x)\<ie+ie
and this holds for all values of x in (a,
Thus the series converges uniformly
67. 1. Weierstrass's
The
Uniform Convergence. The

uniform convergence is due to Weierstrass:
M-Test
following simple test for
u x (x)
series
will converge uniformly in
positive constants
for
+u 2 {x) +u s (x) +
(a, b),
^M
. .
if there is
M +M +M
such that, no matter what value x

\uJx)\
b).
in this interval.
may
a convergent
+ ...,
have in
(a, b),
positive integer n.
n for every
series of
66-67. 2]
M +M +M
Since the series
3 -f
149
. .
convergent, with the usual notation,
is
M +M
when n ^
<
n+1
n+2 + ...+M n+p
j/,
+ K+ 2 0*)| +... +\u n+p {x)\.

IAW|^Mw+1 -blfn+2 + +M n+3
But
|,fl(3)|^K +1 (a;)|
Thus
. i .
<e, when
n^

the inequality holding for all values of x in (a, b).
Thus the given
v,
uniformly convergent in
series is
For example, we know that the
(a, b).
series
l+2a+3a2 + ...
convergent,
is
when a
any given positive number
is
less
than
unity.
It follows that the series
l+2x+3x 2 + ...
uniformly convergent in the interval
is
Ex.
is
1.
Show
- a,
a).
that the series
x cos + x 2 cos 20 + x3 cos 30 + ...

uniformly convergent for any interval (x , x x ), where is any given number.
Ex.
2.
Show
x cos
+ x2
cos 20
+ x?
cos 30
+ ...
x cos
+ -=
cos 20
+ -=
cos 30
are uniformly convergent for

less
< x < xx < 1
all
values of
0,
when
. .
\x\ is
any given
Uniform Convergence.
67. 2. Further Tests for
convergence
Abel's Test
is
not required in the following
and
In the M-Test
Abel's Test.
-\-u 2 (x)
(a, b)
v t (x),
and
is
+u 3 (x)
(1)
the sequence
v 2 (x),
monotonicfor every (fixed) x in

function fn (x)
v 3 (x),
(a, b)
(2)
and uniformly bounded*
said to be uniformly bounded in
an
interval,
number K, independent of x and n, such that fn (x) <K,

the interval, and every positive integer n.
positive
x in
But absolute
usually called
Let the series
u x {x)
*A
tests,
Dirichlet's Test.
converge uniformly in
be
positive
than unity.
the series converges absolutely and uniformly.
I.
and
that the series
and
number
when
there
is
for every value of
INFINITE SERIES
150
Then
Let
r n (x)
WHOSE TERMS
[ch.
the series
Ui(x)v x (x)
is
+u 2 (x)v 2 {x) +u 3 (x)v 3 {x) +
(3)
(a, b).
be the partial remainder for the series (3) and s n (x),

the sum of n terms and the partial remainder for the series (1).
v
R n (x)
Then
R n {x) = u n+1 (x)v n+1 (x) +u n+2 (x)v n+2 (x) + ...
+ u n+v (x)v n+p {x)
= !T n {x)v n+1 (x) + 2 rjx) -
r n (x)] v n+2 (x)
+ [r n (x)- p- 1
= ^ n {x) [v n+1 (x) - v n+2 {x)] +
+ ...
r n (x)]v n+p (x)
+ ,-l^n() [+n-l(*)
Now
all
known
it is
. .
W^M + ^K+1^)"'^)
that
being fixed; and that there is a positive

such that \v n (x)\<K for all values of x in (a, 6), and
have the same
number
sign,
a;
every positive integer n.

Also, since the series (1) converges uniformly, when the arbitrary
that
positive number e is chosen, there is a positive integer v, such
ll*(s)|.
are each less than ej2>K
values of x in
It follows
I2*n(a0l
X )I
when w^y, the same
serving for
all
(a, 6).
from
(4)
that
AW < 3
n+lO*0
_ V n+,(s) +
3JE
n+()
<l+ie
3 e_r 3<
<e, when n^v,
the same
i/
Thus the
Ex. 1
values of x in (a, b).

all
series (3) is
Let a + a x + a 2 +
Then
serving for
. . .
(a, b).
n
be a convergent series of constants and vn (x) = x
O^x^
a n n () converges uniformly in
1.
=
1
Ex.
Let a x + a 2 +
2.
Then
2 Vn(z)
be a convergent
converges uniformly in
series of constants
a;
and
v n (x)
^ 0.
Ex.
a2
...
a lt
Let a + a 1 + a 2 + ... be a convergent series of constants and a
series
the
Then
numbers.
positive
of
be a monotonic ascending sequence
a e- a o x + a x e,-^ + a 2 e~ a zx
,
3.
converges uniformly in x
= 0.
67. 2]
Let
II. Dirichlet's Test.
= u 1 (x) + u 2 (x)+...
s n (x)
Then
u n (x).
the series
u^v^x) +u 2 (x)v 2 (x) + u 3 (x)v 3 (x) +

s n (x) is
(i)
151
...
provided that
(a, b)
uniformly bounded in
(a, 6)*
and
v x (x), v 2 (x), v 3 (x),
(ii)
formly
in
to zero
is
...
(a, b).
With the same notation

p
a monotonic sequence converging uni-
as above,
R n (x) = u n+1 (x)v n+1 (x) + +u n+v {x)v n+v {x)

= [*i.+l(*) - Sn(x)] V n+1 {x) H S n+2( X ~ S n+l(%)] V n+2 {x) +
...
+ [*+*() "
Sn
. .
V
+J>-l(x)} n+P (%)
+ 8 n+p _ 1 (x) [Vn+p-lW - V n+P (x)]

+s n+p (x)v n+Jt {x) - s n {x)v n+1 (x).
Then we have
\
at once
R n (x)\<K{\v n+1 (x)-v n+p (x)\ + \v n+p (x)\
+j n+1 (*)|}.
But the sequence v^x), v 2 (x), ... converges uniformly to zero.

we know that, however small the arbitrary positive
number e may be, there is a positive integer v, such that
Therefore
< 37r
v n (x)
the same
And
serving for
v^x), v 2 (x),
all
n-
v>
are
etc.,
wtien
all of
the same sign.
Therefore
\
the same
serving for
R n (x)\<
all
b;
when n^v,
values of x in
oo
SU
Ex.
1.
The
series
converges uniformly
Cf. footnote
on
(a, b).
l+x 2
when x
p. 149.
n( x ) v n{%)
-^j
0.
k +
+ x2 'S^ + x i
(a, b),
and the
series
. . .
Ex.
WHOSE TERMS
INFINITE SERIES
152
The
2.
x + \ sin 2x + 1 sin
sin
series
+
+ ...
3a;
|ch.
. .
cos x + \ cos 2x + % cos 3x
Ex.
3.
The
x - \ sin 2x + sin 3x
cos x- J cos 2x + ^ cos 3x
sin
Ex. 4.
The
< a < b < 2tt.
when
(a, b),
series -
- a,
a),
where
sin x + sin
series
< a<
...
...
ir.
3# + ^ sin
5a;
. . .
+
cos x + \ cos 3a: + 1 cos 5x+ ...
\ cos 2x + \ cos 4a; + \ cos 6a; +
(a, 6), when
< a < 6 < it.
J sin 2x + 1 sin 4a; + $ sin
6a;
. .
. .
30
Ex. 5.
The
series
00
2 cos nx
< x < x < 2tt,
and
a w sin nx
TC
provided that the

Xj) when
converge uniformly in {x
x
constants a lt a 2 ... form a monotonic sequence and lim an = 0.
Uniform Convergence of
68.
>ao
whose Terms are Con-
Series
In the previous sections dealing with

of the series have not been assumed
terms
uniform convergence the
continuous in the given interval. We shall now prove some
tinuous Functions of x.
properties of these series

I.
when
this condition is added.
Uniform convergence implies continuity in the sum.
If the terms of the series
u x (x) +u 2 (x) +u 3 (x)

are continuous in
_(a, 6),
and
+...
the series converges
uniformly
in this interval, thenf(x) is a continuous function of x in
we know
Since the series converges uniformly,
number
small the positive
may
be, there is
to
f(x)
(a, 6).
that,
however
a positive integer
v,
such that
\f(x)-s n {x)\<ie, when n^v,

the same
serving for
all
values of x in
{a, b).
Choosing such a value of n we have

}
f(x)=s n (x)+R n (x),

where
|22 w (sc)|<Je,
Since s n (x)
tinuous in (a,
is
for all values of
the
sum
of
x in
(a, b).
n continuous
functions,
it is
also con-
b).
Thus we know from 31 that there is a positive number r\ such that
when
x, x' are
\x'-x\^t].
\*n&) -*(*) \'<fc>

of x in the interval
any two values
(a, 6)
for
which
07. 2, 68]
f(x')=s n (x')+R n (x'),
But
where
Also
/(') -/()
Thus
!/(*')
En (x')\ < Je.
= [ajp) - s n (x)] + R n {x') - Rn (x).
-/(x)|^|* w
M-*WI + l*(*')l
<e, when
Therefore /(ic)
//*
II.
is
series,
continuous sum,
|a?'
continuous in
its
+|*.(*)|.
x\^rj.
(a, 6).
whose terms are continuous functions, has a

it
dis-
cannot be uniformly convergent in an interval
which contains a point of discontinuity.

For if the series were uniformly convergent,
that
153
sum must be continuous
we have
just seen
in the interval of uniform con-
vergence.
III.
Uniform convergence
continuity of the
sum
thus a sufficient condition for the
of a series of continuous functions.
It is not
since different non-uniformly convergent

known, which represent continuous functions in the interval
a necessary condition
series are
is
of non-uniform convergence.
For example, the
series discussed in
are uniformly convergent in
x^ a>0,
Ex. 2 and Ex. 3 of
for in
#~a>0.
wnen
\R n v(%)\<
" nx~na
>
63
both cases
Thus |jK n (a;)|<e, when n>l/ae, which is independent of x.

But the interval of uniform convergence does not extend up to
and include x =
even though the sum is continuous for all
y
values of x.
This
is
clear in
that \R n (x)\<e,
Ex.
2,
nx
where RJx) = ^
when n^v, the same
k-~, for if it is
serving for
all
asserted
values of x
^ 0, the statement is shown to be untrue by pointing out that

when n ^ right through
for x=\\v, R (x) = \, and thus R n (x)
in x
the interval,
if
<fc
e,
v,
e<|.
mini
Similarly in Ex.
3,
where
R n (x)= x ~\~n3x2
if it is
asserted that
\R n (x)\<e, when n^v, the same v serving for all values of x in

xÔ, we need only point out that for x=llJvs Rv (x) = \Jvz
Thus \R n (x)\<e, when n^v, right through the interval, if e<|.
,
INFINITE SERIES
154
There
both
in
is,
cases,
WHOSE TERMS
[en. v
integer
positive
which
for
R n (\jm)<<\, when n^ but this integer is greater than 1/m.

Thus it is clear that the convergence becomes infinitely slow
i>,
as x->0.
IV. If the terms of the series are continuous in the closed interval
then it must
(a, b), and the series converges uniformly in a<x<b,
convergence
the
converge for x = a and x = b, and the uniformity of
will hold for the closed interval (a, b).
Since the series
a<x<b, we
is
uniformly convergent in the open interval
have, with the usual notation,
\sjx)-s n (x)\<e, when m>nî>,

the same
(1)
serving for every x in this open interval.
Let m, n be any two positive integers, satisfying this relation.

Since the terms of the series are continuous in the closed interval
that
(a, b), there are positive numbers r] 1 and rj 2 say, such
,
\sjx)-s m (a)\>le,
\s n (x) - s n (a)\ < Je,
and
Choose a positive number
r\
when 0^(x-a)^^
when ^ (x - a) ^ 2
l5
r}
not greater than
r\
or
r)
and
2,
let
0^(x-a)^rj.
Then
\sja)-s n (a)\
sja) - sjx) + sjx) - s n (x) +
1
when
<t,
similar
s n (x)
- s n (a)
m>wê
(-0
argument shows that
(3)
m (b)-s n {b)\<e, when m>nî>
and
a
x
for
=
converges
series
the
that
see
we
From (2)
(3)
x = b, and, combining (1), (2) and (3), we see that the condition
\s
and
uniform convergence in the closed interval
for
If
the terms of the series
are continuous in
(a, ),
(a, b),
where a < a
u 1 {x) + u 2 {x) + u 3 {x) + ...

and the series converges uniformly in every
< p < b,
However we
x=a
the case
But
+ 2x + 3x 2 +
. .
where a < 1, but

x - 1 or x = l.
a, a),
shall see that in the case of the
or x = b,
may be.
interval
the series need not converge for x = a or x = 6.
E.g. the series
for
(a, b) is satisfied.
the uniform convergence in

(Cf. 72.)
this property is not true in general.
it
does not converge for
Power
(a,
(>)
Series,
if it
converges
extends up to a or
b,
as
68, 69]
The
series of
continuous functions
u x (x) + u z (x) + u 3 (x) +
may
155
converge uniformly for every interval
(a,
...
fi)
within
(a, b),
and converge
x = a or x =6, while the range of uniform convergence does not extend up

to and include the point a or 6.
for
x + lx3 -}2 x 2 + lx 5
E.g. the series
formed from the logarithmic
+ Ix 1 - \x*+
(1)
series
x - } x2 + lx z
(2)
by taking two consecutive
<x~
positive terms
and then one negative term,
is
sum, when x = l, is :} log 2.*

Further, the series (2) is absolutely convergent when \x\ < 1, and therefore
the sum is not altered by taking the terms in any other order. (Cf. 22.)
It follows that when \x\ < 1 the sum of (1) is log (1 + x), and when x = 1 its
convergent
sum
is
l\
Hence
when -
1,
and
its
log 2.
(1) is
discontinuous at #
and therefore the
interval of uniform
convergence does not extend up to and include that point.
Uniform Convergence and the Approximation Curves.
69.
continuous functions be uniformly convergent in
Let a
series of
(a, b).
b
FIG. 13.
Then we have,
as before,
|
the same v serving for
In particular,
and we
all
|
<
e, when ra> n~v,

n (x)
m (x) -
m (x) -
v (x)
<
e,
when ra>
v,
suppose v the smallest positive integer which will satisfy this

condition for the given e and every x in the interval.
Plot the curve ys v {x) and the two parallel curves y=s v (x) t, forming a
strip
(t
*Cf.
shall
of breadth 2c,
whose central
line is
Hobson, Plane Trigonometry (7th
y=s v (x).
ed., 1928), 251.
(Fig. 13.)
INFINITE SERIES
156
WHOSE TERMS
All the approximation curves y =sm {x),
y=f(x), where f{x)
is
the
reaches its boundaries.
sum
(Cf.
of the
66
m>
v, lie
series, also lies
[ch.
in this strip,
within the
and the curve

most
strip, or at
(ii).)
the corresponding smallest positive integer

uniform convergence be /. Then / is greater
first strip, and
than or equal to v. The new curve y=sv '(x) thus lies in the
the first
outside
goes
if
it
before,
as
formed
the new strip cr' of breadth 2e',
only
concerned
are
for
we
out,
blotted
portion
this
have
can
part,
strip in any
increases from
with the region in which the approximation curves may lie as m
Next choose e'
less
than
and
let
satisfying the condition for
the value
v.
In this way,
if
we take the
set of positive
> i! > e"
. . .
numbers
where lim
e()
= 0,
and the corresponding positive integers

v
we
obtain the set of strips
Any
= v = v"
<r,
strip lies within, or at
</,
cr", ...
most reaches, the boundary
of the preceding
number increases.
one, and their breadth tends to zero as their
the boundary of
Further, the curve y=f(x) lies within, or at most reaches,
the strips.
continuity of the sum

This construction, therefore, not only establishes the
an interval of uniform convergence,
of the series of continuous functions, in
terms
shows that the approximation curves, as the number of the
of the curve for the sum right
shape
the
to
guide
a
as
used
be
may
increase,
but
it
through the interval.*

70. 1
Series
A sufficient Condition for Term by Term Integration of a
whose Terms are Continuous Functions of x
series of
When
the
continuous functions
u^x) +u 2 (x) +u 3 (x) + ...

seen that
uniformly convergent in the interval (a, b), we have
It follows that f(x) is
(a, b).
its sum, f(x), is continuous in
b.
<x
x
a
when
x
^
and
1 m
Q
integrable between x
v
that the
examination,
further
without
But it does not follow,
is
series of integrals
l
f\(z)dc + [
J,
is
convergent, and, even
without proof, that
its
+\
u z (x)dx
+ ...
J.'o
if it
sum
*Of course the argument of

graphically represented.
u 2 (x)dx
J>
is
be convergent,
it
does not follow,
f(x)dx.
this section applies only to
such functions as can be
69, 70. 1]
The geometrical treatment
157
approximation curves in
of the
when the given
suggests that this result will be true,
69
series is
uniformly convergent, arguing from the areas of the respective

curves.
We
shall
now
demonstration
theorem more precisely and give
state the
Let the functions %(#), u 2 (x), u 3 (x),
and
let
its
be continuous in (a, 6),
...
the series
+u 2 (x) +u 3 (x) +...
u x (x)
be uniformly convergent in (a, b)
and have f(x) for
its
sum.
Then
pi
I
where
pi
pi
pi
f(x)dx=\ u 1 (x)dx +
u 2 (x)dx +
u 3 (x)dx + ...
a^x^x^b.
Let the arbitrary positive number e be chosen.

Since the series is uniformly convergent, we may put
f(x) = s n (x)+R n (x),
where
the same
n^
R n (x) < a when

o
serving for
values of x in
all
Also f(x) and s n (x)
j-
v,
(a, b).
are continuous
in
(a, b)
and therefore
integrable.
Thus we have
pi
f(x) dx
s n (x)
dx
R n (x) dx,
Jx
Jx
Jx
where
pi
f*h
a^x <x ^b.

J*i f(x)
dx -
pi
s n (x)
dx
J x
Xo
pi
R n (x) dx
^ xo
X x - XQ
b-a
<e, when n-^v.
s n (x)
dx = ^\
JX\f(x) dx ,
^
1
pi
I
Jx
u r (x)
dx.
Jx
I
u r (x) dx <e, when n ^
v.
pi
Thus the series of
integrals
is
convergent and
its
sum is
f(x) dx.
J Xq
Corollary
and
WHOSE TERMS
INFINITE SERIES
158
be continuous in (a, b)
U t (x), n 2 (x), u 3 (x),...
Let
T.
[ok.
u^x) + u 2 (x) + u 3 (x)'+...
the series
converge uniformly tof(x) in
(a, b).
the series of integrals
Then
u 2 (x) dx +
u (x) dx +
f(x)dx in
u 3 (x) dx +
...
when a ^ x
{a, b),
<x ^ b.
Jx
This follows at once from the argument above.
Corollary
and
Let ujp),
II.
converge uniformly tof(x) in
Also
integrable in (a, b).

^
Cx
fxf(x)g(x)dx=^\
X
where
a^x <x^b,
. .
(a, b).
and
g(x) be bounded
let
+ u 3 (x) +
u x (x) + u 2 (x)
the series
...be continuous in (a, b)
u 2 (x), u s {x),
u n (x)g(x)dx,
J *o
the convergence of the series of integrals is
and
uniform in
Let the arbitrary positive number
(a, b).
bound
the upper
of \g(x)\ in (a,
Since the series
û
be chosen, and
let
converges uniformly to f(x) in
n (x)
we may put
where
fix)
the same
values of x in
all
f(x)g(x)
dx=
s n (x)g{x)dx
J*o
where
f(x)g(x)dx-
R n (x)g(x)dx,
J *o
= x <x ^ 6.
X
I
(a, b).
we have
Jx
Thus
n ~"'
'
X
\
= s n (x) + R n (x),
R ^\< M{b - a) when
serving for
Therefore
[* s (x)g(x)dx\
n
= \ R n (x)g(x)dx\
\
And
f(x)g(x)dx-^\ u r (x)g(x)dx\
<
e,
which proves our theorem.
be
b).
when n ^
v,
(a, 6),
70. 1]
159
many
these integrations can be repeated as
It is clear that
times as we wish.
Ex.
To prove that
-2ycosx + y 2 )dx = 0, when |y|<l,*
log(l
-rrlog?/ 2 when
\y\>l.
We know that
--cosx-ycos2x-y eos3x +
y~ cosx
when
(1)
...,
-2y cos x + yl
J2/|<1.
Also the series
converges uniformly for any interval of y within
(1)
1, 1)
(67.1).
Therefore
-2ycosx + y 2
Therefore \ log
But the series

positive number
Thus
* *
-2y
(1
less
^~-yn
when
values of
all
< 1.
\y\
< l.f
\y\
when
x,
(2)
\y\ is
some
1).
-2y cos x + y )dx= 2
cos wa;da;,
when
|?/|
< 1.
#. In
Therefore
when
+ y2 = - f
than unity (67.
y^ dy,
nx
cos
converges uniformly for
(2)
log(l
h\
cos x
=_
dy
a
log(l
-2ycosx + y 2 )dx = 0, when
log
- 2y cos x + y 2 )dx =
|y|
< 1.
'o
But
[log y 2
+ log
( 1
--
cos
a;
Therefore
log
- 2?/ cos x + y 2 dx = it log y2 when

)
^ )~J
***
> 1.
\y\
Jo
Ex.
Prove that,
2.
if
m is a positive integer,
cos ma; log (1 - 2y cos x
according as
<1
or
*lt follows from Ex.
and
4, p.
>1
+ y 2 )dx = - 7r y
y-m
or -ir -
131, that
we may
replace the symbols
<,
>
by
:fr
2: respectively.
flf
a;
not zero or an even multiple of
is
converges
the series on the right-hand of (2)
from Abel's Theorem on the Power Series
It follows
I-
Again,
ir,
when y 1
if
when y = -
is
log 2 ( 1
- cos x)'= - 2
not an odd multiple of
-rr,
when x
72, VII) that
or 2nr.
the series on the right-hand of
Then, as above, we have

x
|log2(l+cosa;)=2( -l)- 1
i
COS
710C
,
when x^ (2r + l)v.
Jit follows from Ex. 5, p. 131 that these results hold also for
= 1.
(2)
converges
WHOSE TERMS
INFINITE SERIES
100
Th following extension
70. 2.
sometimes
of the
[ch.
theorem in the preceding section
is
useful.
Let u^x), u 2 (x)
and
be functions integrable in (a, b)
...
let
'
n
s n (x)=^
f(x)
u r (x) and
x.
= û r (x).
1
Thus if(i)f(x)
f(x) in
(a, c),
bounded in
in (a, b), and (ii) the series converges uniformly to

any number between a and b, and (iii) s n (x) is uniformly
is integrable
where c
is
(a, b),
f(x)dx = ^\
J
Since the
sum s n (x)
there
(a, b),
a positive number
is
{x)\<K
and every positive integer
for every x in (a,b)
n.
be chosen, and
let c
be taken so that
IEb
Then
f(z)dx-\
Ja
Ja
b
And
u n (x)dx.
such that
\sn
C<
is
K (independent of x and n),
b -
\\
=[
s n (x)dx
Jc
s n (x)dx\^\
But
Therefore
f(x)
-sn (x)\dx +
\f(x)
Ja
sn (x)
sn (x)dx.
(1)
\sn (x)\dx.
(2)
Jc
b
Ja
&
-sn (x))dx + \ f(x)dx-
(f(x)
f(x)dx-\
\)a
Ja
\f(x)\dx
\
+\
Jc
Jc
< K.
lim sn (x)
^ K\
But the series converges uniformly in (a, c).

Thus there is a positive integer v, such that
(
\f(x)-s n (x)
It follows
from
(2),
all
< 2{b _ a
values of x in
when n
v,
(a, c).
that
|jy^)^-|^M^)^|<2(5 !
- a )+ 2
)^
when n ^
v.
Thus, under the conditions stated in the theorem,

6
/()<fo =
a'
This
may
be extended as follows:
Let the integrable function f(x) be the
u x {x), u 2 (x),
finite
f&
S Ja ur (x)dx.
...
and
let
sum
this series converge
number of sub -intervals,
the
sum
of the series of integrable functions

uniformly in a^x^b, except for a
of whose lengths can be made
less
given number.
Also
let
sn (x)
= 2 u r (x)
be uniformly bounded in (a,
b).
than any
70.2,71]
Then
71.
f(x)dx = l\ u r {x)dxj
Term by Term
sufficient Condition for
If the series
converges in
(a,
and each of
b), and if the
v>i(x)
has a differential
Differentiation.
+u 2 (x) +u 3 (x) + ...
Ul (x)
(a, b)
continuous in
161
its
terms has a differential
coefficient,
series of differential coefficients
+u 2'(x) +u 3 '(x)+...
thenf(x), the
(a, b),
coefficient at every
sum
point of
of the original series,
(a, b),
and
= u 1 '(x)+u 2 '{x)+u 3 '(x)+....

<p(x)-u 1 '(x) +u 2 '(x) +u 9 '(x)+...
f'(x)
Let
Since this series of continuous functions converges uniformly

in (a, b), we can integrate it term by term.
Thus we have
Xl
p(x)dx=\u 1
(x)dx
'
u 2 (x)dx +
,
u 3 (x)dx
+ ...
a^x <x x ^b.
where
Therefore
<f>(x)
dx = [u^xj - u x (xQ )]
f(x1 )=u 1 (x 1 )
But
and
f(x
Therefore
= u x (x
+ [u^xj - u 2 (x
+u 2 (x 1 +u 3 (x 1
+ u 2 {x + u 3 (x
)
<j>(x)dx=f(x 1 ) -f(x
)]
. .
+...
+.,...
).
Jx
Now
put
=x
x = x+Ax.
and
Then, by the First Theorem of Mean Value,
<p()Ax=f(x+Ax)-f(x),
where x
^ ^ x + Ax.
Therefore
(f> (
>
In this paper by the author on

Gazette, 13 (1927), 438.
Integration of Infinite Series" further information on this subject
and a proof is given of the theorem due to Arzela (1885) that when
will be found
function
the series of integrable functions u (x), u 2 (x) r ... converges to the integrable
*Cf.
"
The Mathematical
Term by Term
;
f(x),
and the sum 2u r {x)
is
b
[
)a
f(x)dx = Z
i
-'a
(a, b),
u r (x)dx
then
WHOSE TERMS
INFINITE SERIES
162
But
lim
</>()
[ch.
= </>(x),
A.i-->0
since
(p(x)_ is
continuous in
(a, b).
Therefore f(x) has a differential coefficient f'(x) in
and
f'(x)
(a, b),
= (/)(x)
= u 1 '(x)
-\-u^(x)
+ w3
(^) +
must be remembered that the conditions for continuity, and

term by term differentiation and integration, which we have
It
for
obtained are only
They
sufficient conditions.
are not necessary
We
have imposed more restrictions on the functions

than are required. But no other conditions of equal simplicity
have yet been found, and for that reason these theorems are of
conditions.
importance.
It should also be noted that in these sections we have again
been dealing with repeated limits (cf. 64), and we have found
that in certain cases the order in which the limits are taken may
be reversed without altering the
result.
In term by term integration, we have been led to the equality,

in certain cases, of
and
lim s n (x)dx
Similarly in term
by term
lim
s n (x)dx.
differentiation
we have found
that,
in certain cases,
Umhhn( S ( X+h)
- S W
)}
fanTlim (
and
'<*+> -'<*)
,)
are equal.
The Power
72.
The
Series.
properties of the
are so important,
and
it offers
Power
Series
+ a 1 x + a 2 x 2 + ...
so simple an illustration of the results
just obtained, that a separate discussion of this series will

I.
is
If the
convergent for x
\x\<\x Q
But
= xQ
it
is
+ a x x + a 2x2 +
is
. .
absolutely convergent for every value of
convergent for x = x
|
an x n
there
is
x such
a positive number
< M, when n ^ 0.
K* = KVM
n
we have
given.
\.
Since the series
that
series
now be
that
M such
71, 72]
Therefore
= c < 1,
if
163
KI + \a
1 x\
\a 2 x 2
+ ...
are less than the corresponding terms of the convergent series
M{l+c + c 2 + ...},
and our theorem
series does not converge for
If the
II.
follows.
of x such that \x\> \x \.

This follows from (I), since
\x\> |#
|,
it
(i)
The
from
(I)
and
The
does not converge for
x,
any value
such that
that only the following three cases can occur:
x=0 and
no other value
of x.
converges for
series
it
+ \\x+2\x 2 + Z\x* + ...,

l+x + 2 2 x2 + 3 3 x3 + ....
E.g.
(ii)
(II)
converges for
series
the series converges for a value of
if
must converge for-#=a;
III. It follows
x=x Q
values of x.
all
x2
l+x + ~ + ....
E.g.
(iii)
There
some
is
number p such that, when \x\<p, the

when x > p, the series does not converge.
positive
converges, and,
x-%x2 + lx*-....
E.g.
The
Also
- p<x< p
interval
it is
called the interval of convergence of the series.
is
convenient to say that, in the
the second,
It will
infinite.
following three series
is
first case,
is
zero, and, in
1, 1):
+J+
should be noticed that the
X2
2+-.
X2
it
the interval
be seen that the interval of convergence of the
l+x + x 2 +...
But
series
first of
these does not converge at the ends
of the interval; the second converges at one of the ends;
and the third con-
verges at both.
In the Power Series there cannot be first an interval of convergence, then

an interval where the convergence fails, and then a return to convergence.
Also the interval of convergence
We shall
but
it
denote
may do
its
so;
symmetrical with regard to the origin.

series need not converge at L' or L,
must converge within L'L.
is
ends by L\ L.
and
it
The
IV. If the series converges for a value

l 2
l<hl>
2
|
ofxÔ,
then the sequence
KIV-. KK-..
>
bounded above: and if lim |an

>x>
is
--I
1
l
<x <-.
H-
n
|
=Ac>0,
the interval
of convergence
If lim
n->oo
an
\n
=0,
the series converges for all values
of x.
is
WHOSE TERMS
INFINITE SERIES
164
[ch.
0.
value of x
are given that the series converges for some
take greater than
may
we
number
, which
positive
a
is
there
(I),
in
as
Then,
n <M, for all values of n.
unity, such that \a nx
We
Thus
\a
1
|<-^
"'
i,
for all values of n,
and the given sequence
is
bounded
above and below.

l
By
17. 2,
lim \a n n exists.
\
??><
l
Now
lim
let
\<*n\
= P> -
n-*oo
definite P oint * between 1*1

Take any x for which |x| <l//* and choose a
.
and l//x.
upper limit of mdeterminaThen fi< Ux and, from the properties of the
.
tion, there is
a positive integer
such that
v,
Kl n <*r>
Xq
\anx
Therefore
\.<\,
\a nx
And
when n =
= \anx n \x^\
n
\
?a nxn converges
series
Again, take any
Then a|>
|
Thus
And
\a nx
the
n
|
a;
for
tî, for
\x\
>
series
which
an
when n^v.
X
< _
Thus the
absolutely
|zj
>
infinite
wî.
<1, when
when
\x\ <l//x.
Ijfx.
number
of values of n.
of n.
for an infinite number of values
â nxn cannot converge when x > 1 //*.
1
It follows that wfcen ,x> 0,
*/*e
intertwZ of convergence
of the
series is
-l/jU.<X<l//A.
1.
n =0.
Finally let lim |a n
|
w->oo
Take any value of x other than zero.

Then, by the properties of the upper
is a
limit of indetermination, there
Thus
K*n l<^>
when n
=v
>
all values of x.
this case the series converges for
now show that
we
(III),
of
notation
the
to
Returning
and in
V. The series
is absolutely
convergent in the open interval
-p<x<p.
72]
VI. The series
is absolutely
-p + 8 = x=p-8,
where 8
is
165
and uniformly convergent in the closed

any assigned positive number less than p.
L'
interval
Fig. 14.
To prove (V), we have only to remark that if iVis a point x where - p<x <p,
between
and the nearer boundary of the interval of convergence, there are
values of x for which the series converges, and thus bj(I)it converges absolutely
,
x=x
for
\-
>;
1
1
11
FIG. 15.
To prove
We now
The
(VI), let M',
choose a point
series
M correspond to x = -p + 8 and x=p-8 respectively.

N (say x between M and the nearer boundary L.
)
converges absolutely for x = x
by
(V).
Thus, with the usual notation,
+ an+i xon+1 + <, when n ^ v.

*" +1 + ...
\anx +
+1
above for every point in M 'M including the ends M', M.
I
But
is less
a n xo n
than the
and uniformly convergent
It follows that our series is absolutely
closed interval (M', M).*
And
the
sum
in the
of the series is continuous in this
closed interval.
It
remains to examine the behaviour of the series at the ends of the interval
and we shall now prove Abel's Theorem: f
of convergence,
VII. If the series converges for either of the ends of the interval of convergence,
up
the interval of uniform convergence extends
continuity off(x), the
sum of the
and includes that

up to and includes
to
series, extends
point,
This follows at once from Abel's test for uniform convergence given in
Let the series converge for the end p of the interval of convergence.
Then
We
in Abel's test, take
u n an pn and vn = l-\
(Cf.
Ex.
1, p.
and
the
that point.
67. 2.
150.)
thus establish that the series
a +ajX + a 2 x2 +
converges uniformly in this case in
^= x
p.
...
But we know from (VI) that the
series is
-p + 8^x^0,
when
is
any
positive
number
less
than
p.
When the interval of convergence extends to infinity, the series will be absolutely
convergent for every value of x, but it need not be uniformly convergent in the
infinite interval.
However, it will be uniformly convergent in any interval ( - b, b),
where b is fixed, but may be fixed as large as we please.
E.g. the exponential series converges uniformly in any fixed interval, which may
be arbitrarily great, but not in an infinite interval [cf. 65].
f
Journal fiir Math.,
1 (1826), 311.

INFINITE SERIES
166
'
WHOSE TERMS
[ch.
It follows that the series is
-p + 8^x^p.
And
that f(x), the
sum
continuous in this closed interval.
of the series, is
In particular, when the
converges at x = p,
series
2
f(x)=f(p)=a + a lP + a 2 p +
lim
...
x>p -
In the case of the logarithmic
series,
\og{l+x)=x-^x 2 + lxz -
... ,
< <
1.
x
the interval of convergence is - 1
Further, when x = l, the series converges.
It follows
from Abel's Theorem that
limlog(l+aO = l - + -
s->l
log2 = l-i + i-...
i.e.
Similarly, in the Binomial Series,
(l+x) m = l+mx +
when -
<z<
is
it is
^-|
...
1.
_
And
m(m1) o
'x2
known* that
conditionally convergent
+m+
m(m1)
~-j
+
when - 1<
" "
m < 0, and absolutely convergent when
ra>0.
lim
Hence
(1
+ z) w = l
2w
i.e.
in
both these
On
1)
+m+ m(m^,
+
ra(ra-l)
v
= l+m+-
'
2!
..-,
cases.
the other hand,
if
we put x = l
in the series for (1
The uniformity
which does not converge.
of the
+x)~\ we
get a series
convergence of the series
l-x + x 2 -...
is
for the interval
-l^x^l, where
is
any given positive number less than
1.
VIII. The intervals of convergence of the series
+ a x x + a 2x2 + a z xz + ...
2
a +2a 2 x + 3a i x + ...
and
are the same.'f
131.
*Cf. Chrystal, Algebra, 2 (2nd ed., 1900),
tlf
1
we know that lim
^*
-n+1
an
exists, this result follows
ratio -test for convergence, since in
both
lim |^L+i|<l,
immediately from the
series
when x
|
<
lim
-^-
72]
From
(IV),
be seen that we need only prove that
it will
lim
n><x>
are the same.
since lim n n
167
This
an n
and
lim
na n n
n><x>
a special case of the theorem established in
is
17.
3,
= l.
n><x>
Or we may proceed
We
as follows:
are given that
a
I
converges
+ a l X + a 2 x2 + '
Take x so that
1 + 2 x + 3
x
Xq\Xq
[
Xq
convergent, because the ratio of the nth term to the preceding has for
limit
I
If
xo
which
is less
its
than unity.
we multiply
the different terms of this series by the factors

[
a xx
[,
a2x
|,
which form a bounded sequence (by V),

thus obtain, namely
a 3x
it is
...
\
clear that the series
\a 2 x\+3\a s x 2
1^1+2
is
< x < p.
Then
is
when \x\< p.
which we
+ ...
'\
convergent when \x\<p.
We have yet to show that this last series diverges when x > p.
where \x > p, the same
But if it converges when x=\x
|
\,
would hold
for the series

|
and
a xx
+2
a 2x
+3
a zx*
+ ...
also for the series
\a 1 x
\a 2 x
2
\
\a 3 x
s
\
...
since the terms of the latter are not greater than those of the former.
But this is impossible, since we are given that the interval of convergence of
the original series
is
- p<x<
p.
+ a x x + a 2 x 2 + a 3 x3 + ...
a x + 2a 2 x + 3a 3 x 2 + ...
and the
series
obtained by differentiating the
first
term by term, have the same interval
convergence.
IX. Term by term

Let the power
differentiation
and
integration of the
Power
Series.
series
have - p<x< p for its interval of convergence.

Let its sum hef(x) in this interval.
It follows from (VI) and 70 that
f(x)dxa
*!0
(x
-x
+2
i
(x n+1 - x^*
n-l-l
1
),
when - p < x
< x < p.
of
WHOSE TERMS
INFINITE SERIES
168
Also from (VIII) and
71
we
[oh. v
see that
f'(x)=a 1 + 2a 2 x + 3a 3 x + ...
the open interval -/><*</>, and these integrations
in
point
any
x
is
where
and differentiations niay be repeated any number of times.
2
73. Extensions of Abel's

I.
We
have seen in
Theorem on the Power
72 that
+ a\ + a 2 +
converges, the
Power
...
Series
a + a 1 x + a 2 x2
is
Series.
the series
if
uniformly convergent,
^ x ^k
when
1;
+ ...
and
that,
if
2
f(x) =a + a x x + a 2 x +
f(x)-a + a 1 + a 2 +...
. . .
lim
x->l-0
The above theorem

Let the series
such that
of Abel's
a special case of the following:
is
2 an converge, and a
0â < ai <a
a v a 2 ,...bea sequence of positive numbers
Then
the series
an e-"<
is
uniformly conco
vergenL when
^ 0, and iff(t) = V ane -<*, we /iave

o
lim /(*)
r->+o
This results immediately from Abel's test of 67. 2
(cf.
= 2 no
Ex.
3, p. 150).
00
II.
In Abel's Theorem and
its
extension stated above, the series
2 an
are
dealing
supposed convergent. We proceed to prove Bromwich's Theorem
shall adopt the
with series which need not converge.f In this discussion we
following notation:
=a + a 1 + a 2 + ...+a n
Sn = s + s 1 + s 2 ...+s n
sn
and we write &n

*0> *1 S 2> *
for the Arithmetic
sum
<r n
can be shown
is
s,
of the first
n terms
of the sequence
'
Thus
It
Mean
(cf.
=-
102) that,
if
the series
then, with the above notation, lim
n>-o
<r n
2
=s.
converges and
its
But the converse
does not hold.
*If aQ , Oj,
...
are functions of x
a given interval,
it
and the
series
follows from Abel's Test of
2 a n converges uniformly
to F(x) in
67. 2 that lim 2 on e-~' converges
uniformly to F(x) in this interval.
*Math. Annalen, 65 (1908), 350.

Math. Soc. 25 (1019), 258.
See also a paper by C. N. Moore in Bull. Amer.
-)
72, 73]
Means may converge, while the sum
of Arithmetic
The sequence
169
S a
fails to
converge.*
2 an converge to
series
Means
Let the sequence of Arithmetic
Bromwich's Theorem.
Also
cr.
let
un
be
a function of t with
for the
<r n
the following
proper
when
ties,
> 0:
1n A
(a)
un
< At
fp, q any positive integers; K, a positive\

J'
\ numher independent of p, q and t
lim nun = 0,
((3)
n>ao
(y)lim un =
Then
\.
2 an un converges when > 0,
the series
We
~'
<-++o o
=a
$o =5 o
have
2 anun =z
and lim
# x -2^0 = 5! -s = 1}
2 - 2S 1 +S =s 2 - 5 X -a 2
S 3 -2S 2 + S 1 =s 3 -s 2 = a 3
etc.
Thus
ian wn =ôMo + (>S\-2^
K+
(^ 2
-2^ 1 + ^
)^ 2 + ...+(^n
-2^n_ 1 +
>S'
n_ 2 )wn .
Therefore
anu n = /S A 2 w + /Si A 2 ^ +
+ Sn
. . .
AX + 2Sn un+1 - Sn un+2 - Sn_ un+V
of Arithmetic
But the sequence
cr
and lim
converges,
o~ n
= (r.
Means
o- 3 , ...
o- 2 ,
n>-o
that there
follows
It
#n < (w + 1)C
is
a number G,
not
than
less
|<r|,
such that
for every integer w.
Also from
(/?) it is
clear that
lim (8n u n+1 )=\im
*If
a.n
= -
^ 0,
But
2 an is
see the
(2)
"summable (C,
V a n is not con-
>-oo
Hardy-Landau Theorem,
often said to be
obvious that lim (r n \, and the series
it is
vergent.
series
l) n ,
w nf2 )=lim (^n _ 1 un+1 )
(i8f
)"
1
102, II.
and its sum
(C,
When
1
is
lim
<x
said to be
<r.
the
For
<x.
a discussion of this method of treating series, due to Cesaro, reference may be made
to Whittaker and Watson's Course of Modem Analysis (5th ed., 1928), p. 155.
Knopp, loc. cit., English transl., Ch. XIII.
Hobson, loc. cit., 2 (2nd ed., 1926), Ch. I.
Also see below,
fA 2 u n
//
is
101-103, 108.
written for (u n
-2u n+1 +v n+s ).
Since
all
the terms in
the
series
are positive this condition (a) implies the convergence of this series.
WHOSE TERMS
INFINITE SERIES
170
2 (+ 1) A 2 w n
Further, the series
[ch. v
converges, since, from
(a),
the series
2 w A 2 wn
|
converges,
|^%J<C(n + l)|A\|.
Also
Therefore the series
X Sn i\ 2 u n
from
It follows
and
(1)
(2)
that
iv=Â
=1
Taking the special case
at
n = w + l
we have
3)
= a2 =
and
= 0,
= S(w + l)A
(4)
ax
(5)
Thus, from
(3)
and
(4),
iv-ô=i (-(*+
!)-)
Now
n-=<r.
lim
n
_^
M+
Therefore, to the arbitrary positive

integer
such that
n+
n~=
\S n -{n+l)<r\<-g(n +
Thus
Also
< 4i
7^?, when
- cr
number
Sn <(n+l)C,
\
It follows,
r.
when ra^v.
l),
and
for every positive integer,
from these inequalities and
\ta n u n -
c,
(5),
|<r|=
that
\^\2\s n -(7i+l) T)û n + \y(Sn -(n + l)<r)û n
rn
\.
t<
+ l)\Ah< n
<2CX(n + l)\ 2 un \+^Z(n
^^
(6)
\
i(n + l)|AX|<2Vn|AX[
But
<2#, by
And
lim
A 2 u n = 0,
since
(a)
It follows that, v being fixed, there is a positive
|A 2 "nl<o
Thus from
(6), (7)
(8),
X a n w n - <ru
we
And,
< |e + Je < e, when
lim (â n u n -<ru )=0.
since,
finally,
from
(y),
>/
see that
Therefore
number
such that
i^> when &<*=*/ and n^Lv-\
*
,
and
(7)
lim wn = l, by (y).
lim
X n w n = o-,
lim w
= !*
< t^
>/.
8)
73]
III. Let the sequence of Arithmetic
Means for
the series
â n
171
converge
to <r,
and
let
un
be e~ nt (or
e-"
Then
').
the series
ânun
converges,
lim *Ea n u n = o\
t-^+o o
This follows at once from Bromwich's Theorem above,
the conditions (a), (/S) and (y) of that theorem.
It is obvious that
that
(/?)
and
when
> 0, and
if
e- (or e~ nZt ) satisfy
(y) are satisfied, so it only
remains to establish
(a) is satisfied.
(i)
un = e-"t, t>0.
tfun = u n -2u n+1 + u n+2
Let
Then
= e-t{l-e-tf.
Therefore
A 2wn
positive.
is
n\A 2 un \=n& 2 u n
Also
= nu n+2 -(n+l)u n+1 + Ul

^n\^ 2 u n =e~t, and
Therefore
the condition
(ii)
Let u n = e-n 2
In this case
where < Q <
(a) is satisfied,
>0.
A 2 u n = e-( n + Wz(4:(n + 0) 2
-2t),
2.*
Therefore the sign of

It follows that
it is
/\ 2
u n depends upon that of (4(n + d) 2 t 2 -21).
positive or negative according as
2(n + 0)t-V(2t)<0.
Also
(ft
and
+ 0)>
n + Q)
(
<
V(2ty
V(2tf
w hen n>
"
when
v(2ty
n + 2<
^
V(2)'
Therefore b?u n cannot change sign more than three times for any positive
value of t.
But
it
follows at once from the equation
nA 2 un = ne-(n+e)*t(4 n + Q)2t 2 -2t)

:
that a positive number, independent of
can be assigned such that n\il 2 u n

than this number for all values of n.
Hence
can be chosen so that the condition (a) is satisfied, provided that
the sum of any sequence of terms, all of the same sign, that we can choose
from 2nA 2 u n is less in absolute value than some fixed positive number for
t,
is less
all
values of
t.
Let y,ni\ 2 nn be the sum of such a set of consecutive terms.
*This follows from the fact that
/j )-jy(-+*)+/()
where
(Cf.
O<0<2,
Goursat,
provided that f(x),
loc. cit., 1 (4 e ed.,
=r( , +>A)
f'{x), f"{x) are
1923), 21.)
continuous from x to x + 2h.
WHOSE TERMS
INFINITE SERIES
172
[ch. v
Then we have
in&Hn = re-^-(r-l)e-(>+W-(s + l)e-(*+W + se-(*+V
2t
,
which
differs
from
r ( e -rH-e-(r-n)H)-( s
by
at
+ l)(e-(s+W-e-(s+Wt)
most unity.
when n is a positive integer and t > 0,
<n (e-n 2 t-e-(+W) = 2n(n + d)te-("+Wt...(0<6<l)
But,
<2(n + 0)He-(n+W*
<2e~\
Therefore, for the set of terms considered,
2\nA 2 u n <4e- l + l.
\
Then the argument above shows that the condition
(a) is satisfied.
00
IV. Let the sequence of Arithmetic Means for the series
Then
n
the series ^,anx will converge
<x < 1,
when
2 n
converge to
<r.
and
o
00
lim
x->\
This follows from the
first
-0
a nz n = <r.
part of (III) on putting x = e~t.
V. Let the terms of the series
be functions of x,
and
the sequence
of
Arithmetic
Means for
ina^x^b.
tr(x)
v
this series
Then lim
t-^+o
'
provided that
to
the
bounded function
I.anu n converges uniformly to <t(x)
m this %n-
a function of
of BromwicKs Theorem, when t>0.
terval,
(y)
un
converge uniformly
is
satisfying the conditions (a), (0)
and
argument of (II).
This follows at once by making slight changes in the
the solution
The theorems proved in this section will be found useful in
equation, which
differential
the
when
Mathematics,
Applied
of problems in
The solution has to satisfy
corresponds to the problem, is solved by series.
really need is that, as
we
What
initial and boundary conditions.
certain
shall
the boundary, or as the time tends to zero, our solution
the boundaries, or at
have the given value as its limit. What happens upon
the instant * = 0, is not discussed. (See below 123.)
we approach
74. Integration
of
Series.
Infinite
Integrals.
Finite
Interval.
finite integrals.
In
the
We shall now
ordinary
70 we dealt only with
when the integrand
examine the question of term by term integration, both
integration, supposed finite,
has points of infinite discontinuity in the interval of
but the interval of
and when the integrand is bounded in any finite interval,
shall deal with the
we
section
this
In
infinity.
to
extends
itself
integration
discussion of
first
of these forms,
when
and
it
will
be sufficient to confine the discussion to the case

(a, b), say x=b.
the infinity occurs at one end of the interval
73, 74]
u 2 (x),
Let u x (x),
I.
uniformly tof(x) in
Also
g(x) have
let
and
...be continuous in (a, b)
173
the series *2u n (x) converge
(a, b).
an
Cb
x = b and
infinite discontinuity at
g(x)dx be absolutely
convergent*
Cb
Then
J
Cb
oo
f(x)g(x)dx = ^\
un (x)g{x)dx.
.'a
i
00
From
the uniform convergence of 'Eu n (x) in
we know
(a, b),
its
sum
( 61,
VI).
that
f(x)
is
continuous in
and thus bounded and
(a, b),
integrable.
[b
(b
Also
f(x)g(x)dx
is
g(x)dx
absolutely convergent, since

"
so
is
Ja
Ja'
[b
\g(x)\dx=A.
Let
a
Then, having chosen the positive number e, as small as
we please, we may put
f(x)=sn (x) + Rn (x),

where \Rn {x)\
< -?, when n ^
[b
But
Cb
Cb
It follows that
J
Cb
J
Thus we
all
values of x in
(a, b).
f(x)g(x)dx and
J
r,
both exist
sn (x) g( x )dx
Rn (x)g(x)dx
f{x)g{x)dx =
and that
also exists,
Cb
Cb
sn (x)g{x)dx
Rn (x)g(x)dx.
see that
n
Cb
u r (x)g(x)dx
f(x)g{x)dx-?\
a
lift
= \\ Rn (x)g(x)dx\
|
Ja
.'a
\9( x )\ dx
<j\ a
<, when
which proves that the
series
u n( x ) a x dx
2 )a
is
n^ v
convergent and that
its
sum
cb
f(x)g{x)dx.
is
J
Ex.
This case
1
is
illustrated
by
oo
j"l
loga;log(l+a;)efo:=2(-l) n ~ 1
i
-s>
= S-^
\\n
ri
t^>
2
fw(w+l)
'
'f^~
xn
logxdx
]
-
^Ln
xn
since
\osxdx=
s
n+ \
(n
1
,-..,
(n
1)
+ l) 2 J
2 - 2 log 2 - ys 71 2 using the series for To 71 2

"
"
'"!*
* It is clear that this proof also applies
discontinuities in (a, 6)
f Cf.
and
g(x)dx
is
when
g(x) has
finite
absolutely convergent.
Carslaw, Plane Trigonometry (2nd ed., 1915), 279.
number
of infinite
WHOSE TERMS
INFINITE SERIES
174
Here the
series for log (1
+ x) converges uniformly
[ch.
^ x ^=
in
1,
and
log x dx
Jo
converges absolutely (as a matter of fact log x

< x 1), while log x ->oo as x-+0.
On
cases
always of the same sign in
we may
the other hand,
still
apply term by term integration in certain
conditions are not satisfied, as will be seen from the
when the above
following theorems
is
00
II.
u x (x), u 2 (x),
Let
and
be continuous
...
converge uniformly tof(x) in the arbitrary interval
Further,
let
and
positive
g(x) be positive,
n,
where a
(a, a),
the series
^Zu n (x)
< a < b.
(a, a).
[b
f(x)g(x)dx = 2,\ u n {x)g(x)dx,
Then
)a
provided that either the integral \f{x)g{x)dx or the series
^ûn (x)g(x)dx
converges.
"b
f(x)g(x)dx converges.
Let us suppose that

J
In other words, we are given that the repeated limit

b
lim
~*
[
lim
2 %(#)] g(x)dx exists.
Since the functions u x {x), u 2 (x),
from the convergence
...
are
all positive,
as well as g(x), in
(a, a),
of \'f(x)g(x)dx there follows at once the convergence

Ja
of
b
\
J
Again, let
&
Then
(
J
f(x)
Rn (x)g{x)dx
b
"\
u r (x)g(x)dx
=u x {x) + u z (x) +
also converges,
...
2, ...).
+u n (x) + Rn {x).
and
f(x)g{x)dx-t\ ur (x)g(x)dx =
Ja
Ja
But from the convergence

positive
= l,
(r
Rn (x)g(x)dx
(1)
f(x)g(x)dx
of
it
follows that,
when
the arbitrary
number e has been chosen,
as small as
we please,
there will be a positive
number such that
0<[Jbij f(x)g(x)dx<e.
0< _ Bn (x)g{x)dx<ie,
Afortiori,
and
this holds for all positive integers n.
Let the upper bound of
The
series
S(^
g{x) in (a, b
J)
be
M.
(a, b
-f).
()
74]
Keeping the number

v
we have chosen above,
such that
0<Ji-W <
all
2l(M'
values of x in
< -*Rn {x)g{x)dx < 2M
Thus
"
whe
{a, b
).
j*~
results (2)
and
UX
<|e, when w
(3),
0< Un (a;)gr(a;)tfa;<e,
Then, from
(3)
we have
[b
V*)**
< 2(6oTi
a) Ja
x
Combining these
175
when n^v.
(1),
0<
n rb
[b
f(x)g(x)dx-^Z\ u r (x)g(x)dx<e, when ftî'.

i
Ja
f/>
Jtt
Cb
f(x)g(x)dx =
Therefore
J (t
y.
1
u r (x)g(x)dx.
The other alternative, stated in the enunciation, may be treated in the

same way.
Bromwich has pointed out that in this case where the terms are all positive,
as well as the multiplier g(x), the argument is substantially the same as that
employed in dealing with the convergence of a Double Series of positive terms,
and the same remark applies to the corresponding theorem in 75.*
Ex.
Show
1.
og %
that
o I
dx
= f (* xn log xdx\
o
- " 7T6"*
Show
Ex.2.
that
(;
og
;^f = 2|^J-I? = ^.
The condition imposed upon the terms u x {x), u 2 {x) ... and g(x), that they
may be removed, and the following more general theorem stated
i
are positive,
00
III. Let
u x {x), u 2 (x),
...
be continuous
and
the series
un x
(
converge uniI
00
formly in
* Cf.
Bromwich,
(1906),
f
the
the arbitrary interval (a, a),
where
a<a<b.
Infinite Series (2nd ed., 1926), 496,
Also
let 2)
and Messenger
un ( x f) x )>
)
of Math., 36
1.
The
first
interval (0, 1) has to be broken
we
72, VII.
use
Theorem
and
up
into
in the second
two parts, (0, a) and (a, 1). In

Theorem II. Or we may apply
WHOSE TERMS
INFINITE SERIES
70
Further,
g(x) be bounded
let
and
*>
that
either
la
integral
the
la
00
[b
f(x)g(x)dx^^Z\
la
provided
integrable in (a, a).
[b
Then
[ch.
un (x)g(x)dx,
^Jx)
g(x)
1
dx or
series
the
[b
2\
un( x )
la
9( x )
d% converges.
I
This can be deduced from
using the identity
(II),
u n g = {u n +\u n \}{g+\g\}-{u n +\u n \}\g\ -
Ex.
\un \{g+ \g\}+ \un
\g\,
theorem can be applied to each term on the right-hand
since that
Show
1.
i^^-dx = 2( - lA*
that
lo
L+%
side.
xn logxdx
Jo
+ l) 2
(n
12'
Ex.
Show
2.
^ iog,&=i(-i)-i;^-Mogxrfx=!;i^,
T x
>0.*
(;
when p +
that
Infinite Integrals.
75. Integration of Series.
For the second form

the theorems proved in
u x (x), u 2 (x),
Let
I.
2 un (x)
of infinite integral
Interval Infinite.
we have
results corresponding to
74.
...be continuous
and bounded in x
converge uniformly to f(x) in x==:a.
Further,
= a, and
let
g(x) be
let
the series
bounded and
integrable in the arbitrary interval (a, a), where
a<a, and
may
be chosen as
x
large as
we please
and
g(x)dx converge absolutely.
let
la
coj-oo
(*
f(x)g{x)dx = 2\
r
Then
la
The proof
of this
-a
un (x)g(x)dx.
theorem follows exactly the same
lines as (I) of 74.
'
e xdx
r^<fa=ir
x
"tli (x + n-l)(x + n)'
ex.
This follows from the fact that the series

1
x(x+l)
+ (x+l)(x+2) + '"
x^
1.
converges uniformly to 1/a? in
But it is often necessary to justify term by term integration
*If
If
p>0, Theorem
0>p>
In the
72,
first
VII.
either
III can be used at once.

the interval has to be broken up into two parts (0, a) and (a, 1).
use Theorem I and in the second Theorem III. Or we may apply
1,
we
when
74, 75]
177
,,;,
(00
\g{x)\dx
û n (x)
divergent or
is
~a
can only be shown to converge uniformly
where a can be taken as large as we
in the arbitrary interval (a, a),
Many important
spond to
II.
(II)
and
74
(III) of
u x (x), u 2 (x),
Let
...be continuous
and
positive
and
we
Also
let
X un
the series
a may
verge uniformly to f(x) in the arbitrary interval (a, a), where

large as
please.
cases are included in the following theorems, which corre-
x ) con ~
be taken as
please.
g(x) be positive,
f(x)g(x)dx = 'Z\
Then
(a, a).
un (x)g(x)dx,
I fi
(00
f(x)g(x)dx or the series
.a'
2
1
w n (z)9(%)dx
.'
converges.
Let us suppose that
f(x)g(x)dx converges.
In other words we are given that the repeated limit

n
{a
[lim
a n>co
ûr(x)]g(x)dx
exists,
00
Since the terms of the series y,u r(x) are
all positive, as
well as g(x), in
x>a
from the convergence
of
f(x)g(x)dx there follows at once that of
.'a'
u r (x)g(x)dx
(r=l,
2, ...).
Again
Then
let
f(x)
Rn (x)g{x)dx
u x (x) + u % (x) +
also converges,
oo
positive
u r (x)g(x)dx =
Rn (x)g(x)dx
(1)
.'a
.'a
f(x)g(x)dx,
number has been chosen,

number a such that
0<(
it
follows that,
as small as
we
when the arbitrary
please, there will be a
f(x)g(x)dx<%e.
<
A fortiori,
and
roo
But from the convergence of

positive
+ u n (x) + R n (x).
poo
f(x)g(x)dx-'2\
a
Rn (x)g(x)dx < e,
and this holds for all positive integers n.

With this choice of a, let the upper bound
of g(x) in (a, a) be
M.
00
The
series
ûn (x)
(a, a).
Keeping the number
we have chosen above,
v such that
0<R n (x)<x1-r
all
values of x in
-,
when n^v,
(a, a).
WHOSE TERMS
INFINITE SERIES
178
0<
Thus
Rn {x)g{x)dx<y, whenw^v.
Jo
0<
But
[ch.
Rn (x)g(x)dx<\e,
.'a
and this holds for all values of n.

Combining these two results, we have
Rn (x)g(x)dx<e, when n^v,
0<
J
and from
(1),
<
Therefore
J
The other
ur (x)g(x)dx<e, when n~v.
f{x)g{x)dx-^\
f(x)g(x)dx = Z\
i
u r (x)g{x)dx.
Ja
alternative can be treated in the
same way.
Mn r
~
ax in
L e ax cosh bx dx =^2n\\ e- x dx
oo
Ex.
-a 2 *2 cosh bxdx = Z
2^1
Further, the condition imposed

shall
be positive,
< \b < a.
if
- a2x2 x2n dx.
e
J
upon u x (x), u 2 {x),
and
...
be removed, leading to the theorem
may
g{x), that
they
00
III. Let
formly in
u x {x), u 2 (x), ...be continuous and
the arbitrary interval (a, a),
where a
21^(^)1 converge uni-
the series
may
be taken as large as
we
please.
oo
Also
let
2,u n (x)=f(x).
i
Further,
let
g{x) be

a
f(x)g{x)dx = t\
I
provided that either the integral j^
2 .a
|
(a, a).
u n (x)g(x)dx,
Jet,
\u n (x)
\g(x)\dx or the series

\
\un {x)\\g{x)\dx converges.
This
is
deduced, as before, from the identity
u ng = {a n +\u n \}{g+\g\}-{un +
since the
Ex.
1.
Ex.
2.
Theorem
\un \}\g\
II can be applied to each
\u n \{g+ \g\}
2 2
e~ * cos bx dx =
that
]
76. Certain cases to
which the theorems
\un
âx^ n dx,
Show that jj e-cos6sdk = - ~2î
Show
\g\,
term in the right-hand

if
side.
0<|6| <a.
e~^ x^dx.
2
2n[
of 75
do not apply are covered by
the following test

Let u^x),
u 2 {x),
...be continuous in
xâ, and
uniformly tof(x) in the arbitrary interval

as we please.
(a, a),
let the
series
where a
may
2un (x)
converge
be taken as large
75, 76]
Further,
let
the integrals
u 1 (x)dx,
Ja
converge,
and
the series
Ja
[X
[x
u 1 (x)dx +
and
u 1 (x)dx +
a
,00
f(x)dx =
Also
u^dx +
}a
jet
sn (x)
u^dx+l u
u 2 (x)dx+....
(x)dx +
X
...
Ja
Then we know that lim
we know
p30
Ja
Ja
Also
u 2 (x)dx+...
Ja
f(x) dx converges.
the integral
/--O
Let
u 2 (x)dx+...
Ja
of integrals
Ja
converges,
etc.,
xâ.
the series
u 2 (x)dx,
of integrals
Ja
Then
179
un (x)dx.
)a
sn (x) exists in
that lim sn (x) exists, and
^ a, and we denote
we denote
it
by
it
by F(x).
G(n).
->oo
We shall now show that

From
lim F(x) and lim G(n) both exist, and that the two
n ~^ <x
x~>x
limits are equal.
this result our
theorem as to term by term integration
will follow
at once.
I.
To prove lim F(x)
exists.
X-a>
Since lim sn (x) converges uniformly to F(x) in x
^ a, with the usual notation,

n^
n>oo
we have
F {x) - sn (x) <
when
v,
the same v serving for every x greater than or equal to
Choose some value of n in this range.
|
i e,
Then we have
lim sn (x)
a.
G{n).
x-><*>'
Therefore, we can choose

\
X so that
sn( x")- sn( x ')\<h>
But
wnen x">x' = X>a.
F{x")
F(x")
- sn (x") +
F{x')
|
<, when
Thus F(x) has a

II.
sn {x)
- sn (x') +
|
x">x'^X>a.
limit as x-><x>
To prove lim
G(n) exists.
n*x>
Since lim sn (x) converges uniformly to F(x)

I
sn"{x)
- sn '(x)
< y, when
inxâ,
n"
>n'^ v,
the same v serving for every x greater than or equal to
a.
sn {x')
- F{x')
|
INFINITE SERIES
180
WHOSE TERMS
[oh.
Km 8n(x)=G{n).
But
x>oo
Therefore
X X
we can choose
|
lt
(?(')
such that
Sn'(x)
when x ^ X x > a,
<K
^ X > a.
when x
Gin") - sn'{x) <
X2
X
than
less
x
not
of
value
x or
Then, taking a
|
\G(n")-G(n')\
Gin")
- sn'(x) +
|
sn'(x)
- 8n'(x) +
|
sn '{x)
- G{ri)
<fc+h+h
<
when n">n'^
c,
v.
Therefore lim G(n) exists.

prove lim F(x) = lim G(ra).
n-oo
xkjo
Since lim G(n) exists, we can choose v x so that
III.
To
n*<o
2
Again
2 Jo ur{x)dx
< e,
,()*
n^ v r
when
converges uniformly to F{x)
inxâ.
Therefore
v 2 so that
we can choose
< K when n^ v 2
r(a;)cfa;
n+i
equal to a.
the same v 2 serving for every x greater than or
or
v
Choose v not less than v x
2
.
From the convergence

I
l-'x
of the integral
2 ur{x)dx < K
i
^ can
2 !<*)**
choose
when x ^ X > a.
Vk\\^)dx-F(x)
But
CO
If* V
Vx
fz
<, when
uT {x)dx
v+l
00
fx
00
x-+ao
,oo
=2
1
ur (x)dx= lim
J
G(n).
n_*x>
are given that the series
u 1 (x)+u 2 (x) +
...
interval
converges uniformly to f(x) in any arbitrary
Therefore
ur {x)dx
v+l3 *
x^X>a.
limjF(z)
Therefore
we
oo
ur{x)dx
2M*)^+2+ l<* ur(x)dx-2\
v+l a
^\\j:u r(x)dx
IV. But
J"
(a, a).
we have
X
\
Jr.
f{x)dxJ\ u l (x)dx + \* u 2 {x)dx+... inxâ.

)
X so that
76]
181
Thus, with the above notation,
= \* f{x)dx.
F{x)
Jo
poo
It follows
f(x)dx converges, and from III that
from I that
.'a
f{x)dx =
.a
roo
fao
I-oo
u 1 (x)dx+\ u 2 (x)dx+....
)a
REFERENCES.
Bromwich, loc. cit. (2nd
De la Vallee Poussin,
ed., 1926),
loc. cit. 1
Ch. VII- VIII and App. III.
(3e ed., 1923), Ch.
XI.
Dini, Lezioni di Analisi Infinitesimale, 2 (Pisa, 1909),
Goursat, loc. cit, 1 (4e ed., 1923), Ch. II.

Hobson, Theory of Functions of a Real Variable,
la
Parte, Cap. VIII.
(3rd ed., 1927), Ch. VI;
2 (2nd ed., 1926), Ch. Ill, V.
Knopp, loc. cit., English translation, Ch. V, XI.

Kowalewskt, loc. cit., Kap. VII, XV.
Osgood, loc. cit., 1 (4 Aufl., 1923), Kap. III.
Pierpont,
Stolz,
loc. cit.,
loc. cit.,
Bd.
2 (1912), Ch. V.
I,
Absch. X.
And
Brunel, loc. cit, Enc. d. math. Wiss., Bd. II, Tl. I (Leipzig,
Montel-Rosenthal, loc. cit, Enc. d. math. Wiss., Bd. II, Tl.
1899).
Ill, 2 (Leipzig,
1923).
Pringsheim, "Grundlagen der allgemeinen Funktionenlehre," Enc.
d.
math.
Wiss., Bd. II, Tl. I (Leipzig, 1899).
EXAMPLES ON CHAPTER
V.
UNIFORM CONVERGENCE.
.1.2
sin -
sin
1.
#(*) =
Let
Y~
~~T
sin2 ~ + x cos 2 -
Show
that
lim
n>oo
and
lim 0(#)=2,
Also show that
any
(x)
(f) n
positive number.
2. If (f)(x) is
(x)=0, when #<;0,

<f> n
when x=0.
n>oo
converges uniformly to zero, when x
where x
<fr(x)
is
exists
n_>a0
Also show that x n <f>(x) converges uniformly to

it
0>
~ x ^ 1, show that lim x n
in that interval.
that
^x
i=s
its limit
1,
in
only
2]
^x <
if </>(!)
=0.
1,
and
3.
WHOSE TERMS
INFINITE SERIES
182
Examine the convergence
of the series
*
l)x + ll(n+2)x + iy
+
(nx+l)l(n
f
and by
its
means
continuity of a
on page
[ch.
xÔ,
non-uniform convergence upon the

x represented by an infinite series. (Cf. Example
illustrate the effect of
function of
144.)
2 cosech2 nx is uniformly convergent in any interval
Show that the series

% = xQ where x >0.
4.
5.
Prove that the
series
+ (l+z2 + (l+z2
is
convergent for
all
values of x, but
is
2
)
not uniformly convergent in an interval
including the origin.

6.
Prove that the
series
*+*&
(i)

7. If
k w (s) = ,
interval x
^x
all
(ii)
? i+^2
values of x.
v show that
2 **()
is
uniformly convergent in any
> 0.
where x
00
Also show that,

the convergence
8.
Find
for
is
if
'
/ 1 \
m is any positive integer, 2 u n
-J
>h
not uniform in any interval including the
what values
of
x the
series
and deduce that
origin.
2 un converges where
i
xn _ --!
(Xn
+ X~ n (X^ + X-n- 1
1
~{x-\){xn + x-n
Find also whether the
(i)
(ii)
9.
(x-l)(xn+ 1 + x-i n + 1 ))
series is
uniformly convergent through an interval including

continuous when x passes through the value + 1.
+1
of the
Discuss the uniformity or non-uniformity of the convergence
series
whose general term
is
i-(i+s) w
Un ~l+{l+x)n
10.
+ !
Let
be an absolutely convergent
l-q+s)"-1
~1
i+{i+-x)n
'
+ ...
series of constant terms,
and
be a set of functions each continuous in the interval

comprised between certain fixed limits,
A^frW^B,
where A,
B are constants.
r=0,
1, ...,
let
a^x^fi, and
each
v]
Show
that the series

ao/o(*) +<*i/i (*)
represents a continuous function of x in the interval a =5 x =S
Show
11.
183
(3.
that the function denned by the series
^_
v
2
f n(l+nx
is finite
and continuous
uniformly convergent for
Show
12.
is
that
a^x^b,
and
y, v
series is
+ u 1 (x) + ...
(x)
each continuous and having no roots in the interval
if
u n (x)
where
Examine whether the
such values.
ail
if
series of functions
for all values of x.
do not depend on
x,
r
I
then the given
series is
uniformly convergent
in this interval.
Apply
this test to the series
l+xa + x(x-l) + x(x-l)(x-2) + ...,

'
0<a<
where
1.
13. Using the inequality
show that
s n (x)
if
Also show that the series
k> 1
and that
^ ^
if,
A;
>-, when 0<x<tt,
= I **
1, it
then
,n
(|_)
>h
^ ^
a;
71-,
when
converges, but not uniformly, in this inter-
val of x.
oo
14.
Let the
a<x<b,
,
and
open
converge uniformly to j(x) in the open interval
for every positive integer let lim

*-*>
00
of the
û n {x)
series
Then the series
interval.
u n (x) = l n where
,
c is
a point
00
n converges,
and lim f(x) = 2 l w

x>c
00
15. Let the series
a<x<
b,
and
2,u n (x) converge uniformly to f(x) in the open interval
for every positive integer let lim
u n (x) =a n and lim u n (x) = b n
x-*a+0
00
Then the
series
x>b-0
00
2> n
converges, and lim /(z)
xâ+0
= 2a,
with a similar result
lim f(x).
x-+b-0
for
00
16. Let the series 1,u n (x) converge uniformly to f(x) in the infinite interval
xâ,
and, for every positive integer,

00
and lim f(x) = 2 n

a;>oo
let lim
X->oo
u n (x) = Z n
Then
2
1
n converges,
WHOSE TERMS
INFINITE SERIES
184
THE POWER
SERIES.
x
,
From
17.
the equation
sin
x=
[oh.
dx
wn- X
'
2
)
1
that the series ior sin- * is
show
Prove that the expansion

18.
From
Within what range
Show
_iL
x=
dx
[
\
+x2
does this hold
a;
X
\
Jo
1. 3...2n-i
^dx
and
Jo
|
Show
in the
Han" 1 a;
'
< 1.
,
|*|<1, prove that

CX
"
tan liC
and tan_1 a:
Xt
^dx
*
and, from the integration by parts of

series also represent
20. If
1
substitute the series for sin- *
and integrate term by term, when

Also show that
-....
integrals
isin- 1
tan_1 a;=a;-a;3 + a;5
of
x
^
we may
that
when x =1.*
also valid
tan
the equation
obtain Gregory's Series,
19.
is
X*
x2
and
g,_ 1)W
^^, prove that these
x%
^ = IT2""3T4 + 5T6"--
that the result also holds for x =
1,
and deduce that
l-J-l + l+J-...=043882....
21. If |*|
< If
prove that
2
log(l
Does the
result hold for
ic
**
ic
+x)dx = j-2 -273 + 3T4" ""
x- 1
22. Prove that
]]og<i-*)=-ij o<-<i.
a:
/l+g\<fa
When 3 = 1,
Goursat,
loc.
lim
W n+1
-
= l,
*2n-1
0<z<l
but Raabe's test
ciC p.404) shows that the
series
(cf.
Bromwich,
Zoc.
ci*.,
p. 35, or
converges for this value of *.

v]
Express the integrals
[>(,-*) (>g(1+
185
j>g (}-f)<|:
and
as infinite series.
Show
23.
that
- 1 - f + Jf +
^ - TV - ,\ +
...
=ilog(2 + V3).
24. Prove that,
^ ^ 1,
when
#|log
re
^ e_1
a;|
Hence show
(i)
a2
l+a(xlogx) + (xlogx) 2 + ...
that
=1
a^*^ = 2(-l) n
(ii)that
Jo
25. Prove that,
when a >
X^ ~
a'
(w
+ l)+ r
the series
1,
TX
and
1,
^ x tk
in the interval
where
a; ,
a;
<
1.
Hence show that

l
Jo
and deduce that

1 X a-1
oo
da;
+x
-
=2
i \n
/
-
+x
'
a> 1.
when
n +r9
a
n+a
(See also Ex. 36 below.)'

v
INTEGRATION AND DIFFERENTIATION OF SERIES.

oo
2(ae~ waa: -
26. Prove that the series

(c, y),
where
and
a, /?,-y
c are positive
fie-nfa) is uniformly convergent in
and
< y.
Verify that
oo
oo
$y 2
c
(cte-rcaz
fie-nfr) dx =
fy
Is
Jo
27. If it
CO
00
fy
J
fy
S (ae-* - fie-nfix) dx = ^\Jo

1
(ae-naz _
~nfix) dx.
(ae~nax _ fo-nfix) dx
be given that for values of x between

,
7T
f}e
( 1
coshax=2smha7riH- 12a
a cos a;
.
1+a2
and
tt,
a cos 2a;
)
+ 2oa2 2 -. h
+a
J
,
prove rigorously that

.
IT
sin
a;
2 sin 2a:
3 sin
3a;
WHOSE TERMS
INFINITE SERIES
18G
28.
Show that
equal to -
29.
2z
/(^S-tAd'
+ rfx'
if
then
[ch.
haS a differential coefficient
{t
ri
^pjj^pj.
f or a11
values of x'
a stands for a positive number, then the series
When
-1 or*
Z[^l) r
*7\iF+*' T rV
all
values of x
~
a r
~ 2r
+ x2
and,
if
'
their
sums are/(s) and
F(x) respectively,
30.
Find
the values of
all
a;
/ (-l)
for
e x sin
converges.
Does
of x can the
series
it
ar r
which the
r ~ r>
\
series
x + e2x sin 2a; +
. .
converge uniformly for these values
be differentiated term by term

'
31. Let
J
[
for
any
=I
uJx)=x 2 (x*n ::1 -x*n *)sm-x
"
For what values
for
xÔ,
* n (0)=0,
positive integer greater than unity
and
u 1 (x)=x*sm~ for
xÔ,
\ %(0)=0.
Show that
2 **(*)
converges for aU values of x to /(*), where
i
1
/(a;)=a: 2
Also show that f'(x)
is
sin- for
xÔ
discontinuous at
and/(0)=0.
x=0;
that
2u n '(x)
is
not uni-
origin ; and that/'(z) = 2 '(*)

formly convergent in any interval including the
for all values of z.
32.*
Show
that the series
any two finite limits.

can be integrated term by term between
between the limits
integrated
be
series
the
by
defined
function
If so, is
the series
the value of this integral given by integrating
term between these
Ex.
limits
32-38 depend upon the theorems of
74-76.
Can the
and oo ?
term by
[v
187
33. If each of the terms of the series
u^ + u^x)^...
is
a continuous function of x in x
xK
satisfies
the M-test
term from a to
Show
34.
oo
~a>
and
u 1 (x)+x u z (x) + ...
( 67. 1),
then the original
the series
if
(k>1)
series
may be integrated term by-
that the series
can be integrated term by term between any two positive finite limits. Can
and oo ? Show
term by term between the limits
that the function defined by the series cannot be integrated between these
this series be integrated
limits.
35.
Show
that the function defined by the series

1
(l+xf
can be integrated from
term integration of the
36.
to oo
and that
a?- 1
11
l+x
f--fe=i
value
its
is
given by the term by
series.
Prove that
f1
'+...
+ x)
(2
Explain the nature of the
a+
+
l
*_...
a+2
difficulties
a >0).
involved in your proof, and justify
the process you have used.

37.
By
expansion in powers of
e-X( i
Jo
a,
_e
prove that,
-ax) = log
if
< 1,
+ a),
= hr sinh-1a,
tan-1 (a sin x) -.
'sins ^
examining carefully the legitimacy of term by term integration in each case.

bx\ 2n
38.
Assuming that
(bx)
= '
o
/OO
show that
Jo
e~axJQ{bx)dx=
J
.
a2 + 2 /
when a>0.
NOTE
A valuable collection of Examples in Infinite Series with Solutions, by Francis and
Iittlewood (Cambridge, 1928) will be found useful by the student
special study of Infinite Series and Integrals.
who makes a
CHAPTER VI
DEFINITE INTEGRALS CONTAINING AN ARBITRARY
PARAMETER
Finite Interval.
77. Continuity of the Integral.
In the ordinary
ra'
definite integral
<p(x,
y)dx let
a,
a'
be constants.
Then the
J a
integral will be a function of
y*
found to correspond
very closely to those of infinite series whose terms are functions
Indeed this chapter will follow almost the
of a single variable.
same lines as the preceding one, in which such infinite series were
The
properties of such integrals will be
treated.
I.
If
a continuous function of (x y) in
<p(x, y) is
the region
b^y^b',
a^xâ',
ra'
then
c/)(x,
y)dx
is
a continuous function of y in the interval
(b, b').
Ja
Since
</>(x,
37, it is
a continuous function of
y) is
(x, y)-f,
as defined in
also a continuous function of x and a continuous function
of y.
Thus
(p(x, y) is
integrable with regard to
x.
ra'
Let
f(y)=\
<t>(x,y)fa'
Ja
*As already remarked
in 62, it
is
understood that before proceeding to the

which the integral is required,
limit involved in the integration, the value of y, for

is
to be inserted in the integrand.
a function of two variables x, y is continuous with respect to the two

37, we speak of it as a continuous function of {x, y).
It will be noticed that we do not use the full consequences of this continuity in
the following argument.
f
When
variables as denned in
188
189
We know
that, since
in the given region, to
we
as
continuous function of
y) is a
(f>(x,
any positive number
please, there corresponds a positive
(x, y)
chosen as small
e,
number
such that
rj
\<p(x,y+Ay)-cp(x,y)\<e, when |Ay| ^17,

the same
serving for
rj
Let Ay
all
values of x in
(a,
a')*
and write
satisfy this condition,

Ca'
f(y+Ay)=\
f(y +Ay) -f(y) =
Then
<f>(x,y+Ay)dx.
[ [<p(x, y+Ay)- <p(x,
y)] dx.
Therefore
|
f(y
+ Ay)
-f(y) =s
|
y+Ay)- cj>(x,
<p(x,
y)
dx
Ja
<(a'-a)e, when \Ay\^rj.
Thus f{y)
II.
If
is
(6, &').
a^x^ a',
a continuous function of (x, y) in

and \p>(x) is bounded and integrable in (a,
<p(x, y)
b^y^b',
is
a'),
then
CO,'
<p(x, y)\fs(x)dx is
is
a continuous function of y in
Let
f(y)=\ <p(x,y)^(x)dx.
The
integral exists, since the product of
(b, b').
two integrable functions
integrable.
Also, with the
same notation
f(y +Ay) -f(y) =
Let
J"
as in
+ Ay) - <f>(x,
[$(x, y
M be the upper bound of
(I),
y)] xf,(x)dx.
|\^(z)| in (a, a').
Then
f(y +Ay) -f(y)\<M(a'
Thus f(y) is continuous in (b, b').
|
when
a)e,
\Ay\
rj.
78. Differentiation of the Integral.
fa' <p(x, y)dx, where

of (x y) in
y
y) is
(j>{x,
<-
a^xâ', b^y^b', and
exists
and
satisfies the
same
condition.
Thenf'(y) exists and
is
equal to
^-dx.
*This follows from the theorem on the uniform continuity of a continuous

function
(cf.
37, p. 87).
DEFINITE INTEGRALS CONTAINING
190
Since
to
is
any positive number
e,
sponds a positive number
r\
Ay
Let
serving for
all
in the given region,
(x, y)
chosen as small as we please, there correrj,
d<p(x,y+Ay)
dy
the same
[OH. VI
such that, with the usual notation,
d<t>(x,
y)
dy
<e, when |Ay| ^r),
values of x in
(a, a').
satisfy this condition.
Then
f(y+Ay)-f(y) =
Ay
a'
#e, y+Ay)-<p(x,
[
J
y)
Ay
Ja
=r
J L
3y
J a
3y
Thus we have
f^^
_r
f(y+Ay)-f(y)
Ay
d<p(x,y+6Ay) _ d^x,
9y
9y
L
f' r
y)
00^
..
<(a'-a)e, when \Ay\îr).
And
-~
equal to
II.
that lim
Ayô
ox at any point in
this establishes
Letf(y)=
</>(#,
y)\[s(x)dx,
Ay ^H
^-
?'
^(z)
is
where
exists
and
is
equal to
Let the upper bound of |^(#)| in

Then wje find, as above, that
/(y+Ay)-/(y)
p0
Ay
Ja ^y
And
is
<p(x, y)
and
are as in (I),
<3y
Then f\y)
and
(6, &').
Ja
awe?
exists
\p-(x)
{a, a').
~-
\js(x)
(a, a')
be
dx.
M.
dx <M(a'-a)e, when \Ay\mrj.
the result follows.
The theorems
of this section
show that
if
we have
to differ-
Ca'
F(x, y)dx, where F(x, y)
entiate the integral
is
of
the form
Ja
<p(x, y)
or
<p(x, y)
^(x)> and these functions satisfy the conditions
named above, we may put
the symbol of differentiation under
AN ARBITRARY PARAMETER
78, 79]
may
In other words, we
the integral sign.
191
reverse the order of
the two limiting operations involved without affecting the result.
we
It will be noticed that so far
The
integrals.
are dealing with ordinary
interval of integration
and the function
is finite,
has no points of infinite discontinuity in the interval.

79. Integration of the Integral.
Letf(y)=
'
(p(x, y)\f/(x)dx,
where
cp(x, y) is
Ja
in
of
(x, y)
in
(a, a').
a^xâ', b^y^b', and

Cy
Then
where y
* a
Let
$(x,
y)^(x)dy,
<p(x,
J Vo
y are any two points in
Cy
fa'
f(y)dy=\ dx
Jy
\Js(x) is
(b, b').
y)=\
y)dy.
</>(x,
Jb
Then we know that

and
easy to show that (x, y)
it is
(x, y)
in the region
[ 49],
y)
cf>(x,
is
a 2 xp?a\ b^y^b'.
Ca'
Now
let
g(y)
<fr(x,
y) \fs(x) dx.
</>(x J
y)\fr(x)dx
Ja
From
78 we
know
that
==
Ja
Also g'(y)
where y
is
continuous in the interval (6, b') by 77.

(V
Cy
g'(y)dy=\ dy
<t>(x y) \[r(x)dx,
J'yy
'Vo
dy
Vo
(b,
).
<p(x, y) \fs(x)dx
Ja
=g(y)-g(yo)
[*(, y)
Ja
=J
\
Ja
$te y
)]
H x dx
)
<P(x,y)dy-y<f>(x,y)dyj\!,(x)dx
[J
Cy
Ca'
-1
dx
Jy
<p(x,
y) \jr{x)dy.
192
[ch. vi
Thus we have shown that we may invert the order of integration

with respect to x and y in the repeated integral
\ dx
F(x, y)dy,
I
when the integrand satisfies the above conditions and in particular,

since we may put \fs(x) = l, when F(x,y) is a continuous function
;
of (x, y) in the region
with which the integral deals.
80. In the preceding sections of this chapter the intervals (a, a')
and
in
(b, b')
have been supposed finite, and the integrand bounded

The argument employed does not apply
a^xâ', b-^y^b'.
to infinite integrals.
infinite integral
For example, the
foo
ye-^dx
when y -^ 0, but
when y>0, and/(0) = 0.
converges
foo
converges for
it is
simrye-^^dx
values of
all
discontinuous at y = 0, since f(y) =
but
y,
it is
discontinuous for every
and negative integral value of y, as well as for y = 0.

Under what conditions then, it may be asked, will the infinite
positive
integrals
/
Co.'
and
F(x, y)dx
if
convergent when
(6, 6') ?
F(x, y)dx,
Ja
Ja
b^y^b',
And when can we
sign of integration
define continuous functions of y in
differentiate
and integrate under the
In the case of infinite series, we have met with the same questions
and partly answered them [cf. 68, 70, 71]. We proceed to
The discussion
discuss them for both types of infinite integral.
integrals
infinite
of
requires the definition of the form of convergence
which corresponds to uniform convergence in

81.
Uniform Convergence of
with the convergent
infinite series.
Infinite Integrals.
We
deal
first
infinite integral
(00
F(x, y)dx,
where F(x,
number
y) is
bounded
a' being arbitrary.
in the region a
^ x ^ a', b^y^ b',
the
79-81]
The
I.
F{x, y)dx
integral
193
said to converge uniformly to
is
its
.'a
value f(y) in the interval

chosen, as small as
we
(b, b'), if,
any
number e having been

number X such that
positive
please, there is a positive

X
F(x,y)dx <, when

f(y)-\
J
x^X,
same X serving for every y in (b, b').

And, just as in the case of infinite series, we have a useful
the
test
for uniform convergence, corresponding to the general principle

of
convergence
II.
15)
necessary and sufficient condition for the uniform conver-
F(x, y)dx in the interval
(b, b')
is
that, if
any
number e has been chosen, as small as we

a positive number X such that
positive
shall be
<e, when
F(x, y)d3
please, there
x">x'^X,
J x'
the same
serving for every y in
The proof that

for the
(II)
(b, b').
forms a necessary and
uniform convergence
exactly the same lines as the proof in
theorems in
of the integral, as defined in (I), follows
66 for the corresponding
infinite series.
/QO
Further,
it
will
be seen that
F(x, y)dx converges uniformly
if
J a
in
to
(b, b'),
number
positive
number
there corresponds a
X such that
F(x, y)dx <e, when
x^X,
same X serving for every y in (b, V).

The definition and theorem given above correspond exactly to
the
those for the series
u x (x)
uniformly convergent in a
\R n (x)\<e, when
and
R n (x)\<e,
the same
+u 2 (x)
^x^b
+...
namely,
n^v,
when n ^
v,
serving for every value of x in the interval
(a, 6).
194
82.
Unifdrm Convergence of
We now
[oh. vi
Infinite Integrals (continued).
consider the convergent infinite integral
\F(x y)dx
}
where the interval

in the region a
(a, a') is finite,
is not bounded
more complex than
but the integrand
^ x ^ a', b^y^b'.
This case
is
the preceding, since the points of infinite discontinuity can be
more or
distributed in
We
region.
less
complicated fashion over the given
shall confine ourselves in our definition,
and
in the
theorems which follow, to the simplest case, which is also the most
important, where the integrand F(x, y) has points of infinite
discontinuity only on certain lines
x = a1 a2
,
and
is
bounded
...
an
(aâ x <a z
...<a n â')
neighbourhood
in the given region, except in the
of these lines.
This condition can be realised in two different ways.

infinities
may
be at isolated points, or they
may
The
be distributed
right along the lines.
(i)
()
In the
first of
whenOS2'- 6
vra*
r xv
"1
e"xdx
these integrals there
region, at the origin
when 0^y<\.
is
a single infinity in the given
in the second, there are infinities right along
the line x =
from the origin up to but not including y 1.

In the definitions and theorems which follow there is no need
for any distinction between the two cases.
Consider,
first of all,
the convergent integral
F(x,y)dx,
J
J a
where F(x, y) has points of infinite discontinuity on x=a', and

bounded in aîxâ' - g, b^y^b', where a<a - <a'.
For this integral we have the following definition of uniform
r
is
convergence
fa'F(x, y)dx
is
said
to
converge uniformly
to
a
its
value f(y) in the interval
(b, b'), if,
any
positive
number
having
82]
been chosen, as small as we please, there
is
195
a positive number
r\
such that
F(x,y)dx\<e, when 0<=g?/,
|/(y)Ja
same
the
r\
serving for every y in
(b, b').
And, from this definition, the following

vergence can be established as before
test for
uniform con-
II.
and
necessary
for the uniform conver-
F(x, y)dx in the interval
gence of the integral
(b, b')
is that,
if
J a
any
number
positive
the
has been chosen, as small as we please, there
a positive number
shall be
rt
r\
such that
'~ r
'
J a'
F(x,y)dx <e, when 0<"<'^?7,
same r\ serving for every y in (b, b').

we see that if this infinite integral
Also
in
(b, b'),
positive
number
to the arbitrary positive
number
f
is
there will correspond
such that
r\
a'
F(x,y)dx <e, when 0<^7),
Ja'-
same r\ serving for every y in (b, V).

The definition, and the above condition, require obvious modifications when the points of infinite discontinuity lie on x = a,
the
instead of x = a'.
And when they

by the
lie
on
lines
x = a lt a 2
definition of the integral it
others in which
one of the
F (x,
...
a n in the given region,
has points of infinite discontinuity only at
y)
limits.
In this case the integral
when each
can be broken up into several
is
said to converge uniformly in
of these integrals converges
And, as before,
if
(b, b'),
uniformly in that interval.
the integrand F(x, y) has points of infinite

a 2 ... a n and we are dealing with the
discontinuity on x = a x
F(x, y)dx, this integral must be broken up into several
integral
Ja
integrals of the preceding type, followed
form discussed in
The
integral
when the
is
by an
integral of the
81.
now
said to be uniformly convergent in
integrals into
which
it
(6, b')
has been divided are each uni-
formly convergent in this interval.
196
The simplest
Uniform Convergence,
83. Tests for
[ch. vi
test for the
F(x, y)dx, taking the
uniform convergence of the integral
first
type of infinite integral, corresponds to Weierstrass's M-test for

the uniform convergence of infinite series (67. 1).
I.
a^xâ',
Let F{x, y) be bounded in
b^y^ b'
and
in (a, a'), where a' is arbitrary, for every y in (b, V).
F(x, y)dx will converge uniformly in
integral
(b, b'),
integrable
Then
the
if there is
Ja
a function
independent of
jj,(x),
(i)
(ii)
(x)
[jl
^ 0,
\F(x,
jbi
(x)
dx
such that
xâ;
when
xâ
when
y)\ -=il(x),
Jew
y,
and
b^y^gb'
exists.
For,
by
(i)
and
(ii),
when
x>x'â
and b^y^b',
F(x, y)dx hi
x'
and, from
there
(iii),
is
a positive
/Li(x)dx,
J
x'
number
X such that
f/Lt(x)dx<e, when x">x'
^ X.
x'
These conditions
x
^ a,
and
be satisfied if xn F(x, y) is bounded when

for some constant n greater than 1.
will
^ y' ^ b'
Cokollaey. Let F(x, y)=^(p{x, y) \fs(x), where <p(x, y) is bounded

in xâ and b^y^b', and integrable in the interval (a, a'), where a'
is arbitrary,
for every y in
Also
(b, b').
F(x, y)dx
is
yj,(x)dx be absolutely
let
(b, b').
Ex.
1.
r -i^-m
x i+ y
ji
Ex.
2.
<rx v dx converge uniformly in y
^ y > 0.
jo
dx converges uniformly in
< y ^ Y,
where
7 is
an arbitrary
Jo
positive number.
uniformly for
II.
Let
all
<j>(x,
values of y, where
y) be
bounded
function of x for every y in
n > 0.
inxâ,
(b, b').
b^y^ b'
Also
let \js{x)
and a monotonic
be bounded and
83]
not change sign
more than a
interval (a, a')*
and
let
finite
number of
197
times, in the arbitrary
\Js(x)dx exist.
Ja
Then
<p(x, y) \f/(x)dx
coywerges uniformly in
(b, b').
Ja
This follows immediately from the Second Theorem of

Value, which gives, subject to the conditions
<p(x, y)
where
But
\^(x)dx =
satisfies
bounded
is
<p(x, y)
a<x'
\j/(x)dx
i/i(x\ y)
named
Mean
above,
\Js(x)dx,
<p(x'\ y)
x"
in
xâ
and b^y^b', and
\\/{x)dx
J a
converges.
Thus
V
r
it
follows from the relation
(J)(x,
y) \js(x)dx
Iff
0(^'
that
0(#,
It is evident that
\J,(x } y), if
\fr(x,
'
|Jf
in this
\f/(x)
\j/(x)dx
</>(x",y)\
\]s(x)dx
?/)
|-'z'
lp"
\j/{x)dx
2/)
theorem
(6, 6').
may
y)dx converges uniformly in
be replaced by
(6, b').
J a
f^
Ex.
e~ xy
J0
III. Ze
sin
oc
dx,
(/>(#,
y) be
and tend uniformly

Also
and
let \j/(x)
be
e~ xy
cos
3?
dx (a>0) converge uniformly in y
a monotonic function of x for each y in (b, b'),

as x increases, y being kept constant.
to zero
not change sign more than a finite
interval*
Further,
converging as x->oo
Then
~0.
let
\f/(x)dx
be
(a, a'),
number of times in such an

bounded in
xâ,
without
(p(x, y) \j/(x)dx is
(6, b').
Ja
This follows at once from the Second Theorem of
This condition
Mean
is
Value,
borrowed from the enunciation in the Second Theorem of

If the more general proof is taken, a corre 50. 1.
Value, as proved in
sponding extension of
Mean
(II)
and
(III) follows.
198
Also it will be seen that \^{x)
as in (II).
if
bounded
y) dx is
\ls(x,
in x
[ch. vi
may be replaced by \p-(x, y),
^ a and b^y^b'*
Ja
poo
r~-*>
e-w
Ex.1.
Ex
y^
It
^ y > 0.
Q
Jo
!E^ dx and f Ê*^ dx both converge uniformly in y ^ >
l
x
*/
Jo
Jo
and
e~xy cos x dx converge uniformly in y
x dx,
sin
Jo
-y <0.
can be
to the reader to enunciate
left
and prove
similar
fa' F(x, y)dx.

a
useful test for uniform convergence in this case
The most
corresponding to
Bx.
1.
(I)
xv- 1 dx,
2.
84.
above.
xv- 1 e~ x dx converge uniformly in
^^ dx converges uniformly in
that
> y ^ yQ > 0.
Jo
Jo
Ex.
is
<y=y <
1.
We
F(x, y)dx.
Continuity of the Integral
now
shall
F(x, y)dx, where
consider, to begin with, the infinite integral

Ja
F(x, y)
bounded
is
in the region a
we
^ x ^ a', b^y^b',
form
shall return to the other
arbitrary.
Later
integral in
which the region contains points
a'
being
of infinite
of infinite discon-
tinuity.
Let
F(x,y)dx, where F(x,y)

f(y)=\
J
either
a continuous
is
function of
the
(x, y)
ina^xâ', b^y ^ b',
form
\{s(x) is
a' being arbitrary, or is of

continuous as above, and
<f>(x, y)^(x), where cj>(x, y)

is
F(x, y) dx converge uniformly in
{a, a').
(b, b').
Thenf(y) is a continuous function of y in (b, b').

Let the positive number c be chosen, as small as we please.
such that
Then to Je there corresponds a positive number
foo
F(x, y) dx
<h, when x^X,
the same
X serving for every y in
*In Examples
and 2
(b, b').
of 88 illustrations of this
theorem
will
be found.
83-85]
in 77 that, under the given conditions,
But we have proved

F(x, y) dx
continuous in y in
is
199
(6, b')-
Ja
Therefore, for
F(x,
I
some
number
positive
y+Ay)dx-\
Ja
rj,
when
F(x, y)dx <e,

a
\Ay\^rj.
00
JX F(x
y)dx +
Thus
F(x,y)dx.
/(y+Ay)-/(y) = r \*F(x, y+Ay)dx- \* F(x, y)dx

TOO
foo
Jx
x
r 00
Also
F(x,y)dx.
**(a;,y+Ay)<fo;-
we have
|Jz
F(x, y) dx <ie,
I
poo
I
and
Ux
jF(jc,
+ Ay) da? <Je.
Therefore, finally,
|/(y+Ay)-/(y)|<fc+'fc+i
<e, when |Ai/|^?y.
Thusj^y)
is
continuous in
(b, b').
poo
F(x, y) dx.
85. Integration of the Integral

Ja
Let F(x, y) satisfy the same conditions as in

30
Cv
Then
dy
Jy
where y
f*>
Cv
J a
Jy
'
Let
84.
F(x y)dy,
F(x,y)dx=\ dx
Ja
(6, b').
f(y)=[F(x y)dx.
9
Ja
Then we have shown that under the given conditions

continuous, and therefore integrable, in
Also from
79, for
any arbitrary interval
{a, x),
be taken as large as we please,
dx
y
F(x,
y)dy=\ dy
y being any two points in
(6, b').
f(y)
is
(b, b').
F(x, y)dx,
where x can
200
[oh. vi
Therefore
dx
F(x,
y)dy=\im
dy
F{x, y) dx,
provided we can show that the limit on the right-hand
exists.
But
X
\
Jy
dy
[
J a
F(x, y)
dx= P dy
J
F(x, y) dx
'
-\dy\
F(x, y) dx.
J x
J y
J a
Thus we have only to show that

F{x, y)dx = 0.
dy
lim
x^x J y
J x
Of course we cannot reverse the order of these limiting processes
and write
this as
dy lim
for
we have not shown that
F(x, y) dx,
this inversion
would not
alter the
result.
/oo
But we
F(x, y) dx
are given that
is
Ja
in
{b,
V).
Let the positive number e be chosen, as small as we please.

Then take e/(6' - 6). To this number there corresponds a positive
such that
number
F(x, y) dx
the same
<TTZh> wnen x
X serving for every y in
=X
>
(b, b').
It follows that
if
F(x, y)dc <e,
dy
when
x^X,
JX
JVn
b &')
(
y y lie
In other words,
,
>
\im\fdy
And from
F{x,y) dx -0.
the preceding remarks this establishes our result.

F(x, y)dx.
86. Differentiation of the Integral

Ja
F(x, y) either be a continuous function of (x, y) in the

region a^xâ', b^y^b', a' being arbitrary, or be of the form
is bounded
X} y) \ls(x), where $(x, y) is continuous as above, and \ls(x)
Let
<p(
85-87]
and
Also
integrable in the arbitrary interval (a, a').
partial differential coefficient
which
satisfies the
201
let
F(x, y) have a
same
conditions.
F(x, y)dx converges tof(y), and the integral
oy
Ja
dx converges uniformly in
coefficient at every
point in
Ja
dx
cy
is
has a differential
dF
'
dx.
dy
from
f(y)
and
(b, b'),
00
We know
(b, b'),
84 that, on the assumption
a continuous fvmction of y in
named above,
(b, b').
Let
g(y)=\
^dx.
a
Then, by . 85,
g(y)dy=\ dx
|
where y
Let
y x are any two points in
yQ = y
+ Ay
Jyy
and
g(y)dy=
Therefore
[F(x,
dy,
(b, b').
y x = y+Ay.
y+Ay)-
F(x, y)] dx.
Ja
g()Ay=f(y + Ay) -f(y),

Too
ytk~y+Ay
where
Thus
g(z )=
But lim g()=g(y),
F(x, y)dx.
and f(y)=
J a
I
fkM=m
since g(x)
is
continuous.
It follows that/(?/) is differ entiable,
and that
dx,
where y
is
any point
in
(b, b').
87. Properties of the Infinite Integral
'
F(x, y) dx.
The
results
J a
of 84-86
integral.
The
can be readily extended to the second type of infinite

be sufficient to state the theorems without proof.
It will
argument are in each case parallel to those in the

preceding discussion. As before, the region with which we deal is
steps in the
a-^x^La',
b^y^.b'.
202
[ch* vi
(V
I.
Continuity of
F(x, y) dx.
J a
Let f(y)
E{x
y) dx,
F(x, y)
where
has points
of infinite
J a
discontinuity on certain lines
and x a', and
x = av a2
(e.g.,
...
a n ) between x = a
a continuous function of (x, y), or the product

of a continuous function cj>(x, y) and a bounded and integrable
function \p~(x), except in the neighbourhood of the said lines.
Then,
is either
F(x, y) dx
if
is
(b, b'),
f(y) is
J a
a continuous function of y in
(b, &').
F(x, y) dx.
II. Integration of the Integral

J a
Let F(x, y) satisfy the same conditions as in
Then
where y
dy
F(x,y)dy,
F(x, y)dx=\ dx
(I).
(b, b').
ra'
F(x, y) dx.
III. Differentiation of the Integral

.'a
where
F(x, y) dx,
F(x, y)
has points
of infinite
discontinuity on certain lines
and x a',
and
is
x = a lf a 2
(e.g.,
...
either a continuous function of
a n ) between x = a
(x, y),
or the product
and a bounded and
integrable
of a continuous function <p(x, y)
lines.
function \js(x), except in the neighbourhood of the said
7)F
Also let F(x, y) have a partial differential coefficient -, which
satisfies the
Further,
same
conditions.
F(x, y)dx converge, and
let
formly in
J a
Ja
oy
dx
converge
uni-
(6, 6').
Thenf'(y)
exists
and
is
equal to
dx in
(b, &').
88. Applications of the preceding Theorems.
Ex.1.
To prove
|
(i)
Let
F(a)
= \\-^^dx
2
(ai=0).
*For other proofs, see Ex. 11 and Ex. 12 on pp. 135-6, and Ex. 10 on
p. 213.
87, 88]
(Cf
0,
This integral converges uniformly when a
e- aX Jx is a monotonic function of x when x > 0.
Thus, by the Second Theorem of Mean Value,
203
83, III.)
for
ex" e
-ax
e-o.x' r
x dx =
sin
)x'
'
sin x dx
)x'
e-^"
x"
sin
it
x ax,
)(
where 0<x'î^=x".
Therefore
>
\} x
<4
sm xdx\~
sm * dx
-aX If*...
7r-
'
sin
a;
g-a*
r-
<
It follows that
provided that
f
)
since a
sinx
x"
e-o-x
rf
X > 4/c,
sin
since
x dx
'
rfx
= 2 for all values of ^ and g,
dx
<e when x">x'^X
this holds for every
and
rc
If
a greater than or equal to 0.

and it follows from
This establishes the uniform continuity of the integral,
84 that F(a)
is
inaÔ.
continuous
t^/^v
F(0 )=
Thus
hm
i>
sina;
e-**
a
.
[
JO
(ii)
is
^dx=
X
dx,
a-^0J0
limfe dz.
X
O _*,jo
e-^smxdx
Again, the integral
uniformly convergent in a ;> a > 0.

This follows as above, and is again an example of
Thus, by 86, when a > 0,
83, III.
-K( *>
-C
e-* x 8mxdx.
e-*z(cosx + asina;) = -(a

dx
But
Therefore
+ l)e-*sinx.
e-*sina:to = -j
2
Jo
Thus
J"(a)=
And
^ (a) =
since lim
a>-oo
+l
1
aa +
- tan" 1 a +
>
F{a)=0*
* If a formal proof of this is required, we might proceed as follows:

Let the arbitrary positive e be chosen, as small as we please.
204
from
x
It follows
Ex.
To
2.
'
^dx = -ê~ a
1+rc2
JO
Jo
- rr+ x 2
x(l
)
and
/(a)=j
integral
is
continuous for
dx
-e-)
(l
s
2V
Jo
Let
The
dx =-5
2
prove
cos
(i)
that
(i)
v
[cm. vi
(a
'
= 0).
'
y^s-^-
uniformly convergent for every a, so that, by
we can
values of a, and
all
84, f(a) is
integrate under the sign of inte-
gration (85).
(ii)Let
= \f(a)da.
(f)(a)
'o
Then 0(a)
is
a continuous function of a for
0(a)
Also
sin
(iii)
lies
Again,
( 49).
ax.
)
that /'(a) will be equal to -
we know
2
JO * ~T~ X
this
uniform
convergence
of
integral.
within an interval of
But
+X
Since
Jo
5 is
monotonic when x
i=
this
Also
3-).
-/3
number
a,
so that
,
M sin x ax
ea-%
("*
e-*
But
)x
:==
sin
a:
da)
= e-ô
<ie.
f-rsina;.
dx,
[fsinz-,
)xn
dx + e-**\
~ X.
And we know
that
9
I
*dx \<7riorq>p>0- (cf.

I
sin
e-*
Therefore
I
Thus we can choose
^4
J*o
'
dx
p. 222).
< 2we -**<>.
'
so large that
:
-^x^ldx\<\e, when a>i.
*o
It follows that
II
Jo
I
dx
e-*?^
x
<K+i
[
ax dx,
if
a positive number
a.
independent of
is
a;
independent of
is
sin
= 0.
+ X2
e-az^êfa; converges uniformly in a ^0, there

*
we can choose x
where x
and lim
1,
Ijr
and
ax
+* 2
e(l
values of a
y^f da
<Zz
J
all
+ l*f when
a>A.
88]
Thus we have, when x" > a?' i==
205
1,
where a/î^=a;".
It follows that
^
*'
1+a;2
sin aa;
fc-sfiri*)
**
If*"
^ j^
Therefore
Thus
t-^-:, sin az dx
C-
is
sin
8in
a*
&+ s?r*i*5 C
L-
tfa;
|
to
4a;'
^ ^j^when a ^ a
/,(a)=
8in
> 0,
2Sinaa;rfar '
"LrTa7
a
(iv)
Now
<f>(a) b=
/(a)rfa.
Jo
Therefore
</>'{a)
=/(a)
< =/'(a) =
and
j^-2
ax dx.
sin
fT,
-!^^
^HoL
1
Thus
sin
ax j
o*(l
+ *2
*-!+)
This result has been established on the understanding that a
(v)
From
(iv)
we have
(f)(a)=Ae*
But
<f)(a)
continuous in
is
+ Be~ a +2> when a>0.

0(0) =0.
aÔ, and
lim#(a)=0,
Therefore
a-0
A + B + ^=0.
and
Also
<f>'{a) is
continuous in af|
0.
and
<f>
(0)=-^
A-B-^=0.
Therefore
It follows that
.4=0
and
B= - ^
Thus
0(a)=|(l-e-), when a>0.
And
/(a) = 0'(a)
Both these
=| e~, when
results obviously hold for
a=0
as well.
a>
0.
> 0.
and
206
Ex.
Gamma
The
3.
Function Y(n)
e~ xx n
~x
Jo
To prove T(n)
(i)
N'may
be
is
When n Ss
1,
N^n^n
uniformly convergent when
and however near
n may
zero
e~ xx n
the integral
~x
n > 0, and
dx,
[ch. vi
its derivatives.
>0, however
large
be.
dx has to be examined for convergence
Jo
When < n< 1, the integrand becomes

In this case we break up the integral into
only at the upper limit.
x = 0.
r<*>
e -xx n-l
e -x x
fa +
first
n-i dXt
Ji
Take
infinite at
e~ xx n
the integral
~x
dx.
Jo
When
<x<
z n_1
1,
-1
e~xxn
Therefore
It follows
xn<r\
if
n ^ nQ > 0.
=i e~ xxniT\
if
n^no >0.
from the theorem which corresponds to

when n ^ n > 0.
83, 1,
e~ xx n
Again consider
-1
e~ xx n
that
~x
dx
converges uniformly
n>0.
dx,
Ji
x>l, x -1 ^?*-!,
When
11
e -x x
Therefore
It follows as
above (from
n-l^ e -x x N--l
0<n^N.
if
o<n^N
e~ xxn
83, I) that
0<n^N.
if
f
~x
dx converges uniformly for
Jl
Combining these two
we
results,
see that
e~ x x n
~ 1 dx converges uniformly
Jo
N i^ n = n > 0,
when
may
however large
may
be and however near zero n
be.
poo
To prove T
(ii)
(n)
e- x x n
~x
log
n>0.
x dx,
Jo
We know that
lim (x r log
x)
= 0, when
> 0,
x-+Q
x=0
so that the integrand has an infinity at
when
0< n^L
But when
0<x<
And we have
seen [Ex.
!fra;"o l|loga;|,
7, p.
Jo
It follows as
e~ x x n
above that
When
-1
\ogxdx
is
uniformly convergent when
>0.
Also for
wi^?i >0.
if
x n o~ 1 logxdx converges when
133] that
n ^
<n^l.
<
e-a;a;n-1 log
a:
cfo,
x>l,
'
e~ xx n -^
Therefore
we proceed
xn-i
as follows
<xN -\
log x ^
if
<n^N.
e x x N ~ 1 log a;
<e~ xx N
since
^- < 1
a;
But
n only
if7i^n >0.
a?*-l<a;o-l,
1,
e-&xn l|loga;|
Therefore
n>n
for positive values of
1.
e~ xx N dx
is
convergent.
when x >
1.
88]
~
e~ xxn 1 log x dx
Therefore
is
'
Combining these two

convergent
zero
when
n may
We
we
results,
see that
207
< n rl N.
when
r e- xxn-
\ogxdx
is
uniformly
N ^ n ^ n > 0, however large N may be and however near
be.
are thus able to state, relying
'(
n)
e~ xx n
on
-1
that
86, 87,
n>
log x dx for
0.
Jo
It can be shown in the same way that the successive derivatives of T(n)
can be obtained by differentiating under the integral sign.
Ex.
4.
(i)
log(l-2y cos x + y2 ) dx
To prove
uniformly convergent for any
is
Jo
interval of
(e.g.
h^y^b'); and
(ii)
to
deduce that
log sin
xdx
-\i: log 2.
2
Since 1 -2y cos x + y 2 = (y -cos :r) 2 + sin ;z, this expression is positive for
m
= 0, 1, 2, etc., and
values of x, y, unless when x = nnr and y = ( - l) ,
for these values it is zero.
(i)
all
It follows that the integrand becomes infinite at

one case when y l, and in the other when y= -1.
We consider first the infinity
at
x=0.
bounded
in
any
As the integrand
is
interval of y which does not include y =
x = ir
^ x ^ X, where X <
strip
1,
x=0 and
we have only to examine
ir,
in the
for
any
the integral
log (1 - 2y cos x + y 2 ) dx
neighbourhood of y = 1.
a and a
Put y \+h, where h
to be fixed more definitely later.
in the
is
some
positive
number
less
than unity,
ex
log
Since
- 2y cos x + y 2 dx
)
Jo
h2
log(l-cosx +
it is
clear that
we need only
Jdx + xlog2(l+h),
discuss the convergence of the integral
h2
^(l-cosz+gu-^r*.
Take a value
X>
Then
Now
of a(0
<a<
a
2(1^)
1)
a2
such that h71
2(1 -a)
,
< 1.
h
S 2(T+A)
a0 *"W^*
'
a2
8 = cos -1 ^-^
r
r
2(l-a)
let
It will
be seen that, when h

|
^ a,
h2
< 1 - cos x tk 1 - cos x + 2(\ + h\ ==

provided that
< x ^ p.
*>
208
same
Therefore, under the
\lg{
But the
conditions,
h2
[ch. vi
- C08X + 2t
l
+ fy
\~\ l g( l cosx )\'
shows that the integral
/x-test
og
- cos x) dx
Jo
converges.
log
It follows that
converges uniformly for
And
- cos x +
^ a.
|/^|
therefore
( 1
\og{\
dx
+fe) )
[Cf. 83, I.]
-2y cos x + y 2 )dx
Jo
converges uniformly for any interval (b, b') of y.

The infinity at x = ir can be treated in the same way, and the uniform con-
vergence of the integral

log(l
-2ycosx + y 2 )dx
Jo
is
thus established for any interval

(ii)
(6, b')
of y.
2
f {y) = {"10* {I -2y cos x + y )dx.
Let
Jo
We know
from
70. 1 that
when |y|<I,
f(y)=0,
f(y)=7rlogy
and
when \y\>l.
2
,
just seen that the integral converges uniformly for
But we have
any
interval of y.
It follows,
from
87, I,
that
/(l)=lim/(y)=0
y-*l
/(-!)= Hm/(y)=0.
and
/(I)-
But
log 2(1
-cos
a:)
Jo
= 2tt log 2 + 2r
Jo
= 27r log 2 + 4
"
log sin ^
*
da;
log sin x dx.
Jo
'
Thus
log sin x dx =
\tt log 2.*
Jo
From/( -
1)
=0, we find in the same way that
log cos
xdx=
-\tv log
2,
Jo
preceding.
a result which, of course, could have been deduced from the
This
integral
was obtained otherwise
in
Ex.
4, p.
131
finite
88, 89]
89.
dx
The Repeated Integral
f(x, y) dy.
la
It is
209
not easy to determine
lb
general conditions under which the equation

p
/.CO
(30
dx
f{x,
.o
rCO
y)dy=\
f(x, y)dx
dy
la
J&
lb
is satisfied.
The problem
infinite series
what
We
infinite limits.
shall discuss only a case
similar to that in infinite series given in
Let f(x, y)
term by term integration
closely analogous to that of
is
between
be
of
an
some-
76.
in
(x, y)
x jg
a, ysz. 6,
and
let
the
integrals
fOO
(00
f(x, y)dx,
respectively, converge
lb
uniformly in the arbitrary intervals
b^yTib',
let
the integral
Then
dx
(iii)
converge uniformly in y
f(x, y)dy
la
= x~a'.
ry
reo
Also
lb
b.
the integrals
and
f(x, y)dy
dx
and are
we
Since
f(x, y)dx
dy
la
lb
lb
exist
f(x, y)dy,
(ii)
.a
equal.
are given that
f(x, y)dx converges uniformly in the arbitrary
interval 6
= y ^ b'
we know from
dy
f(x,
y)dx=\
la
85 that
dx
la
f(x, y)dy,
when y>b.
lb
It follows that
f(x, y)dx =lim
dy
dx
y--s>la
'a
f(x, y)dy,
lb
provided that the limit on the right-hand side exists.
To prove the
existence of this limit,
arbitrary positive
number
[y
dx
<
f(x, y)dy
ly
But from the uniform convergence
b,
we can choose
f(x, y)dy
number
X such that
f(x,y)dy <|-e, when
dx
the same
when y"> y / ^ Y.
lb
the positive
e,
x^X,
lb
X serving for every y greater than or equal to b.

j-CC
Also
is
we
are given -that
f{ x ,y)dy
uniformly convergent in the arbitrary interval

c.i.
show that to the

number Y such that
to
of
dx
la
in y
it is sufficient
(a, a'),
(1)
210
choose the positive number
y)dy\<
f(x,
the same
serving for every x in
Thus we have
y)dy =
'f(x,
dx
"f(x,
Therefore from
dx\
y)dy -[' dx [ f(x, y)dy.
'f(x,
)x
lb
lb
y)dy\<
)dy\
&+& +h
f(x,
)y
'a
<
when
y"
> y' ^
Y.
have thus shown that
*V ["/(^ y)dx = \im
[*
'6
It
(2)
we have
(2)
dx
}x
)y
and
(1)
y)dyS
}a
.V
We
>y'=Y,
y">y'~Y
y)dy\<\t, when
f(x,
y
\
y"
clear that
it is
dx
dx\
when
so that
X).
(a,
Cy"
Ij-.V
But
~ -,
3(X-a)'
[ch. vi
f(x, y)dy
dx
[
(3)
lb
|/->co.'o
.'
remains to prove that

f(x, y)dy =
dx
lim
y^ x
Let the limit on the left-hand side be
Then a being any positive number,

number 7 X such that
Also,
f(x, y)dy.
.'&
I.
as small as
we
please, there is a positive
f{x,y)dy\<\e, when y
dx
\l-\
dx
la
)b
la
)b
i^
Y1
(4)
from the uniform convergence of

ry
CO
f(x, y)dy,
when yô,
a positive number
X such that
dx
a
we know
that there
x
ry
is
lb
la
Jb
la
(5)
serving for every y greater than or equal to

same
> a.
Choose a number X' such that X'
tlie
X'^X,
dx\ f{x,y)dy\<^, when
dx\ f{x,y)dy-\
a
fy
r.\'
b.
^X
f(x, y)dy in
Then, from the uniform convergence of
any arbitrary
interval,
)b
we know
that there
is
72
a positive number
f(x, y) dy
f(x,
y)dy\<
72
the same
a;
dx\ f(x,y)dy-\
Thus
I
Now
serving for every
la
J&
take a number
Equations
(4), (5)
and
3(Z '_. g
when
*/
i=
72
in (a, X').
dx
< Je,
f(x, y) dy
)b
greater than
(6)
such that
7 and 7 2
X
hold for this number 7.
when y ^ 7 2
(6)
89]
211
But
l
~] dx \_f{ x>y) dy\

r
dx
f(x, y) dy
'
'&
Jo
rfx ('
J&
jf'
rr
<fo
/(*,
from
(4), (5)
*/)<&/
- {"dx
and
['/(*,
Ja
Jb
rx'
r*>
y)dy-\ dx\
Jb
'a
< h + h + h,
/(*,
Ja
2/)
<fy
f(x, -y) dy
Jb
(6),
This result holds for every number X' greater than or equal to X.
Thus we have shown that
Z= lim
->oo
Also,
from
(3),
rfz
J
/(, y)
.'&
%= Ja
f{x, y) dy.
da;
Jb
we have
f(x,y)dy =
<*
J
Jb
dy\ f{x,y)dx
Ja
J&
under the conditions stated in the theorem.

It must be noticed that the conditions we have taken are sufficient, but not
necessary.
For a more complete discussion of the conditions under which the
integrals
dx\ f{x,y)dy,
Ja
when they both

de
Jb
dy\ f(x,y)dx,
Jb
exist, are equal, reference
la Vallee Poussin,* to
whom
cussion of the whole subject
Ja
should be
the above treatment
made
is
due.
to the works of
valuable dis-
also given in Pierpont's Theory of Functions

question is treated in Hobson's Theory of Functions
is
of a Real Variable. The

of a Real Variable, but from a more
difficult
standpoint.
REFERENCES.
Bromwich, loc. cit., App. III.
De la Vallee Poussin, loc. cit, 2
(4e ed., 1922), Ch. I.

Dini, Lezioni di Analisi Infinitesimale, 2 (Pisa, 1909), la Parte, Cap. IX, X.
Goursat, loc. cit, 1 (4e ed., 1923), Ch. IV and V.
Hobson,
Osgood,
loc.
Pierpont,
Stolz,
cit,\ (3rd ed., 1927), Ch. VI; 2 (2nd ed., 1926), Ch. V.
loc. cit, 1
loc. cit,
Brunel,
Bd.
loc. cit,
Kap. III.
XIII-XV.
(4 Aufl., 1923),
loc. cit, 1
(1905), Ch.
Absch. X.
I,
Enc.
d.
math. Wiss., Bd. II, Tl. I (Leipzig, 1899).
And
Montel-Rosenthal,
loc. cit,
Enc.
d.
math. Wiss., Bd. II, Tl. Ill, 2 (Leipzig
1923).
* His investigations are contained in three memoirs; the first

in
Bruxelles, 16 (1892); the second in Journal de math., (4), 8 (1892);
Journal de math., (5), 5 (1899).
See also Bromwich, Proc. London Math. Soc,
(2), 1 (1903),
176.
Ann. soc. scient

and the third in
212
EXAMPLES ON CHAPTER
Prove that
1.
er^dx
is
[ch.
VI.
= ao >0,
and that
Jn
dx
is
['rw^^fo
Prove that
2.
^ 0, when 6 > 0.
y^y
is
>0, and
2/
Te- ax
that
^^
11
Jo
3.
is
Prove that
e~ ax x n
^ a > 0, when n j
4.
Prove that
> 0, when
a -^ a
5.
uniformly convergent in y
^ 0, when a > 0.
2/
1,
- x cos
^ 0,
~ x sin-x
show that
Sin
1.
is
^ 0, when
< w < 1.
and
~Vo>^
aX
C S
^ 0, when < rc<
da; is
^^- Ac
Jo
is
in the interval a
and
Using the fact that
dx
in the interval a
and
e- ax x n
a:
dx
is
y=z-yo<>
uniformly convergent in any interval of y
Jo
which does not include y a.

75
6.
Show
that
(l
(i)
sin x
-*v)
<fo,
00
f
e-^v
(ii)
(cos
-
Jo
are uniformly convergent for y

7.
Jo
ax - cos
x
&:r)
'
dx,
= 0.
Discuss the uniform convergence of the integrals

f
(iv)
tan" ay
Jo
1
f
(vj^v-^log
f'^Jll^.
'
log *
a,
=f=
^_
(Wz.
(iii)
^>0.
*)&:,
'o
integral sign is allowable in the

8. Show that differentiation under the
are given opposite each
following integrals, and hence obtain the results that
:
f"
(i)
(ii)
e-a*dx =
dx
tv
r xê~a^dx
9n
zi = ^
/ir
l^-a
1.3...(2n-l)
2an +i "
rt>
FT^ 2^
n
***=-v *>
dx
1.3...(2n-l)
_ 7r
<^TaJ^
2nn\a*n
2-
(m)
^+irf
mr,
s=
V."
T
t^-*
.C
f
-i
if**^ 0<M<1;
2 cos -g-
*"<
l0
'I'
J,
'x n
f
\ogx
7r
n7r
ttn '^
'
vi]
9.
213
Establish the right to integrate under the integral sign in the following
integrals
(i)
e~ ax dx
interval a
> 0.
Ss:
Jo
e~ax cos bx dx
(ii)
> 0,
or
any
interval of
b.
> a > 0,
or
any
interval of
6.
interval a }> a
Jo
e~ ax smbxdx
(iii)
interval a
Jo
10.
Assuming that
Jo
1.
z-fx-e-gx
fa: cte
= tan -1
(i)
....
=-=,
62
qr
sin
sin&z
(ll)
sin bx
> 0,
show that
- tan -1 , g
t
dx = tan -1
>/> 0.
-?.
= i7T, 6>0.
rfx
Jo
Show
e~aa; sin 6x efc = -=

2
00
Deduce that
11.
>
xa dx; interval a :> a
(iv)
that the integrals
CMhxdx =
__
a
f"
9
b
^*mbzdx=^,
a>0,
can be differentiated under the integral

and hence obtain the values of
sign, either
xe ax cos bxdx,
with regard to a or
xe~ ax sin bx dx,

Jo
*2 e-a*cosW:r,
a&r^sin&ardfe.
Jo
12.
Jo
f<ll) jriz2S**!i
Jo
Let
Show
that/'(y)
sin
xy cr* dx for
all
r, ta
values of
/(*/) = i log (1+*/ 2
13.
Let
e-a:
/(y)=
y,
and deduce that
).
cos2a:2/^.
Jo
Show
On
that /'(?/)
integrating
From
-21
by
this result,
parts,
it will
2a^
for all
values-of y.
be found that
f'(y)+2yf(y)=0.
show that f{y)=\sjê~v\ assuming that
V^^sJtt.
Also show that

J
and deduce that
are-*2 s i n
dx\ e-%2 cos 2xy dy =

\f(y) dy,
o
Jo
[^
x2
^^
dx =
^[
V
e -y-
dy.
b,
214
14.
n>
> 0,
where a
du
^=
Then show that
d
it
u = T(n)
tt
Also show that,
cos nO
= I
sin nO.
K-l(n)
< * < 1, Mm U = Jo a*"
if
+ n2 w - nu.
# = !>)^-,
Thus
Vrn = v.
dv
= n14
d0
-.
follows that
Deduce that
where -|7r<0<|7r,
Urn = u,
rx = y,
these
6=rsin0,
a=rcos0,
From
er^x"' 1 sin bx dx,
0.
the following substitutions
Make
V=
Uê-^x^cosbxdx,
Let
cos
r
6a; cfe,
-o
lim
F=
a->0
And deduce
nir
0"
cos x
Jo
sin
a;
ax.
ZWwe cm the Calculus (2nd ed.,
Prove that
o
where
lir
[Compare Gibson,
15.
n_1 sin 6s dz
that
rl-n
00
a:
.'o
o
f
6 is
.'ft
any positive number.
1906), 471.]
CHAPTER
VII
FOURIER'S SERIES
90. Trigonometrical
Series
and Fourier's
We
Series.
have
already discussed some of the properties of infinite series whose
terms are functions of
x,
confining our attention chiefly to those
whose terms are continuous functions.

The trigonometrical series,
+ (a 1 cosx+b 1 smx) +(a 2 cos2# +& 2 sin2z) + ...
(1)
where a a v b lt etc., are constants, is a special type of such series.

Let f(x) be given in the interval ( - 7r, it). If bounded, let it
,
be integrable in this interval
unbounded,
if
f(x) dx be absolutely convergent.
integral
J -
the infinite
Then
IT
f(x') cos nx' dx'
and
exist for all values of n.
The trigonometrical
the coefficients a
1
=^
f*
I
/(#') sin nx' dx'
J IT
J IT
let
61, VI.)
be a Fourier's Series, when
series (1) is said to
a v b lt
etc.,
/(#') dx' and,
by
are given
when n
1,
(2)
If"'
an = -
f(x') cos nx' dx'\
Tf J - n
These
bn
f""
7T J
f(x') sin nx' dx'
-n
coefficients are called Fourier's Coefficients or Fourier's
Constants for the integrable function f(x)\ and the Fourier's Series
said to correspond to the function.
is
This nomenclature
is
used quite independently of any assumption
as to the convergence of the series (1)
substituted for a
av bv
etc.
215
when
Fourier's Constants are
FOURIER'S SERIES
216
[ch. vit
The most important thing about Fourier's

f(x) satisfies
sum
Series
very general conditions in the interval
is
(
when
that,
- 7r,
it),
the
of this series is equal to f(x), or in special cases to
H/(*+o)+/(*-o)],
when x lies in this interval.
If we assume that the arbitrary function
f(x), given in the
interval (
it, it), can be expanded in a trigonometrical series of
the form (1), and that the series may be integrated term by term
after multiplying
both sides by cos nx or sin nx, we obtain these
values for the coefficients.

For, multiply both sides of the equation
f(x)
=a
+(a 1 cosx+b 1 sinx) + (a 2 cos2x +b 2 sin2x) + ...
-ir^Xîr,
by
cos nx,
and integrate from -
ir
to
f(x) cos nx dx == ira n
Then
.cos
since
J
when m, n
mx cos nxdx=\
sin
IT
are different integers,
fcos
(3)
it.
mx
cos nx dx = 0,
IT
and
nxdx = 7r.
-IT
Thus we have
1 (
a n = -\
And
f(x') cos nx' dx',
when n ^
similarly,
1 f"
b n =-
7Tj
f(x') sin nx' dx'
ao=%- ._/(*')
<**'
Inserting these values in the series
(3),
the result
may
be
written
-TT^X^TT
This
If
is
(4)
the Fourier's Series ioif(x).
the arbitrary function, given in
x,
-*), is
an even function
~FOURIER'S SERIES
90]
if f(x)
=f( becomes the Cosine Series
in other words,
x),
217
<x<tt the
when
Fourier's Series
f""
= -\
f( x )
f{x')dx'
^h
"
+ - y^cosnx\
*r
fix') cos nx' dx'
Jo
O^X^TT
Again,
it
if
0<<7r the
i.e.
if
f(x)= -/( -x), when
Fourier's Series becomes the Sine Series
=-
/()
an odd function
is
y) sin woj
(5)
/(') sin nx' dx',
0^x~tt
(6)
Jo
717
The expansions in (5) and (6) could have been obtained in the
same way as the expansion in (3) by assuming a series in cosines
only, or a series in sines only, and multiplying by cos nx or sin nx,
as the case may be, integrating now from
to ir.
Further, if we take the interval (
I) instead of ( - x, it),
1,
we
find
and
(6)
the
following
expansions,
corresponding to
(4),
(5)
~[ fV)M+\%l
/(*)
n
Ax')C0B ^(x'-x)dx'
-l^x^l
/(*) =
f(x)
{/(*')
W+jX cos
x /(*') cos
= ^sm-J-xfy(x')sm~~x'dx',
(7)
f x' dx',
O^x^l
(8)
O^x^l
(9)
However, this method does not give a rigorous proof of these

very important expansions for the following reasons
:
(i)
We
have assumed the
function in the
(ii)
possibility of the expansion of the
series.
We
have integrated the series term by term.

This would have been allowable if the convergence of
the series were uniform, since multiplying right through
by cos nx or sin nx does not affect the uniformity but
this property has not been proved, and indeed is not
generally applicable to the whole interval in these ex;
pansions.
(iii)
The discussion does not give us any information
as to
the behaviour of the series at points of discontinuity,
FOURIER'S SERIES
218
if
such
nor does
arise,
conditions to which /(as)

is
give
it
[ch. vii
any suggestion
must be subject
if
as to the
the expansion
to hold.
Another method of obtaining the coefficients, due to Lagrange,*

may be illustrated by the case of the Sine Series.
Consider the curve
y = a1
We
can
sinx+a 2 sm2x + ...+a n
_.
sm(n-l)x.
obtain the values of the coefficients
a2
a lt
...
#-i>
of the curve
so that this curve shall pass through the points
y=f( x )>
at which the abscissae are
In this
way we
find a l9 a 2
to the limit as n->oo
the infinite series
n
,
...
a n . x as functions of n. Proceeding
values of the coefficients in
we have the
f(x)
= a 1 sin x + a 2 sin 2x + ...
an infinite
But this passage from a finite number of equations to
can
results
the
before
examination
number requires more complete
be accepted.
The most
satisfactory
method
of discussing the possibility of
interval ( expressing an arbitrary function /(z), given in the

series
by the corresponding Fourier's Series, is to take the
a +(a 1
cosx+b 1 sinx)+{a 2 cos2x+b 2 cos2x)+...
tt, tt),
and sum the

where the constants have the values given in (2),
limit of this
the
find
to
then
nx)
sin
terms up to (a n cos nx +b n
;
sum,
if it
In this
has a limit, as n->co
way we
shall
show
that,
when f(x)
satisfies
very general
converges to f(x) at every

conditions, the Fourier's Series for f(x)
that it converges to
point in ( - tt, tt), where f(x) is continuous
;
[f(x+0)+f(x-0)]
at
every point of ordinary discontinuity;
+0) +/(tt - 0)] at
x= tt, when
periodic in x with period 2tt,
when the sum
also that it converges to J[/(
tt
these limits exist.
Since the series

is
known
in
is
tt, tt), it is
-Lagrange, (Euvres,
also
known
1 (Paris, 1867), 553:
for every value of x.
Byerly, Fourier's Series, etc. (1893), 30.
FOURIER'S SERIES
90, 91]
219
in which f(x) is
the values of the coefficients in the corresponding
more convenient to take the interval
If it is
defined as
(0, 27r),
expansion would be
2*
I*
o=2^J o /(^)^
1
an = It
2ir
f(x') cos nx' dx' >
1 C
= -\
bn
/
,
2rr
f(x') sin nx' dx',
n^
1.
need hardly be added that the function f(x) can have different
analytical expressions in different parts of the given interval.
And
we can obtain any number of such expansions

will hold in the interval (0, ir), since we can give f(x) any
value we please, subject to the general conditions we shall establish,
in particular
which
( - x, 0).
The following discussion of the possibility of the expansion
of an arbitrary function in the corresponding Fourier's Series
depends upon a modified form of the integrals by means of which
in the interval
Dirichlet* gave the
first
rigorous proof that, for a large class of
With the help
functions, the Fourier's Series converges to f(x).
Second Theorem of Mean Value the sum of the

be deduced at once from these integrals, which we
of the
series
can
shall call
Dirichlet's Integrals.
91. Dirichlet's Integrals (First Form).
Mm
J^W
0<a<6.
When we apply
sin
Km
<**=!/( +0),
f /(,)
/* <&=<),
where
integral
we
.,
see that
p'sinz
I
Jfe'
where
b'
'sinz,
sin
o )h
ic
f*
.
,,
Mean Value
to the
0<6 <c,
dx,
dx = r
x dx
1
-,
c'
f
I
sin
x dx,
c Jj
>
~ ~ c'.
m,
Thus
sin X
J//
-,
dx
-~
/1
1\
\b
cJ
,,+
-,
* Journal fiir
152.
Math., 4 (1829), 157, and Dove's Repertorium der Physik,
1 (1837),
FOURIER'S SERIES
220
[ch. vii
It follows that the integral
sin
is
Its value
convergent.
The
integrals
88 to be
tt.*
J^dx
x
and
J q
ax
has been found in
^ dx
&m
Io
J a
can be transformed, by putting px^x' into

,
sin
f^
sin x
J pa
dx,
f*o
ax,
respectively.
It follows that
^^^
0<a,
f^sinz,
ax ,
dx = \ir,
dx=\
X
X
b8
ba
\im\ ^dx=\im\*
x
x
fi->co J
^ôo J a
a
lim
im->oo
sin
-.
Jo
^dx =
and
'
0<a<b.
0,
fx.a
These results are special cases of the theorem that, when f(x)
given below,
satisfies certain conditions,
limÂxf-^^dx^tt+O),
L
X
0<a,
Môo Jo
b
]im\ f(x)^J^dx=0,
0<a<b.
In the discussion of this theorem we shall, first of all, assume

that /(a) satisfies the conditions we have, imposed upon <j>{x) in
viz., it is
our notation for the Second Theorem of Mean Value
in the
integrable)
therefore
(and
monotonic
and
to be bounded
;
interval with
which we are concerned.
sm
It is clear that
satisfies
the conditions imposed upon
theorem as proved in 50. 1. It is bounded and

integrable, and does not change sign more than a finite number of
in the
\!/(x)
times in the interval.
We
I.
shall
remove some
of the restrictions placed
Consider the integral
?
J
See
also the footnote
a.
on
*JZta,
p. 202.
0<a<b.
upon/(z)
later.
FOURIER'S SERIES
91]
From
221
Mean Value
jW-H^? <&=/( + o) ^ ix +/(& - 0)

where |
is
some
Since/(ic) is
x in a
definite value of
^ ^ 6.
a;
a^x^b, the limits /(a +0)
monotonic in
&,
and/(6 -
0)
exist.
And we have
hand
are zero
seen that the limits of the integrals on the right-
when
II.
nx
! dx = 0, when 0<a<6.
sin
lim
ix-^rao
/^->oo
under the conditions named above,
It follows that,
f(x)
Ja
3J

j.,
sin
ax
*
f f(x)^dx,
v
0<a.
Put f(x) = cf>(x) +/( +0).

ss x ^ a.
monotonic in
Then
<p(x) is
Also
[7(a)
As
We
The
monotonic and
&=/( + o)
//->oo the first integral
shall
now show
<p(
+0)
limit /(
exists, since f(x) is
+0) = 0.
-i^
dx
+ f # (x>
<fa.
on the right-hand has the limit
\ir.
that the second integral has the limit zero.
To prove this, it is sufficient to show that, to the arbitrarynumber e, there corresponds a positive number v such that
positive
'*fc)Mf* < e
Let us break up the interval
where a is chosen so that
when
(0, a)
into
/u,
two
parts, (0, a)
and
(a, a),
|0(a-O)|<
We
there
can do
is
we are given that ^>(+0) = 0, and thus

number a such that
this, since
a positive
\<j>(%)\<
e/2-7r,
when 0<icâ.
Then, by the Second Theorem of
^M
since
<j>(
/2x.
Mean
Value,
w
fc=#(a _-o,j
+0) = 0, | being some
M^
definite value of
a;
in
^zâ.
FOURIER'S SERIES
222
[oh. vii
But, in the curve

sin
a;
the successive waves have the same breadth and diminishing
and -w is greater than that

amplitude, and the area between
between ir and 2ir in absolute value: that between 7r and 2-w is
greater than that between 2tt and 37r, and so on; since |sin x\ goes
through the same set of values in each case, and \jx diminishes
as x increases.
^-dx^\
Thus
Jo
whatever positive value x

f
Also
sin x
than
3'
sin
ax,
positive
is
sinic
dx <7r, when
It follows that
^ dx
<p(x)
and
independent of
But we have seen
7T
fi.
dx=O
<p(x)
Therefore, to the arbitrary positive
number
0^p<q.
in (I) that
8mfi
lim
0<a<a.
number
Je,
there corresponds
such that
dx <e,
<p(x)
when
[i
j/.
i:
Also
\Ca
sin ax ,
sin ax
w sin wx
$(x)S-dx\^\\f tfx)f-dx\+\\f $(x)--^dx
.
Therefore
a
r/>(z)
)
Jo
sin wcc
-,
!-dx\<\e+\
<e, when
Thus
less
<J-X7T
a positive
0<p<q.
ff
this is
a# -
on the right-hand
Therefore
and
of the integrals
for
ir
( sin
have.
aaj=
}v
and each
may
dx<ir
lim
d(x)
/-t
Sm
i/.
^ dx=0.
223
FOURIER'S SERIES
91, 92]
And,
finally,
under the conditions named above,

S
Km\*f(x) J*?dx=?f(+0).
A
x
M _+ x Jo
we have assumed that f(x) is

We
intervals
in
the
monotonic
(0, a) and (a, b).
and
bounded
relaxed.
somewhat
be
may
restrictions
these
that
shall now show
the
I. Birichlefs Integrals still hold when f(x) is bounded, and
interval of integration can be broken up into a finite number of open
partial intervals, in each of which f(x) is monotonic*
This follows at once from the fact that under these conditions
92. In the preceding section
we may
write
= F(x)-G (x)
f( x )
where F(x) and G(x) are
positive,
creasing in the interval with which

or
bounded, and monotonic inwe are concerned [cf. 36. 1
36. 2].
This result can be obtained, as follows, without the use of the theorems of
36. 1
or 36. 2.
Let the interval
(0, a)
be broken up into the n open intervals,
K,
(0, c^),
in each of
which f(x)
Then, writing a
Jo
The
first
integral in this
the limit zero
when
It follows that,
a 2 ),
..., {a n _ lf a),
bounded and monotonic.

and an =a, we have
is
//,-^qo
r = l-r-i
sum
has the limit |/(
+ 0), and
the others have
under the given conditions,

lim [f(x)
dx
=|/(
+0),
< a.
M-
The proof
that,
under the same conditions,

b
\im[ f(x)
[XX
dx^0,
0<a<b,
Mis
practically contained in the above.

It will
intervals
is
be seen that Ave have used the condition that the number of partial
is finite, as we have relied upon the theorem that the limit of a sum
equal to the
sum
of the limits.
The integrals still hold for certain cases where a finite number
points
of infinite discontinuity of f(x) (as defined in 33) occur
of
II.
in the interval of integration.
when f{x) is of bounded variation ( 36. 2) in the interval, since

bounded variation, it can be replaced by the difference of two positive,
bounded and monotonic increasing functions.
*They
if
f(x)
is
also hold
of
FOURIER'S SERIES
224
We
when
shall suppose that,
[ch. vii
arbitrarily small neighbourhoods of
these points of infinite discontinuity are excluded, the
up
interval of integration can be broken
into a finite
partial intervals, in each of which f(x) is bounded
remainder of the
number of open
and monotonic*
Further, we shall assume that the infinite integral

absolutely convergent in the interval of integration,
is not a point of infinite discontinuity.
We may
take
the case
first
when an
know
we
are given that
that
f(x)
And
this
f(x)
is
sm ^ x
and the same
But
j
J\a
s/
(I)
ux
sin
serves for
r\
m^
And, by
dx=
ry
0<= v
\<h> when
values of
all
uniform.
f( x )lT dx
we
^_ m
To the arbitrary positive number e there corresponds

number rj, which we take less than (b-a), such that
IP
and
dx,
absolutely convergent,
is
is
discontinuity-
Ja
^ dx also converges, for

x
convergence
that x
smuX
J a
(a, b).
f(x) dx
and
infinite
occurs at the upper limit b of the integral

only there.
Since
\f(x) dx
(x)
(1)
ju.
^
X
J a
a positive
dx+
J\b~7]
f(x)
^
X
dx
(2)
above,
limp-yêô.
X
>~Ji
ju.
(t
It follows that there is a positive

*
6_
sin
V
f( x )^ dx
ua?
(1), (2)
and
|j|/
number
such that
<\ r when ju^v

,
(3)
*>
From
(3),
we have
fc
at once
<le+i .
|
<e, when fi^
Thus we have shown
f(x)
Ja
^dx =
sin
lim
m-^-jo
v.
that, with the conditions described above,
ux
*The integrals still hold when the function

mainder of the interval of integration.
0<a<b.
0,
is
of
bounded variation
in the re-
225
92, 93]
FOURIER'S SERIES
argument applies to the case when an
similar
infinite dis-
only
continuity occurs at the lower limit a of the integral, and
there.
When
there
is
infinite discontinuity at
an
-^
fV(x)
Ja
fa= [7(a)
J a
33
a and at
b,
and only
from these two, since
there, the result follows
^*
x
P/Iî,
^
<*<*.
discontinuity occurs between a and b we

proceed in the same way; and, as we have assumed that the
number of points of infinite discontinuity is finite, we can break
When an
infinite
up the given
interval into a definite
number
of partial intervals,
to which we can apply the results just obtained.

Thus, under the conditions stated above in (II),
b
Um\ f(x)^^dx = O when 0<a<b.

i
Further,
we have assumed that x =
into
and
satisfies
not a point of infinite
Thus the interval (0, a) can be broken up

and (a, a), where f(x) is bounded in (0, a),
discontinuity oif(x).
two
is
intervals, (0, a)
the conditions given in
(I) of
this section in (0, a).
It follows that
liml" f(xf^-^dx=Z- /( +
jn-wc
and we have
just
Jo
0),
shown that
M ->oo J
(X
Therefore, under the conditions stated above in (II),

a
\im\
^-xx)
si
f(x)
Jo
which we have obtained

92 can be conveniently expressed in terms of what we
93. Dirichlet's Conditions.

in 91,
= ^f( + 0).
^^dx
X
The
results
shall call Dirichlet's Conditions.*

* If the functions of bounded variation of 36. 2 are included in the class of
Conditions
functions available for discussion, f(x) may be said to satisfy Dirichlet's
(ii) when it has a
or
interval;
whole
the
in
variation
bounded
of
it
is
when
(i)
and it is of bounded
finite number of points of infinite discontinuity in the interval,
neighbourvariation in the remainder of that interval, when the arbitrarily small
hoods of these points have been excluded

1
provided that the
infinite integral
FOURIER'S SERIES
226
[ch. vii
function f(x) will be said to satisfy Dirichlet's Conditions* in
an interval (a, b), in which it is

of the two following conditions:
(i)f(x) is
into
it is
subject to one
bounded in
finite
which f(x)
(ii)
when
defined,
is
(a, b), and the interval can be broken up

number of open partial intervals, in each of
monotonic;
f(x) has a finite number of points of infinite discontinuity in

the
interval.
When arbitrarily small neighbourhoods of
these points are excluded, f(x) is
of the
interval,
number of open
monotonic.
and
bounded in the remainder

can be broken up into a finite
this
partial intervals, in each of which f(x) is
Further,
the
integral
infinite
f(x)dx
is
to
Ja
be absolutely convergent.
We may now
When
and
say that:
f(x) satisfies Dirichlet's Conditions in the intervals (0, a)
(a, b) respectively,
where
0<a<b, andf( +0)

^
,u->ao
x->oo J o
and
lim
It follows
sin
I*
f(x)
then
"
^ dx =
from the properties
that except at the points,
exists,
of
0.
monotonic functions
(cf
34)
any, where f(x) becomes infinite, or

oscillates infinitely, a function which satisfies Dirichlet's Conditions, as defined above,
But we have not assumed

a finite
which
is
number
if
can only have ordinary discontinuities.

% that the function f(x) shall
have only
bounded function
can have an infinite number
of ordinary discontinuities.
monotonic in an open interval
of ordinary discontinuities in that interval [cf. 34].
Perhaps
*But
it
should be added that the conditions which Dirichlet
see footnote on p. 225.
fThe conditions
in the text can be further
extended so as to include a finite number

which the function is
bounded [e.g. sin l/(a--c) at x = c], or of continuity, with an infinite number of
maxima and minima in their neighbourhood [e.g. (x-c) sin \j(x-c) at x = c].
This generalisation would also apply to the sections in which Dirichlet's Conditions are employed.
of points of oscillatory discontinuity in the neighbourhood of
^The same remark applies to the case when

variation.
f(x)
is
a function of bounded
FOURIER'S SERIES
93, 94]
himself imposed
upon the function f(x)
227
in a given interval (a, b)
were not so general as those to which we have given the name

He contemplated at first only bounded
Dirichlet's Conditions.
number of ordinary disnumber of maxima and minima.
functions, continuous, except at a finite

continuities,
and with only a
finite
Later he extended his results to the case in which there are a finite
number of points of infinite discontinuity in the interval, provided
Cb
f(x) dx
that the infinite integral
is
J a
Form).
94. Dirichlet's Integrals (Second
where 0<a<6<7r.
In the discussion of Fourier's Series the integrals which we shall
meet are slightly different from Dirichlet's Integrals, the properties
of which we have just established.
The second type of integral and this is the one which Dirichlet
himself used in his classical treatment of Fourier's Series is
J o
sin x
Ja
smx
where 0<a<b<7r.
We
shall
When
now prove
that
f(x) satisfies Dirichlet's Conditions (as defined in 93)
in the intervals
andf( + 0)
(0,
and
a)
b)
(a,
respectively,
where
0<a<b<7r,
exists, then
fi.-+ J
and
lim
fix)
J a
Let us suppose that f(x)

given in
f(x)
up
SllijUX
r
smx
.
dx = 0.
satisfies
the
93 as Dirichlet's Conditions
first* of
the two conditions
bounded, and the intervals (0, a) and (a, b) can be broken

number of open partial intervals, in each of which
monotonic.
is
into a finite
f(x)
is
*Or, alternatively, that f(x)

(a, b).
is
of
bounded variation
in the intervals (0, a)
and
FOURIER'S SERIES
228
Then, by
or
36. 1
we can
36.
f(x)
[ch. vii
write
= F(x)-G(x),
where F(x), G(x) are positive, bounded and monotonic increasing

in the interval with which we are concerned.
mi
.sinux
r
Thus
But
(0, a)
f(x)
J
ic/sin
or
sin x
may
F (x)
v
'
sm x
and G (x)
from
fJL
Hsin^x
^-.
-.
sm x
-i
-+y
-.
siM
are both bounded, positive

r
in the interval (0, a) or (a,
be, provided that
It follows
G(x)
-.
sm x
when 0<a<6<7r.*
and monotonic increasing

case
>
bounded, positive and monotonic increasing in
is
(a, b),
Therefore
r-,,
=\ F(x)
L
b),
as the
0<a<6<7r.
91 that
sm X
j[)
= ?/(+0),
Km
and
Next,
and
+0)
dx = 0, when
let /(a?) satisfy
let /(
We
f(x)
the second f of the conditions given in
lim [/(a)
93,
z/sin x
dx 0, when
(-
is
that
92, II,
sm ic
For we are given that
f(x)dx
is
0<a<6<7r.
absolutely convergent, and

Cb
It follows that
x
fix)
J
is
exist.
can prove, just as in
we know that
0<a<b<7r.
b).
dx
-.
sm
in (a,
a?
absolutely convergent; and the preceding proof
[ 92, II]
applies
to the neighbourhood of the point, or points, of infinite discontinuity,
We
when we
write f(x) -
assign to xjsin x the value
sin x
1
in place of /(#).
at x = 0.
is of bounded variation in the remainder of the

f Or, alternatively, that f(x)
interval, when the arbitrarily small neighbourhoods of the points of infinite dis-
continuity are excluded.
FOURIER'S SERIES
94]
lim
Also, for the case
>y?
/a
we need
and
f(x)
Sill
J q
229
dx,
up the interval (0, a) into (0, a)

bounded in (0, a) and from the results
only, as before, break
(a, a),
where f(x)
is
we have already obtained
in this section the limit
found as
is
stated.
obtain the second form of Dirichlet's Integrals for the cases
If it is desired to
stated below without the use of the theorems of
may
proceed as follows
(i)
36.
or
2 the reader
36.
Let f(x) be positive, bounded and monotonic increasing in
(0, a)
and
(a, b).
Then
is
sin x
But, by
and
so also,
(lAx)
i
i
\
=f(x)
j \ i
sm x
is
so also,
91,
\
im
ft>x>
s
' = Zf(+0).
( $( x J^dx = l<f>(+0)
2 r\
2
)
'
.'0
Therefore
lim
i,
<**=!/( +o),
V(*)lf
X
_
bill
_*.cn In
Also
b
!5_?fc = \ f(x)
Therefore
(ii)
lim
J^L dx _ rf{x) !!!_<&

s
/(x)^'-i dx=0.
Let/(x) be positive, bounded and monotonic decreasing.
Then for some value of

and monotonic increasing.
the function
-f(x)
is
positive,
bounded
Also
"
ri
sin
Using
(iii)
(i),
a;
sin ua:
sin fix
1
sm
sin
J
a;
/xa:
sm
a;
the result follows.
bounded and monotonic increasing, but not positive all the

by adding a constant we can make it positive, and proceed as
and a similar remark applies to the" case of the monotonic
If f(x) is
time,
in
(ii)
decreasing function.
(iv)
When f(x)
bounded and the interval can be broken up into a finite

open partial intervals in which it is monotonic, the result
follows from (i)-(iii).
number
(v)
And
is
of
if f(x) has a finite number of points of infinite discontinuity, as

stated in the second of Dirichlet's Conditions, so far as these points
are concerned the proof is similar to that given above.
FOURIER'S SERIES
230
[ch. vii
Proof of the Convergence of Fourier's Series. In the opening

we have given the usual elementary, but
quite incomplete, argument, by means of which the coefficients in
95
sections of this chapter
the expansion
f(x)
= a + (a x cos x + b x sin x) + (a2 cos 2x + b 2 sin 2x)+...
7r 7L
X~ir
are obtained.
We now
we approach
return to this question, which
in quite
a different way.
We
take the Fourier's Series
sm
a o+( a i cos x +î

1 f
= o~
where a
1 C
a n =-\
7TJ
We
We
f(x')dx' and,
2x+6 2
sin 2x) +...
when n^l,
f(x') cos nx' dx',
6n
=1
sum
^
f
f(x
7TJ
-n
find the
sin nx' dx'
_-
terms of this
of the
and we then examine whether
sin nx,
nôo
x ) +( a 2 cos
ir
series
up
to cos
sum has
this
nx and
a limit as
shall
prove that, when f(x)
is
tt, tt),
and satisfies Dirichlefs Conditions in that interval* this sum has a

It is equal to f(x) at any point in -ir<x<ir,
limit as nâo
.
where f(x)
is
continuous
and
to
i[/(*+o)+/(*-)]>
when
there
is
an ordinary
ify^^ + OHyV-O)]
f(
it
discontinuity
x=ir, when
at
at
the
point
the limits f(ir
and to
-0) and
0) exist.
Let
s n (x)
where a
=a
av bv
Then we
s n (x)
+(a 1 cos x +b
etc.,
sin x) +...
-\-b n
sin nx),
have the values given above.
find,
without
=~ T
f(x')[l
=--f*
/(
'
difficulty,
bounded variation
that
+ 2cos(x'-x)+...+2
sm
\(x
cos n(x'-x)]dx
- x)
* In this and later sections, when reference

be understood that these can be modified,
functions of
+(a n cos nx
is
if
made
to Dirichlet's Conditions,
desired,
by the introduction
as explained in the footnote on p. 225.
it
will
of the
FOURIER'S SERIES
95]
sinU2n + lH X '-x)
dx
sin l(x - X)
f_/
2ir)
sin |(2n
If*
2ttjV
231
,
+ !) -a)
sin |(x -a;)
Thus
il^****
on changing the variable by the substitutions x' - x + 2a.

If -ir<x<7r and /(#) satisfies Diri chiefs Conditions in the
interval
integrals
(-x,
of
x), f{x
(1)
+ 2a)
considered as functions of a in the
Conditions in the intervals
Dirichlet's
satisfy
\ir-\x) respectively, and these functions of

a have limits as aÔ, provided that at the point x with which we
are concerned /(# +0) and/^-O) exist.
It follows from 94 that, when x lies between - ir and it and
(0,
\k +\x) and
(0,
f(x - 0) and f(x +0) both exist,

lim
s n (x)
= l }f(x - 0) +lf(x +0)1

\
= *[/(*-0) +/(*+<>)],
giving the value f(x) at a point where f(x)
We have
yet to examine the cases x
In finding the
value for x in
ml _
sum
a?
= r, we must
insert this
before proceeding to the limit.
s n (x)
.
Thus
of the series for
continuous.
is
x.
S n (7r)
4 a X-
.sin(2^+l)a
1 f" -,
= -\
f{ir- 2a)
,-
since the second integral in (1)
<fo,
sin
7tj o
is zero.
It follows that
n
tt)
= -1 f
vSin(2w+l)a
o
/(V-2a)
*r/
sin
7T J o
da
,
sin(2w+l)a
7T
.1 ff
sin a
= i r j/(^2a)
sin ' 2 "
+
where |
is
sin
1
7T
+1)a
sina
ttJ
fV(-x+2a)
JO
any number between
and
tt.
&
(2w+l)a
a a,
sua.
FOURIER'S SERIES
232
We
can apply the theorem of
satisfies Dirichlet's
/(
- 7T +0) exist.
Thus we have
Conditions in
lim
sn
[ch. vii
94 to these integrals,
x,
and the
ir),
W = K/( ~ * +) +/(*"-
A similar discussion gives the same
value for the
if
f(x)
limits /(7r
0),
0)].
sum
at x = -tt,
which is otherwise obvious since the series has a period 27r.

Thus we have shown that when the arbitrary function f(x) satisfies
Dirichlet's Conditions in the interval ( tt, it), and
U\jw
a =
dx' and,
f(x') cos wx'
TT J _
when n ^ 1.
6n
rfx',
= I
f(x') sin
was'
dx'
7T J -Ti-
tt
the Fourier's Series
a +(a 1 cosx
+ 6 1 sinaj) +(a 2 cos2x +b 2 s'm2x) + ...
converges to
i[/(*+'0)+/(*-0)]
at every point in
at
x=
7r
-7r<x<7r ivheref(x+0) andf(x-0)
exist
and
converges to
it
i[/(-^+0)+/(x-0)],
when /( There
7r
+0) andf(ir -
of course,
is,
by the same
0) exist*
no reason
why
analytical expression in
the arbitrary function should be denned

all
the interval
[cf.
Ex. 2 below].
sum the series, and then let x

approach a point of ordinary discontinuity x we would obtain f(x + 0) or
f(x -0), according to the side from which we approach the point. On the
other hand, if we insert the value xQ in the terms of the series and then sum
the series, we obtain Hf(x + 0) +f(x -0)].
We have already pointed out more than once that when we speak of the
sum of the series for any value of x, it is understood that we first insert this
value of x in the terms of the series, then find the sum of n terms, and finally
Also
should be noticed that
it
if
we
first
obtain the limit of this sum.

Ex.
Find a
series of sines
- tt
e x in the interval
2 sinh
and cosines
<x<
of multiples of
-7r,
will represent
ir.
tt
*If the reader refers to 101, he will see that,

(
x which
7r)
s nv
n( x
(V,.
)=-\
/(a?
ttJo^
'
In this form
interval
if
f(x)
defined outside the interval
is
by the equation f(x + 2ir) =f(x), we can replace

-
we can apply
ir,
ir),
lR
-da + -\
,sin(2w + l)a
-2a)
\.
sin a
the result of
,Tr
^/
f(x
t.'o
(1)
by
,
o \Sin(2ra + l')a
2a)
-da.
sma
94 at once to every point in the closed
except points of infinite discontinuity.
233
FOURIER'S SERIES
95]
What
is
sum
the
f{x) =-^-e*
Here
Integrating
by
x= -1
of the series for
and
an
=^L f
e^ coswx(Za;
-
Therefore
we
1.
parts,
(l+w 2 )l
Also
*S
cos nx to,
f_
Similarly,
7r
-e- )cosn7r.
7r
(-l) n
n~ l +n
=h
find
= (e
2 sinh
ex
(-l) n
1+w 2
sin waj
dx
"1
.
Therefore
7T
_ +
2
2 sinh
7T
-t
when - 7T < x <
When =
a:
+ 12
COS X
+ jp sinz
tt cos
~~ 2* - ^-^ sin 2x
+2
,l+2 2
. .
it.
ir, the
sum
of the series is
/( -
77
W coth
rr,
+ 0) +/(7T - 0) = TT COth
since
7T.
FIG. 1G.
In Fig.
V
y
16,
the curves
e^,
r-r:
2 sinh
7T
x + tt sin
cos ~
1+1
^H:v _ itp wo
2
'
are
drawn
for the interval
- tt,
a;
+ (...) +
J-J32
cos 3 *
+ y+3 2
sin 3 *
tt).
It will be noticed that the expansion
we have obtained converges very
and that more terms would have to be taken to bring the approximation curves [y=sn {x)~\ near the curve of the given function in - ir < x < tt.
slowly,
FOURIER'S SERIES
234
[ch. vii
At x = 7r, the sum of the series is discontinuous. The behaviour of the

approximation curves at a point of discontinuity of the sum is examined in
Chapter IX.
Ex.
Find a
2.
sent/^)
and cosines
series of sines
<x<
- it
in the interval
Here
f*
= iTTX,
<X<
dx =
\ttx
7T
&
f(x)
=
]
when
-7r
\ 7rx cos
nxdx=
sin
-k^x
*
nxdx= -
- cos
When z=
7r the
sum
(cos W7r
.-
An
^ sin
of the series is \tt 2 ,

Tj-2
g-
Ex.
cos
1),
W7r.
-^
a;
sin
2x +
... ,
<x <tt.
result,
sent x
a;
7T.
tt.'o
Therefore
will repre-
lb
^7rJ
an =
which
-7r<x^0,)
/(s)=0,
f(x)
of multiples of x
when
tv,
3.
Find a
+ x2
in the interval
series of sines
Here
&_
<
and we obtain the well-known
+ ....
of multiples of
+ x2 )dx = -\ x2dx =
we
x which
will repre-
rr.
\[ n
(x
wa; er
=-
fw
IT
-
ir ,
x2 cos
wa; dx,
7Tj
find that
4
o
w2
COS W7T.
n
=-
(x
Tj -r
which reduces to
\[ n
2
(# + # ) cos
TT)-w
Also,
and cosines
and, after integration by parts,
"n
= l+ 32 +
- it < x
= o1
a =
+ x2
sin
nx dx = -
f"
1
a:
sin wa;
da;,
7rJ
2
= - l) n-1 w-.
(
Therefore
x + a; 2 =-~-+ 4
when - 7r < x <
When x
- cos x +
sum
of the series is
^-2
96.
in
-4
^ cos 2a; + - sin 2a; J +
...
ir.
tt the
result that
and
sin an
The Cosine
Series.
=1+
22
7r
2
,
and we obtain the well-known
+ 3"2+ "
Letf(x) be given in the interval
(0, ir),
satisfy Dirichlet's Conditions in that interval.
-tt~x<0 by
defined for
Define f(x)
The function thus
the equation f(-%)=f(x).

satisfies Dirichlet's Conditions in this interval,
ir, ir)
and we can apply to
it
the results of
95.
FOURIER'S SERIES
95, 96]
But
clear that in this case
it is
= J"
235
= ~ f* f{x') dx',
leads to a
f(x') dx'
I"'
/.
an = -
/(x
cos
toa w =
leads
rac'da;'
ir
f{x') cos
nx'rfa;',
*"
and
bn
=1f
Thus the
= 0.
/(#') sin nx' dx' leads to 6
sine terms disappear
from the Fourier's Series
for this
function.
Also,
from the way in which f(x) was denned in -tt^x<0,
we have
[/(
+0 +/(-0)]=/( +0),
)
and
-H/(-^+0)+/(7r-0)]=/(7r-0),
provided the limits /( +0) and/(7r - 0) exist.
In this case the
X = 7T
sum
x=
of the series for
is/(+0), and for
/(7T-0).
it is
It follows that,
when f(x)
an
is
arbitrary function satisfying
Dirichlet's Conditions* in the interval (0,
tt),
the
sum
of the Cosine
Series
Vcosra
f(x')dx'+
tt
7rJ
is
f(x') cos nx' dx'
Jo
\ [f{x +0) +f(x
equal to
and
and, whenf(+0) andf(ir-Q)
at every point between
tt
- 0)]
where f(x+0) and f(x-0) exist;

sum is f(+0) at x = and
exist, the
at X=7T.
Thus, when f(x) is continuous and satisfies Dirichlet's Conditions in the interval (0, tt), the Cosine Series represents it in
/(7T-0)
this closed interval.

Ex.
1.
series of cosines of multiples of
Find a
the interval
will represent
ao = ~\ xdx = %Tr,
Here
an
and
x
Therefore
Since the
x which
(0, it).
sum
|- -
xcosnxdx=x+
j^cos
^ cos
of the series is zero at
8
*See footnote, p. 230.
-l +
32
3a;
(cos
. .
.."],
mr
-1).
%x^
tt.
x=0, we have again
52
....
x in
FOURIER'S SERIES
236
In Fig.
17,
= x 7=îr,\
y = x,
the lines
[CH. VII
y=-x,
-7r~x^0,j
and the approximation curve
y=n
C0S ^
*
+ 02 C0S X
"*"
K2 COS
^X
7T
"" *"
" X 3=
^
7T-
are drawn.
Fig. 17.
It will be seen
yz=
in the
how
closely this
approximation curve approaches the
lines
whole interval.
Since the Fourier's Series has a period
2ir,
this series for unrestricted values
Fig. 18.
of
x represents the ordinates of the lines shown in Fig. 18, the part from the
ir, it) being repeated indefinitely in both directions.
(
interval
The sum
Ex.
2.
is
continuous for
Find a
in the interval (0,
all
values of

it),
x which
where
= x = ttA
f(x)=lirx,
f{x)=lir{ir-x),
Here
x.
1 ["*"
lw<x
1 [n
iKxdx+
z\
7T
IT JO
%ir{Tr
in
IT.
-x)dx = -^TT*,
will represent f(x)
a n = - ["Ittx cos nx dx
and
7T
an -
-x) cosnxdx
\\-n-
(n
- x) cos nx dx,
Jin
In + cos mr - 2o cos Jmr]

in = -g2 cos ^7T
-^
<j-2 [1
Thus a n vanishes when n

Also
\tt{tt
7T
xcosnxdx + U
)o
which gives
+-
Jo
=|
i
237
FOURIER'S SERIES
96]
odd or a multiple
of 4.
/(a;)=^-2ricos2a; + icos6a; +
...],
is
,>:';*
B8S|SSSSgS5SS55
S3
SkkUI^
: :
V:xV:
:::::::r:?::::: ::
mmm
PlKisîfini
W7T
-j2
sin-
O^z^tt.
l:|
::::::::::::
":i:i::::::::i!h:
:::::::::::::::
hliLiyiiiiiilHi
lis m
ltl!!II:nIIIII!ilfli!!;I!l!l!!II
(1)
!i
(2)
Fig. 19.
In Fig.
19,
the lines
y = \irx,
y = lw{ir-x)
x~
ir,
and the approximation curves

y = ^\TT 2 -cos2x,
2
J cos 6x,.}.
y Jg 7r - \ cos 2.r are drawn.
It will
be noticed that the approximation curve, corresponding to the
Fig. 20.
terms up to and including cos 6x, approaches the given lines closely, except
at the sharp corner, right through the interval (0, 7r).
For unrestricted values of x the series represents the ordinates of the lines
shown in Fig. 20, the part from - tt to 7r being repeated indefinitely in both
directions
The sum
is
continuous for
all
values of
x.
1
FOURIER'S SERIES
238
Ex.
3.
Find a
in the interval (0,
rr),
[CH. VII
x which
will represent /( x)
where
/(*)=&,
0^*<Jir.j
f(x)=%TT,
\tV<X =
TT.)
11
!lîy!illHi!ii!!i
(1)
M
\ III!
-hjJHj^+rtf
ggj-5
r+rf
Ss i*
5"
4#
44-44-
wtt
Kwi
muni
tefet:
Bffl)
Ittf"
tTTT
(2,
H^fjjJH^:
+tttf
::
wm
:::::::
::::!;:ii!;;is;s:9K::iUH:a::::
flffir
HI"
ii
iiiliiijf
!!"!
iSfflffi
iiiiiiiiiiiii! iiiiiiii
-.
:
mi in sis:::::: ::;:::::::::::
:::: ::
:;
:: ::
(3)
:f:i:::::!:::::::;;i
:::::!::::::: ::H::i:::
flflfltf"
:
fflti]jfr
"TmTrf
:
r|l||
IIIHLil
II
:JJJJJjHJ;J
-
[jjlllfff--
tup, in
TrliT
rrr T
HI!
f nfMii
|]||||||||||:|| !HiS!:
Fig. 21.
If*
Here
7T
Also
Thus
since,
an =
f(x)dx
If*
\w dx = \-n.
f""
\
f*
f(x) cos nxdx=\
cos
nxdx= -
/(jc)=^r-[cosa;--?. cos 3x + i cos
when # = |tt, the sum
6a; -...],
sin Jwr.
0â;=7r,
of the Fourier's Series is [/( \ it
+ 0) +/(|tt - 0)].
FOURIER'S SERIES
96]
From
239
and x = tt, we have the well-known
the values at x =
result,
In Fig. 21, the graph of the given function, and the approximation curves
y = %7T - cos
x,
j
i
y = 7r -co$x + J cos 3a;,
y=zTr- cos x +
0~x~ir
cos 3x -
cos 5x,
are drawn.
The
points
x=0 and x = ir
the point x = %ir
are points of continuity in the
sum
of the series
a point of discontinuity.
The behaviour of the approximation curves at a point of discontinuity,
when n is large, will be treated fully in Chapter IX. It will be sufficient to
say now that it is proved in 117 that just before x = \rr the approximation
is
curve for a large value of n will have a minimum at a depth nearly 0-14 below
y=Q: that it will then ascend at a steep gradient, passing near the point
0-14
(!tt, \tt), and rising to a maximum just after x = \k at a height nearly
above
Ex.
\ir.
4.
Find a
in the interval (0,

ir),
f(x)=0,
lir<xrlir.
f(x)==-l7T,
^"M.
=4
Jo
f*
and
.'o
3n
Thus an vanishes when n
sin
is
rfa;==0
.'*
efc - S
cos na; ~~
n\
= [sin
f{x)
/(r,7r)=-i7r.
/(M=7r,
Also
And
will represent f{x)
^X<lTT,
l-rr<x<^,
f{x)=l7T,
Here
x which
where
?.
n7r
-f
cos wa; da;

71
+ sin W7r]
\mr cos ^mr.
even or a multiple of
= -~- [cos x - } cos 5a: +
J-
cos
7a;
- ,\ cos
3.
1 la;
. .
%, x
.],
<
ir.
The points a; = and a;=7r are points of continuity in the sum of the series.
The points x = \ir and x=.$ir are points of discontinuity.
Fig. 22 contains the graph of the given function, and the approximation
curves
2^3
y=-\- coax
\
>
y = ^ [cos a; - 1 cos 5a:],

o
2V3 [cos
t/
>
a;
cos
5a;
+1
cos
7a;],
1 [cos x-l cos 5a; + } cos 7a; - t\ cos 1 la;],

= 2V3
0;
240
FOURIER'S SERIES
[CH. VII
Fig. 22.
(1)
FOURIER'S SERIES
96, 97]
Ex.
Find a
5.
multiples of x which will represent
series of cosines of
log (2 sin \x)* in the interval (0,
241
tv).
fw
1
= ~\
Here
77
log (2 sin \x)dx
Jo
= log 2 +
fin-
log sin x dx
=0, by Ex.
2 fw
an = -
And
= 4-
cos nx log (2 sin
Thus
log (2 sin |x)
from
this
log (2 cos \x)
|a;)
cite
i"i ff
cos 2nx log (2 sin x) dx
= -i
It follows
4, p. 131.
by Ex.
[cos
5, p. 131.
x + \ cos 2x + % cos 3x +
. .
[cos x - \
cos 1x
+\
cos
3x-
.. .],
when
70.
< # î
=x<
when
These expansions have been obtained otherwise in
on
.],
or may be obtained independentlythat

1.
ir.
tt.
[See footnote
p. 159.]
The Sine Series. Again let f(x) be given in the interval

Define
(0, it), and satisfy Dirichlet's Conditions in that interval.
function
The
equation
by
the
in
x
<0
7r ^
f(-x)= -f(x).
f(x)
97.
- ir, it) satisfies Dirichlet's Conditions

and we can apply to it the results of 95.
thus defined for

interval,
But
it is
bn
clear fchat in this case
If
-
17
f(x') sin nx' dx' leads to h n
and that a n when nÔ.

Thus the cosine terms disappear from
Since
all
in this
~2
Cw
f(x') sinnx' dx',
this Fourier's Series.
biSmx +b 2 sin2x + ...

when x =
vanish
and x = ir the sum
of the series is zero at these
points.
when
It follows that,
f(x) is
an
arbitrary function satisfying
Dirichlet's Conditions^ in the interval
(0, it),
the
sum
of the Sine
Series,
2 x
-
7T
is
equal
This
f(x') sin nx' dx',

Jo
ilf{ x +0) +f(x ~
to
function
y^sin nx
is infinite
fSee footnote, p. 230.
at x
= 0,
but
it satisfies
0)]
FOURIER'S SERIES
242
[oh. vii
and it where f(x +0) and f(x-0) exist;

and x = ir, the sum is zero.
be noticed that, when/(x) is continuous at the end-points
at every 'point between
and, when x =
It will
and x = 7r. the Cosine

The Sine
Series gives the value of the function
Series only gives the value of f(x) at
at these points.
these points
Ex.
1.
interval
if
f(x)
is
Find a series
< x < tt.
Here
of sines of multiples of
bn
Therefore
a;
zero there.
'\
is
will represent
x in the
x&mnx dx - l) n-1 -.
{
=-2 [sing - J sin
At x = - the sum
x which
2a:
+I
sin 3a;-
...],
0~x<rr.
discontinuous.
V
Fig. 23.
In Fig. 23, the
line
y = %-
-tt'^x'~tt,
and the approximation curve

\!y
2 [sin x-\ sin 2x + 1 sin 3a: - } sin 4x + 1 .sin 5x],
-tt^x~
tt,
are drawn.
The convergence of the series is so slow that this curve does not approach
between -tt and tt nearly as closely as the corresponding approximation
y-x
FOURIER'S SERIES
97]
curve in the cosine series approached y =
x.
243
If
is
taken large enough,
the curve y = s n {x) will be a wavy curve oscillating about the line y x from
- it to + it, but we would be wrong if we were to say that it descends at a steep
gradient from x = -
it to the end of y x, and again descends from the other

y = x to x = it at a steep gradient. As a matter of fact the summit of
the first wave is some distance below y = x at x - it, and the summit .of the
last wave a corresponding distance above y = x at x = it when n is large.
To this question we return in Chapter IX.
end
of
Fig. 24.
Since the Fourier's Series has a period

of x represents the ordinates of the lines
open interval (-it,
it, 3it,
points
Ex.
2.
it)
.
lit,
this series for unrestricted values
shown
in Fig. 24, the part
At
are points of discontinuity.
Find a series of sines

= x~it, where
of multiples of
these the
O^X^^lt,
f(x)=iTtx,
J
f(x)=i7t(7t-x), \-^x^_it.
2f 7r
2fi w
=
\t?x
sin
nx
dx
x)
+
l-7r(7r
sin nx dx
&
"if
= \\
which gives
Thus
/(a;)
xsm.nxdx + \\
= sin
a:
rg sin 3a;
+^
y = \itx,
y = lit{it-x),
and the approximation curves
a;,
y = sin
a:
- K2 sm
y = sin x -
-x)
sin
nx dx,
nit
Fig. 25 contains the lines
y sin
(it
sin
3a;,
^ sin 3x +-g sin
5a:
...
0^-a:^l7r,
it
5a;,
sum
The
is zero.
x which will represent f(x) in
the interval
Here
from the
being repeated indefinitely in both directions.
= x ^r-it,
^
1
I
FOURIER'S SERIES
244
It will
closely,
[CH. VII
be noticed that the last of these curves approaches the given
lines
except at the sharp corner, right through the interval.
(2)
IHMttffllfifffflUllinffl lfffi
Fig. 25.
For unrestricted values
shown
in Fig. 26, the part
of x the series represents the ordinates of the lines
from -
it
to
+ tt
being repeated indefinitely in both
directions.
The sum
is
continuous for
all
values of x.
Fig. 26.
Ex.
3.
Find a
in the interval (0,
series of sines of multiples of

tt),
where
f(x)=?0,
0~x<tt,
f{x)=\-K, \tt<x<tt,
/0r)=0.
x which
will represent f(x)
FOURIER'S SERIES
97]
Here
bn
nx dx
sin
2 niv
1 /
= - cos
\
I
2
3W7T
=
sin
.
when n
Therefore 6 n vanishes
And
W7T
sin -r-.
.
a multiple of
is
/(a;)=sin x -sin 2.r+^ sin
3a;
245
sin 5a:
-^
4.
sin
6x+
...
a;t==-7r.
graph of the given function, and the approximation
Fig. 27 contains the
curves
y sin
t/
x,
= sin
a:
-sin 2x,
y = sin x - sin 2x + ^ sin
3a:,
y sin x - sin
3a;
x \k and
2a:
sin
+1
sin
5a:,
x tt
are points of discontinuity in the sum of the

The points
The behaviour of the approximation curves for large values of n at
these points will be examined in Chapter IX.
series.
Ex.
4.
Find a
the interval
series of sines of multiples of
(0, 7r),
<ar<
f(x)=l7r,
Also
x which
will represent f(x) in
where
?.7r,
f(x)=0,
lir<x<f Tt,
f{x)=-\TT,
lir<X<TT.
f(0) =f(7r)
f{ltr)
Here
= }tt
sin wa;
f(fr)
dx -
vr.
sin wa;
r.\
dx
'.'It
= q [1
8
And
/(a:)=sin 2x +
The points x = 0,
sum
a:
when w
sin 4a:
-?>7r,
+ COS
odd or a multiple
of 6.
^/17T
_ ?i7r
T Sm
= 3n COS
Therefore a n vanishes
^?l7T
-COS
is
+ ^ sin Sx
x = tt and
COS
W7r]
mr
T*
l sin 10a:
...
Q~.x1=kiv.
a:=7r are points of discontinuity in the
of the series.
Fig. 28 contains the
graph of the given function, and the approximation
curves
y=sin
2x,
y = sin 2a: + \
y sin 2a: + \
sin
4a:,
sin 4a:
y = sin 2x + \ sin
4a:
+]
+\
sin
8a:,
sin 8a:
+i
sin 10a;
FOURIER'S SERIES
246
(1)
(2)
(3)
V
:
+5i ^f T rmPff
:Hi|
ss iiSii
(4)
s
;:ffi
In
Fig. 27.
[CH. VII
217
97]
Hn
--TT
r|if
58
TTtt
hi. :
nSr
tt
"ffff
Tm
ito
'
SH;
(1)
lp
(2)
//
ji
mi
n-jrr
ffif
tfrr
iiis
I
st?i
xU^ i+tt
Hit HE
g|
Hff
in Hi |S as M SIT aj f
t^j
m 51
55
jtmjTOT
(3)
cr5
gp
fiS
TR?1
tftp
Hi
S5
g:
gr
S
ptf
g:
#5
Hf
s H^~ TH BJ
^H
5j wrr
l+f
^X
-tfrff
ffr"
Px
Ml
Sffl-
-4*
jt:
*fl
TT-
(4)
FOURIER'S SERIES
248
Forms of
98. Other
function
to
is
Fourier's
Series.*
[ch. vii
When
7r, 7r)
In this way we
already obtained
may
deduce the following expansions from those
'
~ x)dx
~l=x
=
''
Whenjf (#)
>
-W
*>
O^x^l
f{x') sin =- x' dx',
Jo
0xrsJ,...(2)
f(x')dx'+j'Zoo8 ^x['f{x')cos^x'dx',
\[
l
Jo
1
Jo
x
T y) sin -fI
arbitrary
this interval
^_/(*W+|Sf_/K) cos t (x
1>
the
- I, I), we can change

by the substitution u = irxjl.
(3)
f>
Conditions in
I,
the
I),
sum
of
+ 0) +/(-0)]
(1) is
where f(x+0) and/(x-0) exist; and at x 1 its sum
![/( - Z +0) +/(Z - 0)], when the limits/( -1+0) and/(Z - 0) exist.
at every point in
equal to \[f(x
the series
-l<x<l
is
When
f(x) satisfies Dirichlet's Conditions in
0<x<l
sum
is
its
sum
is/(Z
- 0), when these
the
sum
of
limits exist.
Conditions in
f(x) satisfies Dirichlet's
(0,
I)
the
sum
of
equal to i[f(x+0) +f(x-0)] at every point in

where f(x+0) andf(x-O) exist; and at x = and x = lita
the series
0<x<l
I),
(2) is
is/( +0), at x l
When
(0,
equal to i[f{x+0) +f(x-0)] at every point in

where f(x+0) and f(x-0) exist; and at x = its sum
the series
(3) is
zero.
sometimes more convenient to take the interval in which

the arbitrary function is given as (0, 2tt). We may deduce the
corresponding series for this interval from that already found
It is
for
-7T,
7r).
Consider the Fourier's Series

n
^-[
where F(x)
Let
F(x')dx'
u = w+x,
Then we obtain the

1
s-
u'
Conditions in
it, ir)~
and f(u) = F(u-7r).
ir-\-x'
series ioif(u),
2,r
f(u')du'
A7TJ0
for the interval (0, 27r).
See footnote, p. 230.
F(x')coan{x'-x)dx',
+- JCT
x^f
+1 2j\
7T
2 ""
Jo
f(u') cos n(u' -u)du',
FOURIER'S SERIES
98]
On
changing u into
we have the
x,
249
series iovf(x),
O^x^Zir
^[ y(x')dx'+~j^\Jo J(x')cosn(x'-x)dx\
^7TJo
7T
The sum
between
and x = 2-7T
(4)
of the series (4) is J [/(a; +0) +f(x-0)] at every point

and 2x where f(x+0) and/(a;-0) exist; and at x =
its
sum
is
H/(+0)+/(2*-0)],
when
In
these limits exist.

(4), it is
the interval
assumed that/(ic)
Conditions in
(0, 2-7r).
This reduces to a Cosine Series
if
/(aj)=/(2x-jc) and to a Sine
Seriesif/(z)=-/(27r-a).
If
the interval
we have
is (0, 1),
instead of
(4)
dx'
\ f(x')
+2 V) P/fc')
Jo
cos 2 *^ (
- a?')
*',
==
a;
^ 1,
....(5)
Jo
a series with period unity.
sometimes convenient to take the interval in which

the function is defined as (a, b). We can deduce the corresponding series for this interval from the result just obtained.
Again,
it is
Taking the
i
we
series
2U
w=a +
write
Then
+- S [*F
F(x') dx'
(b-a)x
2
in the interval
-L P /Y w '\
b-a] aJK
'
On
fa'
changing u into
X ') cos *(' - )<&'.

-,
and f(u) = Fn
,
V PW)
6-aVJ/
a;,
we obtain
CoS
= x = 2?r
f27r(^-a)l
^-
aû^b we have the

+ -JL
>
series for/(w),
_
&- U M
'
) rf
the series for f(x) in
w '.
a^x^b,
namely,
^fV^^+^SjVjcosg^-,)^
The sum
in
its
of the series (6) is
i[f{x+0) +f(x-0)] at every point

exist
arid at x a and x = b
a<x<b where/(z+0) and/(x-0)

sum
is
|[/( +0) +/(& -
Of course f(x)
interval
(a, b).
is
0)],
(6)
when
these limits exist.
again subject to Dirichlet's Conditions in the
FOURIER'S SERIES
250
The corresponding
Cosine
[ch. vii
and Sine
Series
Series
are,
re-
spectively,
b
'
(X,) C0S
W^' - a) dx
T^T \J {x )dx '+6^ ? C S h (x ~ 0) |/
'
a^x^b,
and
-7
b-a 4^
sin -z
(x-a)
b-a
fix
sin
)a
(x
s
b-a
-a) ax
(7)
a^x^b
Ex.
1.
Show
equal to
Ex.
2.
^V
when < x <
Show that the

-=-
when
equal to c
3.
Show
+v a + v
ZlZ^lZJfl
+
v
is
2c /
Ex.
4.
TV
7ra
sm
<x<a
27nc
C0S
-r
-r + -J
<x<
I.
that the series

(
2?wr
sin -^- sin
.
x)
vx
when - Z <
v2
when
v3
when
a;
Tivra;
,_
+ (2v a -i^-*,)
rnrx\
coa -jp*j
< - ^,
l
-o<
a;
<q'
<#<Z.
that the series
L
(7r
sin
and to zero when a
mr n
12 1.
i2_
s in-_ \2{v z -v
Show
+s
27ra
7T.E
cos
-j-
2
represents
...
series
equal to
Ex.
I.
ca
is
(8)
that the series

.TX
3ttx
-x 1
4 f
sin
sm
+ sin -y +
is
''
- x) in the interval
99. Poisson's
3x
_+ sin
"3~ + -J
sin 2a;
~]
sma; + -2
Discussion
< x < 2tt.

of
Fourier's
Series.
As has been
mentioned in the introduction, within a few years of Fourier's

discovery of the possibility of representing an arbitrary function
by what is now called its Fourier's Series, Poisson discussed the
subject from a quite different standpoint.
He began
with the equation

00
-2r
cos (x
-x)
+r
+2y^rn cosn(x' -x),
'
FOURIER'S SERIES
98-100]
where
[<1, and he obtained, by integration,
If
251
l_ r 2
2r cos (x
f{x')ds
- x) +r 2,
7r
f(x')dx'+-y,r n \
=<rT
f(x')cosn(x'-x)dx'.
Poisson proceeded to show that, as
/->!,
the integral on the
left-hand side of this equation has the limit f(x), supposing f(x)
continuous at that point, and he argued that/(se) must then be the
sum
of the series
on the right-hand
when r=l.
side
Apart from
the incompleteness of his discussion of the questions connected
with the limit of the integral as r-l, the conclusion he sought

draw is invalid until it is shown that the series does converge
to
when r=l, and this, in fact, is the real difficulty. In accordance

with Abel's Theorem on the Power Series ( 72), if the series
converges, when r=l, its sum is continuous up to and including
r=l. In other words, if we write
F (r,
we know
x)
that,
if
f(x')
dx
+-
2f ^
f(x') cos
n (x' -
x) dx',
F(l, x) converges, then
= F(l,
lim F(r, x)
x).
r->l
But we have no
right to assume, from the convergence of
lim F(r,
x),
that F(l, x) does converge.

Poisson's method, however, has a definite value in the treatment
now give a presentation of it on

we have followed in the discussion of
and we
of Fourier's Series,
shall
the more exact lines which
and
series
integrals in the previous pages of this book.
The
100. Poisson's Integral.
integral
r
li', N2r ..'s( i '-) + ^ f(X

is
'
)dx ''
r
I
<
'
called Poisson's Integral.
We
shall
assume that f(x)
is
either
in the interval
frr
(
-7r,
or that the infinite integral
7r),
Now we know
that
f(x)dx
_ ,2
=
^
2r cos
is
qo
s-
+ r2
+2
>] r
n cos
nd
when r < 1, and that this series is uniformly convergent

0, when r has any given value between - 1 and + 1.
|
for
any
interval of
FOURIER'S SERIES
252
^277
1-r 2
fir
It follows that
[ch. vii
^
ST- 77-777
dx' =
_^1 - 2r cos fa;' - a;) + r 2
and that
l-r a
1 it
-2rcos(x'-a;)+r
2irl-irl
= o"
/(*') <&'
n
\
2tt) tt
k + ~-
/( a
;/
Tr
2 r
cos n(#' -#)<&
/(a;')
it'
under the limitations above imposed upon f(x).
(Cf.
70.
1,
(1)
Cor. II,
and
74, I.)
Now
sum
let
number x between - 7r and tt for which we wish the

what is the same thing, the value of Poisson's Integral,
us choose a
of this series, or,
27r]_,r l-2rcos(:r'-.r)
+ r2
this sum, or the integral, by F(r, x).

Let us assume that, for the value of x chosen,
Denote
lim [f{x
+/(*-')]
exists.
Also, let the function
be denned when -tt^x'
(f)(x
</>(x')
tv
by the equation
=f(x')-% lim [f(x + t)+f(x-t)].
Then
F(r, x) - I lim
1
I f(x
fir
2tt)
-n
= ^r.
But we
+ t)+f(x-t)]
- 2r cos
r2
[f(x')
-x)+r
(x'
- 1 lim [f(x + 1) +f(x -
l-2rcos(y^) + r^
(X
'>^
lim [f(x + 1) +f(x -
are given that
1)])
dx'
t->0
<2)
1)]
exists.
to ê there will correspond a positive
\f(x
be chosen, as small as we please.

>/ such that
Then
number
+ t) +f(x-t)- \im[f(x + t)+f(x-t)]\<h,
(3)
<
t
when
>/.
The number y
fixed
upon
will
be such that (x -i],x +
?;)
does not go beyond
Then
1
1-r 2
[x+y,
fcr]*-,
-2r cos(z'-:r)+r 2
1
9{X )aX
"^^
= 2tt}o
* \1
+
v
o cos / + r 2 [(f>(x
- 2r
=<rP
-i
2ttJo 1
2
o
-2rcos
+ r {/(*
""^/^
t)
+ <!>(x-t)W
7
Km [/(* + *) +/(*- <)]}#.

+ ') +/(*-*)^
^^
FOURIER'S SERIES
100]
253
It follows that
\2Tr\ x -r
l-2rcos(x'-x)+r
2<t>{X
]< 4v)o l-2rcos* + r 2
1-r 2
^jj*
47TJ _ w 1
Also,
Itt
_-
.!*+,, /
<
l-2r
1-r 2
- 2r cos
27r 1
1-r2
=
s
- 2r
cos
1
x> x
But
77
cos
rv
+ r2
-a;)
(a;'
dt
(4)
'
*"
+ r2
.xi, say
J
+ r2
(5)'
l
'
77
1-r 2
- 2r cos + r 2
<h
< r < 1,
when
dt
2(1 -r)
,<
.
2
...
'
(1
-r)
5
z
l-2rcos7j+r
+ 4r
.,
if0<r<l,
,
.
sin
, t
2
1-r
<
2r sin 2 ^
1-r
a
And
i
2r sin
<
2.4
77
provided that
>
l
+z si^
It follows that
1
/ fz-i
S(
..
1-r2
L J T=S^^T^+^
l>r>
if
(4)
and
,,
>
,1
|
<2
Combining
.,
*(*
(6)
+ isin!
be seen that when any positive number c

we please there is a positive number p such that
a)-* lim \.f(z + t)+f(x-t)]\<e,
(6),
it will
has been chosen, as small as

I
F(r,
when p^lr < 1, provided that for the value of x considered lim [f(x + t) +f(x -t)]
t^
exists.
We have thus
established the following theorem
Letf(x), given in the interval
-it,
7r),
6e
bounded and
absolutely convergent infinite integral, in this range.
x in
integrable, or have an
Then for any value of
~7r<x< tt for which

\im[f(x + t)+f(x-t)]
f->0
exists,
Poisson's Integral converges
to that limit
as r-> 1
from
below.
,:
FOURIER'S SERIES
254
In
particular, at
[ch. vii
a point of ordinary discontinuity off(x), Poisson's
converges to
Integral-
Uf( x + 0) +f{x - 0)]
and, at a point where f(x)
is
continuous,
converges tof(x).
it
no conclusion can be drawn from
It has already been pointed out that
this
as to the convergence, or non-convergence, of the Fourier's Series at this
But
point.
we know that the
if
from Abel's Theorem that
it
Fourier's Series does converge,
must converge to the
it
follows
which Poisson's
limit to
Integral converges as r->l.
We
have thus the following theorem

- -k, it), which
is any function, given in
:
Iff(x)
or has
an
is either
absolutely convergent infinite integral
bounded and
integrable,
f(x) dx, then, at any point x
J IT
in -7r
<x<7r
at
which
the Fourier's Series is convergent, its
lim [f(x
provided that this limit
With
It follows
-7r,
If
- it and
ir
as well as points between -
tt
and
it
made
to apply
immediately from this theorem that

Constants are zero for a function, continuous in
the Fourier's
7r),
be equal to
exists.
the interval
then the function vanishes identically.
the constants vanish but the function only satisfies the conditions
ascribed to f(x) in the earlier theorems of this section,
the function must vanish at

points where lim [f(x
Further,
p,
sum must
-t)-\,
certain obvious modifications these theorems can be
to the points
If all
+ t)+f( x
if (a, b) is
points where
all
+ 1) +f(x - 1)],
an interval
in
corresponding to the arbitrary
in the interval
in (5),
and thus
(a-,
b)
e,
it is
exists, this limit
which f(x)
may
is
we can
only infer that
continuous, and that at
must be
zero.
continuous, the same
be chosen to serve
for this is true, first of the
number
all
number
the values of x
y in (3), then of
finally of p.
It follows that as r-*-l Poisson's Integral converges uniformly to the value
f(x) in
any
interval (a, b) in
which f(x)
is
continuous.*
This last theorem has an important application in connection with the
approximate representation
of functions
by
finite
trigonometrical series.f
101. Fejer's Theorem.f
Let f(x) be given in the interval
*It
is
Also f(x)
tt,
tt).
If bounded,
let it
be
in this that f(a - 0) =f(a) =f(a + 0) and f(b - 0) =/(&) =f(b + 0).
subject to the conditions given at the beginning of this section. Cf.
assumed
is
107.
t Cf.
e
1 (2 ed., 1905), 275; Bocher, Annals of Math., (2),
Hobson, Theory of Functions of a Real Variable, 2 (2nd ed., 1926),
Picard, Traite d' Analyse,
7 (1906), 102;
637.
tCf.
Math. Annalen, 58 (1904), 51.
FOURIER'S SERIES
100, 101]
integrable
unbounded,
if
ilM
TAen a every point x

andf(x-0) exist,
lim
With the above

1
(x)
of the (n +1) terms
/(*') c s
*) *'
++**-!
.
-tt<x<tt
in the interval
cr n
f(x)dx be
= [f(x +0)
at
which f(x +0)
+f(x-0)].
notation,
ir
=o
sum
s n the
a n (x) = *o+*i
integral
infinite
fww t~ s r
let
5n
the
Denote by
2*
let
255
r/
27rJ_/f(v
,,cosn(x' -x)-cos(n+l)(x' -x)

)
y
- cos
(a?
z)
^
7
Therefore
f
;/
(x) = n
rdx
f(x)-,
nW
t~,
l-cos(x
2W7rJ_/
-x)
r
, x
ar n
-2n7r)- n+ /
if
f(x)
is
-cosw(#' -x)
[X)
/(as
(x, 7r+a;),
the
Now
range
by the equation
it, it)
of
integration
x'
x-â
(~w+x,x)
into
in the
first,
ri
2
,sin wa
suppose that
cc
is
a point in
f**
-w,
it)
sin 2 wa
than
\tt
x'
and
= x+2a
to e there corresponds a positive
at which f(x +0)
/(# - 0) exist.
Let e be any positive number, chosen as small as
Then
and
we obtain
i,r
(1)
>
+2x) =/(*).
and substituting
in the second,
dX
sm*i(x'-x)
denned outside the interval
Dividing
'
number
we
?;
such that
|./(*+2a)j-/(a+0)|<e when (Xa^rj.
and
please.
chosen
less
FOURIER'S SERIES
256
[ch. vii
Also
J^
ft*
r/c*>s=*-ir.w^-/f.s?*
n-rr.
in gi
to *
+/t"+4
(**'
n 2 na
/ (x +2a) -r-= a
n7r,
**
sin2 na
da
n-jr
= z 1 +/ 2 +/ 3 + /4>
,
fln
and
3)
C n _ 1 = J+cos2a+cos4a + ... + cos2(n-l)a,
Putting
we have
sa y-*
(cos2(n-l)a-cos2na)
4(Tôs"2a)
(l-CQs2n a ) _ 1
1
r o+
+r
G n-i-2
6
c i + --+ r
_ cos2a -2
(1
_
Cn _ x = \
'
sin2nat
sin 2 a
r^% .2r
Thus
sin2 a
'
c0+ c 1+ ... + c_ 1 )^
= JW7T,
since all the terms
on the right-hand side disappear on integration
except the
each of the G"s.
first in
h = lf(* +)-
It follows that
^ -t\y
Ais
'
(x+2a) - f{x+o)
sin 2 >?a
<
2
sura
wttJo
|/ 3
da
act
(4)
^^
sin^M/Tf
f 1>,T
I^J, l^ +2a)l
<
-,
<le
Further,
>>-
sinâ
n-7rj
sin 2na
t\f(x+2a)\da
+
<,
This
is
any
discussion also applies
positive
number
less
than
*
2
.-w
i/mi^'
when the upper
limit of the integral
on the
left
\ir.
fThis integral can be obtained at once from
(1)
by putting f(x) = l
in
-ir,
tt).
FOURIER'S SERIES
101]
But we
interval
the
outside
and we have
\f(x')\dx' converges,
are given that
denned f(x)
257
by
tt, it)
the
equation
f(x+2 7r)=f(x).
Let
\f(x')\dx'
= 7rJ,
say.
Then we have
|/<*l<
*>l
Om oin
2n
Also
[Z 4 1<
sm^
Combining these
1
l/(x+0)[
results, it follows
sin 2na
//
f'-'
from
(3)
that
w/
sm 2
n-TrJ
(6)
nv
+ \X*+)\}.
<h+ttttV
2w sin
2
?y
Now
let
i/
be a positive integer such that

1
+ l/(^+0)|}<e
'^
2-{^
sin 2 ^
i'
(7)
'
Then
|
sin
"-7TJ q
a.
<e, when wî/.

In other words,
1
M-
1im ^Z
_vc 'iTTj
n
when f(x + 0)
h"
//
bill
>
exists.
In precisely the same

l
Iim-J-(
n _>oo
when /( - 0)
sin 2 na
^
w/ A
/(*+2a)r-Ai = l/(aJ+0),
way we
find that
/(-2)^&
=i/(-0),
U
t
/&7TJ q
Dill
exists.
Then, returning to
(2),
we have
\im<r n (x) = i[f(x+0)+f(x-0)l
when f(x0)
exist.
This proof applies also to the points
x=
ir,
when
/(tt
/(-7r+0) exist. Since we

- 7r, it) by the equation f(x +2?r) =/(#), it is clear that
/(-7r+0)=/(7r+0) and /(-x-0) =/(*-<>).
In this way we obtain
0)
and
have defined f(x) outside the interval
]im<r m (x)=H/(-x+0)+/(x-0)I
when/(-7r +0) and/(7r-0)
exist.
258
Corollary.
when the
FOURIER'S SERIES
[ch. vii
a~x^b,
including the end-points,
continuous in
If f(x) is
number
arbitrary positive
same
chosen, the
is
values of x from a to 6, including the end-points.
will
7/
Then, from
do for
all
follows
(7), it
that the sequence of arithmetic means
sumf(x) in
converges uniformly to the
assumed
It is
is
continuous at x = a and x = b
f(a-0)=f(a)=f(a + 0), and

f(b-0)=f(b)=f(b + 0).
as well as in the interval
(a, b)
Two Theorems on
102.
the interval (a, b).
statement that f(x)
in this
i.e.
the Arithmetic Means.
Before applying this very
we
important theorem to the discussion of Fourier's Series,
shall
prove two
theorems regarding the sequence of arithmetic means for any series

1 + 2 + tt 3 +...
.
In this connection
we adopt
the notation
sn
= u + u 2 + ...+u n
1
Th kokem
I.
sequence of arithmetic means

First,
(i)
we assume
tr
+ y2 + u3
nx
//' the series
+s 2 + ...+s n
_'s 1
(Tn
...
converges
and
sum
its
is s,
then the
n also converges to s.*
that lim
sn
?<->*>
0, and we prove that lim cr n =0.

n*->
Take the arbitrary positive number e.

Then there is a positive integer N, such that
\s
Also
< |e, when n^N.

kv +l +
+ ---+s.vl
a \g K+*2
n
But we can choose
> N,
S
i-S\
1
so that
+ S2
<rn
+ ...+|gw
+S
\
1
<
n
Therefore
kv+-jl
1
i
>
when n ^
t.
\<h + l { l -n)
< when n ~ v.
e>
v.
when n ~
lim
Thus
cr
= 0.
n
(ii)
lim s n =s-Q.
Let
*If the series ^w r diverges
+ 00
to
(or to
00
),
then lim
<r
= + ac (or - 00
may
For, in the case of divergence to + 00 , however large

such that s n >K, when n~^N.
there is a positive integer
be,
).
we know that
on
But
The
(
first
i
)
Sx
S2
part tends to zero
K, which tends to
+ ... + Sv + S.y+i + S\ f 2 + + 8n
n
when n -> 00 and the second part
,when n ->
00
is
greater than
101, 102]
FOURIER'S SERIES
Then
= (ii1 -s)+u z + us + *..
The arithmetic mean for n terms

sn
= 2> r
and thus
- ( 2> n - ns j, where
n V i
J
of this series is equal to
equal to (v n -
it is
259
s).
But by
means
the limit of this sequence of arithmetic
(i)
zero.
is
Therefore limo- n = s.
Theorem
Let the sequence of arithmetic means
II.
ux + u2 + us +
converge
to o-
and
<K
either n{sn >-8 n+1 )
of the series
cr n
or n(s n+1
-s n
< K,
where
K is some
positive constant.
lims n = cr.*
Then
We
may, as above, without

o-=0,
K=
Suppose we are given
and n(s n -sn+x )<\.^
l,
lim
It is clear that lim s n
take
loss of generality,
cr
and
-s n+x
n(s n
not equal to +oo (or -
is
oo
<
1.
because
),
if
this limit
n><x>
were +
(or
oo
oo
we would have
lim
= + oo
(or
oo
[Cf. footnote
).
on
p. 258].
(i)
lim s n = A, where
If possible, let
Then,
But
to
or a finite positive number.
is
> A,
M M M
for
such that
o- n
Let
+oo
any positive number < A,

an infinite number of values of n, say
x
the arbitrary positive number e, there corresponds a
if
sn
A is
M be
<
2,
3,
....
positive integer
/x,
greater than
/x,
when n ^ fi.
c,
M M M
the first of the sequence
lt
2,
3 , ...
which
is
and such that ilO^ an even positive integer.

Let 2p be the largest even positive integer not greater than
A < 4p.
2p ^
Then
MA.
This proof of the Hardy-Landau Theorem is due to Professor A. E. Jolliffe.

This theorem, in a less general form, was given by Hardy in Proc. London Math.
In the earlier edition of this book Littlewood's proof, given
Soc. (2), 8 (1910), 302.
followed.
Cf. also de la Vallee
in Whittaker and Watson's Modem Analysis is
e
loc. tit. (2nd ed., 1926), 151,
Bromwich,
and
93
ed.,
1922),
cit.
loc.
2
Poussin,
(4
;
where further references
tForifweput
we have
be found.
=~^,
S
- <?
n
S n = %->
n(S n
Thus
In the other
will
case,
we put
C7 2
= /<'
,
and
-S n+1 )<l,
etc "
and S n = ^ U r,
1 ^ c _ &n &
~Z^r~
jf~
if
n{s n -s n + x )<K.
Ux = ^, U 2 = ~ T[>
etc
FOURIER'S SERIES
260
Also
5 /+1>5tf
,
__
\
^ 2>Sji+1 -mT\ >Sm ~\m
\M + m+i)
M+l) >s
/ 1
and each
of these
[ch. vii
'i
~m'
3
Sjl+i
>
> &M - jiff
Hjf+p
>
> Sjf - -^,
> A - ~.
"+* "
But
WTp~
> M+p' crjl + p

<,
>^~j(A-a)-e
p\
W+p'\ A
~M)
where a = ^^\A.
Now A-a = \A
and ~ > -r
Q> -J-
__
Therefore
and
If
+
(il
since
^=a>\A.
JxL
_)>__,
, >^4j-.
we take c =
we have
Thus
(ii)
o-.,
/+v,
> e,
which
is
impossible.
lim 5n <0.
We
shall
now prove
that lim s n i= 0.
Take
=A
any positive number
Then
sn
<- B, for an infinite number of values of n,
For,
if
possible, let lim sn
< ,
when n ^
the sequence
Nv N N
And
Let
|o-
be the
first of
such that between -
^-^
and
integer.
< 0.
< - A.
Let the integer next above
-z
2,
3 , ...
+N2&
=-=,
j-^ there shall

*
De
<7>
an(i write
Then
__5< , <T__,
since
\B 
v an(i
be at least one positive
N
= 1 + b.
q
261
FOURIER'S SERIES
102]
But
Sv-i
<s
N-r
i
F^2^
vr
11
<
Therefore each of these

Also
-q
iV
B + -- <-B + b.
cry
V^
< -|
<r.j
5 + 6)
< _ rl_ (5 _ &)+ .
>-^
and
-&>i5
But
j^ (B -b) > g^,
Therefore
B2
* <"" 2(^+4) + C
and
If
we take
n,
m>
+ 4)
we have
lim
Thus
< - which
s n ^ 0.
o\v
4(J5
"
is
impossible.
n*
But we have seen that
lim s n
^ 0.
n->c
lims n =0.
It follows that
Corollary
Let the sequence of arithmetic means
I.
ul + u2 + u3 +
converge
to
series
n for the
o\
If a positive integer n
where
<r
...
exists
such that \u n
<
-,
when n
~n
K is a positive number independent of n,
then the series
û n
converges
and
its
sum
is cr.
This
is
a special case of Theorem II.
Corollary
ofx, and
an
let
Let
II.
the sequence
u x (x)+u z {x) + ... be a series whose terms are functions

of arithmetic means cr n(x) converge uniformly to cr(x) in
interval (a, b).
Then,
where
n^ n
or n[s n+1 (x) - s n (x)] > K, when

n[s n (x) - s n+1 (x)] <
independent of x and n, and the same n Q serves for all values of x in
if either
K is
(a,b),
lim
n<
s n (x)=<r(x)
uniformly.
As before we may, without
K=
Then we have
the same v serving
\,
loss of generality,
and
(r(x)=0,
\
puo
-sn+1 (x)]<
= v,
e, when n
n
values of x in (a, b).
\cr
for all
<
n[s n (x)
1.
FOURIER'S SERIES
262
does not converge uniformly to zero, there must be a positive

such that an infinite number of the set
If s n (x)
number
[ch. vii
\s x {x)\,
are greater than or equal to
must be an
\s 2 {x)\,
1*3(3)1,....
each for some value of x in
infinite set of positive integers
N N N
2,
lt
3,
(a, b)
. .
that
\Sy2 (Xy2 )\, ...

kv^.v,)!,
are each greater than or equal to e a^,, xN<l? ... being points of
x,
2,
there
(a, b), corre-
N N
sponding to
is,
such that
....
N N
Let
be the first of this set
... which is greater than
2
l9
such that Nt
an even positive integer, say 2p.
,
i/
and n
and
Then
+ j(Xy)
> S A (Xy) - ^,
)>Ss {Xjf )-.
8s+p(xs
And
the argument proceeds as before.
We
Fejer's Theorem and Fourier's Series.*

Theorem (101) to establish the convergence of
103.
limitations imposed in our previous discussion;
WJienf(x)
= -\
"")
bn
the
sum
is 2
that
= -\
ffj
f{x')dx\
IT
n f(x')
an =
sin nx' dx'
we
is
Dirichlefs Conditions in the interval
satisfies
shall
now
use Fejer's
Fourier's Series under the
shall
-7r,
7r),
show that
and
f(x') cos nx' dx',
"J IT
(w^l),
of the series
a + {a x cos x + b x sin x) + (a 2 cos 2x + b 2 sin 2x) + ...

+ 0) +f(x - 0)] at every point in - tt < x < it where f(x + 0) and f(x - 0)
and at x= tt the sum is \ [/( - it + 0) +f(ir - 0)], when these limits exist.
First, let /(x) be bounded in ( - it, ir) and otherwise satisfy Dirichlefs
lfi x
exist
I.
Conditions in this interval.!

If the interval ( - ir, 7r) can be broken up into a finite number (say p) of
open partial intervals in which f(x) is monotonic, it follows at once from the
Second Theorem of Mean Value that each of these intervals contributes to
in ( - ir, ir).
a n or b n a part less than 41f/nir, where \f(x) <
a n cos nx + b n sin nx < SpMjmr,
Thus we have
where
is independent of n.
It follows from Fejer's Theorem, combined with Theorem II, Cor. I of
|
102,
that the Fourier's Series
a
converges,
* Cf.
and
+ (a x cos x + b x sin x) + (a 2 cos 2x + b 2 sin 2x) + ...

sum is %[f(x + 0)+f(x-0)'] at every point in -tt<x<tt
its
Whittaker and Watson,
flf the
loc. cit.
(2nd
ed., 1915), 167.
more general condition that the function
is
of
bounded variation
in
- 7r, 7r) is taken, then f(x) is the difference of two positive, bounded and
monotonic functions, and a similar argument applies.
(
FOURIER'S SERIES
102, 103]
at which f{x 0) exist, and at x = tt its

provided that /(-tt + 0) and/(-7r - 0) exist.
II.
Next,
- 7r,
number
there be a finite
let
when
sum
263
is
\[f{ -7r
+ 0) +f(ir -0)],
of points of infinite discontinuity
neighbourhoods of these points are

excluded, letf(x) be bounded in the remainder of the interval, which can be
broken up into a finite number of open partial intervals in each of which f(x)
in
is
7r),
but,
arbitrarily small
In addition,
monotonic.
f(x')dx' be absolutely
the infinite integral
let
convergent.
let re be a point between - -k and tv at which f(x + 0) and f(x - 0)
Then we may suppose it an internal point of an interval (a, b), where
b-a<7T, sindf(x) is bounded in (a, b) and otherwise satisfies Dirichlet's Con-
In this case,
exist.
ditions therein.
Let
ira n '
and
7r6 n
while
f(x
cos
nx dx'
f(x
sin
nx
'
fd
'
dx'
)a
'
27ra
[b
[b
i
f{x')dx'.
Then, forming the arithmetic means for the
+ (a/ cos x +
aQ
we
have, with the notation of
^) =
+S
^-
(a 2 'cos
series
2x + b 2 'sm 2x)
101,
+Sn - 1
8in*n(x'-x)
f(
J{X)
-2n^) a
6/ sin x)
Bm*Ux -x)
([W-a)
((hix-a)^
^
fix
Jy - 2a
L
{
V
sin 2 na
wa
,
7
da +
'
**
sm 2 a
2
na ,
sm
-.*- da
f(x
Jy + 2a) -sm
2
a
[l(b-x)
f*(&-*)_
a
^
\
where |(x - a) and |(6 - a;) are each positive and less than |7r.
But it will be seen that the argument used in Fejer's Theorem with regard
to the integrals
1 fi w r,
f(x
*
mr)
applies equally well
less
than
o
2a
,
sin2 na
.
sm2 a
'
when the upper
da
limits of the integrals are positive
|-n\*
Therefore, in this case.
lim
<r n (x)
And, as the terms
(a n cos
satisfy the condition of
ao
+ ai
(
converges and that
= [f(x + 0)+f(x-0)].
'
Theorem
cos x + bi
its
sum
is
Cf. footnote, p.
256.
II of
sin nx)
102, it follows that the series
sm x + a 2
lim Sn (x).
n><*>
s
nx + b n '
X
cos % x
+ ^2' sm ?x) +
and
FOURIER'S SERIES
264
[ch. vii
But
(a
- aQ ') + 2((n _ a n) cos nx + (n -
& n') sin
nx)
-1/
a
[
+
sin*(2n + l)(s'-s)
+ f* *\ fl^
sin^'-z)
&
sin(2n + l)a
+ l)a
1 f**
n .sin(2n
r
n
-.
.
da.
da + -\
fix + 2a)
fix -2a)
sin a
-27r\]_ 7r+a;
=-lfi"""J i(x
By
/^
..
a)
7T.1
1(6
sin a
94 both of these integrals vanish in the limit as n-^
<x>

00
(a
-a
')
+ 2{( n _ a n)
cos
nx + (K - K')
sin
nx)
converges, and that
its
sum
is zero.
But we have already shown that

n
+ Z(fl n
<Lq
'
nx + &' sm wa:
COS
converges,
and that
its
sum
is
hlf( x + 0)+f(x-0)}.
It follows,
by adding the two
that
series,
re
+ 2 ( n
cos
nx + b n
sin nx)
converges,
and that
its
sum
is
H/(*+o) +/(*-<>)]
at
any point between - ir and 7r at which these limits exist.

the limits /( -7r+0) and/(7r -0) exist, we can reduce the discussion
the sum of the series for x = tt to the above argument, using the equation
When
of
f(x+2ir)=f(x).
We can then treat
x=
7r as inside an interval
(a, b),
as above.
REFERENCES.
Bocher, "Introduction to the Theory
(2),
of Fourier's Series,"
Annals of Math.,
7 (1906).
De
la Vallee Poussin,
Dini, Serie di Fourier
loc. cit,
e altre
2 (4e ed., 1922), Ch. IV.
rappresentazioni analitiche delle funzioni di
una variabile reale (Pisa, 1880), Cap. IV.

Fourier, Theorie analytique de la chaleur, Ch. III.
Goursat, loc. cit., T. I (3e ed.), Ch. IX.
Hobson, Theory of Functions of a Real Variable, 2 (2nd ed., 1926), Ch.
Jordan, Cours a"'Analyse, 2 (3 e ed., Paris, 1913), Ch. V.
Lebesgue, Lecons sur les series trigonometriques (Paris, 1906), Ch. II.
Neumann,
Tiber d.
nach Kreis-, Kugel-
enden Entwickelungen (Leipzig, 1881), Kap.

* f(x) is
defined outside the interval
w,
ir)
VIII.
u. Cylinder- Functionen fortschreitII.
by the equation f(x + 2ir)
=f(x).
RlEMANN, loC. Cit.

Weber-Riemann, Die
Physik,
partiellen Differential-gleichungen der mathematischen
Braunschweig, 1910).
1 (2 Aufl.,
(5th ed., 1928), Ch. IX.
loc. cit.
EXAMPLES ON CHAPTER
1
< x < =,
f(x)
-^l-x,
-<x<l,
f(x)
= x--l.
In the interval
in the interval
,
Prove that
67TZ
2tTX
21 {
=
cos -j- f 9 cos
~T
2
The function f(x)
2.
COS
25
denned as follows
is
IOttx
f(x)
VII.
and
265
FOURIER'S SERIES
103]
~~T~
'
for the interval (0,
tt)
- \x, when
5S x %. \tt,
f(x) \ir, when \tt < x < 7r,
f(x) = %(Tr-x), when jJTr^x^Tr.
f(x)
Show
that
/W^gl
/V
7T
Expand f(x)
3.
'"^ 2 "- 1 '" 8!"' 2 "" 1 ^, when

2
(2n-l)
in a series of sines of multiples of irxja, given that
f(x)
when
;fS x ^ a,
when |aâ:â.
= mx,
/(a;)=ra(a-a;),
4.
OSig,.
Prove that
,-
j
S111
ao
H-*=-2
Z
2n7rx
~~p"
-
>
7T 1
when 0<a;<Z,
2wn-a;
2 jj.
M cos
and
of
(hi
- x? =y^ +
x?
when
_ I.
5. Obtain an expansion in a mixed series of sines and cosines

x
x which is zero between - tt and 0, and is equal to e between
give
its
6.
of multiples
and
tt,
and
values at the three limits.
Show
that between the values -
tt
and + tt
of
x the following expansions
hold:
sinmx =
/ sin
2 sin 2x
p-^-2^r^2
Sinm7T^
7r
ra cos
I
raa; _ 2
2+
sinhraTr~^\2^~l 2 +ra
7.
Express
a;
a;
ra cos 2a;
22
between -
for values of x
22
2a;
and
--
ra cos 3a;
+ m2 +
^ +m
ra cos 3a;
+ ra 2
tt
3a;
3^^ +
/Ira cos 2- ra cos
cosraz=-smra7r(^ +
cosh
3 sin
32
tt
+ ra 2
as the
sum
of a constant
and a series of cosines of multiples of x.

Prove that the locus represented by
oo
_ \ \n\
V
!
is
two systems
of area
tt
2
.
/
i
^2
sin nx sin ny =
7*
of lines at right angles dividing the plane of x,
y into squares
FOURIER'S SERIES
266
8.
[ch.
Prove that
V
2 c3
" 4d (
o
3 + 2, -s
=o3d
n3 7r z
sin
mrc
nrrc cos
-j
mvc\
mr
-^- cos =- x
d J
d
represents a series of circles of radius c with their centres on the axis of x

at distances 2d apart, and also the portions of the axis exterior to the circles,
one
having
circle
9.
centre at the origin.
its
A polygon is inscribed in a circle of radius a, and is such that the alternate

6=0
sides beginning at
Prove that the
subtend angles a and
third,
first,
at the centre of the circle.
/3
pairs of sides of the polygon
. . .
may
be represented,
except at angular points, by the polar equation
r=a 1
TfO
sm^^ + a
.
2tt0
sin^
T?+ ...,
__ în^^-^cos -^
where
mr(2a +
cos^sec^
$[&
fi)
mr<h
Find a similar equation to represent the other
sides.
regular hexagon has a diagonal lying along the axis of x.

Investigate a trigonometrical series which shall represent the value of the ordinate
of any point of the perimeter lying above the axis of x.
10.
< x < 2tt,
11. If
12.
prove that
tv
sinha(7r-a: )
sinhaTr
sin
Prove that the equation
2 sin 2x
3 sin
+ l 2+ a 2 + 2 2
~~a 2
a2
'
3a;
+ 3 2+
""
in rectangular coordinates
2
3ttx
f
^fc+êos-^cos
+ 3- co*_-.
*,
wx
h^ 4h
represents a series of equal
and
similar parabolic arcs of height h
2k standing in contact along the axis of
and span
x.
13. The arcs of equal parabolas cut off by the latera recta of length 4a are
arranged alternately on opposite sides of a straight line formed by placing
the latera recta end to end, so as to make an undulating curve.

the equation of the curve can be written in the form
7r
irx
=sin TA
64a
4a
.
3772-
+ sin3
4a
Prove that
5irx
(---.sin
3
4a
+....
14. If circles be drawn on the sides of a square as diameters, prove that

the polar equation of the quatrefoil formed by the external semicircles, referred
to the centre as origin, is
4^72 =1 +
where a
15.
is
On
Is cos
40 -
:;
cos 80
,]
..
cos 120
...
the side of the square.
the sides of a regular pentagon remote from the centre are described
FOURIER'S SERIES
vii]
segments of
circles
7rr
= 5atan
a being the radius of the
and
circle
In the interval
in the interval
C}
lie
between
by the two
and
<x<
I,
f(x)=0.
series of sines
Draw
and
also
sums
are the
by means
of a
showing the functions
figures
x not restricted to
series respectively for all values of
What
I.
rr-1 J'
f(x)=x 2
of cosines of multiples of -=-.
represented
Ll-22(-ir
Express the function by means of a

series
prove
is
circumscribing the pentagon.
0<x<
=
pentagon
of the
which contain angles equal to that
that the equation to the cinquefoil thus obtained
16.
267
of the series for the value
x=
17. A point moves in a straight line with a velocity which is initially

and which receives constant increments each equal to u at equal intervals
Prove that the velocity at any time t after the beginning of the motion is
2wr
1
ut
u 1
v = n u-\
1,
h- Z, -sin
u,
t.
and that the distance traversed
18.
curve
is
(*
W=1
is
ur *
ut
ut
27
7T
+ r) + j2 - 2T2 ^
2mr
cos
n2
formed by the positive halves of the
See
Ex
4 a o v e.]
circles
+ l)a) 2 +y 2 =a 2
(x-(4:n
and the negative halves
._
ra
*
of the circles
(x-(4n-l)a) 2 + y 2 = a 2
n being an integer. Prove that the equation for the complete curve obtained
by Fourier's method is
,
y =?
9
19.
T
K
=i
(
v
ir
'
2/ a
Having given the form
of the curve y=f(x), trace the curves
w=-2sinra:l
ff
M\in (k - 2/1) a V;?^ &-.
si ( *
f(t)sinrtdt,
Jo
TT
w=-5sin(2r-l)4V(<)sin(2r-lU^,
T1"
Jo
and show what these become when the upper

20.
Prove that for
all
values of
5
y\ - sin
l
is
for all values of
tttx
I
and
x between b -at and
instead of "^
-= sin j- sin
I
is
and - the sum
between
rirb
zero for all values of x between
is -r
limit
of the series
rirat
;
I
- at and between at\ b and
at
+ b, when
<
I,
and
FOURIER'S SERIES
268
Find the sum
21.
of the series
"
u=- \
tt ~
"
sin
2n -
'
-1 )ttx
sin (2a
K
Z.
7r
2mrx
n
and hence prove that the greatest integer in the positive number x
[See Ex. 1, p. 250, and Ex. 4 above.]
sented by x + u - Sv 2
repre-
is
22. If x, y, z are the rectangular coordinates of a point which moves so

2
2
that from y Q to y x the value of z is K{a -x ), and from y = x to y = a
and
2
2
the value of z is k(cl - y ), show that for all values of x and y between
z may be expressed by a series in the form
a,
^,, S inf(^) Sinf(^),

find the values of
and
AP
for the different types of terms.
= \ir sin
f(x)=^,
and
when
when
x,
f(x)
23. If
prove that, when
^=
f(x)
= \tt
sin
a;
where
~ x ^ \tt,
<x^
7r >
< 7r,
a;
+ Sx + S2 - S3
#i = j^
7r
sin 2a:
^^ sin
4a:
+ g^
sin
2 =sin x + l sin Sx + l sin 5a:+ ...

S 3 = sin 2a: + J sin 6a: + 1 sin 10a: +
8 V S 2 and S 3 separately for values
6a;
and
find the values of
the assigned interval.
[Cf.
,
f{x)
24. If
Ex.
4
sin
3a:
x lying within
52
sin 2a:
+ J^""*--2- +
/(a;) is
of
sin 5a:
sin
x2
p. 250.]
1,
= -s
2
show that
continuous between
sin 3a:
-S
and
7r,
and that
show that f\x) has a sudden change
of value
at the
/(tt
7T
point ^.
-0) = 1.
[See Ex.
Also
1, 2,
pp. 242-3.]
25.
,,
Let
" sin3(2n-l)a:
^) = t
x
when = x ^
Show that
2^1
sin (2n
2n-l
l)a:
sin &(2n
^f
1)tt sin
(2n - l)s
(2^H^
'
7r.
/(
+ 0) =/(tt - 0) = - ir,
/(iT + 0)-/ftff-0)=-iir,
/(** +0) -/(*-(=**;
/(0) =/( Jir) =/(*ir) -/(tt) =0.
the graph of f(x) in the interval (0, tt). [See Ex.
also that
Draw
above.]
1, p.
250,
and Ex. 2
CHAPTER
VIII
THE NATURE OF THE CONVERGENCE OF FOURIER'S

SERIES AND SOME PROPERTIES OF FOURIER'S
CONSTANTS
The Order of the Terms.
104.
Before entering upon the dis-
cussion of the nature of the convergence of the Fourier's Series for
a function satisfying Dirichlet's Conditions,

certain cases the order of the terms
I.
Iff(x)
bounded and otherwise
is
in the interval
ir, ir),
may
we
show that
shall
be determined
in
easily.
Conditions
the coefficients in the Fourier's Series for
fix) are less in absolute value than Kjn, where
is
some positive
number independent of n.
If the interval ( - ir, ir) can be broken up into a finite number of
open partial intervals (c r) c r+1 ) in which/(x) is monotonic, it follows,
from the Second Theorem of Mean Value, that
7ra n
= i;j
f( x ) cos nx dx
some
is
f{cr+1
- 0)
cos
nxdx\,
Sf
definite
number
in
(cr , c r+1 ).
^|a.|<^S{|/(c r +0)|+|/(c r+1 -0)|}
<
where p
is
of
|
pM
number
the
f(x)
Therefore
where
4-
'
Thus
bound
nx dx
Cj.
v.
where r
cos
fi cr +0)
(*<V+1
C*r
=2
n
of partial intervals
in the interval {-it,

|
an
and
it).
< Kjn,
K is some positive number independent of n.
is
the upper
THE CONVERGENCE OF FOURIER'S SERIES
270
And
[oh. viii
we obtain
similarly
\<iqn.
\b n
We may
speak of the terms of this series as of the order 1/n.

When the terms are of the order 1/n, the series will, in general,
be only conditionally convergent, the convergence being due to
the presence of both positive and negative terms.
II. If
we
are given that f(x)
result follows at once, since/(x)
of
is
bounded variation
= F(x) - G(x), where
in
-ir,
7r)
the same
F(x) and G(x) are bounded
and monotonic functions.
If f(x) is bounded and continuous, and otherwise satisfies

Conditions in -7r<x<7r, while f(ir-0)=f(-ir +0),
III.
DirichleVs
and
iff'(x) is
Conditions
in the same interval, the coefficients in the Fourier's Series for f(x)
are less in absolute value than K/n 2 where
is some positive number
independent of
In this case
n.
we can make
- 7r, it) by giving to

and 7r respectively.
(
Then
7ra n
it
continuous in the closed interval
f(x)
the values /( -
tt
+0) and/(-7r -0) atic= -
?r
f(x) cos nx dx
J IT
= i r f(x)
=
But we have
J _n
sin
nx
f'( x ) sin
if"
nx dx
f'(x) sin
nx dx.
just seen that with the given conditions

f'(x) sin
nx dx
i:
J IT
is of
the order 1/n.
It follows that
where
A
f(ir
\a n
<K/n 2
similar argument, in
- 0) =/( -
7r
+0)
is
which
it will
be seen that the condition
used, shows that
\K\<Kjn*.
Since the terms of this Fourier's Series are of the order 1/n 2
it
follows that
it
is
absolutely convergent, and also uniformly
convergent in any interval.
The above result can be generalised as follows If the function

th
are bounded,
f(x) and its differential coefficients, up to the (p-l)
:
AND PROPERTIES OF FOURIER'S CONSTANTS
104, 105]
and otherwise
-7r<x<7r, and
continuous
interval
/W(-T+0)=fW(x-0),
and
if the
Dirichlet's
th
Conditions
DirichleVs
satisfy
271
in
the
[f=0, l,...(p-l)],
differential coefficient is
satisfies
Conditions in the same interval, the coefficients in the
Fourier's Series for f(x) will be less in absolute value than
K/n p+1
where
The Riemann-Lebesgue Theorem,* and
105.
its
Consequences.
Let f(x) be
[b
iff(x) is unbounded,
(a, b), or,
f(x)dx
let
be absolutely
Ja
convergent.
Then
lim
f(x)
cos
n>c*Ja
nxdx = 0.
be less than A in (a, b), and e the usual arbitrary positive

(a) Let \f(x)
number.
There is a mode of division of (a, b), say a = x Q x v x 2 ... xm_ 1 xm = b, such
that S - s for it is less than e ( 42).
|
Cb
Thus
nx dx = 2
fix) sin
\f( x r)
\
*
[f( x r)
\Cx r
=2
frr
+ (f( x ~f( xr)] sm nx dx

)
sin
I
nx dx\ +
l-aV-l
rxr
la
-
V-l
\f( x )
~/{ x r)
]sin nx\
dx
\
J
< 2m^4 2 (MT - m r

1-
2ra^4
(a; r
- x r -i), with the notation
of 39
2mA
<
+*
,
<
e,
4m^l
when w =
-
r&
Hence
/(x) sin
lim
>-oo
nxdx=0.
[b
And
in the
same way,
lim
f(x) cos nx
dx=0.
n>ocio
r&
(/?)
f(x)dx
If
the definition of
tinuity in
51,
an absolutely convergent
we have only
finite
infinite integral,
number
according to
of points of infinite discon-
(a, 6).
As we can
the case
is
treat
when a
or b
these separately,
is
it is
clear that
we need only
We
discuss
take the latter
alternative.
* This theorem was proved by Riemann [Math. Werke, 1 (2 Aufl., 1892) 254] for
the functions stated in the text, and extended by Lebesgue to functions with a
Lebesgue Integral [Ann.
sc. de,
V^cole normale
(3),
20 (1903), 471],
272
In this case, there
a point
is
an<
between a and
(5,
10
bounded
is
in (a,
/?)
'.'a
<
|e
when n
i\
< e + |c,
f(x) sin nx dx
a
.'
I
f{x) dx.
and thus, by the above, we know that
/(.r) sin ??x c?x
\\
?ix eta
J/3
'
[./3
Therefore
f(x) sin
f(x) sin nx did
!a"
But /(a;)
f(x) sin nx dx\ =\\ f(x) sin nx dx
jl
such that
\f(x)\dx<l*.
r
Also
b,
[ch. viii
when nî',
and the theorem is proved.

The following results can be deduced almost immediately from the RiemannLebesgue Theorem. In all of them x is a point of the interval ( - ir, ir) and
it is bounded and
f(x) is subject to the conditions named in that theorem
;
integrable in
unbounded, the integral
or, if
ir, tt),
f(x)dx
absolutely
is
IT
convergent.
The Fourier
(i)
Constants a n and b n off(x) tend
The behaviour of
(ii)
the
Fourier's Series corresponding
vergence, divergence, or oscillation at a point x
f(x) in the neighbourhood of x

Here, with the notation of
where
(.r
is
as
to f(x),
con-
to
depends only on the values of
we have
+ 2T, fir
95,
/fa-o-Si?
+ 2>/)
2/;,
when n^co
zero
to
f-r
s in
a neighbourhood of x
\{2n +
-x
\){x'
Thus
i^TJii;
sin(2w
^7t-t-i;u
sin
+ l)a ,
1
ll'^+ô)^
n
da
/(b-2o)
n(*o)=~
sin a
l[l(n-or
l[*i
sin(2n + l)a,
,,
/(s + 2a)
rfa + /(%+2a)
+)
^
sin a
7rJ^
By
the Riemann-Lebesgue Theorem, the
when
the limit
(iii)
??.-
and the
+ l)a
sina
first
and second
da.
integrals vanish in
result follows.
The behaviour of the Fourier's
-~ <a<b < 7r,
sin (2n
TJ-rj
Series corresponding to f(x) in
an
interval
as
to
convergence, divergence, or oscillation, depends
only on the values of f(x) in (a -
S,
(a, b), ivhere
+ 8),
ivhere 8 is
an
arbitrarily small positive
n umber.
This
(iv)
is
proved as in
Iff(x)
is
(ii).
of bounded variation in a neighbourhood ofx
atx to%[f(x + 0)+f(x

As above,
and the
hm sn (x =-1
/
result follows
the series converges
-0)].
from
v
lim
T?
("
f(x
+ o2a)
,
sin(2n+l)a
Dirichlet's Integral
-j
( 94).
da,
105, 106]
lim [f(x
If
+ h) +f(x - h)]
and 2f(x
exists,
taken equal to this
is
273
limit,
two
important sufficient conditions for the convergence of the Fourier's Series

They are usually called
at x to the value f(x Q ) are given in (v) and (vi).
In both
Dini's Condition and Lipschitz's Condition.
<*>()
The Fourier's
(v) Dini's Condition.*

its sum when x = x
for
=/(b + 2 ) +/(*o - 2 ) - 2/MSeries corresponding to f(x) has f(x
if there is a positive
number
da
such that
rj
is
convergent integral.
For,
we
see
from
lim
C { f(x,
n _MJo ^
\f(x
for
its
j/ ,
<f>(a)
<C
In this case, there
And
sin
is
\t\~k
a value of
//
IX^da,
sin a
...
if
^
A<
V <|tt.
C and
to f(x) has
k exist such that,
<2
fc
Ca
fc
,
oj-nf
fc
da < 7r2 C
I
'
=a^
when
*
ak
7/.
ij
~^da,
Jo
?/
small enough.
a special case of the preceding.
106. Discussion of a case in
has an infinity in
da =0
some fixed positive number.

which
which can be made as small as we please by taking

is
"
for
,sin(2w+l)a
</>(a)
Y
fn |<f>(a)|
numbers
if positive
when
Fourier's Series corresponding
T/ê
,
This condition
II^jl/
"
sin a
[/(av+2a) +f(x - 2a) - 2f(x Q )

*
si
^ + l)a
c?a=0.
sin(2w + l)a
[<-,
dal^lrr
\.
n
sum when x = x
+ t) -f(x Q
provided that
),
sin (2ra
0(a)
Condition.!
(vi) Lipschitz's
f(x
fi
lim
ll"
=f(x
n^Jo
Also
+ 2a) +/(*. - 2a) - 2/(*)
..
.,
if
i.e.
that lim s n (x
n>oo
(ii),
- tt,
tt).J
which f(x) satisfies Dirichlet's Conditions and

have seen that Dirichlet's Conditions include
We
the possibility oif(x) having a certain
number of points of infinite discontinuity
in the interval, subject to the condition that the infinite integral
f(x)dx
is
is
Let us suppose that near the point x

such that
where - 7r
<x <
it,
the f unction/(#)
where < v < 1 and cj>(x) is monotonic to the right and left of xu while
do not both vanish.
In this case the condition for absolute convergence is satisfied.
,
<f>(x
*Dini, Serie di Fourier e altre rappresentazioni analitiche delle funzioni di

variabile reale (Pisa, 1880), p. 102.
tLipschitz, Journal fiir Math., 63 (1864), 296.

%
See also Ex.

c.i.
5, p.
241.
s
0)
una
274
[oh. viii
Then, in determining a n and bn where

,
iran
we break up
and
f{x) cos nx dx
irb n
f(x) sin nx dx,
the interval into
(a,x ), (a%, ft, and (ft ir),

(-7T, a),
where (a, ft is the interval in which f(x) has the given form. In ( -tt, a)
and {ft tt) it is supposed that f(x) is bounded and otherwise satisfies
Dirichlet's Conditions, and we know from 104 that these partial intervals
and b n contributions of the order 1/n.

The remainder of the integral, e.g. in a n is given by the sum
give to a n
of
rL
cos nx dx
lim
{X-X )Y
SÔJa
limits being known to exist.
and
these
We
of
lim
nx
dx,
fioJro + a (*-*<>)
take the second of these integrals, and apply to
Mean
cos
the Second Theorem
it
Value.
Thus we have
3
</>(a)
where x Q +
8 tk
P
)
Xo+6
,,#.[*
-dx + 4j{ft\
,
cos
nx j
dx,
obtain
cos nx dx =
J*
{x-x y
Q
COS
But
when
nx
= ft
n(x-x )=y, we
Putting
Also
cos
cj
,,
T^-t-r cosnxdx = ch (xQ + 8)\
f'
a, b
d?/
0)
C S iy
+ nX
X'~*
y
cos (y + nx
MftW
ni-
)nS
= cosna;
dy
dy-sin?ia:
dy.
are positive.
ir&cosj,
\P*jny
IJfl
l.'a
are both less than definite numbers independent of a and &, when
Thus, whatever positive integer n may be, and whatever value 8
subject to
< 8 < ft - x
<v<
may
1.
have,
IJ*
where
i' is
some
+s(z-V
positive
cos
nxdx\< K'\ri^ - ",
number independent
of
n and
It follows that
lirnT
cos
nxdx
< K'\n\
ls-M)Jiro+s(-ffo)
similar
argument applies to the integral

<f>{x)
(x
-x y
cos
w#
dx.
It thus appears that the coefficient of cos nx in the Fourier's Series for the
is some
given function f(x) is less in absolute value than Kjnl-y, where
positive number independent of n ; and a corresponding result holds with
regard to the coefficient of sin nx.

It is easy to modify the above argument so that
when the
infinity occurs at nr.
it will
apply to the case
106, 107]
107.
The
Uniform
shall deal, first of
f(x),
when
f(x)
Convergence
all,
number
Fourier's
Series.*
We
with the case of the Fourier's Series for
bounded
is
finite
of
275
in ( - tt, tt) and otherwise satisfies

Later we shall discuss the case where a
of points of infinite discontinuity are admitted.
It is clear that the Fourier's Series for /(a?)
cannot be Uniformly
convergent in any interval which contains a point of discontinuity
since uniform convergence, in the case of series whose terms are
;
continuous, involves continuity in the sum.
Letf(x) be bounded in the interval
tt), and otherwise satisfy

Then the Fourier's Series
for f(x) converges uniformly to f(x) in any interval which contains
neither in its interior nor at an end any point of discontinuity
of the
(
tt,
Dirichlet's Conditions in that interval.
function.^
As before the bounded function f(x), satisfying Dirichlet's Con- tt, ?r), is defined outside that interval by the equation
ditions in
/(* +2*-) =/(*).
Then we can express/^) in any intervale.g. (-2tt, 2tt) as

the difference of two functions, which we shall denote by F(x)
and G(x), where F(x) and G(x) are bounded, positive and. monotonic increasing.
They are also continuous at all points where
f(x)
is
continuous
[ 36. 1
or
36. 2].
l>etf(x) be continuous at a
where, to begin with,
we
shall
and 6t and at all points in a<x<b,

assume - 7r<a and 6<tt.-
let x be any point in (a, b).

Then with the notation of 95,
Also
Lit)
_/
sin h
(x
l(x
f(x+2a).
sin a
*See footnote,
-x)
x)
da,
where
m = 2w+l.S5
p. 230.
be seen from 105, (ii), that if (a, b) be any interval contained in

-tt, tt) such that /(a) is continuous in {a, b), including
(
the end-points, the answer
to the question whether the Fourier's Series converges uniformly
in (a, 6), or not,
depends only upon the nature of f(x) in an interval (a', b'), which includes
(a, b)
fit will
in
its interior
and exceeds
it
in length
JThus/(o + 0)=/(o)=/(a-0)
by an
arbitrarily small
amount.
and f(b+0)=f(b)=f(b -0).

have replaced the limits -tt, tt by -ic+x, ir + x in the integral before
changing the variable from x' to a by the substitution z' = x + 2a.
Cf. 101.
We
276
[ch. viii
Thus
-If*
tt)
We shall now
G(x
+ 2a)^^da
discuss the
**
first
integral in (1),
,,.
ma
sin
Fix +2a) .
\n
ma
^
,
J(a+2a)
f*"
i.e.
sma
Jo
Let
/Li
da
da,
sin
sin
(1)
sin
-in
f**
F{x-2a)
sm ma
da.
sin
Jo
be any number such, that 0<^<iTr.
Then
ma,
sma
sin
v
'
^f
j7r
sin
ma
sm a
c?a
= I 1 +I 2 +I 3 ,say
We
(2)
can replace Fix +0) by F(x), since J(*)
But
P'si n(2n+l)a rfa= f*

sm a
Jo
(1
+2
continuous at
{F(x +2a) - F(x)}
increasing
in
any
J
interval;
(3)
bounded, positive and monotonic
is
and
-=
sm a
0<a^j7r.
is
also
bounded, positive and
monotonic increasing in
Therefore we can apply the Second Theorem
the integral
(>
0()
Jo
0(a)
where
ma
sin
of
Mean Value
<*a,
= {F(x +2a) - ^)}
^.
It follows that
^mx^-F^^l^^
where
0^1 = ^.
x.
cog 2m) ^^
I 1 = iirF(x)
Thus
Also
is
to
107]
277
But we know that

!
ma
P* sin
Jf
m
f ^ sin
da
ii
\J
(91.)
sin
I t = {F(z+2p)-
Finally
da <7T.
+ 2^) - F(x)} --
\I 2 \<{F(x
Therefore
mt
(4)
ix
da
F(x)}f'^^
sin a
Jp
where /u^
But,
if
'
+ {J(x+tt )-F(x)}\
Jf
da -sm
sma
1
Je
sm ma da +- If*.
sin ma
sm 0j x
f
^
0J
r-
mi
da,
O<0<0^j7r,
f^sinma
where 6
.
sin
7r.
cfa,'
^ ^ = 0,
Inereiore
ma da
sina
P* sin
|Jj
<2 {cosec 6 + cosec rJ

,
m
<4 cosec 0.
m
<}
It follows that
|Zsl
< m~4n^ [{F{X +2 ^ ~ F
< 4Z
^ +^
x +7r )
" *(*]
(5,
where Z is some positive number, independent

pending on the upper bound of \f(x)\ in ( - tt, it).
Combining (3), (4) and (5), we see from (2) that
sm a
)7r.F
<{F(x
similar
+2//)
of
m, and de-
'
iK
F(x)}J^+
"sin/*
m-n-sm/u
(6)'
v
argument applies to the integral
^(sc
- 2a)
but in this case
it
The corresponding
f^zv
-.
sma
da,
has to be remembered that F(x -
tonic decreasing as a increases
I*
from
to
result for this integral

N
sin
ma
_.,
<\F(x-2fi)-F(x)\
2a)"*is
mono-
Jtt.
is
that
-J^+^_,
(7 )
278
some
as before being
number independent of m, and

( - tt, tt).
we can take K the same in (6) and (7).
positive
depending on the upper bound
Without
From
loss of generality
and
(6)
we obtain
(7)
of \f(x)\ in
at once
2^c+2a)
7TJ-J7T
da-F(x
sma
8Z
< sm {1^+2^-^)1+1^-2^-^)1
/'
m-TT
/*
we
Similarly
G(x
Thus, from
(1),
da -
G(x)
+ \G(x - 2//) -
G(z)
+2a)
6r(ic
/)
-^
sm
*""
/*+2a)
sm ma
sin a
+\F(x-2/i) -
+ \G(x +2/*) - 6?M|

16Z
m7r sm /^
Now .F(se) and

# = a and x = b.
+|fl?(a!-2/i)
G(x) are continuous in
\F(x +2/0 - F(x)\<e,
the same
we know that
/u
Choose
ju
passes from
fi
8Z
W7r
sm /i
(9)
J(sc)|
- G(*)|}
a<x<b, and
e,
/ll
Also
(10)
Thus, to the arbitrary positive number

(which can be taken
a positive number
to \ir as
da -f(x)
{|J(s +2/i) - F(x)\

< -^sin/t
\ju\
(8)
/i
7TJ -itr
when
sm /^
find that
7T J
sm
[oh. viii
also
when
there will correspond
less
than
Jtt)
such that
G(x +2//) - 0(a;)|<e,
oixmaîx^b.
^ serving for all values
from unity
continuously
increases
^
cosec
to
\ir.
as above, less than Jtt,
and put p = p
in the argu-
allowable,, as the only restriction
ment of (1) to (10). This is

upon fx was that it must lie between and
Then the terms on the right-hand side
16Z/m7r
sin
\ir.
of (10),
are together less than 8e for
all
not including
values of x in
(a, b).
it
except that
So far nothing has been said about the number m,
(2n
integer
+1).
is an odd positive
107, 108]
Let n
be the smallest positive integer which
279
the in-
satisfies
equality
(2w +1)tt sin
As K,
and
fj,
m (i.e. 2n +
Then
it
jli
are independent of x, so also
is
We now choose
n^n
from (10) that
so that
follows
W~/WI<9f; when n^n
the same n serving for every x in a

x ^ b.
In other words, we have shown that the Fourier's Series converges uniformly tof(x), under the given conditions, in the interval
6).*
-7r+0)=/(7r-0), we can regard the points tt, 2tt, etc., as points

beyond (-tt, tt) by the equation f(x + 2ir)=f( x), is
continuous, for we can give to/(7r) the common value of
/( -tt + O) and
/(t-0).
If /(
at which f(x), extended
108. The Uniform Convergence of Fourier's Series

f (continued). By
argument similar to that employed in the preceding section, it can be proved
that when/(x), bounded or not, satisfies Dirichlet's Conditions in the interval
( - tt, rr), and/(x) is bounded in the interval (a', V) contained within ( - tt, tt),
then the Fourier's Series for f(x) converges uniformly in any interval
(a', b'), provided f(x) is continuous in (a, b), including
the interior of
{a, b)
its
in
end-
points.
But, instead of developing the discussion on these

* It
may
lines,
we
shall
help the reader to follow the argument of this section
if
now show
we take a
special case
-TrrgaÔ, ^
E.g. f(x)=0,
_,
/(*)
= !,
O^X^TT.
then we nave:
Interval.
-2tt< X
<
j\x)
TO
-7T
- 7T^ X^_
0<x^v
ir<
<a~x=b<
If
tt,
x^2tt
the interval
(a, b) is
<-Xr)
an interval
in
which f(x)
is
continuous.
The argument of the preceding section will then apply to the case
- tt or tt is an end-point of the interval (a, b) inside and at the ends
f(x)
t
is
continuous.
See footnote, p. 230.
in
of
which
which
280
how the question can he treated by Fejer's Arithmetic Means

we shall prove the following theorem
(cf.
[oh. viii
101),
and
Letf(x), bounded or not, satisfy DirichleVs Conditions in the interval ( -rr, ir),
and let it be continuous- at a and b and in {a, b), where -irâ and bîr* Then
the Fourier's Series
for f(x) converges uniformly
-8) contained within
Without
to
f(x) in any interval {a +
8,
(a, b).
loss of generality
could be treated as the
sum
we may assume
two such
of
'=
27ra
Let
-a
it, f or
a greater interval
intervals.
cb
f(x')dx',
a
f(x') cos nx' dx'

a
and
ir
V=
f(x') sin nx'dx',
}a
continuous at a and b and in (a, b),

and we can use the Corollary to Fejer's Theoremsequence of Arithmetic Means for the series
Since f(x)
is
it is
bounded in (a, b),

and assert that the
also
101)
co
'
+ 2 K' cos nx + b n
sin nx)
'
converges uniformly tof(x) in

Also
But /(a:)
(a, b).
a n cos nx + b n ' sin nx %, {a n

'
is
bounded
in (a, b)
Thus we can write
'2
+ &n
'2
and satisfies Dirichlet's Conditions therein.

f{x)
= F(x)
- G(x),
where F(x) and G(x) are bounded, positive and monotonic increasing functions
It follows that we can apply the Second Theorem of Mean Value
in (a, b).
to the integrals
*
F{x')
C0S
*nx'dx>,
an
.o
G(x')
)a
sm
nx'dx>,
2
2
and we deduce at once that (a n + b n )% < K\n,
on the upper bound
depending
number
positive
some
is
K
where
'
'
of \f{x)\
in (a, 6).
Then we know, from Theorem

a
'
II, Cor. II of 102,
+ 2( n
'
cos
nx + b n
'
that the series
sin nx)
converges uniformly to f(x) in (a, b).

Let us now suppose x to be any point in the interval (a +
the interval
8,
(a, b).
With the usual notation

f(x')dx',
T
Tra n
irb n
f(x') cos nx' dx'
IT
5^1
f(x') sin nx' dx'.
- 8) lying within
108]
281
It follows, as in 103, that
K-a
is
o)
+ 2 { (an - a n ')
cos
nx + (bn - bn ')
sin nx)
equal to
sma
irJKx-a)
f(x) being denned outside the interval
- tt,
it)
sin a
'
TrJ^^-a;)-
by the equation
f(x + 27r)=f(x).
Now/(a;)
is
supposed to have not more than a
of infinite discontinuity in
- 7r,
it),
and
\f{x')
number
finite
of points (say
m)
dx' converges.
We
can therefore take intervals 2y 1 ,2y 2 ,...2y enclosing these points, the
m
intervals being so small that
\f{x')\dx'
<2ts\n\8
[r
=l
2,
...
m]
J2y r
being any given positive number.

[*""
ft
/(a;
sin(2w + l)a
,
' da,
sina
o
- 2a)
v
H(x-a)
x,
+ 8, b - 8).
we may meet some
as already stated, being a point in (a
As a passes from %(x-a)
to
^7r,
- 2a).
of discontinuity of the given function in f(x
centres of the corresponding intervals
Also the smallest value of (x - a)

j..
Thus
/(3:
,
...
points
y TO
is 8.
- 2a) sin(2n + l)a

sina
.
y lf y 2
or all of the
Let these be taken as the
..
vl
* <E3SJJ- fa>l*'
<.
When these intervals,
such of them as occur, have been cut out, the integral

fi"-
-,
/(a?
^ + l)a
da
.sin(2w
-2a)
will at
most consist
of
(m+1)
sina
Ji(x-o)
separate integrals (Ir ).
[r
= l,
2,
...
m + 1.]
In each of these integrals (Ir ) we can take f(x - 2a) as the difference of two
bounded, positive and monotonic increasing functions
F(x-2a)
and
Then, confining our attention to (7r ),
G(x-2a).
we
see that
sin(2 m -M)
sin
= \(F - ){cosec
1
|8'- (cosec \8
~\J X + \J 2 + \J*\ + \J*\,

\
- cosec
a)} sin (2n
l)a da
282
where
J x = cosec
\8 j
F sin (2n + 1 )a da,
J 2 = cosec
JSl
G sin (2n + l)a da,
J {cosec |8 - cosec
1
J4 =
{cosec
}.8
- cosec
a} sin (2n
+ l)a da,
a} sin (2
+ l)a da.
[oh. viii
integrals,
But we can apply the Second Theorem of Mean Value to each of these
monotonic.
is
I)a
(2n+
sin
multiplies
since the factor in each integrand which
rr
It follows that
Ir
<^^\ cosec i S
'
number independent of n and a;,- and depending

in ( -tt, it), when the intervals 2 yi 2y 2 ... have
where K is some
only on the values of f(x)
been removed from that interval.
positive
Thus
If*'
LW-a//(
when
(2n
+ 1) > X
Similarly
we
-^
sina
'
|.S.
is
when x
independent of
lies in
a;,
the integral converges uniformly
the interval (a + 8,b-
8).
find that
fin-
_.
f(x + 2a)
sin(2w + l)a
s ina
)i(b-x)
converges uniformly to zero
Thus the
< (2w + l) sm 8 -M*
<(2ra + l)e,
cosec
Since this choice of

to zero as n-><*>
tS+^
sin(2 " + 1)a
*_2a)
when x
,
da
lies in this interval.
series
K-O+2
{(
- <) cos nx + (b n - bn ')
sin nx)
converges uniformly to zero in (a + 8, b But we have shown that the series
8).
GO
'
+ 2 ('
cos nx
+ bn
'
sin nx)
converges uniformly tof{x) in (a, b).

uniformly,
Since the sum of two uniformly convergent series converges
we
see that
CO
+ 2 (a n cos nx + b n sin nx)

l
converges uniformly to f(x) in (a + 8, b 109. Differentiation

Differentiation.
term by term
8).
and Integration of Fourier's
Series.
From the worked out examples
in 95-97
it is
clear that
differentiation of Fourier's Series is not in general permissible,
as the terms do not tend to zero sufficiently quickly.

to the
This difficulty does not arise in the application of Fourier's Series
solution of the Equation of Conduction of
often written
Heat
as
^j-*^*-
108, 109]
283
Here * the terms in the appropriate series are multiplied by a factor (e.g.
g-kuw/d?^ w hich may be called a convergency factor, as it increases the rapidity
of the convergence of the series, and allows term by term differentiation both
with respect to x and
t.
Again we have seen that under certain conditions the Fourier's

Series for f(x) converges uniformly to f(x) in the interior of any interval in
which the function is continuous. In this case we know that we may integrate"
the series term by term within such an interval, and equate the result to the
Also this operation can be repeated
integral of f(x) between the same limits.
Integration.
any number of times ( 70. 1).

But such a simple series as the Sine Series
4/
1
_
1
,
e
1=
is
sin
oj
+^
sin
sm
3x +
5a;
However
where x
+ .--),
as its
tt,
sum
is
discon-
and x = ir.
it
tt
namely
tt,
~x
not uniformly convergent in the interval
tinuous at
<x<
0<x<7r
for unity in
and x,
can be integrated term by term between the limits
This can be verified at once by comparing the Cosine
(cf . 70. 2).
namely
1
4/
cos x + n92 cos 3x +
Series for x in (0,
tt),
ir
x=
7T\
1
K
_, cos
5x+ ...
oa
==
tt
with the series obtained from the above by integration.

In the days when Fourier's Series were first used, term by term integration
was employed without any hesitation, both in the case of the Fourier's Series
and when the series considered was that obtained by multiplying the
Later it was seen tb
Fourier's Series term by term by another function.
such a step required justification. Hence the importance attached to t;
for/(x),
question of the uniform convergence of Fourier's Series in certain cases.
Bi
the theorem that follows shows that the presence or absence of uniformity of
convergence has
Fourier's Series
little
and
or no bearing on the subject of the integration of the

that,
even the convergence of the
series, is of
secondary
interest.
Lei f(x) be bounded
and
integrable in
- tt,
tt), or,
if unbounded,
let
f(x),
+ (a x
cos x
+ bx
cos 2x + 6 2 sin 2x)
sin x)
rx
J
+ (a 2
+ ...
Then
-tt
F(7r)
and
(n sin
nx + b n (cosn7r -cos
nx)),
when
Let
f(x)dx=aQ(x +7r)-f2~
y = F(x)
IT
the Fourier's Series corresponding to
converges or not,
\
Then, whether
namely
a
f(x)dx
be absolutely convergent.
f(x) dx - a
= x=i7r.
x.
f(x)dx-a Tr=a
Tr,
since
27ra
IT
f(x)dx.
J
F(-tt) = F(it).
Carslaw, Conduction of Heat (2nd
ed., 1921),
Ch. IV.
30.
TT
284
Also F(x)
And
it is
is
of
continuous in - ir, tv) ( 49).

bounded variation in this interval
[ch. viii
for,
with the notation of
36. 2,
S Wr+i-Vr
Hence, by
^1
2
\[
xr + l
n -1
:-2
f(x)dx 1^2
(xr
I
i-\
|/(*)|<fe^
\.m\dx^
xr
95, the Fourier's Series for J\a:),
which we write as
+ (.4! cos x + B x sin a:) + (J 2 cos 2x + B 2

its sum at every point of - tt ^ x ^ ir.
^,,
has
^(a;) for
when n
Also,
>
^4 n
=-
.F(a:)
7TJ
f*
mr)- n
("
=~ 1
raH J
...
cos nx dx
-F"(#) sin
nx dx
*
~ tto) Sin WX dX
_K
n
Bn = -
Similarly
sin 2x)
it
|/(*)|tf*
jP(a:)
sin
nx dx = -*.
GO
Therefore
jP(a;)
=^4
2-nJ (o w sin na; - bn cos
4-
?i:r).
But
since ^(tt)
= F{ -7r) =a
GO
7T
we nave
7r >
= ^4 + S w- ( n sm W7r ~ ^n
cos niT )'
Hence, on subtraction,
-n
f(x)dx=a Q {x + ir) + ^,1
It also follows that,
f{x)dx = a Q (x 2
-j-i
Again since i^z)
(a n sin 7ix
+ b n (cos mr -cos
when - tr < x x < x 2
<
nx)),
-tt^.x^tt.
ir,
-x 1 + 2 (on (sin'wa; a -'Snna; 1 )+6n (coBna: 1 -'cosa; s )).

n
l
is continuous in -tt^lxÎ it and JP(ir) = F( -7r), we know
)
that the Fourier's Series for F(x) converges uniformly to F(x) in this interval.
Hence term by term integration of f(x) can be repeated any number of
times.
110. Parseval's
Theorem on
Fourier's Constants.
In this section, as usual,

sn
=a + 2(r
cos rx
+ b r sm
rx )>
and a
a lt
...
bv
...
are Fourier's Constants for the function f(x).
*It is assumed that the rule for integration by parts can be applied, and that
This makes the condition attached to f(x) less general than in the
F\x) =f(x) - a
statement of the theorem. For another proof, see 110, IV.
.
109, 110]
We shall
285
prove that under certain conditions
M*
7T
[f(x)]*dx = 2a
+ %(a n* + b n %
rr
,,
a result usually called Parseval's Theorem.*

I.
Let f{x) be hounded and integrahle in
if unbounded,
tt), or,
-tt,
let
[f(x)]*dx
be convergent.
Then
2o
i r*
2
+i( + V)
TT
[f(x)fdx.
we have
[/(*)" 'J" ** = [*
I"
J" W
IT
[/(^)]
""
at once
Substitute for sn on the right-hand side, and
^-27r[2a 2 + i(r2 + V)]

1
+ 7r[2a 2 +
2K
+ V)]
It follows that
^ Jr
[/(*)
^ = - r
- *j 2
7T
TT
la*)]
^ - ô
+ (r 2 + &r 2 )^M
n [/(*)]
7r
-.
And
2a
U)
b r 2 )i
J
7r
2a
Thus
IT
UWfdx.
+ & n2 )=^r
J.(2)
had converged uniformly to the continuous

both sides of the equation
multiplied
have
function f(x), we could
But
if
this Fourier's Series
=a + (! cos x + b x sin x) +
f(x)
byf(x), and, integrating term by term,
[f(x)fdx = 2a
TT J
. .
we would have
2
obtained
+ ^(a n^ + b n %
it
in this case.
proceed to prove that this equality holds
We
attached to /(a;)
II.
is
that
it is
Letf(x) be bounded and integrable in

1
Then
TT
Since
*If the
Any
tr
sn
[/(*)]
dx = 2a
-tt,
+
1
TT
= J- (*
Air
Lebesgue Integral
,1
in
- tt,
it).
K+
2
& n2 ).
used, the following theorem also holds
a + (% cos x + &! sin x) + (a 2 cos 2x + b 2 sin 2x)
+b n2
integrable (L) in (-7T,

is
'gonometrical series
*
for which 1 (an
This
condition
tt).
f(x')(l+2% cos r(x'-x)) dx',
it
is
when the only
known
+ ...
converges is the Fourier's Series of a function whose square

TT).
as the Riesz -Fischer
Theorem.
is
286
we
have, as in
101,
%\rP\n{x' -x)
we know that
Also
ffif"*W =2ir(cf.p-256).
['
n -f(x) =
Therefore
<r
J-
[*
[/(*') -/(*)]
kn -/(*)
and
where
[ch. viii
ilf,
are the upper
^-fe^1^'
das',
(3)
^Jf-m,
(4)
and lower bounds oif(x)
in
-ir,
tt)
and
a;
is
any
point in this interval.
Now
and k be any arbitrary positive numbers.

from 42, III, that there is a mode of division
let e
We know
finite
number
the oscillation oif(x)
Let
is
of
- 7r,
7r)
into a
such that the sum of these intervals for which

greater than or equal to k is less than e.
of partial intervals
is
denote the intervals in this
and
greater than or equal to k,
mode
of division for
which the
oscillation
8 the other intervals.
from the intervals of 8 at each of their ends a part, so that the sum
segments cut off is less than e.
Let 8" denote the segment cut off, and 8' the inner parts of 8 which remain.
Cut
off
of these
Then
n -f(x)fdx = {j:\+^\+'2\
we mean that
where, by this notation,

the intervals of
Now
8',
let
8"
-f{x)f dx,
(5)
the integrals are taken respectively over
and A.
be one of the intervals of
let (a, b)
8,
and
(a',
V) the corresponding
cosec 2 A (x
8'.
interval of
Also
}[<rn
[a-
f*
be a point of
a;
Then from
(a', b').
(3),
Taking these three integrals separately, we have
[/ ( x
2n
-/(*)]
< -^
r dx
- x) dx
K
<,
n
where
K is a positive constant depending upon the position of (a, b) and (a',
Similarly
\2mr)b
we
[f( x
l
b').
find that
-/(*)]
J
sm291%{x'-x)r
.
dx
\
< -=2mr
cosec2 h (x -x) dx
)b
K
<,
n
where
later
it is
we may take the same

value we please.
clear that
by any
larger
value for
K in both, and replace
it
110]
287
Finally
2^r
I
[f{X
~ f{X)]
sin^-(V-4
)a
<
dX
2^ l
**
sinHK-^)
_*_[* siy?ln{x' -x) dx

2mr J-7T sin2 ^^'-^)
Thus, from
(6),
K-ZWI^ + sf)
But the sum
Therefore in
of the intervals of S' does not exceed
Also
(4),
[cr
&t
2f
since the
2-rr.
we have
(5),
x\
by
{7)
sum
h --f(x)fdx^27r^K +
-f(x)7dxê(M-m)\
[cr n
than
of the intervals 6" is less
And
S[
A
(8)
2^-J
-/(^)]
[cr w
(9)
e.
^ê(ilf-m)2,
(10)
.'A
for a similar reason.

It follows
from
and
(5)
(8), (9), (10),
limf"'
n>rj
since k
we
and
[<T n
that
-f(x)fdx = 0,
(11)
it
which can be chosen as small as
are arbitrary positive numbers,
please.
vn = a +
But
Therefore, as in
i r_
[<T
n 1/fi f\
\-j{a
cos rx
+ b r am rx).
we have
(I),
- f{x)f dx=
i i_
{x)fdx
2a 2
Y o^K
n
= {- T
%
+ -2 2
But we know from
-
f*
[/(^)]
r2
(r
+ &r2 )]}
that
[f( x )f
dx - [2a
Therefore from (11) and (12),
I*
TJ
Also
(12)
we
+' 2?
(< + &r
)]
^ 0.
see that, under the conditions stated in the
theorem,
-
+ ^ 2)
^-[2 2 + 2
+ &r 2
U{x)f dx = 2a 2 +
lim
(a r 2
+ 6 r 2 ).
(i^rW + VjW
288
III. Let
f(x) and g(x) he hounded and integrahle in
Fourier's Constants for f(x), a

n
Then
(3
f^)9^)dx = 2a
and
On
from
_-
ir)
and an
h n the
+ y(a n a n + hj n ).
-ir
We have
-ir,
n the Fourier's Constants for g[x).
(II),
[AX) ~ g{x)fdx = 2(a " Qo)2 +
[(a
a )2
- -
+ {b n ~
Pnn
subtraction, the result follows.
IV.
We may
put
g(x)
=0
in
x x ) and
-ir,
Thus we have the following theorem

Letf(x) he hounded and integrahle in
(x 2
ir),
where -ir
= x <x
1
îr.
rahle in (x Xi x 2 ) where
Then from
7r
^ x <x
x
-ir,
and
ir)
hounded and
g(x) he
integ-
^1tt.
(III),
f(x)g(%)dz=a
'*!
"
\
g(x)dx+'Z(an
yx-i
*
\
g(x) cos
}x 1
nx dx + hn
*
\
g(x) sin
nx dx
)x!
).
It will be seen that (IV) establishes the possibility of
term by term integrabounded and integrable function f(x), and

can be done whenf(x) is multiplied by another function
tion of the Fourier's Series for the

also
shows that
this
same class.
The argument in
of the
(II) is
taken from Hurwitz' proof of Parseval's Theorem.*
REFERENCES.
BoCHER, IOC. Cit.
De la Vallee Poussin, loc. cit. (2e ed., 1912), Ch. IV.
Hobson, loc. cit., 2 (2nd ed., 1926), Ch. VIII.
Lebesgue, Lecons sur les series trigonometriques (Paris,

Hilb
Enc.
d.
u.
loc. cit.
M. Riesz, "Neuere Untersuchungen
math. Wiss., Bd. II, Tl. Ill,
*Math. Annalen, 57 (1903), 425.
1906), Ch. III.
(5th ed., 1928), Ch. IX.
2, p.
1189
iiber trigonometrische
et seq.
Another proof
in his Cours d' Analyse 2 (2 e ed., 1912), 158.
is
Reihen,"
(Leipzig, 1923).
given by de
In his proof
la Vallee
f(x)dx
Poussin
is
to be
absolutely convergent: but the integration under the sign towards the beginning
of his demonstration is not discussed.
The question of the removal of the restrictions upon the functions involved is now of less interest, as these and other theorems
concerning Fourier's Constants are discussed with Lebesgue's Integral instead of
Riemann's.
CHAPTER IX
THE APPROXIMATION CURVES AND THE GIBBS
PHENOMENON
IN FOURIER'S SERIES
have seen in 104 that, when f(x) is bounded and

continuous, and otherwise satisfies Dirichlet's Conditions in
- ir<x<7r,f{7r - 0) being equal to/( - tt +0), and/'(z) is bounded
We
111.
Conditions in the same interval,

2
the coefficients in the Fourier's Series ioif{x) are of the order 1/n
and the series is uniformly convergent in any interval.
and otherwise
In this case the approximation curves

y = s n (x)
in the interval
- 7r<x<7r
with
will nearly coincide
taken large enough.

Jet f(x) be the odd function defined in (-tt,
-ir^xf--\ir,
as follows:
f{x)= -\ir(ir +x)
- \ir = X -= Jtt,
=
\irX,
f(x)
=
x),
\ir^kx^ it.
-J7r (tt
f(x)
when n
is
As an example,
tt)
The
Fourier's Series for f(x) in this case

1
is
O
K + K-92 sin ox
sm x - ^2 sin 6x
is
. .
which
the Sine Series

.
uniformly convergent in any interval.
The approximation curves
= sin x,
y sin x -
?/
^2
sin
2>x,
y = sinx- q2 sin 3x
+ ^ sin 5x,
are given in Fig. 29, along with y=f{x), for the interval
289
(0, tt),
THE APPROXIMATION CURVES AND THE
290
and
the
it will tyj
sum
seen
how
[oh. ix
closely the last of these three approaches
of the series right
through the interval.
Fig. 29.
Again,
be the corresponding even function
let /(a;)
^ X ^ - f TT,
Jtt < X <
f{x)={ir{7T+x),
TT
f(x)=l7TX,
f( X )
The Fourier's
16
7T
= J-7T (tt - X),
llf^Xti
Series for /(a) in this case
^2 cos 2x
2
|
is
V = iV^
the Cosine Series
+p COS 6iC +2 cos lOx +
which is again uniformly convergent in any

The approximation curves
interval.
cos2#,
y-
IT.
i cos 2#
- TV cos
Gx,
...
Ill, 112]
GIBBS
PHENOMENON IN FOURIER'S SERIES
291
are given in Fig. 30, along with y=f{x), for the interval (0, tt),
and again it will be seen how closely the second of these curves
approaches y =f(x) right through the interval.
0)
(2)
Fig. 30.
It will
of
be noticed that in both these examples for large values

of y = s n (x) nearly agrees with that of y=f(x),
n the slope
except at the corners, corresponding to x=\tt, where f'(x) is

This would lead us to expect that these series
discontinuous.
may
be differentiated term by term, as in fact
112.
When
the function f(x)
f(x)=X,
is
x- |
sin
the case.
given by the equation
-7T<X<7T,
the corresponding Fourier's Series

2 {sin
is
is
the Sine Series
2x +J sin 3x-
...}.
The sum of this series is x for all values in the open interval
- 7r<x<Tr, and it is zero when x 71-.
This series converges uniformly in any interval ( - it +S, w-S)
contained within ( - 7r, ) (cf. 107), and in such an interval, by
taking n large enough, we can make the approximation curves
77-
about y = x as closely as we please.

it was wrongly believed that, for large values
each approximation curve passed at a steep gradient from the
oscillate
Until recent years

of n,
292
point
y=x
7T,
till
0) to a point
near
tt,
steep gradient from a point near
who
tt),
x approached the value x,
(tt, tt)
did not properly understand what
infinite series,
and then
when
is
[oh. ix
oscillated
about
the curve passed at a
And to those
meant by the sum of an
to
(ir, 0).
the difference between the approximation curves
for large values of n,
and the curve y = lim
s n (x) offered consider-
able difficulty.
In Fig. 31, the
line
y = 2 (sin x -
y = x and the curve
} sin
2x
+ }s
sin 3#
-1
sin ix +\- sin 5x)
are drawn, and the diagram might seem to confirm the above
y
FIG. 31.
view of the matter

near one end of the
namely,
line,
that there will be a steep descent

from the point ( - 7r, 0) to near the point
and a corresponding steep descent near the other end of

the line. But it must be remembered that the convergence of this
series is slow, and that n = 5 would not count as a large value of n.
(
x,
7r),
GIBBS
112, 113]
PHENOMENON
IN FOURIER'S SERIES
113. In 1899 Gibbs, in a letter to
the approximation
quite a different
sum.
He
way
at the points of discontinuity
falls
from the point
point very nearly at a depth
above and below y = x
= s n (x),
the axis of
close to this line until
when
it falls
large
then
x,
x approaches
|
it,
in the
for
x, 0) at a steep gradient to a
2*^ dx below
J
oscillates
Nature* pointed out that

behave in
curves for this series do, in fact,
stated, in effect, that the curve y
values of n,
293
from a point very nearly at a height 2

J
ci Tl
Of*
dx
above the axis of x at a steep gradient to (x, 0).

The approximation curves, for large values of n, would thus in
- 7r, 7r) approach closely to the line y xoi Fig. 32 with the lines
(
parallel to the axis of y as
drawn
in that figure.
Fig. 32.
His statement was not accompanied by any proof.
Though
the remainder of the correspondence, of which his letter formed

a part, attracted considerable attention, this remarkable observation passed practically unnoticed for several years.
In 1906
Bocher returned to the subject in a memoir f on Fourier's Series,

and greatly extended Gibbs's result. He showed, among other
things, that the phenomenon which Gibbs had observed in the
case of this particular Fourier's Series holds in general at ordinary
points of discontinuity.
To quote
*Nature, 59 (1899), 606.

f Annals of
Math.,
%loc. cit., p. 131.
(2),
7 (1906), 81.
his
own words:*
294
[ch. ix
Iff(x) has the period 2?r and in any finite interval has no discontinuities other than a finite number of finite jumps, and if it has a
derivative which in any finite interval has no discontinuities other
than a finite number of finite discontinuities, then as n becomes
infinite the approximation curve y = s n (x) approaches uniformly the
continuous curve made
(a) the
up
of
discontinuous curve y=f(x),
number of straight lines of finite lengths parallel

to the axis of y and passing through the points a lt a 2 ... on the axis
ofx where the discontinuities of f(x) occur. If a is any one of these
(b)
an
infinite
points, the line in question extends between the two points whose
ordinates are
f(a-0) +
where
DP
AN
D is the magnitude of the jump inf(x)
-I
BP
f(a+0)
smx
dx=
at
a* and
-0-2811.
Fig. 33.
of
Tins theorem is illustrated in Fig. 33, where the amounts
Both
positive.
and
negative
the jumps at a v a 2 are respectively
Gibbs and Bocher thought they were describing properties previits
ously unknown. However, in 1848, Wilbrahamf had noticed
*i.e.
D=f(a + 0)-f(a-0).
\Camb. and Dublin Math. Journal, 3 (1848),
198.
113, 114.
1]
GIBBS
295
occurrence in the approximation curves for the series
5^ ( _ ir-x!^
^
And
in 1874
and
_
(
irl
du Bois Reymond
conclusion, both for Fourier's Series
cos(2r-r)x
2r -
would have reached the same

and Integrals, had he not made
a curious slip in dealing with the integral
dt,
when n->cc
and ->0 simultaneously. In recent years a number of other

writers have dealt with the matter, and the property in Fourier's
and other series is now well known as the Gibbs Phenomenon.^
114.
of this
1. The series on which Bocher founded

and other extensions of Gibbs's theorem
sin
x+\
which, in the interval
sin
(0, 2ir)
as follows:
2x
sin 3x
+J
+ ...
his
demonstration
is
represents the function f(x) defined
/(0)=/(2tt) -=0,
f{x)
= \(ir-x), 0<x<2tt.
In this case
/
s n (x)
= smx+-1
sm2ic
+ ...
sin nx
1
-\
(cosa+cos2a+... +cos no) da

Jo
^J
The properties
of the
*
sm \a
maxima and minima
of s n (x) are not so
easy to obtain, % nor are they so useful in the argument, as those of
R n {x) = i(ir-x)~s n {x)

sm(n+i)a
fe
la
sm fa
o
memoir Bocher dealt with the maxima and minima
ip
'J n
In his
of
*Math. Annalen, 7 (1874), 241.

fCf. Enc. d. math.
Wiss., Bd. II, Tl. Ill, 2, p. 1203.
referred to on pp. 1203-4, the following
may
In addition to the papers
be named:
Circ. Mat. Palermo, 29 (1910) and 30 (1910).

Cooke, Proc. London Math. Soc. (2) 22 (1928), 171.
Wilton, Journal fur Math., 159 (1928), 144.
And a historical note by Carslaw, Bull. Amer. Math. Soc, 31 (1925), 420.
Weyl, Rend.
JGronwall discussed the properties of s n (x) for this series, and deduced the Gibbs
for the first wave, and some other results.
Cf. Math. Annalen, 72
Also Jackson, Rend. Circ. Mat. Palermo, 32 (1911).
(1912).
Phenomenon
296
R n (x),
[ch. ix
more than once to the fact that

the height of a wave from the curve y =f(x) was measured parallel
This point has been lost sight of in some exposito the axis of y.
and
called attention
lie
tions of his work.
114. 2.
However the
series
2(sin#+^sin 3z+^sin 5x+...)

has certain advantages as an approach to the Gibbs Phenomenon.
This
is
the Fourier's Series for/(ic),
- 2 tt
f{x)=-lir
for
\ir
when
-tt<x<0
0<x<tt
for
/(-)=/W=/(o).=o.
and
Let
Then
= 2 2j
s n (x)
fa
and
nv
Cx sin
'
'
Thus
s nV
n (sc)'
2m;
f
da =
sina
(i)
da.
psin2na
sin x
2nt
sin
sJx)
nV
Also
cos (2r-l)x =
f*
sin
/
2na
/a
(--=
sma\3! r
o
da=\f* sm2na-.
,
nr
ft
<^
sponding positive number

I
2 **
UnW- Jo
I*
r\,
sin
a:
when 0<xî-7r.
number
e,
there
fla|<e
when
_
Oâ;^^
this holds for all values of n.
Thus,
if
we choose n
so large that ~-<r),
\2w/
is
a corre-
such that
and
\,
to \ir as x passes
H îilo**
Cz/n% glTl
s - (a
we take
.+... Ida.
5!
aa.
and 0<^-^+...<o-, when 0<x^|tt.
to \-k
Thus
If
a/
a3
But x cosec x continually increases from

from
1\
Wa
we have
<e;
(2)

GIBBS
114.1-2,115]
/t\
that
fir
fsin a
|,^_j_^_ + |J^_ -&|;|< e
is
297
This shows that the approximation curves y = s n (x)for this series

sum in the right-hand neighbourhood of the origin by
rise above the
da
nearly
\J
ir
is sufficiently large.
^
dx
X
The integral
J
Or,
when n
it
known
is
we may put the matter
as follows
si
y=\
Let
maximum
give
ir
- 0-2811.*
da = <p(x).
when x fir
of this curve occur
The turning points
minimum
n a
a
Jo
multiples of
to be approximately
the odd
ordinates; the even multiples give
ordinates.
The maxima
when zôo
continually decrease from
The minima
when x-ôo
increase from
continuallv
"
da towards
\ir
da towards
\ir
2n,
we
Jo
the abscissae of this curve are reduced in the ratio
If
obtain the curve

[
2nx sin
H. ir da
and,
and
when n
'
increases, the turning points of the curve
closer to x
= 0,
the
first
come
closer
(and largest) ordinate being always at

30
7r
a 7
If sin a ,
da
-da = -x+\\
a
a
2
2w'
|J,
J
By (2), the curve y = s n {x) in the neighbourhood of the origin
C2.nx gj-Q
a
da.
differs by the arbitrarily small e from the curve y=\
7T
x jr- and of height

&
-,
-,
""
sin
"
The Trigonometrical Sum

115.
The discussion
2j
(2r-l)#
sin
2r^-l
in 114. 2 establishes the existence of the
Gibbs Phenomenon in the Fourier's Series 2
V
*Y*
In
>
anc*>
we shall see in 117, it can at once be deduced that the phenomenon will appear at x a in the approximation curves for the
as
*Cf. Bocher, loc.
cit.,
p. 124.
298
[ch. ix
any function f(x) with an ordinary discon(-w, ir).
Fourier's Series for
tinuity at x = a, iif(x) satisfies Dirichlet's Conditions in
But the approximation curves

y = 2 f sina;+o
sm ^
+~
+...
sin(2n- l)x)
= s n (x)
are worth a more detailed examination, for it will be found that

from the properties of their turning points and their graphs, the
Gibbs Phenomenon is exhibited in the clearest possible manner.*
In this section we proceed to obtain the properties of the
and minima
of these
approximation curves
?/
= 2V
when 0<<7r.
I.
any integer m,
Since, for
sin
it
maxima
(2m -
1)(Jtt +x')
= sin (2m - l)(\ir - x'),
follows from the series that s n (x) is symmetrical about x
and x ir
and when x =
\ir,
zero.
it is
When 0<x<tt, s n (x) is positive.

From (I) we need only consider ^<x~\-k.
II.
We
have,
by
114.2(1)
1 C2" x
sin2na 7
da=\
2wJ
Jo sin a
Cx
,
s n(
nv x )=\
.
'
The denominator
increases
successive
may
or
the
in
sin
Q<x--lir.
*
da,
"
2n
in the integrand is positive
interval
By
integration.
of
waves in the graph
may
sin
of
not be completed,
and continually
considering the
sinacosec the
,
An
it
is
last of
which
clear that the integral is
positive.
III.
The turning points of y sjx)

(
3tT
7T
2ra'
7T
x9 =
"2n'
xx
""
'
...
2ft -1
^n-i- 2n
2tt
,
are given by
#o(ra
it
n n-\ it
[maxima),
.
(minima).
* 115-117 are founded on a paper by the author in the American Journal of

The computations for Fig. 34 and Fig. 35 were
Mathematics, 39 (1917), 185.
made by Mr. F. G. Brown, B.Sc.
GIBBS
115]
We
PHENOMENON
299
have
Psin2na-,
y=\
9
dy s'm2nx
-y-=
sina?
dx
-.
and
da,
-.
sina
The
IN FOURIER'S SERIES
result follows at once.
IV.
As we
proceed from x =
continually diminish, and

increase, n being kept fixed.
to
x = \ir,
heights
the
the heights of the
of the
minima
maxima
continually
My
The curve y = sn {x), when n = 6.

Fig. 34.
Consider two consecutive

/Own
namely,
less
sn
(9 *)
and
sn
than or equal to \ (n -
maxima in
(2m +1 \
T?)>
2n
We
1).
0<x = \ir,
the interval
m bein g a positive integer
have
2m
2m- 1
IT
~2n
%n
2wJ (2m +l)ir
sm-
2n
2^|
r2m7r
J (2m(2m-l)ir
sina
sin
x
The denominator
r(2m+i)^
2tott
in both integrands
-v
sin
~~
2n
2w
tinually increases in the interval
sma
is
positive
and
it
(2m- 1)tt = a= (2m + l)7r;
conalso
continually negative and in the second

values for elements at equal
absolute
the
positive,
continually
the numerator in the

distances from
first is
(2m- \)tt and 2mir being
the same.
300
Thus the result follows.

examine the sign of
Similarly for the minima,
(m - 1
m is
where
V. The
(m
[ch. ix
we have
to
a positive integer less than or equal to \n.
maximum
first
height continually
of x =
to the right
x = ^~ and
is at
When n
diminishes as n increases.
its
tends to
infinity, its limit is
S\l\X
dx.
y
^0
U
"^^-^
L.
/Tbcx'^-y'
J^"
>^"~~^'OC>?y
_j2^ V X^^- -^~=^-U ÛZ^/^-N\X

JZZA^5a2s2saS:
=S^v25\/v2iS
ifr^V* ^/v*-*vb^--TRk^r^^^r^vn
tn
ESitzt <>*tzz-7 y^r'^\xyJ\ \\\
^
1-5
tu-4
Wt-t
^~ -
ttf -t
44 tul Z
**JtT vt
M-t
4U
v
5
X--
y*f
vL
JXV
f St
44 t
Vs
-L-
31
CI
V-
5 \\
-J-
M-T
&$tt7
**ttt
A|
\\1
M-Zi
t/
-W-~^-9-b-+-+$-- -Hi-r-29-+i5The curves y = s n (x), when n = l,
2, 3, 4,
yi
VI
5 and
|^
m-
6.
Fig. 35.
We
have
7T
sma
f^sin2wa
n
\2n
Thus
sma
""
.
si
sin
a cosec - da.
In
2ft J
IT
UJ
Sn+1
l-x cosec A
2n
aa = _-
2n+2
2n
x cosec -
2w+2-
Since a/sin a continually increases from
from
to deal
to
ir,
it is
J_
2n
to oo
clear that in the interval with

_a_
COSeC
2n
1_
2n+2
cosec
2^+2
>0.
da.
as a passes
which we have
IN FOURIER'S SERIES
PHENOMENON
GIBBS
1151
7T
Thus
Sn
But, from
\2{n+iy
positive when 0<z<7r.
{^-\ tends to a limit as n tends to
infinity.
can be obtained by the method used by
of this limit
The value
>0.
Sn+1
{2n)~
(II), s n (x) is
It follows that s n
301
[ ^Âl
Bocher for the integral
'"
But
da.*
it
readily
is
obtained from the definition of the Definite Integral.
/7r\(2n
/ 7r \
TT
3t
2w-l
2n
^
S i n n
T
+ (2w-l)7r
T^
2n
2w
&-1)tt
.
MZ
=1
Therefore
lims n = 2
!
\2n) *)
/ 7T \
""
sin
*"
ic
<fo-
sin
Jo
The
its
th
maximum
first
a?
dx.
wave
a special
is
to the right
tends to infinity,
2r-l
of x =
height continually diminishes as
When n
* sin
Jo
VI. The result obtained in (V) for the

case of the following
dx=\f
is at
x 2r _ 1 = ^- x, ana
increases, r being kept constant.
limit is
its
aI)* gm a;
dx,
r.
which
is greater
The
r'
than
minimum
\ir.
to the right
continually increases as
ofx =
is at
increases, r being
S
tends to infinity,
its
x 2r =îr, and
limit is
jo
^-dx,
hep
which
its
constant.
is less
than
height
When
Jtt.
these theorems we consider first the integral

a \ j
1
a
C mn
( 1
da cosec
cosec
sm a
-
To prove
J,
(SS
being a positive integer
0< <tt
2n
2^T2
less
than or equal to 2n-l, so that
2^+2)
in the interval of integration.
* Annals
of Math., (2), 7 (1906), 124.
Variable, 2 (2nd ed., 1926), 494.
Also Hobson, Theory of Functions of a Real
302
Then
F(a)
cosec
2n
in this interval.
^--2n 2n+2
cosec
[ch.
ix
>0
2n+2^
(Cf. (V).)
Further,
C0S
(2^W
=a
where
<p
2^T2
{0 cos
and
But
\jr
cosec 3
2
\Js
cos
cosec 2
x/r
\/r} }
= aj2n.
-y- (0 2 cos
=-
2^2 - <k? C S C0Se
cosec 2 -^ -
= a/(2n+2)
C0Sec2
cosec 2 0)
[0(l=Fcos 0) 2 2 cos 0(0=Fsin
<p)].
And
the right-hand side of the equation will be seen to be

negative, choosing the upper signs for O<0<|tt and the lower
for |7T<</><7T.
Therefore
It follows,
from the expression
cosec 2
cos
diminishes as
increases from
to x.
for F'(a), that l?'(a)>0,
and
F(a) increases with a in the interval of integration.

The curve
^sinzficosec^-^cosec^y,
...
thus consists of a succession of waves of length
above and below the
0<x<m*,
alternately
7r,
and the absolute values of the ordinates

at points at the same distance from the beginning of each wavp.
axis,
continually increase.
It follows that,
sin a f
negative
is
positive.
and,
4,
...
2(n -
1),
the integral
\j
-
cosec 2n
cosec
da
2n+2
2n+2/
1
Jo
is
m is equal to 2,
when
m?r
2w
when
m is
equal tol,
3,
...
2ft-
1,
this integral
Eeturning to the maxima and minima, we have, for the

to the right of x = 0,
S n\ X 2r~l) ~ S n + x
l\ 2r-l)
maximum
2r-l w
2
2nl
>
2r-l
sin
P"
J
sin a
Jo
2na
|?<+i>
1)7r
sma fh
2w
cosec
sin2(^+I)a
sin a
2n
2n+2
cosec ^
- \ da.
,
2n+2J
Therefore, from the above argument, s n (x _ )>s

n+1 (x 2r ^ 1 ).
2r 1
th
115, 116]
GIBBS
th
Also for the r
minimum
C2rn
s n (x 2r )
IN FOURIER'S SERIES
PHENOMENON
W^)=J
( 1
* a [ Tn
a
cosec
303
we have
a
^-^fg
C S6C
-j
2rT^ **'
s n (# 2r )<s n+1 ( 2r ).
and
By an argument
similar to that at the close of (V)
hm* w
(a> )
r
.->
Jo
we have
- dx
for
It is clear that these limiting values are all greater than Jtt
minima.
the
for
than
\ir
less
and
positive
and
maxima,
the
The Gibbs Phenomenon for the Series

2(sin x
116.
From
+1
sin 3x
+ V sin 5x +
the Theorems I-VI of
Gibbs Phenomenon for the

2(sin
+1
sin 3x
115
).
. .
.
all
the features of the
series
+\ sin bx + ...)...,
^ x ^ x,
tt
follow immediately.
It is obvious that
we need
^ x ^tt,
only examine the interval
and that a discontinuity occurs at

For large values of n, the curve
x 0.
y=s n (x),
where
s n (x)
= 2(sin z+gsin 3x + ...
+^T^ sin
a steep gradient from the origin to
its first
very near, but above, the point [0,
~^~
2n ~
l )x
)>
rises at
maximum, which
dx
115, V).
is
The
curve, then, falls at a steep gradient, without reaching the axis

of x ( 115, II), to its first minimum, which is very near, but below,
the point
(o,p
&)
(115, VI).
It
then oscillates above
y = \ir, the heights and (depths) of the waves

we proceed from x = to x = \ir ( 115,
=
to
from
x
and
\tt
x = tt, the procedure is reversed, the
IV);
being symmetrical about x = \ir
interval
0^~tt
curve in the
and below the
line
continually diminishing as
115,
I).
The highest
(or lowest) point of the r
of
For the values
(*"
th
^^ dx, see Annals of Math.,
wave
(2),
to the right of
7 (1906), 129.
304
x=
r
-
values of n, be at a point whose abscissa
will, for large
115, III)
[ch. ix
and whose ordinate
is
is
very nearly
-<Z.x(115, VI).
x
'o
as
By
increasing n the curve for
we
please to the lines
t^x
^ ir
_
0<y^\f
A
x = 0,
71"
sin
x=
dx,
2'
0<y^\
tt,
0<x<tt,
can be brought as close
sin
a;
dx.
Jo
Fig. 36.
We may
(i)
state these results
If e is
any
more
definitely as follows
positive
please, there is
\i7r-s n (x)\<,
when n
^ v e^x^\ir.
,
This follows from the uniform convergence of the Fourier's

Series for f(x), as
denned
in the beginning of this section, in
an
interval which does not include, either in its interior or at an end,
a discontinuity oif(x)
(ii)
(cf.
107).
Since the height of the
tends from above to

'
first
maximum
to the right of x =
dx as n tends to
infinity, there is a
positive integer v" such that
0<s n
C" sin
2n
Jo
dx<e, when n
^ v"
116, 117.
(iii)
1]
GIBBS
Let v" be the integer next greater than
maximum
abscissa of the first

is less
than
It follows
from
(i),
behaves as follows
It rises at
is
Then the
n^ v",
when
e.
positive integers v,
which
305
(ii)
v"
and (iii) that, if v is the greatest of the

and /", the curve y = s n (x), when n^v,
a steep gradient from the origin to
above
J o
its first
maximum
dx and within the rectangle
0<x<
0<y<\
sin x
dx+e.
Jo
After leaving this rectangle, in which there may be

it remains within the rectangle
many
oscilla-
tions about y = lir,
e<X<7T-
Finally,
it
^7T
~<y<^7T
+.
enters the rectangle

T HJT1
7r-e<x<7r,
0<y<\
J o
and the procedure
in the first region
is
OJ
dx+e,
repeated.*
The Gibbs Phenomenon for Fourier's Series
an ordinary discontinuity
117. 1. Let f(x) be a function with
when x = a, which
in General.
Conditions in the interval
- ir^xîr.
Denote as usual by f(a+0) and f{a-0) the values towards
which fix) tends as x approaches a from the right or left. It
will be convenient to consider /(a +0) as greater than f(a -0) in
the description of the curve, but this restriction
is
in
no way
necessary.
1
Let <p(x-a) = 2^p jn~--rvsm(2r-l){x-a).

/
Then
<p(x-a)=
Jtt,
<j)(x-a)= -lir,
0(+O)=
|tt,
<p(0) =
*The cosine
when a<x<w +a,

when -7r+a<x<a,
<M-0)=-!tt,
and
<p{x)
= <t>{x +2tt).
series
cosa;-^cos3a; + -cos5a; + ...
which represents
in the interval
can be treated ih the same
way
^x <
\tt
and
\ir in
the interval
as the series discussed in this article.
\tt
< x ^tr.
306
Now
put
=f(x) - \{f(a +0) +f(a - 0)} -
yf,(x)
^ {f(a +0) ~f(a - 0)}<j>(x - a),
and let/(a;), when x = a, be denned as \{f(a

Then i/x (a +0) = i/r (a - 0) = V' (a) = 0, and
The following
Since
X
I
If
is
r)
for
continuous at x a and
\ls(x) is
(a) is
argument are numbered
distinct steps in the
the sake of clearness
+f(a-0)}.
continuous at
+,0)
= &.
(i)
[ch. ix
number
please, a
I'( X )\
not originally
\[s
r\
(a)
= 0, if e is a positive
exists such that

less
than
we can choose
e,
this part of
r\
for our interval.
can be expanded in a Fourier's Series,* this
(ii) \js{x)
a^xrQ /?
uniformly convergent in an interval
interval which includes neither within
it
nor as an end-point any
other discontinuity off(x) and cp(x-a) than x = a
Let
s n (x),
(j)
(x-a) and
\p-
being
series
contained within an
(cf.
108).
be the sums of the terms up to
n (x)
and including those in sin nx and cos nx in the Fourier's Series

ff)(x-a) and \}s(x).
Then e being the positive number
for f(x),
of
(i),
we
as small as
please, there exists a positive integer v such
that
when n^v',
\*lsn( x )-i!'( x )\<fc>
the same
Also
/?.
r
IV n(*)|ÎV r n()-^()|+|^(aj)|<ie+ie
in |x-a|^>/,
(iii)
=\x :g
serving for every x in a
Now
if
right of x = a
a<a-^<a<a + */</?,
if
is
even, the
7r
is
at a
and
maximum
first
if
and n
is
odd,
= Jc,
= v.
in
it is
</>
n (x-a) to
at a
+ -^w+
the
In
either case there exists a positive integer v " such that the height
of the first
n
(
sin x
Jo
(iv)
This
maximum
7
-.
dx and
dx+ iVrr
T
e
maximum
v
2{/(a+0)-/(a-0)}'
.
Jo
first
between
lies
f" sin
will
have
its
when w>/'.
abscissa between a
and
a +rj, provided that --<rj.
*If f(x) satisfies Dirichlet's Conditions,
that
\f/
(x)
does so also.
it is
clear
from the
definition of <p(x-a)
117.
GIBBS
1]
Let v" be the
307
positive integer which satisfies this in-
first
equality.
a+rj^x^
In the interval
(v)
converges
s n (x)
/?,
{iv)
Therefore a positive integer
to f(x).
uniformly
exists such that,
when
\f(x)-S n (x)\<e,
the same
iv)
j/
serving for every x in this interval.
Now, from the equation
defining
(x),
\[s
/(a+0) ' /(a "
(*)
= H/(i+0) +/(a-0 +
IT
V n(x-) + ,(*).
from (i)-(v) that if v is the first positive integer greater

v" and v {[v) the curve y = s,n (x), when n^v, in the
interval a^x^ f$, behaves as follows
When x = a, it passes through a point whose ordinate is within
i f 2 (/( a +0) +/(a-0)), and ascends at a steep gradient to a
It follows
than /,
v",
point within
e of
{/(<*
+0) +/( -
0)}
/(+0)-/(-0
+
7T
This
may
f-
sin*
Jo
&
be written
/(a+0 )./
!+'')-/(-'
and, from Bocher's table, referred to in
f'E! fc|
115,
we have
X>S
-0-2811.
dx=
x
then
oscillates about y=f(x) till se reaches a + rj, the character

waves being determined by the function <p n (x - a), since the
term \js n (x) only adds a quantity less than k to the ordinate.
And on passing beyond x = a +rj, the curve enters, and remains
It
of the
within, the strip of width 2e enclosing y=f(x)
On
from x = a+r) to
the other side of the point a a similar set of circumstances
can be established.
Writing D=f(a+0)-f(a-0), the crest (or hollow) of the
wave to the left and right of x = a tends to a height
/(a-0)+ ,/(a+0)-^?,
1
7T
where
Px =
.' IT
^^
dx.
X
7T
first
308
[ch. ix
The argument of the previous section can be at once adapted to the

by Arithmetic Means, and it will be seen that, if we can
summation
case of
show that the Gibbs Phenomenon does not occur in the approximation curves
for a single example of Fourier's Series with this method of summation, it
cannot appear in these curves for any function satisfying the conditions of
117. 2.
a point of ordinary discontinuity.*

however, easy to show that it does not occur in the curves y=v n (x)
for the Fourier's Series corresponding to the function in 114. 2.
For this case, we have f(x) = - |tt, when -7r<a;<0, &ndf(x)=$Tr, when
101, at
It
is,
<X<
IT.
But from
101
we know that, when/(x) is bounded and integrable in - tt,

(
1 n,- 1
(*) = -
rn
sin
J*r=&]-/<*
sin2 in(x'
("
-x)
Thus, for the function with which we are
|w {x' - x)
W*W^
.
'
now concerned,
>sm*%n(x '-x)
f*
< In)-^mH(x'-x)
dX
,
'
It follows that
< x < are all positive.

s n when
< a-n (x) < iTr, when < x <
But we know by
By
115, II,
wholly within
the
(0,
sums
tt,
tt.
101 that
o- n (x)
\tz in
any interva
tt).
Thus the Gibbs Phenomenon does not occur when the Fourier's
2(sin x
is
summed by
}.
sin 3x
+\
sin
5x +
. .
Series
.)
Arithmetic Means.'j"
In this example, by
114.
(r 2rHi(-r
It
tt).
)-
2 (1)
sin 2rt
f'isi
2n +
sint
J o~sT
can be shown that, as we proceed from
dt.
to
|tt,
the ordinates of th?
maxima continually increase, and that the same holds of the minima.
\tt, has the greatest ordinate for the whole
The last maximum in
^ x^
and when wôo this tends to \iv from below.

to
The curve y=<r Zn+l (x) for ?* = 6 is given in Fig. 37, and it is interesting
to
compare it with the curve of Fig. 34, which corresponds, with this notation,
y = s 2n (x)ioTn = 6.
n {x) are
In the case of the ordinary sums the approximation curves
brought within the shaded part of Fig. 38 by taking n large enough.
interval,
ys
by Fejer in Math. Annalen, 64 (1907), 273.

"
(2r- 1)3
_
~ 2 si n s
jit should be noticed that with the notation of this section 2 ?i-i
I
2r
^
since the coefficient of sin 2nx is zero.
and that s _! =s
*For another proof see
2n
2n ,
4 of a paper
'
117.2]
In
the
PHENOMENON
GIBBS
case
of
the
arithmetic
IN FOURIER'S SERIES
means
y=(r n (x) are brought within the shaded part
the
of Fig.
approximation
309
curves
39 by taking n large
enough.
Fig. 3:
Fig. 38.
Fig. 39.
The width
we please.
of the
shaded part in these diagrams
may
be taken as small as
THE APPROXIMATION CURVES
310
REFERENCES.
Bocher, "Introduction to the Theory
(2),
of Fourier's Series,"
7 (1906).
Gibbs, Scientific Papers, Vol. II, p. 258 (London, 1906).

And the other papers named in the text.
Annals of Math.
CHAPTER X
FOURIER'S INTEGRALS
118.
When
the
sum
function f(x)
arbitrary
Conditions in the interval
/,
I),
we have
Dirichlet's
satisfies
seen in
98 that the
of the series
f(x')dx'
+]l'
J(x')co S
^(x'-x)dx'
(1)
equal to J [f(x +0) +f(x-0)] at every point in -l<x<l

where f(x +0) and f(x-0) exist; and that at x= 1 its sum
is
i [/( - ^ +0) +f(l - 0)], when these limits exist.
Corresponding results were found for the series
is
jl f(x') dx'
and
2 cos
V sin
-j-
x
\
in the interval (0,
-j-
x\ f( x ') cos
-j-
dx',
'
(2)
f(x') sin -j~ x' dx',
(3)
I).
Fourier's Integrals are definite integrals which represent the
They
arbitrary function in an unlimited interval.
but not established, by the forms these
series
are suggested,
appear to take as
tends to infinity.
If
is
taken large enough,
and we may neglect the

f(x)dx
J -
is
irjl
first
convergent.
may be made
term in the
as small as
series (1),
we
please,
assuming that
Then we may write
00
i
i
as
-
Aa
a
Xr
I
f( X ') C0S
J -I
~T~
f(x')cosAa(x'-x)dx'+Aa\
7T
where Aa = 7r/Z.
311
X ~ X ) d>X
'
'
cos 2Aa(x'-x)dx'
-:...
FOURIER'S INTEGRALS
312
Assuming that
this
sum has
a limit as lôo
which, of course, would have to be defended
be regarded as in any
-
way
and
it
complete,
da\
7T,
7!"Jo
-x
[oh.
its
f(x') cos a
if
an assumption
the proof were to
value would be
(x'
- x)dx'
would follow that

f(x')coaa(x'-x)dx' = i[f(x+0)+f(x-6)l
-f da\
J -
7TJ
oo
when
00
<<00
these limits exist.
In the same
way we
are led to the Cosine Integral
Integral corresponding to the Cosine Series

,
and Sine
Series
/GO
.-Q0
da\ f(x') cos ax cos ax' dx'

Jo
'Jo
poo
a
= i[/(*+0)+/(*-0)],
/.x
2'
,
and the Sine
0<x<oo,
da\ f(x') sin a# sin ax' dx'
=*.[/(* +o)+/(*-)l
when
these limits exist.
must be remembered that the above argument is not a proof

any of these results. All that it does is to suggest the possi-
It
of
bility of representing
an arbitrary function /(a?), given
of x, or for all positive values of x,
We
shall
now show
by these
for all values
integrals.
that this representation
is
possible, pointing
out in our proof the limitation the discussion imposes upon the
arbitrary function.*
119. Let the arbitrary function f(x), defined for
satisfy Dirichlet's Conditions in
any finite
interval,
all
values of x,
and in addition
/OO
let
J -
CO
Then
f{x')cosa(x'-x)dx' = i[f(x+0)+f(x-0)]
-fdaf
J -oo
TTJ
at every point
where f(x +0) andf(x - 0)
exist.
upon the value of x for which we wish to evaluate

can choose a positive number a greater than x
we
integral,
the
bounded in the interval a ^ x'^ b, where b is
is
such that f(x')
Having
fixed
'See footnote, p. 230.
FOURIER'S INTEGRALS
118, 119]
and
arbitrary,
with our definition of
\f(x')\dx' converges, since
number
the infinite integral only a finite

continuity were admitted
It follows that
313
of points of infinite dis-
( 60).
a (x'
/(se')cos
dx'
Ja
converges uniformly for every
a,
so that this integral
is
a con-
x)
da
tinuous function of a (83, 84).

da\ f(x') cos
Therefore
JO
Also,
<x(x'
exists.
by 85,
f(x') cos a (x'
da
- x)
dx'
=
Ja
Ja"
Jo
=
But
-x)dx'
Ja
x'
- x ^ a - x>0
And
in x'
^ a,
dx'\ f(x') cos a (x'

Jo
f{x)
*M^
dx
(1)
we have chosen a>x.
since
\f(x')\dx' converges.
Ja
Therefore
dx' also converges.

x x
-^-r
Ja
sin
It follows that
ff(x')
Ja
converges uniformly for every
^-^) dx
XX
q.
Thus, to the arbitrary positive number

positive number A>a, such that
I
\]
the same
,,
'
f( x
JK
>
there corresponds a
smq(x' -x) ,
dx <|e, when A'
*t
x -x
,
'
)
'
serving for every value of
But we know from
~A>a,
(2)
the
interval
q.
94 that
lim
f /(x <)?iM(z^W
XX
q>-x> J a
sm qu
j{u+x)
= lim
q-^-xj
Ja - x
du
= 0,
since
(a
- x>
f(u+x)
A- x),
satisfies
Dirichlet's
Conditions
and both these numbers are
in
positive.
FOURIER'S INTEGRALS
314
Thus we can choose the positive number
from
(2)
= Q.
(3)
that
-x
'
\} a
(3)
smq(x>-x)
If"
when
and
Q, so that
<Je, when q
x'-x
'
\) a
It follows
q(x'-x)
sin
||* JK
[CH.
dx
<l e +i = e,
= Q.
Thus
lim
q-oo
and from
f( x ') cosa(sc' -a?) dx' =
da
(1),
87,
Ca
Ca
CQ
da\ f(x) cos a [x
- x)
=1
dx'
'
.'o
(4)
J a
>0
But, by
r /K)!in|(^)^ =0)
Cq
Jo
Ja:
/(#') cos a(:r' -a?) da
cfo'
f(x')
^ fq
X)
dx'
sinrm,
*,.,
= P f(u+x)f-du.
u
.
.Jo
Letting g->oo
we have
f(x')cosa(x' -x)dx'
da
when f(x + 0)
Adding
= ~f(x +0)
(5)
Jx
Jo
exists.
and
(4)
(5.),
we have
,
da[ f(x')cosa{x -x)dx
when f(x +0)
= ^f{x+0)
(6)
exists.
Similarly, under the given conditions,

X
Vda\
when f(x - 0)
Adding
(6)
/00
-J
for
every
exist.
f(x )cosa(x'-x)dx' = ^f{x-0),

f
(7)
exists.
and
p
(7),
we obtain
Fourier's Integral in the form
00
da\
point
f(x')coaa{x' -x)dx'
in
= $[f(x+0)+f(x-0)]
-oo<<oo, where f(x+0) and f(x-0)
FOURIER'S INTEGRALS
119, 120]
More general conditions
120.
as before,
the conditions in any finite interval remain
if
and the absolute convergence
replaced by the following conditions
of the infinite integral
f(x)dx
is
For some positive number and to the right of it, f{x') is bounded and monoand lim f(x')0 and for some negative number and to the left ofit,f(x')
x'y&>
bounded and monotonia and lim f(x') =0.
tonic
is
In this section we shall show that
for f (x).*
Fourier's Integral formula holds
315
a,V> vo
Having fixed upon the value of x for which we wish to evaluate the integral,
we can choose a positive number a greater than x, such that /(a;') is bounded
and monotonic when x' "~ a and lim f(x') =0.
By
Mean
rA"
\
- x) dx'.
f(x') cos a {x'
Value,
cA'
cos a(x' -x)dx' +f{A")\
f(x')cosa(x' -x)dx' f(A')\

)a'
)a'
cosa(x' -x)dx',
Jf
< A' = = A".
where a
Thus
M"
f(A') f(f-x)
f(x )coaa(x -x)dx' =

f
)A'
f(A")- C*(A"-x)
cosudu+J
)a(A'-x)
tt
cos udu.
J(f-T)
Therefore
I
M"
II
f(x')cosa(x'
cos a{x'
f(x')
-x)dx'\
x) dx'
DA'
4 *
0.
Qh
But we are given that lim
f(x')
It follows that
f{x') cos a (x'
=0.
- x) dx'
.'a
is
uniformly convergent for
tinuous function of a in a
Also,
by
a^g
^q
rq
rCO
f(x')cosa(x' -x)dx'
dx'
Jo
But
than
x'
this integral represents a con-
85,
,-QO
(q da\
>0, and
-x = a-x>0
f(x') cos a(x'
X'
- x) da
-X
X'-X
in the interval x' =^ a, since
we Jiave chosen a
V
'
greater
x.
And
r /(x
JK
)a
,
}
sing(a/-*)
X'-X.
sing(x'-x)
C
'
la
X'-X
both converge.
These extensions
are due to Pringsheim.

Cf. Math. Annalen, 68 (1910), 367, and
Reference should also be made to a paper by W. H. Young in Proc.
Royal Soc. Edinburgh, 31 (1911), 559.
71 (1911), 289.
FOURIER'S INTEGRALS
316
Therefore, from
(1),
00
[a
da
-x)dx
f(x') cos a(x'
Now
[CH.
*S3g^W - f
/(,-)
^W
sin
/{*')
*>
(2)
consider the integral
From
x - a:
'
Mean
Value,
X -X
\A'
(t
f(AI)
A*
where a
^ +f{A K nq.{x>-x) dx
Bing.Q'-x)
*&
'
Jo
(3)
JC
< /l'^| = ^4".

sin g'oC^'
Also
~~
_ [?l(~ x
sin
da.
the limits of the integral being both positive.

"
sin<7 (a/
Therefore
X'
A'
4"
And
similarly
i
Thus, from
-x)
<7T.
dx'
-X
91.)
(Cf.
smq Q (x' -x)

dx' < TV.
X' X
(3),
aA'
nin q,( x '-x)
JK
" f{x') sfrgofr'

X X
It follows that
<2tt
X'-X
'
g)
(4)
\f(A')\.
la
is
uniformly convergent for q i? 0, and by 84 it is continuous in this range.

X
shiq {x '~ X)
,
Thus
dx'=0,
lim [ f(x')
(5)
X -X
Io ->0 i'a
<
since the integral vanishes
Also from
when
qn
= 0.
(2),
da\ f(x')cosa(x' -x)dx' =
But we have already shown that the

is
sinq(x' -x)
dx'.
f(x')
integral
on the right-hand side of
.(6)
(6)
0.
uniformly convergent for q
Proceeding as in 119, (2) and (3),*
:
f*
follows that
fW V
x
X -X
da
f{x') cosa(x'
hra
Thus, from
it
smg(a/-z)
rt->xJ
..
(6),
dx
=0
-x)dx' =0.
(7)
.'o
*Or we might use the Second Theorem

Infinite Integral.
of
Mean Value
as proved in 58 for the
FOURIER'S INTEGRALS
120, 121]
But we know from
317
119 that
Cdai" f(x')cosa(x'-x)dx' = ^f(x + 0),

when
this limit exists.
Thus
-[ da[
7r
when
(8)
f(x') cos a(x'
-x)dx' = %f(x + 0),
(9)
.'
.'0
this limit exists.
under the given conditions, we find that
Similarly,
-[da[
J
71'
when
f(x') cos a{x'
-x)=\f(x -0),
(10)
-co
this limit exists.
Adding
and
(9)
(10),
our formula
proved for every point at which f(x 0)
is
exist.
We shall now show that Fourier's Integral

121. Other conditions for f(x).
formula also holds when the conditions at oo of the previous section are
replaced by the following:
(I)
For 'some positive number and to the right of it, and for some negative
number and to the left ofit,f(x') is of the form g(x') cos (\x' + fx), where
a(x') is bounded and monotonic in these intervals and has the limit zero
as x'->
oo
Also, (II)
^7-
^~ dx' converge.
and
dx'
We have shown in 120 that when g(x') satisfies the conditions named
above in (I), there will be a positive number a greater than x such that
K
r-ft
da
g(x') cos a{x'
- x)dx' = 0,
-61
But,
if
any
is
number,
positive
rco
da
da
da
|1
\\
J
-A
'
da\
da\
g(x') cos a{x'
da\
.'
g{x') cos a{x'
-x)dx
g{x') cos a(x'

)
-o
- x)dx'
+ \){x' -x)dx' + \\ da\
- x)dx'
o,
da
JA
g{x') cos (a
- x)dx'
)a
- x)dx' + |
g(x') cos a(x'
)a
.'o
da\
-x)dx'+
g(x') cos a(x'
)a
-A
= |1
- x)dx' +
)a
g(x') cos a(x'
g{x') cos a(x'

'a
g{x') cos (a
- \)(x' -x)dx'.
.'
Therefore
da
o
g(x') cos a(x'
- x)dx'
'a
r*
=p
e?a
o
.a
<7(a/)
cos A(:r'-:r) cos a(a/ -x)efa/=0
(1)
FOURIER'S INTEGRALS
318
Again we know, from
-x)dx'
^ A > 0.
uniformly convergent for a
da
It follows that
g(x') sin a(x'
Ja
kx
exists, for
120, that
g{x') sin a(x'

is
[ch.
-x)dx'
Jo
> A > 0.
Also
da
J
An
- x)dx'
g{x') sin a(x'
)C
dx'
g(x
/
)
- x)da
sin a(x'
(by
84)
JA
Jtt
= fV) (SÂoi^z^) _ cos_A^-^)|

'
Ja
00
yv
9( x
'
00
,
,.
dx' -\
'
r
aj'-a;
-a;
a;'
- x)
A 1(a/
-
cos
,,
-a;
a;'
cos A^x' -a;)
P
x'-x
g(x')
yv
'
dx
,
,
...(2)
since both integrals converge.

9(%')
But we are given that

converges
and
it
we know
,,
g(x)
yy
'
uniformly convergent for A
Thus
lim Cg(x')
from
cos
lim
A->oo
and
^(a:
eta
(a;'
p
x'-x
- x) j
-dx
*'
x)
dx
= ['S^l
J
^
dx' x
%
g(x
J
dx'.
~~
,.003 A(a/ -a;)

)
rX
'da/. ...(3)
Ax *^
Ja
imposed upon g{x')*
A
-dx-=0.
,
(3),
sin
a (a; -x)dx =\ g(x

J
da\
gr(a;')
find that, with the conditions

00
Therefore, from (2)
-X
Ja
,.
and
(
,
Ja
we
dx'
when A >0,
(2) that,
Also, as before,
and therefore continuous.
da\ g(x)sma(x' -x)dx' =

Jo
cos
^ 0,
A -^0 â
It follows
that
is
^7
follows that
also converges, so that
dx'
g(x') sin
-,
a (x' -x) dx' =
~~
x
dx'.
dx
....
(4)
(5)
*This can be obtained at once from the Second Theorem of Mean Value, as
proved in 58 for the Infinite Integral; but it is easy to establish the result, as
in 119, without this theorem.
FOURIER'S INTEGRALS
121]
da\
Again, since
we have
da\
sma{x -x)dx
Ja
Jo
exists,
g(x
319
-A
-x)dx' =
g{x') sinot(a;'
(6)
Ja
roo
Joo
da
g(x') sin a(x'
x)dx\
it
follows
that
00
/-OO
(
g(x')sm.a{x' -x)dx' -\ da\ g(x') sin a(x' -x)dx'=Q.
Ja
Ja
Ja
Thus
da
g(x')sin(a
\
+ X)(x'-x)dx'-\
Ja
Jo
da\
da\
(1)
(a- A)(#' -x)dx'=0.
g(x') sin. X(x'
-x) cosa(x' -x)dx'=0
(7)
Ja
Jo
Multiply
g{x') sin
Ja
Jo
rOO
00.
cos (Xx + fx)
by
and
(7)
by
sin (Xx
+ fx) and
subtract.
It follows that
CO
[-00
da
And
in the
+ fx)
cos a (x' -x)dx'
=0
same way, with the conditions imposed upon

20
f
I
C-a'
da\
g(x') cos (Ax'

'
+ fi)
cos a(x' -x)dx'
.(8)
g(x'),
we have
(9)
J-co
Jo
These
g(x') cos (\x'
results, (8)
and
(9),
may
be written
[.so
da
f(x') cos a{x'
- x)dx / = 0,
.(10)
fi x ') cos a { x ' - x)dx' =0,
da
when
f{x') =g{x') cos
But we know
that,
da\
jo
when
-a
Xx'
(10)
and
in (a, oo
and
oo
- a').
Conditions in
(11),
[Cf.
we
119
of a
number
still
hold
it
we replace cos
if
when
in this section.
(
Xx'
+ ll)
of terms of the type
can be proved * that the theorem
by an
(11)
(5).]
a n cos(A n z
It
a),
see that Fourier's Integral formula holds,
It is clear that the results just established
by the sum
-a',
the arbitrary function satisfies the conditions imposed upon

in (I)
f(x')cosa(x'-x)dx' = 1> [f(x + 0)+f(x-0)l
these limits exist.
Adding
+ ll)
when f(x)
is
+ /x n
).
also "valid
when
this
sum
is
replaced
infinite series
00
San cos(A. n* + /zn

,
when
X n
),
converges absolutely and the constants X n so far arbitrary, tend

,
to infinity with n.
*Cf. Pringsheim,
Math. Annalen, 68 (1910), 399.
FOURIER'S INTEGRALS
320
122. Fourier's Cosine Integral
when/(x)
is
and Sine
[ch.
Integral.
given only for positive values of
x,
In the case
there are two forms
which correspond to the Cosine Series and
of Fourier's Integral
Sine Series respectively.

I. In the first place, consider the result of 119, when f(x)
has the same values for negative values of x as for the corresponding positive values of x i.e. /( - x) =f(x), x>0.
:
Then
da\
~\
co
&\ /(#') [cos a (#'+#) +cosa(x' -x)]dx'
tJo
~~
f(x')cosa(x'-x)dx'
J _
7T Jq
Jo
da\ f(x')cosaxcosax'dx'.
7Tj
It follows from 119 that
values
of
and
x,
when f(x)
DirichleVs
satisfies
defined for positive
is
Conditions in any finite
f(x) dx converges absolutely, then
interval, while
'o
f (x) cos ax cos ax' dx' = %[f(x+Q) +f(x-0)],
da\
at every point
where /(x+0) and f(x-0)
value of the integral
Also
may
it
For some
(ii)
For some
shall
II.
the
<7 \
number and
to
the right
of
it,
f{x')
bounded and monotonic and lim /(a/) = 0;
number and to the right of it, f(x')

form g{x') cos (Xx' +p), where g(x') is
monotonic and lim g(x') = 0.
Also
positive
be of the
bounded
f
positive
be
shall
that,
and when x =
be replaced by either of the following

(i)
or,
exist,
/( +0), if this limit exists.

follows from 120 and 121 that the condition at infinity
is
and
j
dx must converge.
i
In the next place, by taking /(- x\ -f(x), x>0, we see
when f(x)
is
defined for positive values of x,
DirichleVs Conditions in any finite interval, while
and
satisfies
f(x) dx
con-
verges absolutely, then,
da\
at every point
integral is zero.
f(x')smaxsinax'
where f(x+0)
dx'
= i[f(x+0)+f{x-0)],
andf(x-0)
exist,
and when x =
the
FOURIER'S INTEGRALS
122, 123]
321
and 121 that the condition at infinity

of those given under (I).
other
or
one
may be replaced by
we express the arbitrary function
when
It should be noticed that,
Also
it
follows from 120
by any of Fourier's Integrals, we must first decide for what

x
value of x we wish the value of the integral, and that this value of
indicated
integrations
must be inserted in the integrand before the
f(x)
are carried out
(cf. 62).
Discussion. In many of the probof which Fourier's Integrals

solution
lems of Applied Mathematics in the
with an exponential factor
form,
different
slightly
a
in
appear
occur, they
In these cases we are concerned with the limiting value
(e.g. e - a2 t) added.
Sommerfeld's
Integrals.
123. Fourier's
as t->0 of the integral
da\ f{x')cosa{x'-x)e-^tdx',
<h(t)=-\
it
Ja
Jo
and, so far as the actual physical problem

for
=0
is
is
concerned, the value of the integral
not required.
was shown, first

hand side exists,
It
of all
by Sommerfeld,*
da
lim
f(x') cos a(x'
that,
when the
limit
on the
- x)e - **H dx' = \[f{x + 0) +f(x
right-
0)],
when a<x<o,
= f(a + 0), when xa,
= 1/(6-0), when x = 6,
the result holding in the case of any integrable function given in the interval
(a, b).
The
it
case
when the
interval
was
also treated
greater detail
by Young.|
is infinite
much
when
has been examined in
by Sommerfeld, but
It will
be
sufficient
the arbitrary function satisfies the conditions

120-122, Sommerf eld's result still holds for an infinite
in this place to state that,
at infinity imposed in
interval.
However,
it
should be noticed that
we cannot deduce
the value of the
integral
- da\ f{x')
K Jo
Ja
from the above
results.
0(0
for
cos a(x' - x)dx'
This would require the continuity of the function

2
da\ f(x') cosa(x' -x)e- tdx'
= 0.
We have come across the same point in the discussion of Poisson's treatment
[Cf. 99.]
Die willkurlichen Fundionen in der

Konigsberg, 1891.
Sommerfeld,
Diss.,
fW. H. Young,
C.I.
loc. cit.,
mathematischen
Proc. Royal Soc. Edinburgh, 31 (1911).
Physik,
FOURIER'S INTEGRALS
322
REFERENCES.
Fourier, Theorie analytique de la chaleur, Ch. IX.
Hobson, Theory of Functions of a Real Variable, 2 (2nd ed., 1926), Ch. IX.
Jordan, loc. cit., T. II (3e ed.), Ch. V.
Lebesgue, Lecons sur les series trigonometriques (Paris, 1906), Ch. III.
Neumann, loc. cit., Kap. III.

Weber-Riemann, loc. cit., 1 (2 Aufl.,
And the papers named in the text.
1910), 18-20.
EXAMPLES ON CHAPTER
Taking f(x) as
1.
when
< x,
X.
< x < 1, and
when
in Fourier's Cosine Integral
as zero
show that
00
1=-2
,
a cos ax
sin
00
ax
sin a cos
2=-]
(0<z<l)
da,
TJo
(*=1)
da,
o
0aa
2f rinaco.a
2.
By
00
in the
Show that the
equal to
4.
is
a;
when
Show that
COSaZ
,=,
e_
' to
(/?
> 0),
prove that
'
Integral, prove that

7T
expression
7T
Jo
00
2a 2
w.
same way, from the Cosine

00
is
ax
sin
7F+F*
J.
3.
'
considering Fourier's Sine Integral for e-fr

f
and
ttJo
fg
a:
a/0-^3
1
< a,
"sin
{6x\dd[e
COS
V 2 COS V
cfo
> a.
g^-^ga|
and to zero when x
^ {* + tan a
sin
the ordinate of a broken line running parallel to the axis of a;* from x=0
and from x = b to x = cc , and inclined to the axis of x at an angle a
to x = a,
between x = a and x = b.
5.
Show
Integral,
that
/(a?)
Ti"
Show that
Integral,
satisfies
the
conditions of
120
for Fourier's
and verify independently that
6.
=p.
cos
era
Jo
f(x)
aa;
cos ax
= when
,
s/x
Jo
satisfies
s/x
a:
> 0.
the conditions of 121 for Fourier's
and verify independently that
JW
sin
a;
cos a (x ',
a:)
= sin x
dx'
APPENDIX
PRACTICAL HARMONIC ANALYSIS AND PERIODOGRAM

ANALYSIS
1.
Let yf{x) be a given periodic function, with period
that, for a very general class of functions,
2ir.
We
have seen
we may represent f(x) by its Fourier's
Series
u Q + a x cos x + a 2 cos 2x + ...\

+ b sin x + b 2 sin 2x + ...)'
... are Fourier's Constants for f(x).
;
where a Q a v a 2
,
...
bv &2,
= x ^ 2tt.
the range of x to be
C 0S
.
nx are replaced by
If
the period
is a,
We may suppose
instead of
2mrxja, and the range becomes
27r,
the terms
^ x ^ a.
However, in many practical applications, y is not known analytically as a

function of x, but the relation between the dependent and independent variables
Or again,
is given in the form of a curve obtained by continuous observations.
we may only be given the
values of y corresponding to isolated values of x,

made at definite intervals. In the. latter case
the observations having been
is drawn through the isolated points

both cases Fourier's Constants for the function
can be obtained by mechanical means. One of the best known machines for the
purpose is Kelvin's Harmonic Analyser.*
we may suppose that a continuous curve
And
in the plane of x, y.
2.
The
in
above can also be treated by
practical questions referred to
substi-
tuting for Fourier's Infinite Series a trigonometrical series with only a limited
number
of terms.
Suppose the value
of the function given at the points
0, a, 2a,
...
(m- l)a, where ma=27r.
Denote these points on the interval

Xq, #i>
and the corresponding values
of y
(0, 2tt)
X2
by
Xrn_ 1
by
yi, y*>
ym-v
Such mechanical methods are described in the handbook entitled Modern

Instruments and Methods of Calculation, published by Bell & Sons in connection
with the Napier Tercentenary Meeting of 1914.
323
PRACTICAL HARMONIC ANALYSIS
324
Let
= a + ax
s n (x)
cos x
+ bx
+ 1 = m, we can determine
If 2ti
sn (xr
)=yr
+ a2
cos
2x+
+l
these 2n
+ bn
nx }
sin
constants so that
when r = 0,
The 2w + l equations giving the values
...+a n cos nx
x + b 2 sin 2x +
sin
[APP.
1, 2, ...
of
2w.
a lf ...a n
follows:
a,
+ a + ...
cos a^
+ a v cos px x + ... +a n cos wa^
+ b l sinx 1 +...+6 P sin^ 1 + ... +bn sin nx x
+ ax
+ ax
cos x 2n
+ &!
sin
-f-
x 2n +
. . .
+ a p cos px 2n +
+ b p sin px 2n +
. . .
Adding these equations, we
since
and
sin
pa + sin 2pa +
when
(2n
. . .
+ an
+ bn
= 2yr
+ cos pa + cos 2pa +
cos
w# 2r^
sin
nx 2n
+ cos 2npa 0,
...+ sin
2npa = 0,
+ l)a=27r.
we know that
+ cos pa
cos ra
+ cos 2pa
cos 2ra
+ + cos 2npa cos 2nra = 0, ^p 4- r

ra + cos 2j9a sin 2ra
p = 1, 2,
f
+ ... +cos 2npa sin 2wa=0,
( r = l, 2,
2
2
1 + cos >a + cos 2^>a +
+ cos 2 2w^?a = \{2n + 1).
. . .
cos j9a sin
And
->
It follows that,
cos
px 2
y 2n
-
see that
(2n+l)a
Further,
=y
+ a.
etc.,
if
...
...
we multiply the second equation by
cos
px lf the
third
by
and add, we have

2n
Similarly,
we
\ (2n
+ 1 )a v = 2 Vr
\{2n
+ l)b v =
cos
# ra
find that
In
trigonometrical series of (2n +
l)
2/r
sin pra.
terms has thus been formed, whose sum
takes the required values at the points

0, a, 2a,
...
2na,
where {2n +
It will be observed that as n-><x> the values of a
\)a
,
= 2ir.
a lf
...
and
b lt b 2 ,
...
reduce
to the integral forms for the coefficients, but as remarked in 90, p. 218, this
passage from a finite number of equations to an infinite number requires more
careful handling
3.
if
For purposes
the proof
is
to be
made
rigorous.
of calculation, there are
advantages in taking an even
AND PERIODOGRAM ANALYSIS
2-4]
number
of equidistant points instead of
an odd number.
325
Suppose that to the
points
0, a, 2a,
we have
(2n-
...
the corresponding values of
yo,Vi, V 2
In this case
we can obtain the
where nair,
l)a,
y,
>
2/2n-l-
values of the 2n constants in the expression
+ a x cos x + a 2 cos 2x+ ... + a n _ x cos (n - \)x + a n cos nx\

>
+ b n _ x sin (n - 1 )x
+ b x sin x + b 2 sin 2x +
the sum shall take the values y y x ... y 2n -i at these 2n
a
so that
points in
(0, 2tt).
It will
be found that
2 " _1
aQ
=- 2 y r
zn r=0
ap
^ yr
n r= o
& 2)
2T r?
cos pra,
if
i=
n
fairln.
2?(-l
an==
yrCOSr7r
S?/rSin|)ra
Runge* gave a convenient scheme for evaluating these constants in the case
This and a similar table devised by Whittaker for
the case of 24 equidistant points will be found in Whittaker and Robinson's
Calculus of Operations (1924), Ch. X.
of 12 equidistant points.
4.
we
may
This question
be looked at from another point of view.
Suppose
are given the values of y, viz.

2/o2/i 2/2. 2/m-i.
corresponding to the points

0, a, 2a,
Denote these values
sn ( x )
Let
...
(m-
+ &x
cos x
sin
where raa=27r.
by
of x, as before,
ao + a i
l)a,
+ a2
cos 2#
x + b 2 sin
2a:
+
+
+ a n cos nx
}
+ b n sin nx
on the understanding that m>2n + l,t the 2n + l

b n are to be determined so that s n (x) shall approximate as closely as possible to y y x ... y m _\ at a* x x ... xm_ 1
For a given value

constants a
ax
...
of n,
an
b lt
...
*Math.
Zeitschrift,
48 (1903) and 52 (1905).
Also Theorie
u.
Praxis der Reihen
(Leipzig, 1904), 147-164.
m<2n + l,
we can choose the constants in any number of ways so that

be equal to y yv ... y m -\ at x xx ... x m _ x for there are more constants
than equations. And if m =2n + 1, wc can choose the constants in one way so that
f If
s n (x) shall
this condition
is
satisfied.
326
The Theory
[app.
Squares shows that the closest approximation will be
of Least
obtained by making the function
regarded as a function of a
The conditions
for a
,a lt
an
...
minimum,
b lf
...
bn ,
a minimum.
in this case, are:
{yr-s n (x r ))-=o
r=0
inI 1
p = l,
Z{y r -s n (x r ))cospx r =
2, ...n.
^(yr-SniXr)) sin pxr =

It will
be found, as in
2 above, that these equations lead to the following
values for the coefficients:
m-l
ma - 2
r=0
^ma^ =
yr
2/r
cos 2jra
-=o
lmb P =
But
if
m is
yr empra
= 1.
1,
JP
2,
... 71,
od(L
even, the coefficient a P (when
p = \m)
is
given by
m-1
ma^ m =
2 yr cos r7r
r=0
the others remaining as above.
In some
cases, it is sufficient to find the
The values
of a
+ ax
terms up to cos x and sin
x, viz.
+ b x sin x.
make this expression approximate most
cos x
a x and b lt which will
closely to
2/o2/i2/2
at
0,
a,
2a,
Vm-i
...
(m -
when
l)a,
raa = 27r,
are then given by:

Wl
ma = X
/r
-=o
!?*!
= 2
2/r
sin ra.
Tables for evaluating the coefficients in such cases have been constructed by
Turner.*
5.
In the preceding sections we have been dealing with a set of observations

The graph for the observations would repeat
to have a definite period.
known
* Tables for
Harmonic Analysis, Oxford University
Press, 1913.
AND PERIODOGRAM ANALYSIS
4-6]
327
exactly after the lapse of the period; and the function thus denned could
be decomposed into simple undulations by the methods just described.
But when the graph of the observations is not periodic, the function may
incommenyet be represented by a sum of periodic terms whose periods are
bodies,
surable with each other. The gravitational attractions of the heavenly
not
are
periods
whose
components
of
up
made
the tides are due, are
itself
to which
commensurable.
But in the tidal graph of a port the component due to each

would be resolvable into simple undulations. A method of extracting these
of the tidal
trains of simple waves from the record would allow the schedule
as these comoscillations at the port to be constructed for the future so far
ponents are concerned.
from a graph of length L a part that repeats

A is to cut up the graph into segments of this
If there are enough
length, and superpose them by addition or mechanically.
segments,
segments, the sum thus obtained, divided by the number of the
The usual method
itself
of extracting
periodically in equal lengths
approximates to the periodic part sought; the other oscillations of different

the sum
periods may be expected to contribute a constant to the sum, namely
of the
6.
mean
The
part of each.
principle of this
method
is
also used in searching for
hidden periodi-
Suppose that a
a set of observations taken over a considerable time.
intervals,
equal
taken
at
are
that
they
and
observations
these
in
occurs
period T
there being n observations in the period T.
cities in
Arrange the numbers in rows thus:
W (m-Dn>
Add
w (m-l>n+i'
n+2'
u (m-Vn+2>
u mn-l'
umn-2>
the vertical columns, and let the sums be
u lt
uv
Un _
...
un _ 2
un _ x
the component of period T will be

...
In the sequence U U x
x
2,
may be
multiplied m-fold, and the variable parts of the other components
horithe
into
phases
different
with
enter
will
expected to disappear, as these
,
The
zontal rows, and the rows are supposed to be numerous.
difference
U x U 2 ... U n _ 2 U n _ x
between the greatest and least of the numbers U
of period T,
component
the
of
amplitude
the
of
furnishes a rough indication
period is indicated by this difference
if such exists; and the presence of such a
,
being large.
Let y denote the difference between the greatest and least of the numbers
If y is plotted
C7 n _! corresponding to the trial period x.
... U _ 2
n
have peaks
will
curve
This
periods."
of
as a function of x, we obtain a "curve
When
exist.
really
which
periodicities
the
to
corresponding
x
of
values
at the
then
the presence of such periods is indicated by the curve, the statistics are
U Q U lf U 2
,
>
analysed by the methods above described.

This method was devised by Whittaker for the discussion of the periodicities
328
entering into the variation of variable stars.* It is a modification of Schuster's

work, applied by him to the discussion of the statistics of sunspots and other
cosmical phenomena.! To Schuster, the term "periodogram analysis" is due,

but the "curve of periods" referred to above is not identical with that finallyadopted by Schuster and termed periodograph (or periodogram).
For numerical examples, and
for descriptions of other
methods
of attacking
may
be made to Whittaker and Robinson's Calculus

of Observations, Chapter XIII, already cited, and to Schuster's papers.
this problem, reference
* Monthly Notices, R.A.S. 71 (1911), 686.

See also a paper by Gibb, "The Periodogram Analysis of the Variation of SS
Cygni," ibid., 74 (1914), 678.
fThe following papers may be mentioned:

Trans. Camb. Phil. Soc, 18 (1900), 108.
Trans. Royal Soc. London, (A), 206 (1906), 69.
London (A), 77 (1906), 136.
Proc. Royal Soc.
APPENDIX
II
LEBESGUE'S THEORY OF THE DEFINITE INTEGRAL

is meant by the lower
bounded linear set of points. The limiting points of
such a set have been defined, and Weierstrass's Theorem, that an infinite set
bounded above and below must have at lerst one limiting point, has been
1.
Introductory.
and upper bounds
In Chapter II we have seen what
of a
proved.
Lebesgue's Theory of the Definite Integral depends essentially on the idea

measure of a set of points. This is a number, positive or zero, associated
an
with the set and depending upon it. When the set consists of the points of
of the
interval,
interval.
of
same as the length of the

which belong to one or other
without common points, should be the sum of the measures of
open or closed, the measure

And the measure of the set
is
to be the
of points,
two sets
two sets.
these
It is this very subtle
forms
and rather
the chief obstacle in the
into Analysis in place of the
difficult idea of
the measure of a set that
way of the introduction of the Lebesgue Integral
Riemann
Integral.
The
discussion which follows
confined to bounded linear sets, though one of the advantages of Lebesgue's

work is that the extension to two and three dimensions is more or less im-
is
mediate.
Among
the alternatives at our disposal the treatment
by de
la
(3c ed., 1914), has been

Vallee Poussin in his Cours $ Analyse Infinitesimale, 1
adopted. The more compact and direct development in the first and second
chapters of his Integrates de Lebesgue, etc. (1916), seems more difficult as an
But much use is made below of the third chapter of that work
introduction.
dealing with the properties of the Lebesgue Integral.

The first step in this extension of the Riemann Theory of Integration was
made by Jordan* (1892), when he introduced the idea of the inner and outer
content of a set of points.
Jordan, Cours d? Analyse,

and aire exterieure.
In 1898 Borelf showed that a much more useful

e
1 (2 ed., 1893), p. 28.
He
uses the terms aire interieure
e
fBorel, Lecons sur la theorie des fonctions (l ed., 1898).
work is both revised and enlarged.
edition (1928), this
329
In the third and
last

LEBESGUE'S THEORY OF
330
[app. ii
concept was what he called the measure of the set. The principles that guided
Lebesgue in his theory of measure were those of Borel, and his definition and
earliest treatment are given in his Paris Thesis
Integrate, Longueur, Aire
published in Ann. di Matematica (3), 7, 1902). Thisjtvas followed by his book,
Lecons sur V integration et la recherche des Jmictions primitives (1904), of which

a second and greatly enlarged edition has just appeared (1928).
2.
We now give such definitions and simple properties of bounded linear sets
of points as will be required in the discussion of
measure and the Lebesgue
Integral.
As before we denote a set by E, and we shall assume that its points lie on the
^ b. The points of (a, b), which are not points of E, form the
complementary set denoted by CE. Clearly C(CE) = E.
A set E is said to be countable (or enumerable), when there is a one-one correspondence between the points of the set and the positive integers. To every point
of E there corresponds a positive integer, and to every positive integer there
interval a tk x
corresponds a point of E.
The terms
of a countable set
can be written as
u lt
The
set
1, J, , ...
l/n,
...
is
u2
u3
...
obviously countable.
The positive rational numbers form a countable set.

For every such number can be expressed as a fraction
are positive integers without
according to the
common
sum p + q, beginning
factor.
We
pjq,
where p and q
arrange these fractions
in each with the fraction of lowest
numerator.
When p + q = 2, we have 1 only, and this is taken as u v

When p + q = 3, we have \ and 2, and these are taken as
When p+q =4, we have and 3, omitting the number 1
these are taken as w 4 and u b
A
1 h
set
is
i s
said
n t
to be
and
a closed
closed, but,
already used: and
so on.
set if it
if
u 2 and u 3
contains
the origin
is
its
limiting points;
included in the
set, it
e.g.
the set
becomes a
closed set.
point
P of abscissa x
to be an interior point of the set, if a neighbourof whose points are jioints of the set.
A point P is said to be an exterior point of the set E, if it is an interior point
of CE: in other words, there must be a neighbourhood of P none of whose
hood of P,
a<x< p,
is
said
exists all
points are points of E.
set
E is said to be an open set,
if all its points are interior points of the set.

have the important property that they are composed of a finite, or
countably infinite, number of not-overlapping open intervals.
Open
sets
To prove this, take any point P of the set, and let its abscissa be x
We
can divide all the positive numbers into two classes A and B as follows: a
number h is put in the lower class A, if all the points x given by x rr: x < x + h
belong to the set; and it is put in the upper class B, if this is not the case.
There are numbers of both classes, and every number in the class A is less than
every number in the class B. If fi is the number separating the two classes,
.
1-3]
x + /a
the right-hand end-point of the interval associated with the point r
is
Similarly
XQ -
A.
Let
by taking negative numbers, we obtain the left-hand end-point

To all the points of the open interval r - A. < x <^ + /* this same
Now
interval corresponds.
They
331
consider the set
of
all
intervals.
such open
are obviously not overlapping.
}, c 2 , ...
that lim e n =0.

n>&
be a monotonic descending sequence of positive numbers, such

There can only be a finite number of the intervals of the set
ex
We can arrange these in order from left
Again there can only be a finite number of those that remain, each
These we now arrange in order after the first group; and so on.
of length = e 2
If the end-points a and 6 are points of the open set, they are to be the lefthand end-point, and right-hand end-point, respectively, of intervals (a, a),
(/?, 6), open at the ends a and /?.
b is a
The complement of an open set E with respect to the closed interval a 7=i x
just described, each of length ^
to right.
closed set.
For
let
P be a limiting point of CE.
Then
wise there would be a neighbourhood of
P cannot be a point of
any point
of
And
P is
a limiting point of CE.

Again, the complement of a closed set E is an open set.
For let P be a point of CE. There must be a neighbourhood of
this is impossible,
E, other-
containing no point of CE.
if
inside
otherwise
it;
would be a limiting point
P without
and
of E,
therefore a point of E.
Two
E and E
Ex = E2
sets
and we
write
Ex
set
and some
is
said
to
are said to be equal, if they consist of the
and we
that
...
En
one of these
same points,
when
sets,
it is
write
if
E1
consists of the points of
E2
E 2 >E V
The
Operations on Sets.
E lf E 2
a set E 2
E > E2
be greater than
other points,
And E X <E 2 means

3.
set
composed
said to be the
is
sum
of the sets
which belong to at
of the points
least
and we write
E = E1 + E2 +
+ E n = XE r
...
E is said to be the difference of two sets E v E 2 when it consists of

E x which do not belong to E 2 and we write E = E X - E 2
En is called
The set of points which are common to all the sets Ex E
The
set
the points of
the product of these sets, and
we
...
2,
write
E = Ei E 2 .... En = II Er
.
Multiplication and subtraction are reduced to addition by the use of com-
plementary
sets,
it will
be seen that
CE X + CE 2 = C(E X E 2 E 1 E 2 = C(CE 1 + CE 2
= E X CE 2
C(E 1 + E 2 = CE 1 .CE 2 C(E 1 -E 2 = CE + E 2 E x \,
),
The Commutative and Associative Laws

and multiplication of sets;
e.g.,
EX + E2 = E2 + E
E + E 2 + E 3 = E + (E 2 + E 3
X
),
Algebra hold for the addition
E X E 2 = E2 E
E X E 2 E 3 = E (E E 3
X,
X,
of
).
332
Law
Also the Distributive
applies;
E3 = E E + E E3
(E 2 +
Since in general (E x -
l)
mentary
is
is
no
must be
care
2,
In practice there
difficulty, for
replaced by multiplication and addition, making use of comple-
The commutative,
sets.
exercised in dealing with subtraction.
subtraction
gives, for example,
E = E E + E E 3 + E E i + E Ei
E + E 2 f E but is actually E + E
)(E 3 +
and the ordinary algebraical process

(E 1 +
[app. ii
and distributive laws may then
associative
be employed.
The operations above referred to are finite; that is to say, carried out on a
number of sets. But addition and multiplication can be extended to an
infinite number of sets.
The set
finite
E=E +E +
X
...
=2E
to
is
composed
The
which belong to at
of the points
E E
of sets
2,
infinite
...
to
oo
least
one of the
infinite
number
product
E = E E2
X
...
to
oo
=11
Er
is
made up
The
of the points
which are common to
all
the sets
E E
lf
...
2,
to
oo
distributive law
(E 2 +
applies again here,
which themselves
and
may
E3 +
C(E 1
and
It is clear that the
finite or infinite
E2 E3
sum
to
...
to
of a finite
oo
of factors,
sums.
oo)=CE1 .CE 2 CE 3
to
oo
number
holds also for the case of a finite
it
be
C(E 1 + E 2 + E 3 + ...
Further
= E E 2 + E E 3 + ...
to ao)
...
=CE
number
...to oo
+ CE 2 + CE 3 +
of countable sets
...
is
to
oo
a countable set,
for we can take all the first points of the sets in order, and then place after
them all the second points of the sets in their order, and so on.
But it is important to notice that the sum of a count ably infinite number of
countable sets
To prove
is also
a countable
set.
this let the sets be
E =a n +0 12 + a 13 + a u + ...,
E 2 = a{[ +a 22 + P23 + E^ 31 +. ^3*2 + a 33 +
1
'
'
^4=4'l'+ a 42
and
+ a 43 +
---
so on.
Then
where the points
Ê
= n +a 2l +a l2 + a 3l +a 22 + a l3 +
in the
sum
...
are taken from left to right along the dotted
diagonals as in the above scheme.
3, 4]
From
we can deduce
this result
333
that an open bounded linear
of points can
set
of not-overlapping closed intervals*
up
For we have seen (2) that a bounded open linear set is composed
or countably infinite set of not-overlapping open intervals.
into a countably infinite
be broken
Let (a,
Divide
set
of a finite
be one of these intervals.
/?)
and take the two middle parts, both closed,

The remaining left-hand quarter and right-hand
quarter are then to be divided into two equal parts, and the closed halves
The outer
lying nearest the centre are to be taken as intervals A 21 A 22
and
on.
so
halved,
again
parts are
We thus replace this open interval a<x<(l by the countable set of notit
into four equal parts,
as intervals
A n A 12
,
overlapping closed intervals
way from each
In this
vals
A n A 12 A 21 A 22 A 31 A 32
,
....
of the finite or countably infinite set of
open
inter-
obtain a countably infinite set of closed intervals, and the result
we
follows.
4.
The Measure of a Bounded Linear
I.
Let
points of
in (a,
Set of Points.
E be a set of points all lying on the interval a 5= x S= b, and let all the
E be enclosed as interior points in a set of intervals o\, 8 2 ... lying
,
b).
This set of intervals can be replaced by a set of not -overlapping intervals

...
such that every point of E is an interior point of one of the
For start with
intervals, or the common end-point of two adjacent intervals.
A lf A 2 A 3
,
o\
and
call it
in 8 V
If
<5
A r Then take
2 lies
82
and suppress the
altogether outside 8 lt
we take
parts,
if
any, of S 2 which
A 2 and also if 8 2
we take for A 2 the part
it
as
lie
abuts
of 5 2
but does not overlap it. If it overlaps,
If 8 1 lies wholly
o\, and we have in addition their common end-point.
within 8 2 we replace 8 2 by the two open intervals outside o\ and the two end-
on
81
outside
points of 8 V
In this process, at any stage when we take in the interval 8 n we add to our
A 2 ... which replaces 6\, 8 2 ... a finite
',
set of not- overlapping intervals Aj,
number
of not-overlapping intervals, and, possibly, the
common
end-points of
adjacent intervals.
Now
let
A lt A 2
...
be a finite or countably infinite
intervals, all in (a, b), or with
a and
point, respectively, such that every point of

intervals, or the
the
sum of the
common
set
of not- overlapping
b as left-hand end-point or right-hand end-
an
is
interior point of one of the
end-point of two adjacent intervals.
lengths of the intervals of the set
The lower bound of 2A for all such

E and denoted by m e (E).
sets
Ax A2
,
...
of intervals
Let
2A
denote
is
called
the
exterior
measure of
* The word overlap is Used here in its natural sense

two intervals overlap if they
have points in common which are not end-points of either. Thus (0, 1) and (f, 2)
The closed intervals (0, 1) and (1, 2) in a more exact sense overlap, as
overlap.
but in the text this meaning of the term is not
there is a point common to both
;
used.
A pair of such intervals
(open or closed)
may
be said to abut.
334
[app. ii
measure of E is defined as (b -a) - m e (CE), where CE is the set

a =i x
b which are not points of E.
The interior measure of E is written nii(E), and is not negative, since m e (CE)
cannot exceed (b -a).
The
interior
of points of
m (E)^m
To prove
II.
{E).
From
a lf a 2
the above definition, there must be a set of not- overlapping intervals

... for E, as above, such that
m (E)^2a<m
e
and similarly a
set
/? x
for
...
/? 2 ,
m (CE)-^2p<m
e
(E)
+ e,
CE, such that

e
(CE)
+ ,
being the usual arbitrary positive number.
The combined
set
a i Pl>
can be replaced as in
by a
(i)
possibly the addition of the
among
the
02*
set of not-overlapping intervals
common
y Xt y 2
...
with
end-points of certain adjacent intervals
y's.
This countable set y x y 2 ... of not-overlapping intervals is such that all the
points of a ^= x
b are either interior points of these intervals, or common end,
points of adjacent intervals.

00
*y r = b -
It follows that
a.
when n
For,
any
is
positive integer, 'y r
<(b -a).
Thus
\im
)l
If possible, let this limit
Form
new
be
>oo
set of intervals
^y r t^b -a.
1
-a - 2e'.
y l9 y 2
...
by adding -^^
to
yr (r = l,2,
at each end.
a~x^b is an interior point of at least one

and by the Heine-Borel Theorem ( 31.2, p. 71)
made up of a finite number of them.
Then every point

intervals
yx
true for a set
y2
Hence
which
is
...
of
-a< Zy r < Xyy r +

^J
<(&-a-2e') + ',
impossible.
00
But
of the
this is
it is
00
00
2y r = 2)a + 2A-
clear that
Thus we have
-a<m
and
(E)
+ m e (CE) + e,
m (E)^m
^ m (E
i
(E).
m (E
If possible let m E < m E
Then there must be a set of intervals A A
for E
as in the definition of the exterior measure, such that the sum of their
lengths is greater than m (E
and less than m (E
But as E contains E this set of intervals A A
will also serve for
E and the sum of their lengths cannot be less than m (E
III.
If
E 2 > Ex
then
x ).
x.
lf
2,
x,
1 ).
2,
lt
...
1 ).
2,
4]
E 2 > Ev
IV. Further if
m,(# 2
then
^ mÊJ.
CE >CE 2
For
and
Therefore
^ (b -a)-m
mi E ^ m E
+ E )^m E + m E
e (E 1
Let a lf a 2 ... be a
such that
,
2)
To prove
V.
(CE x )^m e (CE 2 ).
{b-a)-me {GE 2
Thus
335
{CE x ).
x ).
2.
set of not-overlapping intervals for
Ex
as in the definition
of (I),
me (E 1 )^â<m e (E 1 + ^,
)
and
fi
v p2
...
E2
a set for
m
e
such that
(E 2 )^l^ r
<ê(E 2 + b,
)
being the usual arbitrary positive number.
As
before,
from the
set
a lt
a2
Pi,
P2
...
we form a not- overlapping set y x y 2 y z ... all in (a, 6), such that every point
oi(E + E 2 ) is either an interior point of one of the y's or the common end-point
,
of
two adjacent
y's.
Also
2>/ r
=2 a
and
Thus
(E x +
E 2 )^.fyr
{E 1 + E 2
=s
VI. Definition of Measurable Sets.
measurable
set,
and
It is clear that
The measure
measure
is
if
its
E/
1)
E; 2 )
+ .
{E x )+m e {E 2 ).
If
+ me
(E)
=mi(E), then
is
said to be a
(E) is their common value.

measure
is a measurable set, GE is also measurable.
of a finite
number
of points is obviously zero, since the exterior
zero.
Further, a countably infinite linear
Let
(^ 1 + J57 2 )<w e
Therefore
+ 2r>
be the set of points x x , x 2
Take the arbitrary
positive
x 2 in an interval of length
2,
of points
set
...
is
of measure zero.
and enclose x x
in
an interval
of length -
and so on.
~
1
m {E)<^-<e.
Thus
Therefore the exterior measure of this set

is
is zero,
and the measure
of the set
also zero.
measure is (P~a).
If E is an interval (a, p) in (a, b), it is measurable and its
For we must have m e (E)^p-a, since the interval (a, P) is itself a possible
interval for the exterior measure with the definition given above.
And the Heine- Borel Theorem, as in (II), shows at once that the sign of
inequality
is
impossible.
336
Again
it is
[app.
ii
clear that
m (CE)^(b-a)-(P-a).
e
But the sign
of inequality
can be omitted, for
-a<
and we would have
(b
- a) - m e (CE)
would give
t
m {E)<m {E).
i
it
< m (E),
sum of a finite number of not-overlapping intervals, it is

measurable, and its measure is the sum of the lengths of these intervals.
In the discussion which follows it will be seen that the sum of a finite, or
Similarly, if
countably
is the
infinite, set of
measurable sets
is
are measurable.
Thus
measurable.
we
starting with
open sets
complement of a measurable set is measurable,
measurable. All sets which can be obtained by additions,
countable sets of points and countable sets of intervals,
And
closed sets are also
see that
as the
subtractions and multiplications, finite or infinite, of measurable sets are also
measurable.
It is only for
measurable sets that this theory of measure
a debatable question whether, and,
still
if
so, in
what
studied.
is
sense, sets
It is
which are not
measurable do exist.*
5. A Necessary and Sufficient Condition that a Set E be Measurable.!
A necessary and sufficient condition that a linear set E ina7=x=b be measurable
is that to the
number of
arbitrary positive
intervals
and two
there shall correspond
sets e! ,
set
e" of exterior measures
I consisting of a finite
<
c,
such that
E = I + e' -e".
(i)
To prove that
this condition
is
necessary,
we note that
to the arbitrary
there correspond the sets of not- overlapping intervals a lt a 2

ft, ft,
...
for
...
for
and
CE, such that
mE ;=: Sa < mE + \i,

r
mCE ^ ft < mCE + |e,

i
mE + mCE = b-a.
and
ia r + ft.<(&-a)+.
Thus
But, as in
we
4, I,
from the
set
1>
ft,
a 2> 02*
a n, .fti
obtain a set of not-overlapping intervals
.V
such that
ii
ii
2}V = 2ar + 2ft-2(a r ft),
e
*Cf. Lebesgue, Legons sur V integration (2 ed., 1928), footnote, p. 114.
mE
of
instead
tin this and the following sections, we shall, when convenient, write
m{E) for the measure of E, when E is measurable, and similarly m e E for m e (E),
m,E
for nii{E).
It will
be noticed that, when
is
measurable,
mE = m E = miE.
e
4-6]
337
common
the last term on the right-hand side comprising the parts

a 8 blotted out in forming the y's.
As n tends
to
oo
so does
to an a
and
N and we know that 2 y r = (& - a).
Thus the sum of the lengths of the finite or countably infinite set of notoverlapping intervals blotted out as above is less than e.
Now
take n so large that
Sn and
the points of a n+1 a n+2

,
to
Rn
by
oo
of
If
and S n GE < the points common to

and m e (S n GE) < 2 (a/5) < e.
E)<e
(R
n
e
the condition
all
the a's and
...
a n by
0's.
E is measurable.
E = I + e'-e",
is satisfied,
have
where / consists of a finite set of intervals and e', e" are two
measure < e, which we suppose have no common points.
EI-e".
CE<C(I-e") = CI + e".
m e CE ^m {CI + e") < mCI + e.
me E + me CE < ml + mCI + 2e = (6 - a) + 2e
m E <m E + 2z.
m E î rrtiE.
me E ^ rriiE.
m E = miE.
Also
Thus
Therefore
Jt follows that
And
Thus
But
Hence
If E x and E 2 are measurable, so is E x + E 2

have to show that E x + E 2 can be broken up as in the theorem of
6. I.
We
a lt a 2
Rn E < Rn
Thus
(ii)
and denote the points
E=Sn + Rn E-S n GE.
Then
Also
We
Sa r <,
n+1
Given the arbitrary positive
we have
E l = I 1 + e '-e 1 ",
E 2 -I z + e 2 -e 2 ",
1
where I lt I 2 are
sets of a finite
Thus
where e"
number
E1 + E2 = Ix +
is
contained in
Also
fi
and
me <
e
<
e.
') - e",
2 + (e/ + e z
ex"
+ e 2 ".
me (e x + e 2
m e" ^ m
and
of intervals
{e x
^ m e + m\e <
+m
+ ") ^ m
e
'
ee x
ea
E x + E 2 is measurable.
// E x and E 2 are measurable,
e,
e2"
Thus
so is E x - E 2
II.
C{E -E 2 )= CE X + E 2
We know that
Thus C(E - E 2 is measurable, and E x - E also.
III. If E x and E 2 are measurable, so is E E 2
We know that C(E X E 2 )=CE X + CE 2 and the result
.
follows.
%e, etc.
5.
338
7.
e1
If
I.
*and
E2
are measurable
and without common
m(E 1 + E 2 )=mE x + mE 2
With
the.
same notation
[app. ii
points, then
we have
as in 6,
E2 = I2 + e 2
- e a ",
Avhere ? e e/<|e, etc.
E + E I x + I 2 -e",
Also
where
+ e.
2)
<
e"
"+e
",
C^ + âKC^ + ^-e"^^/^/^".
and
E 2 )=m C(E + E 2 < mC(I x + 7 2 + e,

m(E + E 2 )>m(I x + I 2 )-c
?C'(^ +
Therefore
and
Thus
But
^ 2 )-m(/ + /
|m( E 1 +
,
)|<.
for the finite sets of intervals it is clear that
m(I x +
x
since
as before
Thus
Similarly
Again
Ix
m(I x I 2
^m
+ ex
xe 2
"+e
",
\m(E x + E 2 - mE x - mE 2
)
If
For
let
E E
II.
x,
(I)
+ m(I x +
\
-mE
2)
ml
ml 2
.
\
|<3c.
are measurable, then
m{ E + E + mE x E = mE + mE
E = E E 2 + e and E 2 = E E 2 + e 2
E X E 2 are measurable and without common
E + E2 = e + e + EX E2
mE x = mE E +me
2)
and
points,
x,
mE =mE E +me
2
2.
But by
e.
a special case of the following
ex , e 2
+ me e 2 " <
ml 2 - mE 2
m(E + E 2 )=mE + mE 2
Then
|^-
\m(E x + E 2 )-mE x
It follows that
is
ex"
+ ml x - mE x
This theorem
^| m(E x + E 2 - m(I x + I 2
Thus
^m
+ e 2 ")
+ e 2 ".
(E 2 + e 2
//
E E 2 = 0.
since
(e x "
<E
I2
ex
<e
Therefore
<E + " and

I <E E + E
Ix
It follows that
But
we have
E I - e x ",
CE <C(I x -e ") = CI 1 + e 1 ",
mE x > ml x - |t.
mE - ml x < \e.
mE - ml 2 < |e.'
Also since
And
= ml x + ml 2 - ml x l 2
2)
2,
m(E x + E 2 =m(e x +e 2 + E X E 2 = me x + me 2 + mE x E 2
)
and
7,8]
m(E1 + E 2 + mE x E 2 =mE 1 + mE 2
Therefore
If E x and
III.
are measurable
and
E > E2
x
then
m(E x - E 2 = mE 1 - mE 2
(E x - E 2 + E 2 = E X
We have
and (E x - E 2 ), E 2 have no common
Thus from (I) the result follows.
339
point.
00
If
8. I.
ly
2,
. . .
are measurable
and without common points, then
E = Êr
we
have, for
00
is
measurable, and
mE = ^mE r
Since
any
E + E 2 + ...+ En
is
positive integer (by
measurable and contained in
(a, b),
7),
m{f E r = f mE T <(b-a).
)
2 Er ^ (b - a).
Thus
Taking the arbitrary positive
there
e,
mEn+1 + mEn +
Now
for
En+r
there
2 +
.',
.
is
<
when
e,
n i=
v.
a set of not-overlapping intervals a n+r>1
is
a n+r>2
...
such that
mEn+r ^ 2 an+rt < mEn+r +

^
1,2,....
-~-,
*=i
Also the countably infinite set of intervals
2a
s=l
are such that
(En+1 + En+2 +...)<^ (2a n+r>s )
(mEn+r + ~
<f
r=l\
l
< + .
Now
Rn = Ê r
Sn = '2Er and
let
m E=m
Then
(Sn
+ Rn )'^m e Sn + m e Rn
E>Sn
But
Thus
m E ^ mSn
m E <m E + 2z.
It follows as before that
mE = m,E.
Therefore
^mEr = mSn ^mE< mSn + 2e,
Again
when n ^
v.
ti^mE-mSn <2e,
Thus
m^ = lim %n,Er = 2 m^r
and
II.
when n^v,
7/
the
measurable.
measurable
sets
Ex E
,
2,
...
have
common
points, then
E = %Er
1
is
340
For any integer
we have
n,
E x + E + + En
= E 1 + (E 2 -E 1 + (E 3 -E 1 -E 2 +
. . .
and the
different sets
î = E x
and 2^r
is
...
+ (E n -E 1 -E 2
...
- E^),
on the right-hand side have no common points.
E1 + E2 +
Thus
where
[app. ii
=6 1 + 6' 2 + ...
,
to
...
oo
to go,
and 6 r = E r -E 1 - E 2 ... -Er_ x
r^2,
measurable.
Let E,
9. I.
E E2
lf
E -E x -E 2
Then
be all measurable.
...
...
is
measur-
able.
This follows from

II.
If in addition
and
E E2
lt
6, II.
. . .
are all contained in
and have no common points;
m(E- f Er )=mE - %mEr

i
This follows from
If E lf
III.
...
7,
and
III
8, I.
are all measurable, then
E = E1 E2 E3
.
is also
...
measur-
able.
C{E 1 .E 2 )= CE 1 + CE 2
C(E 1 .E 2 .E 3 = GE X + CE 2 + CE 3f and so on.
C(E 1 .E 2 .E 3 ...) = CE 1 + CE 2 + CE 3 + ...
We know that
Thus
and the
result follows at once.
IV. //
E E2
lt
...
and E x < E 2 < E 3 <

E = E1 + E2 + E3 +
are all measurable
...
then
...
is measurable,
and
(mE n
mE =lim
In this case
).
>x>
E = E 1 + (E 2 -E 1 + (E 3 -E 2 + ...,
)
mE = lim m(E r - Er ^)
and
n *-x>
by
8,
=lim f,{mEr -mEr_ x ) by
7,
III
=lim (mE n ).
n*-oo
V. If
E E2
lt
...
and
are all measurable
E{> E 2 > E 3 >
E E\ E 2 E 3
.
is
measurable and
mE = lim {mEn
n
We have as in
...
then
...
).
>x>
(III),
m(GE) = m(CE l + CE 2 +...)

= lim m(CEn )bj (IV).
n*-oc
mE = (b-a)- m{CE) = (b -a)- lim {mCEn
And
10.
necessary
and
sufficient
integrable in (a, b) according to
=lim (m# n ).
condition that the bounded function f(x) be
Riemann's
tinuity form a set of measure zero.
definition is that its points of discon-
8-10]
We prove first that

Take the arbitrary
Let
Gk
34]
this condition is sufficient.
positive
e,
< ^tt
and k
r.
be the set of points of discontinuity at which the oscillation*
=k
Then the measure of this set is zero since it is a part of a set of measure zero
and it is a closed set.f
Thus the points of Gk can be enclosed as interior points in a finite set of notoverlapping intervals, the
and
m are the bounds of f(x)
The complement
of their lengths being ^= ~-p
sum
in (a,
of this set of intervals is
finite
number
of these can be divided
of closed not-
overlapping intervals, the oscillation at every point of these being
Each
where
r,
6)4
up into a
number
finite
<
k.
of partial intervals,
such that the oscillation in each of these partial intervals
< k.
(Cf. 31. 1,
footnote, p. 71.)
Thus we have a mode
of division of (a, 6) for
$-s<k'(b-a) +
and f(x)
is
integrable in
which
2^
M _ m) x(M-m)<e,
according to Riemann's definition.
(a, b),
We now prove that the condition is necessary.

Let k > k 2 >
and lim k n 0, and G r be the
. . .
oscillation^ k r
set of points for
Let the measure of this closed set be
C > 0, and take e \k r C.
Since f(x)
integrable (R), there
is
a=x
such that
jS
-s<
a point of Gr
in (xs_ lt xs )
kr
If
If it
which the
#!
is
...
mode
of division of (a, b)
#n_i> xn = b$
e.
is
inside one of these intervals (xs__ lt xs ), then the oscillation
common
coincides with the
oscillation in at least
end-point of (xs _ lt xs ), (xs , xs+1 ), the
one of the two must be |& r
*Cf. 29. 4, p. 66.
be a limiting point of G k and not a point of G k

Then the
equal to x < k, and there is a neighbourhood |a;-a;
77 in
which the oscillation
k.
But P is a limiting point of G k so there is a point P'
of G k inside (x -t], x +r)), and the oscillation at P' ^k.
Therefore there is a
neighbourhood of P' inside (x -rj, x +t)) for which the oscillation iS k, which is
fFor
let
P{x
oscillation at
is
<
impossible.
is a closed set of measure zero and CG is thus an open set of measure (6 - a)
k
can therefore be broken up- into a countably infinite set of not-overlapping
\G k
CG k
closed intervals
\, A 2
We can choose the
positive integer
(6
Gk
The
points of
left,
when A v A 2
and
...
-a
2A r = (6-a).
so that
-! A ^ < 2(sbrr
are interior points of the finite set of not-overlapping intervals

... A
n are removed from (a, 6).
342
[app.
ii
Thus these two adjacent partial intervals give to (S-s) a contribution

by the sum of the lengths of these intervals.
x n _ x x n = b, which have a point of
of a=x x x ...
intervals
the
all
Hence
Gr inside them or at an end contribute to S - s an amount = \lc r (their sum).
i= \k r multiplied
But the sum
of the lengths of these intervals i^ C.
mode
Therefore, for this
of division,
S-s^lkr C>,
which is impossible.
and mGr = 0.
O=
Thus
But the points of discontinuity are given by the sum
E=G +G +
2
and
Therefore (by
9,
11.
. .
to
qo
...
m(E) =lim m(G n =0.
IV),
Let
^G ^G
Measurable Functions.
E be a bounded measurable
set
j- oo
of points on the axis of
x.
The function
to be measurable in E, if the set of points

f(x), defined at the points of E, is said
measurable,
for every constant A.
is
A
which
f(x)
E
for
of
>
denote the set of points of E for which f(x) > A by E[f(x) > A], and
A.
similarly E[f(x) i= A] denotes the set of points of E for which f(x)
We shall now show that if f(x) is measurable in E as defined above, the sets
We
E[f(x)Â],
E[f(x)<Al
E[f(x)Â]
are also measurable.

We are given that E[f(x)
Then
its
> A] is measurable.
complement with respect to E is also measurable, that
is
E[f(x)Â]
is
measurable.
Again
is
if
En is the set of points of E for which f(x) >A~ n >we know that E n
measurable.
And
the infinite product
E[f(x)^.A^.E 1 .E 2
is
measurable
Therefore
...
( 9, III).
E[f(x)=A]
measurable.
is
by addition of E[f(x) > A], that E[f(x)Â] is measurable, and

by subtraction from E[f(x) ^4],we see that E[f(x) <A]is measurable.
It is clear too that, if A and B are any constants,
It follows
E[A<f(x)<B],
E[A^f(x)<B],
E[A<f(x)^B]
ElA^f(x)^B]
and
are all measurable.
on Measurable Functions.
measurable, so are a +f(x), af{x) and \f(x)\, where a
12. Operations
I.
Iff(x)
is
We know that
E[f(x)
Thus E[f(x)+a>A]
>A
is
- a]
is
measurable.
measurable.
is
a constant.
10-12]
The others follow from the
fact that
E[f(x)> A /a]
is
343
measurable, and that
Elf(x)>A-\ + Eif(x)<-A}
is
measurable.
II.
E[fx (x) >f2 (x)] is measur-
Iffx {x) andf2(x) are finite and measurable, then
able.
E[fi( x ) >fi( x )^
the
is
sum of the countably infinite measurable
sets
E[Ux)>r].E[f2 (x)<rl
where
r is
any
number.
rational
III. If x (x) and f%(x) are finite and measurable, their sum, difference and
product are measurable.
We know
Eifx{x) f2 (x)>A] = E[f1 (x)>AT f2 (x)].
Thus the sum and

Also
if f(x) is
E[f(x)
Thus
('fjfx)
of/1 (a;),/2 (a;) are measurable by (II).

2
for E[(f(x)) 2 > A] is equal
is (f(x))
difference
measurable, so
to
> VA] + E[f(x) < - VA].
+f2 (x)) 2 and (fx {x) -fz (x)) 2
and the
are measurable,
result
follows.
IV. Letf^x),/^), ...bean infinite monotonic sequence of measurable functions.
Then, for every x in E, lim
w>0
able.
fn {x) exists (finite or infinite) and this limit is measur-
For example, for every x in
Then E[ lim fn {x)
>A]
nmo
this sum is measurable
is
E letf^x) <f%(x) <f3(x) <
the
sum
of
...
E[f1 (x)>A~\, E[f2 (x)>A~\,
...
and
( 8).
be an infinite sequence of measurable functions.

Then
f
lim fn (x) and lim fn (x) exist (finite or infinite), and these limits are measurable.
n-x
r*-x
Suppose <f>x(x) to be the upper bound of fi(x),fz (x),fg{x) ... and < 2 (x) to be
V. Let fx (x),
z (x),
...
the upper bound of f^(x),f3 (x),
Then
<M*)
Also
...
lim
this is measurable
VI. Let
fx (x), fz(x),
lim fn (x) exist

nmo
Then
by
...
on.
= <M*) = 4>3(*)--.
nmo
and
and so
fn(x)
lim
nmo
<$>
n (x),
(IV).
be
an
infinite sequence
of measurable functions, and
(finite or infinite).
this limit is
a measurable function.
This follows from (V), since
lim
fn (x) = lim fn (x) lim fn (x).
nmo
nmo
nmo
We note that every monotonic function is measurable in an interval; and, in
particular, that f(x)
constant and f(x) =x are so. Applying the above results
we see that every polynomial is measurable: and, as a continuous function,
by a theorem due to Weierstrass,* is the limit of a sequence of polynomials,
*Cf.
Hobson, Theory of Functions of a Real Variable, 2 (2nd.
ed., 1926), 22S.
[app. ii
see that every continuous function is measurable.
Then discontinuous
344
we
which are the limits of sequences of continuous functions, are also

measurable; and so to more complicated classes of measurable functions.
functions,
13.
The Lebesgue
Let f(x) be a bounded and measurable function
Integral.
measurable set
for the
contained in
(a, 6).*
be the lower and upper bounds of f(x) in E.

Divide the interval (A, B) on the axis of y into n partial intervals
Let
and
(h,h),...{l n -i,B).
(A,h),
Denoting
by
and
by
we thus have the mode
A=l lt
E for which
of E for which n _
l
Let
er
be the set of points of
and e n the
Then e lf
Form
...l n _ lt
set of points
e2
...
en
the sums
_ x ^f(x) < lr
=f(x) ^k l n
lr
= 1, 2
...
(n
1 ),
common
points.
where
s,
S = ^l r m(er
and
m(e r being the measure of the set
= ^l r_ x m(e r
),
er .
SÂm(E)
Then
..
are measurable sets without
S and
of division of (A, B),
n = B.
s^Bm(E).
and
Thus the sums S and s for all possible modes of division of (A, B) have a
lower bound J and an upper bound /, respectively; and for the same mode of
division
We
shall
s.
now show
Let some or
smaller intervals,
I^kJ.
that
the intervals
all of
_ lf
(lr
lr )
in
Q,
... l n
_x
be divided into
and
A l> A 2>
^0
A A-1>
^1 '
be the new mode

This mode of division is said to be consecutive to the former.
or.
Let its sums, as above, be
Compare, for example, the parts of 8 and 2 which come from
of division of (A,
B) thus obtained.
(Z
).
From 2 we have
A 1 m(e 11 + A 2 m(e 12
)
where
e 12 , ... e lk
...
+l 1 m(e lk ),
are the sets of points of

(l
^f(x)
<
for
X,), ..(XV.1 =i/(*)
which
< h)-
=e 11 -\-e x2 .:.+e xkf

the sets on the right-hand having no common points.
A i(e n + A 2 w e i2)
Thus
+ h m îk) = hwitth
It follows that 2 ==i S; and similarly we have vr ~ s.
Now take any two modes of division of (A, B),
A=l ,l 1 ,... m_ lf m = B, with sums S and s,
A=lQ x ... I'nî, n = B> witn sums $' and/
And
e1
',
(1)
(2)
*In this section, so far as possible, the notation corresponds to that of 39-41,
and the argument proceeds on exactly the same lines.
pp. 91-4,
12, 13]
On superposing
(1) and (2) we obtain a third mode

and (2).
sums be 2 and <f.
345
of division (3) consecutive
to both (1)
Let
its
8^2
Then
But
and
cr^'s'.
2^cr.
S^s',
Thus
and the sum 8
from any mode of division is not less than the sum

arising fr,om the same or any other mode of division of (A, B).
arising
For suppose I > J.

a
mode
since /
Now
in
J is the lower bound

sum 8 to the left of
the upper bound of the sums
s,
there
must be
J I, and
of division giving
impossible as a
is
than any sum s.

be an arbitrary positive number, as small as we please.
sum 8
less
let e
Take a
of division
^8-s=
Therefore
way
S, there
This
which all the partial intervals are less than

For this mode of division it is clear that
I and
sums
the middle point of
of the
must be a mode
the right of this middle point.
s to
cannot be
mode
Since
of division giving a
is
sum
= J.
It follows at once that
(Z x
- lQ )m{e x + (l 2 - Z 1 )m(e 2 +
)
we must have I=J, and
the
...+(l n - ?n_ 1 )m(en )
sums
8, s tend to the
as the number of points of division of (A, B) tends
< em(E).
common
to infinity
value of
in such a
that the largest of these partial intervals tends to zero.
This number I
is called the
function f{x) in the measurable
Lebesgue Integral of the bounded and measurable

E, and we write
set
f{x)dx.
If f(x) = C in E, where
If E consists of all
C is a constant, we define]
the points of an interval (a, b),
f(x)dx as Cm(E).
we use
the
ordinary notation
[b
\
f(x)dx,
and
the integral is
now
Ja
a and b.
Sometimes it
called the Lebesgue Integral off(x) between the
limits
convenient to
is
make
Lebesgue's sense by placing a capital
clear that the integral is taken in
before the ordinary symbol.
a case the Riemann Integral would be written as (R)

Integral as (L)
In such
f(x)dx and the Lebesgue
f(x)dx.
)a
Again,
if
the bounded function /(re)
is
integrable with Lebesgue's definition
we say that it is integrable (L)

with Riemann's definition, we say that it is integrable
for the interval (a,
b),
We shall see below that for

also integrable (L),
integrable (L)
and,
if it is
integrable
(R).
bounded functions, if f(x) is integrable (R), it is

and the two integrals are equal: but that f(x) may be
and not integrable
(R).
3*6
Just as in the case of the Riemann Integral,
Jb f{x)dx for a
a
> 6,
\b
by the equation
of-
it
we define the Lebesgue Integral

f
f{x)dx = -
f(x)dx.
'a
14. Properties
[app.
'b
Bounded and Measurable
the Lebesgue Integral of a
Function.
[.
Cf(x)dx- :C\ f(x)dx, where
This follows at once from

Iff(x)
II.
Since
is
for
any sum S
Thus
result holds for f(x)
A,
it is clear that, if
^ Cm (E).
f(x)dx
then
j
S ^ Cm{E).
S ^ Cm{E).
iorf(x),
Therefore the limit of the sums
A similar
a constant.
is
E andf(x) ^ C,
measurable in
we have
the definition.
C.
are the lower
and upper bounds
of f(x) in E,
then
Am{E) ^
f(x)dx
III. Let fix) be measurable in E, and

measurable and without common points*
Then
\
J
^ Bm{E).
let
E1 + E 2 = E
f(x)dx = \ f(x)dx+\
A B
Let the bounds of f(x) be
'
E1
where
and
E2
are
in
E2
f(x)dx.
Ai
in E, a lt
pt
in
lt
and a 2
fi 2
and
/? 2 .
respectively.
Then
is
the smaller of a x and a 2 , while B

mode of division of (A> J5),
is
the larger of
/? x
Consider any
A=l
(1)
= B.
,l lf ...*_!, ln
/Ts and the

If two of these points do not coincide with the smaller of the
mode of
consecutive
a
have
we
points
these
introducing
by
a's,
the
of
larger
and S is not increased, s not decreased.

a sum S for Et + a sum S
Thus the sum S for (1)
division
^1
J
Therefore
f(x)dx +
sum
f(x)dx^\ f(x)dx + \
IV.
sets
The theorem
with no
of (III)
common
f(x)dx.
JEl
JEi
s,
JE
)E
Hence
f{x)dx.
f{x)dx^\ f(x)dx +
E%
J A'a
A",
Ja'
Similarly from the
for
f(x)dx.
JE2
f(x)dx=\ f(x)dx + \ f(x)dx.

can be at once extended to the
points,
sum of n measurable
two by two.
We now prove that it holds also for a countably infinite number of measurable sets.
*It
is
clear that if f(x)
is
measurable in E,
it is
E U(x) > C] is the product of E [f(x) > C] and Ex

x
measurable in
Et
and
Ev
for
13, 14]
Letf(x) be measurable in
and without common
E and
/fr'
= f# r
where
E E
lt
347
are all measurable
...
2,
points, two by two.
Then
)e
f{x)dx = $\
f(x)dx.
JEr
E = ?,Er + Rn
Then we know that f(x) is measurable in Rn and that m(Rn )^0 as n->oo ( 8).
Let
But
JE
Thus
f(x)dx-% f(x)dx\^Mm(Rn ), by
|[
when If
|
is
f(x)dx = ^\ f(x)dx + \ f(x)dx.

JXn
lJEr
above,
(II),
the upper bound of \f(x)\ in E.
Therefore
f(x)dx=i \
f(x)dx.
)E
\)Er
V. Iff(x) and g(x) are measurable in
E andf(x) ^ g(x),
f(x)dx^[
then
g(x)dx.
Je
jjs
Let (A, B) be the bounds of g(x) in
E and A =l
Z
t ... n
_ 1} ln = B any mode
of division of (A, B).
Let
e lf 6 2 ,
points of
...
be the sets of points as in
E for which
^ g{x) <
^n
_x
which ln _ x :sS gr(#)
of points for
Then
lr
IB
since g(x) i^ J,^ in e r
Thus
lr ,
f(x)dx
is
the set
,._!
^ any sum s for ^(z).
f(x)dx^\
g{x)dx,
Je
VI. Letf(x) and g(x) be measurable in E.
(i)
and e n
Je
1),
ji-
lt follows that
Je
the set of
i.e. e r is
{n
...
f{x)dx^hr-Mer)
Je r
and therefore /(a:)
2,
/()<& = if
1
13 for g(x)
when r = 1,
Then
(f(x)+g(x))dx = \ f(x)dx +
Je
Let g(x) = C in E, and
let
A,
g(x)dx.
je
be the lower and upper bounds of f(x)
as in 13.
Let A
=Z
*!,... ?_!, Zft
= B be a mode of division of {A
B) with sums and s
ioTf(x).
Then f(x) + C has

a mode
If S'
On
of. division
and
s'
4 + C, B + G for its
bounds and
+ C,
+ C,
are the
sums
+ C for this mode

S'=S + Cm(E).
for f(x)
of division,
we have
proceeding to the limit, this gives

(
)x
(f{x)
n+C
... l
of the interval.
+ C)dx = [
J
f(x)dx + Cm(E) = \ f(x)dx +

J
C dx.
n
is
348
(ii)
With the same notation
A=l
be a
mode
h-i =/(*)
as in
l
n_ lt
... l
=B
and let e r be the set of points of E for which

(n - 1), and e n the set for which Zn _ x ^f(x)^l
n
of division of (A, B),
<
when r = 1,
Then
2,
...
= if
(/(a) +(*))**
i^S
(Jr-i'+0(*))<*
(/()+ y(*))cfa
.'<,.
ii
)i
= Z^r-i m e
(
r)
the sums for f(x) for this
g{x)dx
.!>
i'
=*+ 2
5 are
.-
+2
where $,
ii
for f(x), let
(i)
[app.
g(x)dx,
mode
of division.
It follows that
f
JE
(f(x)+g{x))dx^\ f(x)dx+\ g(x)dx.

J /;
Again
f
(f(x)+g(x))dx^\
JJE
and from
this
we
(l r
+g(x))dx,
Je r
see that
'
(f(x)+g(x))dx^\ f{x)dx +
JE
g{x)dx.
JE
Thus we have
(f(x)+g(x))dx=\ f(x)dx +
It is clear that
we
g(x)dx.
)e
)e
JJE
also have
{f{ x )-g{x))dx
f(x)dx-\ g(x)dx.
JE
)E
VII. Letf(x) be measurable in E.
Then
\^f{x)dx\^\^f{x)\dx.
This follows at once from (V), since
-1/(^)1^/(^)^1/(^)1.
VIII. Letf(x) be measurable in
points of
E other
Then
\
Since g(x)
is
and
g(x) be
than points of a component
Ex
bounded and equal

of
E whose
f(x)dx=\ g{x)dx.
E
JE
also measurable in E,
f(x)dx-\ g(x)dx =
J
f(x)
- g (x))dx
(f(^)-g(x))dx
0,
by
(II), since
to
measure
m(E 1 )=0.
f(x) at all
is zero.
14, 15]
The theorem
15.
of this section is of great
349
importance in the application of
the Lebesgue Integral.

I.
Letfx {x),f^x),
...
be a sequence offunctions which are measurable in
and
not negative.
Also letfn {x) be uniformly bounded* in E, and lim
fn (x) = 0,
for every point
X of E.
Then
Take the arbitrary

Let
Ex
positive
lim
n>-oo
fn (x)dx=0.
E for which
be the points of
1 (x),/2
(x)
...
are
r^y
all<2
E 2 the points for which f {x) ^ gÂjy and/ (z), f3 (x),

2
E3
the points for which fz (x)
and so on.
Then i^,
common
J 2 ,
^ ^gj &nd M x )'M x all< 2mjE)

)>
are a11 measurable,
by
9, III,
and no two
them have
of
points.
Everv point
of
Êr is a point of E
"
and, since
limfn (x) =0, every point
of
->*>
is
< ^j^y
all
...
a point of 2-^rt'
i
Now we
are given that/(z)
is
uniformly bounded in E, and that
it is
not
negative.
~fn(x) < K, the same K
such that
Therefore there is a positive number
serving for every x in E, and every positive integer n.
E = ?,Er
But
m{E) = Y.m{E r ).
and
Therefore we can choose the positive integer N, so that
m(E r )<~-
J
But when n
^ N, fn{x) <
Thus
for every point in
/J*)** ^
2^p) ?
E E2
,
...
EN
m( ^ r)
< |e, when n^.N.

*See footnote 67.2,
II, p. 149.
fLet x be any point of E.
Then we know that lim fn (x) = 0, and
there
is
a positive integer
such that
0^fn (x)< 2m {E) ,whenn^.

The smallest integer
question
Then
is
which
will
supposed taken.
this point
a;
is
a point of
Ev
satisfy this inequality for the point
in
350
And
r=A +lJEr
w
m(E T
fn (x)dx^K A+l
< Je,
But
\f
n (x)dx=i
E
r=l
).
for every n.
Er
fn {x)dx.
Q=\ fn x )dx<h + h e whennîV
Therefore
Thus
>
lim
f
n.>x>
fn (x)dx=0.
We can now prove the following more general theorem

II.
fn (x)
let
xofE*
...
Then
be
lim
u-ôo'e
fn (x)dx = Je f(x)dx.
\fn (x)\<K>
Since
the
a sequence offunctions measurable in E.

be uniformly bounded in E, and lim fn(x) =f(x) for every point
w ^
Let fx (x), f2 (x),
Also
Tapp. ii
same constant
K serving for every x in E and every n, it follows that

\f(x)-fn (x)\^2K.
and
\f(x)\^K
Fn (x) = \f(x)-fn(x)\.
Let
Then
(x),
F 2 (x),
...
are uniformly
bounded and measurable
in E,
and not
negative.
limi^n (a;)=0.
Also
n>qo
Therefore,
by
(I),
It follows that
Thus
lim ( \f(x)
h^-ooJe
-fn (x)\dx=0.
lim 1 (f(x)
n-+ao*E
-fn {x))dx=0.
lim
fn (x)dx =
f(x)dx.
The theorem just proved makes the question of the possibility of term by term
much simpler to answer when Lebesgue Integrals
integration of an infinite series

are used.
Let the functions u x {x), u 2 (x), ...be given in an interval
(a, b),
and the series
00
2 un (x) converge tof{x) in that interval.

i
If
u x (x), u 2 (x),
...
are measurable functions,
we know that/(z) is also measur-
able.
If,
in addition,
we
are told that sn (x)
is
uniformly bounded, this theorem
establishes that
Cb
f(x)dx = \im
and term by term integration

*
We know
integrable in E.
12,
(7>
n > oo a
'a
sn (x)dx,
is possible.
VI) that f(x)
is
measurable and bounded, and therefore
sum
16
is
of the integral, then
T: S
Ten
Md
if
of
noint
P
Integral,
we have
add the eondition that
to
be integrable.*
of the series
Iffix)
Riemann
of the
In the case
tlie
351
15.17]
its
set of points in
Memann's definition
acecrding to
(a, b)
Lebesgue Integral also exists and
* .* b for Vhieh/(,) > A
* A*
same as
is the
it
any constant
an rntenor
is continnons there,
and/(x)
is one of these points
the ends x=a and x-b.
with the usual convention as to
measure.
is a point of a set of zero
it
point,
this
at
discontinuous
if /(.x) is
(Cf. 10. above.)
Thus
consists of
an open
and a
set
set of zero measure.
measurable
sets.
the sum of two

is measurable, as it is
Therefore
Riemann's definition it is measurable
Hence if /(*) is integrable according to
(x) is also bounded.
since/
exists,
Integral
Lebesgue
and its
in (a, 6)
Now
a=x x *
let
...
The sum
a,
with
the notation of
s = m {x
^ m r in {xr_
x
Also /()
-x
Riemann's Integral,
+ m 2 (x 2 - xj +
..
is
given by
+ ro(:rn - xn_ x ).
xr ).
it
(L) f'
Therefore
where
*n=& be a mode
*n-x,
denotes that this
/(*>**
~W
' (aV
" Xr~ l)f
interval.
the Lebesgue Integral for this
is
Hence
s^(L)\
Similarly
8^{L)\Ax)dx.
Since the sums

f(x)dx,
"it
it
Riemann
Integral
bounded function integrable (R) the two
integrals
tend to their
follows that for a
common
value, the
integrable (L) and not

give examples of bounded functions
eTy 'to
fc
8 and
f(x)dx.
"SSl^for every irrational value of x in % x%
and/(*) =0 for every
rational value.
Then
its
f{x)dx = l.
Lebesgue Integral
(Cf. 14, VIII.)
'o
But
(R).
this function is not integrable
Function. In 13-16, f(x) has

The Lebesgue Integral for an Unbounded
The definition
set E
measurable
and
been supposed bounded in the bounded
of un-
17
of the
Lebesgue Integral
is
now
modified, so that
it
will include a class
bounded functions.
define an auxiliary function
Take first the case when f(x)^0 in E, and
.
/.(*) as follows:
=/(
^ ^ ^ ^^ ^ where/( ,
=ra, at
*Cf. footnote
on
p. 161.
of
all
points of E, where f(x)
-g
}
> n.
352
The number n
any assigned positive number.

is bounded and measurable
is
Then,/*n (^, for every n,
Also
fn (x)dx
Thus
fn {x)dx
When
and
exists,
and
is finite, the integral
Whenf(x)
in E, then the in tegral
f(x)dx
is defined
is defined
and
as
in E.
f(x)dx
.'
this integral exists
a monotonic increasing function of n.
is
either converges to a definite limit, or it tends to
the limit exists
[app.
as
oo
by this limit.
f(x) \dx, when
is finite.
Again when f(x) does not keep the same sign in E, write
= \f(x)\ +/(*),
= |/(*)| -/(*),
at all points of # where f(x) ^ 0, and it vanishes at all other
2/i(*)
2/.(*)
so that/âj) =/(*),
points of E.
= -f(x), at
other points of E.
Similarly f2 (x)
all
When
j
fx {x)dx
and\
all
f2 (x)dx
points of
E where f(x) ^ 0,
and are finite,
exist
and
it
vanishes at
the integral[ f(x)dx is de)e'
fined by the equation
f(x)dx = \
)
{x)dx-\
Also when a measurable function f(x)
f2 (x)dx.
such that
is
JE
number, f(x)
is
said
to be
summable
measurable function
necessarily so,
When
f(x)
the integral
if it is
is
f(x)dx
18. Properties
is always summable in E,
unbounded.
summable
of
is
in E,
it is
called the
f(x)dx exists as a finite
in E.
if it is
bounded, but not
also said to be integrable (L) in E,
and
Lebesgue Integral of f(x) in E.
the Lebesgue Integral
f(x)dx,
j
when
f(x)
is
not
bounded.
It is
I.
obvious that for a summable function

Cf(x)dx = c\ f(x)dxt
J E
'
and that if f{x)

summable and
g(x)
^0
in
able
Letf(x) be summable in
and f(x)
f(x)dx i=
II.
E and
let
is
summable, then
g(x) is also
g(x)dx.
E = j?Er
where
E E
lf
2 , ...
are measur-
and without common points.
Then
\
.'E
It is only necessary to
f(x)dx = ?\
1
f(x)dx.
)e,.
prove this for the case when f(x) i^
in E.
'
17j 18j
Define fn (x) as in
and
17,
Then
let
number
Ev E
of sets
/(*)<**=(
be
2 , ...
finite,
say
k.
/(*)**
we have
Proceeding to the limit,
f(x)dx=i\
Again,
the
353
when the number
of sets
Ex E2
infinite,
... is
f(x)dx.
that
we know from
14,
IV
.
(
\
fn(x)dx.
fn (x)dx^
fn (x)dx=?\
'-=l
JE
f
*V
Thus, letting
And
- oo
letting s -> oo
)E
-"r
we have
we have
/n (a#r^[
since
/(i)fe,
r\'
F. r
Fr
fn {x)dx^\
'
Thus
f(x)dx^l\
f(x)dx.
Therefore
f{x)dx =
/(*)**
III. Letf(x)
and
g(x) be
Then
I [^
f{x)dx.
r
summable in E.
(f(x)g(x))dx =
f(x)dx
g(x)dx.
functions.
It will be sufficient to take the sum of the two
before:
Let F(x)=f(x)+g(x) and define n (x),fn (x), gn (x) as
Fn {x) = F(x) at all points of E where ^(a:) ^

=, at all points of # where i^(i) > w.
rc,
e.g.
(i)
But,
when
/(a;)
and
F 2n(x)dx^\ Qx)dx + \jjx)dx^ F n (x)dx.
->
oo
we have
F(x)dx =
(
}
(ii)
And
easy to verify
FJx) and
Also /(*), gn(x),

Therefore
it is
F 2n (x) ^fn (x) + gn (x) i? FJx).

^ 2n (z) are bounded and measurable.
that
Letting
both not-negative in E,
g{x) are
f(x)dx
the same result holds for the case
\
'
g{x)dx.
when f(x) and
g(x) are
or negative in E.
(iii)
Now
let f{x)
^ 0,
and
g{x)
0,
F(x)
Then
(
f(x)dx =
and F(x) =/(
r)
+ g(x)
in #.
+ \g(x)\=Ax),
^ yfryb + ^
fif()
4b =
^ F(x)dx - ^ g(x)dx.
both zero

354
|app.
But when f(x) and g(x) do not each keep the same sign in E, we can
break up E into a certain number of measurable sets without
common points
two by two, to which we can apply the results just found.
(iv)
For example, with the notation of 1 1 the points which belong to

come from the following products:
,
F(x)
^ 0]
will
E[F(x)-^0].
E[f(x)>0].
E[g(x) >0],
E[F(x)^0].
E[f(x)>0].
E[g(x)<0l
E[F(x)^0\.
E[f(x)<0].
E[g(x)>0],
E[F(x)^0].
E[f(x)=0].
E[g(x)>0],
E[F(x)^0].
Elf(x)=0).
E[g(x)=0],
E[F(x)^0].
E[f(x)>0].
E[g(x)=0],
and similarly for the points which belong to E[F(x)

Denote these sets by E x E 2 ... E k
,
Th u
F(x)dx=$[. F(x)dx,
r=
A'
Kr
'
= S([^
=
IV.
With the notation
It follows
from
of 17,
+\
g(x)dx)
g(x)dx.
we have
(III) that, iff(x) is
the Lebesgue Integral
f{x)dx
f{x)dx
\f{x)\dx=\
Thus
< 0].
summable, \f(x)\
is also
summable, and
f (x)dx+\ f2 (x)dx.
1
of an unbounded function
is
an
absolutely con-
vergent integral.
The theorem of 15 applies also to the case of unbounded functions, with

some alteration in the conditions as there given. Term by term integration
is permissible in this case also, but for the discussion
of this question and a
fuller treatment of the Lebesgue Integral reference must
be made to other
works.*
19. Fourier's Series, using Lebesgue Integrals.
Before discussing the convergence of Fourier's Series iorf{x), when the coefficients are Lebesgue
Integrals,
we must prove the Riemann- Lebesgue Theorem (Cf. 1C5, p. 271) for the case
when f(x) is summable in ( - it, it), and in doing so we require the following
approximation theorem
Iff{x) is summable in the interval {a, b) and e is an arbitrary positive number,
there is a continuous function cfi(x), such that
:
!/(*)\
*Cf., for
#*)|<.
example, Hobson, Theory of Functions of a Real Variable,

and 2 (2nd ed. 1920), Ch. V.
1927), Ch. VII,
(3rd ed
18, 19]
This important approximation theorem
355
obtained by proceeding from
is
simple measurable functions to the general summable function.*
Let
(i)
f(x)
in the interval (a,
=0,
in the rest of (a,
/?),
a<a< @<b
where
b).
+ e' on the interval (a, b) between x = a, x = a, anda; = /?,

where 0<e'<e, and join these points to the points x = a,
Take a -e' and
fi
x = b respectively,
y = \, and x^, y=\ respectively.
Then
if </>(#)
=0 va.a<x<a-*.' and
ordinate of this broken line
/?
+ e' < x < b, and <f>(x) is equal to the
we have a continuous function
ma-t'^x
\f(x)
-#r) dx<e.
|
finite set of not- overlapping intervals
elsewhere in
be the continuous function obtained in
r {x)
Ax A2
,
A,
(a, b).
A r and zero elsewhere in (a,
b),
(i)
such that, when
we have
\fT (x)-<f> r {x)\dx<

H
Take
<f>{x)=2<j>r (x).
Then we have
b
\f{x)-^x)\dx^?
la
(iii)
la
Let f(x) = 1 in a measurable set
\fr (x)-<f>r (x)\dx<e.
in
(a,
b)
and zero elsewhere
in
(a, h).
Then a
set of not-overlapping intervals
that the points of
A l5 A 2
...
all
in (a,
adjacent intervals, and
m( E) =S
f W A r < m( #) + Je.
(
Also there
is
N such that
a positive integer
m(E)
^ 2 m(A < m(E) + &,

r)
2 m( A r < |
and
Let/A (x) = 1
in
Then by
we can
(ii)
A A2
lf
A A and
\*
Er
be the points of
Hobson,
zero elsewhere in
in
loc. ciL, 1
<f>(x)
lM*)~<K*)\<fr'
A r (r = l,
2, ...).
E = f,Er
Then
*Cf.
find a continuous function
la
Let
b),
exists such
are interior points of these intervals or end-points of
(3rd ed., 1927), 632.
(a, 6).
such that
two
356
b
Also
\f(x)
-Mx)
dx = |[
f(x)
[app.
-fA{x) dx
|
=2 JA \Ax)-M*)\dx+2
iV+lJA
1
But
And
-fA (x) dx=m{A r )-m{Er
f(x)
/(x) -/\ (a:)
^ 2 m(
dx
r ),
ii
),
when
since
= l,
2,
/A (x) =
|/(x)-/^)|^x.
r
...
in
iV.
2Ar
It follows that
f
Ja
|/()
= r^(A
Li
-A() *
I
ra(#)
since
- m( JB)1
r)
_l
- fm(J*r )~| + 2 m(A r

+ [~m(#)
l_
_l
jv f i
2 m( A r < e.
2 m( Jr = 2 m{E r )^TV+l
)
A'+l
Thus we have
\f(x)-<Mx)\dx<e.
l*
J
(iv)
Let
common
E E2
lt
En
...
be measurable sets in
(a, 6),
no two
of
them having
points.
n
Let fr {x) =
in
and zero elsewhere
.Ey,
in (a,
b),
and /(x)
= 2 cr /r (#), where
i
c lf c 2 ,
We
...
cn
are constants.
find a continuous function
"a
*>
"
^'
{x) as in
<$> r
(iii),
such that
<
** <
|e 1
|c. |...
|-
Take
</>(z)
= 2cr
(/>
r ( a0
Then
(v)
\
Ja
Now
\f(x)-<f>(x)\dxÎ
let/(x) be
With the notation

a
mode
\c r \\
Ja
bounded and measurable
of division
,
taken, the largest of
Also
in (a, b).
B are its lower and upper bounas in (a,
of 13, ^4,
A=l h-.
is
\fr (x)-<j> r (x)\dx<e.
e r is
its partial
**-*,"}*=*
intervals being
rj.
defined as in that section.
A function
F(x)
is
defined for
a^x^b
(r = l,2,
as being equal to l r-1 in
er
...n).
A continuous function <(#) is obtained for j^z) as in (iv), such that

6
[
Ja
\F(x)-cf>(x)\dx<y.
6),
and
19, 20]
Then
\f(x)-<j>(x)\dx^[
Ja
- F(x) dx +
\f(x)
F(x) -
Ja
Ja
^S
\f(x)-F(x)\dx +
357
<f>(x)
dx
\F(x)-<f>(x)\dx
<rj(b-a) + li
when
<,
t)-
2(6 -a)
Let f(x) be summable in (a, 6) and unbounded.

With the notation of 17, we have
(vi)
Let
f(x)dx =
\
la
\
Ja
lf
Ux)dx-^\
Ja
=f1 (x), when fx (x) ^ n
n (x)
= n
and
similarly for
/a
f2(x)dx.
when fx (x) > n
n (:r).
Choose N, so that
b
fx (x)dx - Ja
\
\
Ja
/,,
A (z)<fc
< i,
Then obtain continuous functions
and
[*
Ja
<j> x
[*
/a(s)L - Ja
/,,
and
(x) for /,, A (x)
{x)dx
<f> 2
< \t.
(x) for /,,
x (x),
such that
Ja
and
Aa
(x)
<f> x
{x)
dx < %,
and
Ja
let
<f>(x)
<f> x
{x)
<f>
/a,
A (s)
<f>
t(x)
\dx<le,
2 (x).
But
f
Ja
|/(*)-#r) |d*^(
\fx (x)-fu
ja
+f
J
l/^)-/uWI^ + fa
(*
Therefore
*(*)!& + (* IA 4*)-<M*) | 2 (x)\dx.
\f(x)
<f>(x)
\dx
< 6.
Ja
20.
The Riemann-Lebesgue Theorem.
Iff( x )
summable in
Defining/(x) outside
7r, 7r)
- tt,
tt)
8m
<Aen lim I
/(a:)
ii
cos
n->ooJ-7r
by the equation f(x
f(x) sin wa; <& =
Thus
and
Let
2f
2|j*
(a, /?)
/(z)sinru;<fc: =
/(a:) sin
~\
/(âinnarcfal^j'
be an interval enclosing
- tt,
2ir)
dx=0.
=f(x),
nx dx
f\ x *'~) sm nx ^x
(/(a;)
'
v(* + ^)) sin wards
!/(*+*)
7r).
we have
-/(*) \dx.
358
Then there is a continuous function
f
when
continuous in
(j>(x) is
|ft|
Take
such that
(f>{x),
\f(x)-<j>(x)\dx<&,
(a, (5),
\<j>{x
when
the usual arbitrary positive number.
is
Since
[app.
the same
77,
>/
Then
there
?/
and
?r
+ J
>/
such that
in (a,
a:
<
an
is
+ h)-cfi (x)\<~
serving for every
<
v so large that
given
/?).
/?
|/(*+^)V(*)l^^j'j/(* + ^)-^(* + s)l*'
j'_
+r
i</^+|)-^)i^+r
< e, when
lim
Therefore
?i
x) -/(*)i*e
v.
/(a;) sin
n>-oo J
nx dx = 0,
ir
-77
and the proof
f(x) cos nx dx.
applies equally to
J
IT
Thus the Fourier Constants tend to zero

in
-7r,
as
nôo when f(x)

,
is
summable
7r).
Corollary.
Iff(x)
summable in any
is
lim
nxdxQ.
fix)
cos
n_>oo] a
If (a, 6) lies in
proved, for
ir, tt),
this result follows at once
we may put/(x) =0
extends beyond
If (a, b)
intervals ((m
1)tt,
interval (a, b), then
(m+ 1)tt)
from the theorem just

- ir, ir).
in the remainder of the interval
-tt,
in
tt),
we
which
apply, as above, the theorem to the
it lies,
being a positive integer.
21. We now apply these theorems to the discussion of the Fourier's Series
corresponding to the arbitrary function f(x). We replace the conditions
attached tof{x) in 105 by the condition that f(x) shall be summable (cf. 17
above) in
As
-ir,
tt).
before, let sn(x) be the
nx and cos nx.

Then, if x is a point
sum
of the Fourier's Series forf(x)
up to the terms
in sin
The second
in the interval
integral vanishes,
- ir,
tt),
we have
by the Riemann-Lebesgue Theorem, and
it
follows that
I.
The behaviour of
the Fourier's Series corresponding to f(x), as to conver-
THE DEFINITE INTEGRA!,
20, 21]
gence, divergence, or oscillation at
in the neighbourhood
And
ofx
a point x
359
depends only on
<a<6<
where -re
(a, b),
off(x)
similarly
an
interval
to convergence, divergence, or oscillation,
depends
The behaviour of the Fourier's Series corresponding
II.
the values
as
7r,
to
f(x) in
an
arbitrarily
small
which lim [f(x + h) +f(x - h)]
exists.
+ 8), where
only on the values of f{x) in (a -8, b
is
positive number.
Now let
x be a point in
for
tt, tv)
We may give f(x) at the point x

m,
Then
sn (x
where
the value of this limit.
M*
-/(*,)*-]
(a)
sin(2rc
#()
+ l)a d
7
sina
=f(x + 2a) + f(x - 2a) - 2f(x
*>
).
Therefore
,sin(2n+l)a
a
If*" j/
n(ab)-/(ab)=
<()
ir J
<*
Jsin(2*i + l)aâ.
+-( ^Ta)/'-^
rv
'
ttJo
The second
Theorem, as
integral
<h(a)
7 v
And
the
summable
first
is
summable
integral also vanishes
in (0,
J7r).
when w->oo, provided
that
(f>{a)}a is
Dini's Condition that:
for the convergence of the Fourier's Series correfunction f(x), summable in ( - it, tt), to f(x ) at a point x in
where %hm[f(x + h)+f(x -h)] exists and is equal to f(x ) is that
sponding
(-7r,
a)
in (0, \tt).
Thus we again have

III.
when n->oo, by the Riemann-Lebesgue
vanishes
Vsma
'
a)
\sina
to the
7r),
h-*0
f(x +2a)+f(x -2a)-.

a
This condition
xQ
\ is
summable in some
interval (0,
?/).
when fix) is summable in (-ir,ir) and at the point

Condition; namely that positive numbers C and k exist
is satisfied
satisfies Lipschitz's
such that
\f(x
when
\t\
We
+ t)-f(x )\<C\t\K
^ some fixed positive number.
can also show that:
IV. Whenf(x)
is
summable in
neighbourhood of a point x
-ir,
it),
and
is
of bounded variation in the
the Fourier's Series converges there to
H/(*o+0) +/ta> -0)].

Let/(#) be of bounded variation in the interval (x -2)], x
+ 2r)).
Then
* ,,
0(a)
sin(2w + l)a ,
-da
a
To the second
integral
<!>(<>)
9*** da + (*"
si
c/,(a)
"( 2 " +
>
da.
we can apply the Riemann-Lebesgue Theorem; and in
360
the
first
<f>(a)
is
of
app.
ii
bounded variation; so the Lebesgue Integral and the
Riemann integral are the same.

But the Riemann Integral has
(Cf. 16 above.)
zero for its limit
(cf. 92),
and the required
result follows.
To show
22.
full bearing of the Lebesgue Integral on the Theory of

would be necessary to discuss at much greater length the
the
Fourier's Series
it
properties of that integral.
We
have only touched upon these properties in the case of the integrals
f(x)dx and
f(x)dx, where f(x)
is
summable in the bounded and measurable
set E, or in (a, b).
f(x)dx have not been

a
dealt with at
all.
Some of them may be mentioned here without proof. In

named in the list of books at the end of this section a full
several of the works
discussion of the Lebesgue Integral (and other associated integrals) will be
found.
1'x
f(x)dx
is
continuous and of bounded variation in the
.a
when/(x) is summable, whether bounded or not, in

Also F'(x) exists and is equal to f(x) almost everywhere in
interval (a,
6),
(a, b).
(a,
b),
and
certainly at all points of continuity oif(x).
And, further,
a function which has at every point of (a, 6) a differbounded in that interval, then f'(x) is integrable (L)
in the interval (a, x) and its integral differs from/(x) by a constant only.
In the case of the Riemann Integral this last theorem is subject also to the
condition that f\x) shall be integrable (R).
if f(x) is
ential coefficient /'(*),
REFERENCES.
The
literature dealing
Integral
with the Theory of Sets of Points and the Lebesgue
and other associated
reeller
entitled
Veranderlichen " in
found
useful.
Article
II
integrals is extensive.
'
Neuere
the Enc.
d.
Untersuchungen
iiber
Funktionen
2, will be
math. Wiss., Bd. II, Tl. Ill,
It is divided into three parts

" Die Punktmengen," by Zoretti and Rosenthal
II
b, " Integration und Differentiation," by Montel and Rosenthal
and II C 9 c, " Funktionenfolgen," by Frechet and Rosenthal.
II
C9
C9
a,
And the article II C 10 in the same volume " Neuere Untersuchungen iiber
trigonometrische Reihen " by Hilb and Riesz covers the work in. trigonometrical series up to about 1922.
In Pascal's Repertorium der hoheren Mathematik (2 Aufl.) Bd. I, Tl. Ill,

" Neuere Theorie der reellen Funktionen " (XX),'
there are useful chapters
by Kamke, and " Trigonometrische Reihen " (XXV), by Plessner. Both
:
bring their survey
up
to about the date of publication (1929).
21, 22]
The following books are either devoted wholly to the

useful chapters bearing upon it
Borel, Lemons sur
les
361
subject, or contain
la theorie des fonctions (3e ed., Paris, 1928)
fonctions de variables reelles
et les
Lecons sur
developpements en series de polynomes
(2e ed., Paris, 1928).
De la Vallee
Poussin, Cours
$ Analyse,
1 (3 ed., Paris, 1914)
Integrates
de Lebesgue (Paris, 1916).
Caratheodory, Vorlesungen ilber reelle Funktionen (2 Aufl., Leipzig, 1927).

Hahn, Theorie der reellen Funktionen, 1 (Berlin, 1921).
Hausdorff, Mengenlehre (2 Aufl., Leipzig, 1927).
Hobson, Theory of Functions of a Real Variable, 1 (3rd ed., 1927), and
2 (2nd ed., 1926).
Kamke, Das
Lebesguesche Integral (Leipzig, 1925).
Lebesgue, Lecons sur

(2e ed., Paris, 1928)
I'integration
Lecons sur
et
les series
la recherche des fonctions primitives
trigonometriques (Paris, 1906).
Pierpont, Theory of Functions of Real Variables, 1 (1905) and 2 (1912).

Scheslinger u. Plessner, Lebesguesche Integrate und Fouriersche Reihen
(Leipzig, 1926)-.
Schonflies, Die Entwickelung der Lehre der Punktmannigfaltigkeiten, Tl. I

Tl. II (ibid. Erganzungsband,
(Jahresber. D. Math. Ver., Leipzig, 8 (1900))
;
1908).
Schonflies
Erste Hdlfte
Punktmengen
tj.
Hahn, Entwicklung
und ihre Anwendungen,

Mengen und Theorie der
der Mengenlehre
Allgemeine Theorie der unendlichen

(Leipzig, 1913).
Tonelli, Serie trigonometriche (Bologna, 1928).

Young, W. H. and G. C, The Theory of Sets of Points (1908).
Young, L. C, The Theory of Integration (1927), being No. 21 in the
Cambridge Tracts in Mathematics and Mathematical Physics.
series of
INDEX OF PROPER NAMES

The numbers
refer to pages.
Du
Abel, 139, 149, 165.
Bois-Reymond,
12, 38, 90, 110, 126,
295.
Arzela, 161.
Euclid, 29, 30.
Baker, 137.
Euler, 2-5.
Bernouilli, 2, 4.
9, 254, 264, 288, 293-295, 297,
301, 303, 307, 310.
Bocher,
Fatou, 17.
Fejer, 12, 14, 15, 18, 254, 280, 308
Bonnet, 110.
Fischer, 18, 285.
Borel, 71-73, 329, 330, 361.
Fourier, 1-9, 264, 322.
Boussinesq,
Francis, 187.
Bromwich,
8.
32, 54, 168, 169, 175, 181,
Frechet, 360.
184, 211, 259.
Gibb, 328.
Brown, 298.
Brunei, 133, 181, 211.
Gibbs, 293-295, 310.
Burkhardt,
Gibson, 19, 214.
4, 19.
Gmeiner
Byerly, 218.
Cantor, 12, 14, 30, 31, 33.
Gmeiner).
Gronwall, 295.
Caratheodory, 361.
Carslaw, 161, 173, 283, 295, 298.
Cauchy,
(see Stolz u.
Goursat, 32, 54, 67, 88, 91, 133, 171,

181, 184, 211, 264.
9, 10, 90, 139.
Hahn, 361.
Hardy, 15,
Cesaro, 18, 169.
24,
54,
88,
90,
107,
145,
169, 259.
Chrystal, 166.
Harnack, 126.
Clairaut, 4-6.
Hausdorff, 33, 361.
Cooke, 295.
Heine, 12, 13, 30, 71, 72.

Hilb, 19, 288, 360.
D'Alembert, 2-4.
Hobson,
15, 21, 32, 33, 56, 66, 72, 78,

102, 133, 155, 169, 181, 211, 254, 264,
288, 301, 322, 343, 354, 355, 361.
Darboux, 6, 7, 92.
Dedekind, 23, 24, 27-32.
Delambre, 7.
De
Hurwitz,
1,
14, 288.
la Vallee Poussin, 14, 15, 32, 54, 80,
88, 133, 181, 211, 259, 264,
288/329,
361.
Jackson, 295.
JollifEe,
259.
Descartes, 30.
Jordan, 13, 14, 80, 264, 322, 329.
Dini, 13, 14, 32, 126, 133, 181, 211, 264,

273.
Kamke,
Dirichlet, 5, 9, 10, 13, 14, 90, 151, 219,
Kelvin, Lord,
226, 227.
Donkin, 139.
360, 361.
7.
Knopp, 32, 54, 169, 181.

Kowalewsky, 133, 181.
363
INDEX OF PKOPER NAMES
364
Lagrange,
3, 4, 6,
218.
Rosenthal, 133, 181, 211, 360.
Landau, 169, 259.
Runge, 325.
Laplace,
Russell, 30, 32.
6.
Lebesgue, 15-18, 72, 80, 90, 133, 264,

271, 288, 322, 329, 330, 336, 361.
Legendre,
Sachse, 19, 139.
6.
Schlesinger, 361.
Leibnitz, 30.
Schonflies, 361.
Lipschitz, 13, 273, 359.
Schuster, 328.
Littlewood, 15, 187, 259.
Seidel, 12, 13, 145.
Montel, 133, 181, 211, 360.

Moore, 168.
Stokes, 12, 145.
Sommerfeld, 321.
Neumann,
Newton,
Stolz, 133, 181, 211.
Stolz u. Gmeiner, 20, 32, 54.
264, 322.
30.
Tannery, 32.
Osgood, 51, 64, 88, 133, 144, 181, 211.
Tonelli, 19, 361.
Turner, 326.
Parseval, 14, 17, 18, 284.
Watson
Picard, 254.
(see Whittaker and Watson).

Weber-Riemann, 265, 322.
Pierpont, 88, 96, 99, 106, 123, 133, 181,

211.
Weyl, 295.
Perron, 17.
Weierstrass, 30, 31, 36, 90, 110, 148, 343.
Plancherel, 19,
Whittaker, 325, 327.
Plessner, 19, 360.
Whittaker and Robinson, 325, 328.

Whittaker and Watson, 169, 259, 262,
Poincare,
Poisson,
7.
8, 9,
265, 288.
250, 251, 321.
Pringsheim,
20, 29,
181, 315, 319, 361.
32,
38,
54,
88,
Wilton, 295.
Raabe, 184.
Riemann,
Wilbraham, 294.
Young, Grace Chisholm,
10-12, 14-16, 19, 90, 98,

126, 265, 271.
6,
Riesz, F., 18, 285.
Young, L. C, 361.
Young, W. H., 15,
Riesz, M., 19, 288, 360.
Robinson
(see Whittaker
and Robinson).
18, 33, 73, 90,
361.
Zoretti, 360.
18, 33, 73, 315, 361.
GENERAL INDEX
The numbers
refer to pages.
Abel's Test for Uniform Convergence, 149.
Abel's
Theorem on the Power
Absolute Convergence, of
165; extensions
Series,
series,
50
of, 168.
of integrals, 117, 128.
Absolute Value, 36.
bounded
Aggregate, general notion of, 33 ; bounded above (or on the right), 33
below (or on the left), 34 : bounded, 34 ; upper and lower bounds of, 34 ;
Weierstrass's Theorem on limiting points of, 36.
limiting points of, 35
;
Almost everywhere, definition of, 16.

Approximation Curves for a Series, 139.
See also the Gibbs Phenomenon.
Bdcher's Treatment of the Gibbs Phenomenon, 293.

Bounds (upper and lower), of an aggregate, 34 ; of f(x) in an interval, 56
f(x, y) in a domain, 85.
of
Bromwich's Theorem, 169.
Method of summing Series (C, 1), 169, 258-262.

Change of Order of Terms, in an absolutely convergent series, 51
Cesaro's
convergent
in a conditionally
series, 53.
Closed Interval, definition of, 55.

Conditional Convergence of Series, definition
of, 51.
of
Continuity, of functions, 66 ; of the sum .of a uniformly convergent series
continuous functions, 152 ; of the power series (Abel's Theorem)* 165 ; of
f(x)dx when f(x)
involving a single parameter, 188

parameter, 198, 202.
Continuous Functions, theorems on, 67
86
of ordinary integrals
integrable, 106
bounded and
is
of infinite integrals involving
integrability of, 97
of
two
single
variables,
non-differentiable, 90.
Continuum, arithmetical, 29
linear, 29.
Convergence, of sequences, 37 ; of series, 47 of functions, 57 of integrals, 113,12(5.

See also absolute convergence, conditional convergence, and uniform convergence.
;
Cosine Integral (Fourier's Integral), 312, 320.

Cosine Series (Fourier's Series), 217, 234.
Countably Infinite,definition
of, 21, 330.
Darboux's Theorem, 92.

Dedekind's
Axiom
of Continuity, 28.
Dedekind's Sections, 24.
Dedekind's Theory of Irrational Numbers, 23.

Dedekind's Theorem on the System of Real Numbers, 27.
365
GENERAL INDEX
366
Definite Integrals containing an Arbitrary Parameter (Chapter VI) ; ordinary

integrals, 188; continuity, integration and differentiation of, 188 ; infinite
integrals, 192 ; uniform convergence of, 192 ; continuity, integration and
differentiation of, 198.
Definite Integrals, Ordinary (Chapter IV)
the sums S and s, 91
Darboux's
Theorem, 92 ; definition of upper and lower integrals, 94 ; definition of, 94 ;
necessary and sufficient conditions for existence, 95, 340 ; some properties of,
First Theorem of Mean Value, 105
100
considered as functions of the upper
limit, 106
Second Theorem of Mean Value, 107. See also Dirichlet's Integrals,
Fourier's Integrals, Infinite Integrals, Lebesgue Integrals and Poisson's Integral.
;
Differentiation, of Series, 161

of power series, 167 ; of ordinary integrals, 189
of infinite integrals, 200, 202 ; of Fourier's Series, 282.
;
Dini's Condition, 273, 359.

Dirichlet's Conditions, definition of, 226.
Dirichlet's Integrals, 219.

Dirichlet's Test for
Uniform Convergence,
151.
Discontinuity, of Functions, 73 ; classification

and Points of Infinite Discontinuity.
Divergence, of sequences, 41
Enumerable.
of series,
48
of, 73.
See also Infinite Discontinuity
of functions, 57
of integrals, 113, 126.
See Countably Infinite.
Fejer's Theorem, 254.

Fejer's
Theorem and the Convergence
of Fourier's Series, 262, 280.
Fourier's Constants (or Coefficients), definition
of,
215.
Fourier's Integrals (Chapter X); simple treatment of, 312, more general conditions for, 315
cosine and sine integrals, 320 ; Sommerf eld's discussion
of, 321.
;
Fourier's Series, definition of, 215 ; Lagrange's treatment of, 218 ; proof of convergence of, under certain conditions, 230 ; for even functions (the cosire
series), 234
for odd functions (the sine series), 241 ; for intervals other than
Poisson's discussion of, 250 ; Fejer's Theorem, 254, 262, 280
( -7T, it), 248
order of the terms in, 269
uniform convergence of, 275 ; differentiation and
integration of, 282
more general theory of, 271, 358.
;
Functions of a Single Variable, definition of, 55 bounded in an interval, 56 Upper

and lower bounds of, 56 oscillation at a point, 66 oscillation in an interval,
56 limits of, 56 continuous, 66 discontinuous, 73 monotonic, 75 inverse,
76 integrable, 97 of bounded variation, 80 measurabie, 342 summable, 352.
Functions of Several Variables, 84.
;
General Principle of Convergence, of sequences, 38 of functions, 61.

Gibbs Phenomenon in Fourier's Series (Chapter IX), 289.
;
Hardy-Landau Theorem, 259.

Harmonic Analyser (Kelvin's), 323.
Harmonic Analysis (Appendix I), 323.
Heine-Borel Theorem, 71.
Improper Integrals, definition
Infinite Aggregate.
of, 126.
See Aggregate.
Infinite Discontinuity.
See Points of Infinite Discontinuity.
Infinite Integrals (integrand function of a single variable), integrand bounded and

interval infinite, 112 ; necessary and sufficient condition for convergence of,
with positive integrand, 115

114
absolute convergence of, 117; /i-test for
convergence of, 119 other tests for convergence of, 120 ; mean value theorems
;
for, 123.
367
GENERAL INDEX
variable), integrand infinite 125 ;

Infinite Integrals (integrand function of a single
127 ; absolute convergence of, 128.
/i-test and other tests for convergence of,
variables), definition of uniform conInfinite Integrals (integrand function of two
continuity, invergence of, 193 ; tests for uniform convergence of, 193 ;
tegration and differentiation of, 198.
Infinite
See Sequences and Series.
Sequences and Series.
Infinity of
of, 74.
a Function, definition
Integrable (L) and Integrable (R), definition of, 345.

209.
Integration of Integrals (ordinary), 191 ; infinite, 199, 202,
power series, 167 ; Fourier's Series,
Integration of Series (ordinary integrals), 156
Integrate Functions, 97
283
(infinite integrals), 172.
open at one end and closed at the

and not-overlapping, 333 abutting, 333.
Interval, open, closed,
other, 55
overlapping
Inverse Functions, 76.

Irrational
See Numbers.
Numbers.
function, 344
Lebesgue Definite Integral, of a bounded and measurable
summable
of a
function, 352.
Lebesgue Indefinite Integral, 360.

of functions, 56
Limits, of sequences, 37
;
of functions of
two
variables, 85
repeated, 142.
Limits of Indetermination, of a bounded sequence, 43

Limiting Points of an Aggregate, 35.
of a
bounded function,
64.
Lipschitz's Condition, 273, 359.
Lower
Integrals, definition of, 94.
Mean Value Theorems
of the Integral Calculus
123
integrals),
(infinite integrals), 123.
105
(infinite
first
integrals),
theorem (ordinary
second theorem (ordinary integrals), 107
Measure of a Set of Points, 335 exterior measure, 333 interior, 334.

Measurable Sets of Points, 335.
Measurable Functions, 342.
Modulus. See Absolute Value.
integrability
admit only ordinary discontinuities, 76
Monotonic Functions, 75
;
of, 97.
Monotonic in the Stricter Sense, definition of, 43,

Monotonic Sequences, 42.
M-test for Convergence of Series, 148.
yu-test for Convergence of Integrals, 119, 129.
75.
Neighbourhood of a Point, definition of, 58.

irrational,
Numbers (Chapter I) rational, 20 irrational, 21 Dedekind's theory of
development of
Dedekind's Theorem on the system of real, 27
real, 25
23
Dedekind s
the system of real, 29. See also Dedekind's Axiom of Continuity, and
;
Sections.
Open
Interval, definition of, 55.
Ordinary or Simple Discontinuity, definition

Oscillation of a Function at a Point, 66.
Oscillation of a Function in
domain,
an
Interval, 56
of, 74.
of a function of
two variables
85.
Oscillatory, Sequences, 41
series,
48
functions, 58
integrals, 113, 126.
Parseval's Theorem, 284.

Partial
Remainder p R n ),
(
definition of, 48
[#(*)], definition
of, 138.
in a
GENERAL INDEX
368
Periodogram Analysis, 326.
Points of Infinite Discontinuity, definition of, 75.

Points of Oscillatory Discontinuity, definition of, 74.
Poisson's Discussion of Fourier's Series, 250.
Poisson's Integral, 251.
Power Series,
interval of convergence of, 163 ; nature of convergence of, 165

integration and differentiation of, 167.
theorem on, 165
Abel's
Proper Integrals, definition
of, 126.
Rational Numbers and Real Numbers. See Numbers.

Remainder after n Terms (R n ), definition of, 49 [R (x)]
n
;
138.
Repeated Limits, 142.

Repeated Integrals, (ordinary), 191
(infinite), 199, 202, 209.
Riemann-Lebesgue Theorem, 271, 357.
Riesz-Fischer Theorem, 1 8.
;
Sections.
See Dedelcind's Sections.

Sequences
convergent, 37
limit of, 37
necessary and sufficient condition for
convergence of (general principle of convergence), 38 ; divergent and oscillatory, 41
monotonic, 42.
;
definition of sum of an infinite, 47 ; convergent, 47 ; divergent and oscillatory, 48

necessary and sufficient condition for convergence of, 48 ; with
positive terms, 49
absolute and conditional convergence of, 50 ; definition
of sum, Avhen terms are functions of a single variable, 137 ; uniform con-
Series
vergence of, 144

necessary and sufficient condition for uniform convergence
of, 147
Weierstrass's ilf-test for uniform convergence of, 148 ; uniform
convergence and continuity of, 152
term by term differentiation and integration of, 156.
See also Differentiation of Series, Fourier'' s Series, Integration of
;
Power Series and Trigonometrical iSeries.

of Points on a Line
bounded, 33
limiting points of, 35 countable (or
enumerable), 330 open, 330 closed, 330 interior and exterior points of, 330
complement of, 331 operations on, 331 j interior and exterior measure of, 333
measure of, 335
measurable, 335
nepessary and sufficient condition that a
Series,
Sets
set be measurable,
336
properties of rneasurable sets, 337.
Simple (or Ordinary) Discontinuity, definition
of, 74.
Sine Integral (Fourier's Integral), 312, 320.

Sine Series (Fourier's Series), 217, 241.
Summable Functions, 352.

Summable Series (C, 1), definition
Sums S and 5, definition of, 91.
of,
169.
Trigonometrical Series, 215.
Uniform Continuity of a Function, 69, 87.

Uniform Convergence, of Series, 144; Abel's Test
151
for,
149; Dirichlet's Test
for,
of Integrals, 192.
Uniformly Bounded, 149.
Upper
Integrals, definition of, 94.
Weierstrass's non-differentiable Continuous Function, 90.

Weierstrass's ill-test for Uniform Convergence, 148.
Weierstrass's
Theorem on Limiting Points
of a
Bounded Aggregate,
36.
PRINTED IN CiREAT BRITAIN BY ROBERT MACLEHOSK AND CO. LTD. THE UNIVERSITY PRESS, GLASGOW.
INFINITE
SEQUENCES AND
SERIES
by Konrad Knopp
modern mathematics, Dr.
topics that are of partwo
upon
Konrad Knopp here concentrates
One
of the best expositors
ticular
interest
theory of
to
infinite
20th
in
the field of
century
mathematicians.
sequences and series from
its
He develops the
beginnings to a point
where the reader will be in a suitable position to investigate more

advanced stages on his own initiative. The foundations of the theory
are therefore presented with especial care, while developmental aspects are limited to the scope and purpose of the book.
theorems are proved with enough

detail to make them readily comprehensible. In Chapter 1, the author
All definitions are clearly stated; all
begins with
the
of
construction
the
system
of
real
and complex
numbers, and proceeds through such fundamental concepts as sets
and functions of real and complex variables. In his treathe covers arbitrary and null
ment of Sequences and Series Chapter 2
sequences; sequences and sets of of numbers; convergence, divergence;
Cauchy's limit theorem; main tests for sequences; and infinite series.
of numbers,
) ,
Chapter 3 deals with main tests for infinite series, and operating
with convergent series. Chapters 4 and 5 explain power series and
the development of the theory of convergence, respectively. Chapter
6 treats expansion of the elementary functions; and Chapter 7 concludes with numerical and closed evaluation of series.
Translated
5%
by
Frederick
Bagemihl.
Bibliography.
Index.
186pp.
x 8.
SI 53
Paperbound $1.75
THE ANALYTICAL THEORY OF HEAT

by Joseph Fourier
unabridged reprinting of the Freeman translation of
THEORIE ANALYTIQUE DE LA CHALEUR. In this landmark
modern science, the great French mathematician first showed how
This
is
an
Fourier's
of
any discontinuous function could be represented by a trigonometric

series, and advanced many other concepts, methods, and insights
that
opened the way
to
modern mathematical
physics.
Besides being a compendious treatment of problems associated with

heat, both practical
and
theoretical, Fourier's
THEORIE ANALYTIQUE
has served generations of mathematicians interested
in
trigonometric
and their applications. It is still a mine of insights and a

source book for physicists, chemists, mathematical physicists, engineers, and persons who are concerned either with heat or the
mathematical tools which Fourier developed.
series
PARTIAL CONTENTS: Introduction. Heating of closed spaces. Equaof heat movement. Heat propagation in infinite rectangular
tion
solid.
Arbitrary
movement
Movement
functions
of heat
of heat
angular prism.
in
in
in
ring.
cylinder.
solid
Movement
trigonometric
of heat
in
Linear,
series.
Propagation of heat
in
varied
solid sphere.
Propagation of heat
in
rect-
solid cube. Diffusion of heat.
Addendum.
"The ultimate source of much modern work
tions
of a
real
variable,"
E.
T.
Bell,
in
the theory of func-
DEVELOPMENT OF MATHE-
MATICS. "Indispensable in the treatment of nearly every recondite

question in modern physics," Thomson & Tait, ELEMENTS OF NATURAL PHILOSOPHY. "Fourier's discoveries dominate the whole of
mathematical physics," ENCYCLOPAEDIA BRITANNICA.
Translated
by Alexander Freeman, M.A. 77 footnotes, containing

3
-f- 466pp. 5 /s x 8.
Paperbound $2.00
S93
considerable bibliography. 20 figures, xxii
Catalogue of Dover
SCIENCE BOOKS
BOOKS THAT EXPLAIN SCIENCE
THE NATURE OF LIGHT AND COLOUR IN THE OPEN AIR, M. Minnaert. Why is falling snow
sometimes black? What causes mirages, the fata morgana, multiple suns and moons in the
sky; how are shadows formed? Prof. Minnaert of U. of Utrecht answers these and similar
questions in optics, light, colour, for non-specialists. Particularly valuable to nature,
fascinating!"
science students, painters, photographers. "Can best be described in one word
Physics Today. Translated by H. M. Kremer-Priest, K. Jay. 202 illustrations, including 42
Paperbound
$1.95
T196
8.
x
photos, xvi
362pp. 53/s
UNIVERSE, Max Born. New enlarged version of this remarkably readable

Nobel laureate. Moving from sub-atomic particles to universe, the author
explains in very simple terms the latest theories of wave mechanics. Partial contents-, air
and its relatives, electrons and ions, waves and particles, electronic structure of the
atom, nuclear physics. Nearly 1000 illustrations, including 7 animated sequences. 325pp.
T412 Paperbound $2.00
6x9.
THE
RESTLESS
account
by
MATTER AND LIGHT, THE NEW PHYSICS,
L.
de Broglie. Non-technical papers by a Nobel
laureate explain electromagnetic theory, relativity, matter, light, radiation, wave mechanics,
quantum physics, philosophy of science. Einstein, Planck, Bohr, others explained so easily
that no mathematical training is needed for all but 2 of the 21 chapters. "Easy simplicity
.
should make this source-book of modern physcis available to a wide
and lucidity
.
public,"
THE
Saturday
Review.
COMMON SENSE
Unabridged.
300pp.
5%
8.
T35
Paperbound $1.60
OF THE EXACT SCIENCES, W. K. Clifford. Introduction by James NewPearson. For 70 years this has been a guide to classical scientific,
man, edited by Karl

mathematical thought. Explains with unusual clarity basic concepts such as extension of
meaning of symbols, characteristics of surface boundaries, properties of plane figures,
vectors, Cartesian method of determining position, etc. Long preface by Bertrand Russell.
Bibliography of Clifford. Corrected. 130 diagrams redrawn. 249pp. 5% x 8.
THE EVOLUTION OF SCIENTIFIC THOUGHT FROM NEWTON TO EINSTEIN,
A. d'Abro. Einstein's
special, general theories of relativity, with historical implications, analyzed in non-technical
terms. Excellent accounts of contributions of Newton, Riemann, Weyl, Planck, Eddington,
Maxwell, Lorentz, etc., are treated in terms of space, time, equations of electromagnetics,
finiteness of universe, methodology of science. "Has become a standard work," Nature. 21
T2 Paperbound $2.00
diagrams. 482pp. 53/a x 8.
BRIDGES AND THEIR BUILDERS, D. Steinman, S. R. Watson. Engineers, historians, everyone

ever fascinated by great spans will find this an endless source of information and interest.
Dr. Steinman, recent recipient of Louis Levy Medal, is one of the great bridge architects,
engineers of all time. His analysis of great bridges of history is both authoritative and
easily followed. Greek, Roman, medieval, oriental bridges; modern works such as Brooklyn
Bridge, Golden Gate Bridge, etc. described in terms of history, constructional principles,
artistry, function. Most comprehensive, accurate semi-popular history of bridges in print in
English. New, greatly revised, enlarged edition. 23 photographs, 26 line drawings, xvii
x 8.
401pp.
5%
CATALOGUE OF
CONCERNING
THE
NATURE
OF
THINGS,
Society by Nobel
laureate, dealing
scientific background is needed
processes and aspects of modern
Morning Post. 32pp. of photos. 57
No
William Bragg. Christmas lectures at Royal

with atoms, gases, liquids, and various types of crystals.
to understand this remarkably clear introduction to basic
science. "More interesting than any bestseller," London
figures, xii
x 8.
232pp.
Sir
5%
THE RISE OF THE NEW PHYSICS, A. d'Abro. Half million word exposition, formerly titled
"The Decline of Mechanism," for readers not versed in higher mathematics. Only thorough
explanation in everyday language of core of modern mathematical physical theory, treating
both classical, modern views. Scientifically impeccable coverage of thought from Newtonian
system through theories of Dirac, Heisenberg, Fermi's statistics. Combines history, exposition;
broad but unified, detailed view, with constant comparison of classical, modern
views. "A must for anyone doing serious study in the physical sciences," J. of the Franklin
"Extraordinary faculty ... to explain ideas and theories ... in language of everyday
of set: philosophy of science, from practice of Newton, Maxwell, Poincare,
Isis. Part
etc. Modes of thought, experiment, causality, etc. Part II: 100 pp. on grammar,
vocabulary of mathematics, discussions of functions, groups, series, Fourier series, etc.
Remainder treats concrete, detailed coverage of both classical, quantum physics: analytic
mechanics, Hamilton's principle, electromagnetic waves, thermodynamics, Brownian movement, special relativity, Bohr's atom, de Broglie's wave mechanics, Heisenberg's uncertainty,
scores of other important topics. Covers discoveries, theories of d'Alembert, Born, Cantor,
Debye, Euler, Foucault, Galois, Gauss, Hadamard, Kelvin, Kepler Laplace, Maxwell, Pauli,
Rayleigh Volterra, Weyl, more than 180 others. 97 illustrations, ix
982pp. 53/8 x 8.
T3 Vol. 1 Paperbound $2.00
T4 Vol. II Paperbound $2.00
Inst.
life,"
Einstein,
SPINNING TOPS AND GYROSCOPIC MOTION, John

unsurpassed
flexible,
internal
in ships,
for
lucid,
Perry. Well-known classic of science still

exposition. How quasi-rigidity is induced in
falls,
top rises; nature, effect of
Appendixes describe practical use of gyroscopes
transportation. 62 figures. 128pp.
x 8.
accurate,
delightful
motions;
fluid bodies by rapid
fluidity on rotating bodies; etc.
compasses,
monorail
why gyrostat
5%
FOUNDATIONS OF PHYSICS,
Lindsay, H. Margenau. Excellent bridge between semipopular and technical writings. Discussion of methods of physical description, construction
of theory; valuable to physicist with elementary calculus. Gives meaning to data, tools of
modern physics. Contents: symbolism, mathematical equations; space and time; foundations
of mechanics; probability; physics, continua; electron theory; relativity; quantum mechanics;
causality; etc. "Thorough and yet not overdetailed. Unreservedly recommended," Nature.
x 8.
Unabridged corrected edition. 35 illustrations, xi
S377 Paperbound $2.45
537pp.
R.
B.
5%
FADS AND FALLACIES IN THE NAME OF SCIENCE, Martin Gardner. Formerly
entitled "In the

of Science," the standard account of various cults, quack systems, delusions which
have masqueraded as science: hollow earth fanatics, orgone sex energy, dianetics, Atlantis,
Forteanism, flying saucers, medical fallacies like zone therapy, etc. New chapter on Bridey
Murphy, psionics, other recent manifestations. A fair reasoned appraisal of eccentric theory
which provides excellent innoculation. "Should be read by everyone, scientist or nonscientist alike," R. T. Birge, Prof. Emeritus of Physics, Univ. of Calif;
Former Pres.,
Amer. Physical Soc, x + 365pp. 53/8 x 8.
Name
ON MATHEMATICS AND MATHEMATICIANS,
R. E. Moritz. A 10 year labor of love by discerning,

discriminating Prof. Moritz, this collection conveys the full sense of mathematics and
great
personalities of
mathematicians. Anecdotes,
aphorisms, reminiscences, philosophies,
definitions, speculations, biographical insights, etc. by great mathematicians, writers: Descartes, Mill, Locke, Kant, Coleridge, Whitehead, etc. Glimpses into lives of great mathematicians,
from Archimedes to Euler, Gauss, Weierstrass. To mathematicians, a superb
browsing-book. To laymen, exciting revelation of fullness of mathematics. Extensive cross
index. 410pp.
5%
x 8.
OF MATHEMATICS AND PHYSICS, N. G. Parke III. Over 5000

entries under approximately 120 major subject headings, of selected most important books,
monographs, periodicals, articles in English, plus important works in German, French,
Italian, Spanish, Russian, (many recently available works). Covers every branch of physics,
GUIDE TO THE
LITERATURE
math, related engineering. Includes author, title, edition, publisher, place, date, number
volumes, number of pages. 40 page introduction on basic problems of research, study
provides useful information on organization, use of libraries, psychology of learning, etc.
Indices of authors, subjects. 464pp.
Will save you hours of time. 2nd revised edition.
53/s x 8.
of
THE STRANGE STORY OF THE QUANTUM, An Account
for the General Reader of the Growth

Ideas Underlying Our Present Atomic Knowledge, B. Hoffmann. Presents lucidly, expertly,
with barest amount of mathematics, problems and theories which led to modern quantum
physics. Begins with late 1800's when discrepancies were noticed; with illuminating analSchroedinger, Dirac,
Pauli,
ogies, examples, goes through concepts of Planck, Einstein,
Sommerfield, Feynman, etc. New postscript through 1958. "Of the books attempting an
account of the history and contents of modern atomic physics which have come to my
attention, this is the best," H. Margenau, Yale U., in Amer. J. of Physics. 2nd edition. 32
of
tables,
illustrations.
275pp. 53/8 x
8.
T518 Paperbound $1.45"
DOVER SCIENCE BOOKS

HISTORY OF SCIENCE
AND PHILOSOPHY OF SCIENCE
Poincare. Many of most mature ideas of 'Mast scientific
whether
truth,
of scientific
universalist" for both beginning, advanced workers. Nature
by man, logical thought vs. intuition (relating to
imposed
or
order is innate in universe
simultaneity),
psychological time
Weierstrass, Lie, Riemann, etc), time and space (relativity
Newton
Mayer
Carnot,
Hen's concept of force values within disciplines of Maxwell,
147pp. 53/8 x 8.
Lorentz, etc. Mi
THE VALUE OF SCIENCE, Henri
science
L. S. Stebbing. Philosophical aspects of modern
science,
terms of lively critical attack on ideas of Jeans, Eddington. Tasks of
experience,
determinism, probability, relation of world physics to that of everyday
energy levels, inferences
philosophical significance of Planck-Bohr concept of discontinuous
of "becoming" involved in 2nd law
to be drawn from Uncertainty Principle, implications
2 |5f^>.
Laplaceî^dete^mimsm.^
of
discarding
of thermodynamics, other problems posed by
PHILOSOPHY AND THE PHYSICISTS,
examined
in
causality,
5%
8.
METHOD, W.
THE PRINCIPLES OF SCIENCE, A TREATISE ON LOGIC AND THE SCIENTIFIC
to
S.
inremains stimulating contribution

deductive logic, theory of
inductive
conlogic,
s
Boole
number probability, limits of scientific method; significantly advances
of probability in physics, astronomy,
tains detailed introduction to nature and methods
says, [Jevons] continues
everyday affairs, etc. In introduction, Ernest Nagel of Columbia U.
of scientific
logic
the
articulate
to
to be of interest as an attempt
"
pe/bound Ilia
5 3/s x 8.
Jevons. Milestone in development of

vestigation of inferential validity in
symbolic
logic
sciences.
Treats
""J^
Dreyer. Only work in English

A HISTORY OF ASTRONOMY FROM THALES TO KEPLER, J. L. E.
prehistoric times to Kepler. Partial
to give complete history of cosmological views from
Homocentric spheres of
Greeks,
Early
systems,
Near Eastern astronomical
contents:
Kepler, much more.
Euxodus Epicycles, Ptolemaic system, Medieval cosmology, Copernicus,
unsurpassed in
"Especially useful to teachers and students of the history of science
Revised
Kepler.
to
Thales
from
Systems
Isis. Formerly "A History of Planetary
its field,"
by W. H. Stahl. xvii + 430pp. 53/8 x 8.
foreword
HISTORY OF MATHEMATICS, D. Struik. Lucid study of development of ideas,

Ages, Renaissance,
techniques, from Ancient Near East, Greece, Islamic science, Middle
Important mathematicians described in detail. Treatment not anecdotal, but
times.
modern
Rich in conneeded.
training
analytical development of ideas. Non-technical no math
Booklist.
60 illustrations including
Quarterly
U.S.
interpretations,"
in
thoughtful
tent
xix
2 5?RP;
2nd edition,
Greek, Egyptian manuscripts, portraits of 31 mathematicians.
A CONCISE
'
53/a
8.
carefully balanced
covers such matters
It
exoositon of Peirce's complete system, written by Peirce himself.
pragmatism,
signs,
of
theory
logic,
symbolic
law,
vs.
chance
pure
as scientific method,
gives more than adequate evidence
experiment, and other topics. "Excellent selection
of Peirce
Philosophy
"The
entitled
Formerly
Personalist.
of the range and greatness,"
217 raperoouna >i.3o
THE PHILOSOPHICAL WRITINGS OF PEIRCE, edited by Justus Buchler. A
xvi
'
368pp.
of scientific discovery, methodology, exbeing. Most signifiperiment, idea-germination processes by which discoveries come into
of ideas. Chapters cover selection
application
cant and interesting aspects of development,
mathematics and logic; Whitehead, Russell,
reasoning,
chance, mathematical
facts,
of
Cantor, the new mechanics, etc. 288pp. 53/8 x 8.
SCIENCE AND METHOD, Henri Poincare. Procedure
science. How such conin

mechanics develoDed, how modern
as number, magnitude, space, force, classical
modern physics.
of
theories
scientist uses them in his thought. Hypothesis in physics,
the breadth of vision
Introduction by Sir James Larmor. "Few mathematicians have had
Bell. 272pp.
T.
E
exposition
of Poincare, and none is his superior in the gift of clear
S221 Paperuouna $1,35
S3/ x 8
SCIENCE AND HYPOTHESIS, Henri Poincare. Creative psychology

cepts
most
ESSAYS INEXPERIMENTAL LOGIC, John Dewey. Stimulating series of essays by one of
his most mature thoughts on
influential minds in American philosophy presents some of
experience;
and
inquiry
between
Relationship
contents:
Partial
subjects.
wide range of
and
data
dependence of knowledge upon thought; character logic; judgments of practice
meanings; stimuli of thought, etc.
viii
444pp. 53/8
x 8.
WHAT IS SCIENCE, Norman Campbell. Excellent introduction explains scientific method, role
science and
of science,
of mathematics, types of scientific laws. Contents-. 2 aspects
measurement and numerical
nature, laws of chance, discovery of laws, explanation of laws,
laws, applications of science. 192pp. 53/8 x 8.
3
CATALOGUE OF
FROM EUCLID TO EDDINGTON: A STUDY OF THE CONCEPTIONS OF THE EXTERNAL WORLD,
Edmund Whittaker. Foremost
from
western
divisions: Space, Time
Mechanics; Eddington
physical
world with
curvature, etc. 212pp.
losophy
Sir
scientist traces development of theories of natural phiEuclid to Eddington, Einstein, Dirac, etc. 5 major
and Movement; Concepts of Classical Physics; Concepts of Quantum
Universe. Contrasts inadequacy of classical physics to understand
present day attempts of relativity, non-Euclidean geometry, space
53/8 x 8.
British
rediscovery
of
THE ANALYSIS OF
MATTER, Bertrand Russell. How do our senses accord with the new
This volume covers such
topics as
logical
analysis of physics,
prerelativity
causality, scientific inference, physics and perception, special and general relativity, Weyl's theory, tensors,
invariants and their physical interpretation, periodicity and
qualitative series. "The most thorough treatment of the subject that has yet been published," The Nation. Introduction by L. E. Denonn. 422pp.
x 8.
physics?
physics,
5%
LANGUAGE, TRUTH, AND LOGIC,
Ayer. A clear introduction to the Vienna and Cambridge

schools of Logical Positivism. Specific tests to evaluate validity of ideas, etc. Contents:
function of philosophy, elimination of metaphysics, nature of analysis, a priori, truth and
probability, etc. 10th printing. "I should like to have written it myself," Bertrand Russell.
160pp. 53/s x 8.
A.
THE PSYCHOLOGY OF INVENTION IN THE MATHEMATICAL FIELD, J. Hadamard. Where do ideas

come from? What role does the unconscious play? Are ideas best developed by mathematical
reasoning, word reasoning, visualization? What are the methods used by Einstein, Poincare,
Galton, Riemann? How can these techniques be applied by others? One of the world's
leading
mathematicians discusses these and other questions,
xiii
145pp.
T107
5%
8.
Paperbound $1.25
GUIDE TO PHILOSOPHY,
C. E. M. Joad. By one of the ablest expositors of all time, this is

a history or a typological survey, but an examination of central problems in
answers afforded by the greatest thinkers: Plato, Aristotle, Scholastics, Leibniz,
Kant, Whitehead, Russell, and many others. Especially valuable to persons in the physical
sciences; over 100 pages devoted to Jeans, Eddington, and others, the philosophy of
modern physics, scientific materialism, pragmatism, etc. Classified bibliography. 592pp.
5% x 8.
not
simply
terms
of
SUBSTANCE AND FUNCTION,
and EINSTEIN'S THEORY OF RELATIVITY, Ernst Cassirer. Two

books bound as one. Cassirer establishes a philosophy of the exact sciences that takes into
consideration
new developments in mathematics, shows historical connections. Partial
contents: Aristotelian logic, Mill's analysis, Helmholtz and Kronecker, Russell and cardinal
numbers, Euclidean vs. non-Euclidean geometry, Einstein's relativity. Bibliography. Index,
xxi + 464pp. 53/s x 8.
FOUNDATIONS OF GEOMETRY, Bertrand Russell. Nobel laureate analyzes basic problems in

the overlap area between mathematics and philosophy: the nature of geometrical knowledge,
the nature of geometry, and the applications of geometry to space. Covers history of nonEuclidean geometry, philosophic interpretations of geometry, especially Kant, projective
and metrical geometry. Most interesting as the solution offered in 1897 by a great mind
to a problem still current. New introduction by Prof. Morris Kline, N.Y. University. "Admirably clear, precise, and elegantly reasoned analysis," International Math. News, xii
201pp. 53/8 x
8.
THE NATURE OF PHYSICAL THEORY,
P. W. Bridgman. How modern physics looks to a highly

unorthodox physicist a Nobel laureate. Pointing out many absurdities of science, demonstrating inadequacies of various physical theories, weighs and analyzes contributions of
Einstein, Bohr, Heisenberg, many others. A non-technical consideration of correlation of
science and reality, xi + 138pp. 5 3/8 x 8.
EXPERIMENT AND THEORY

and
IN
PHYSICS,
Max Born. A Nobel
laureate examines the nature

physics. Synthetic versus
Bohr, Heisenberg, Planck,
of the counterclaims of experiment and theory in

scientific advances are analyzed in works of Einstein,
Eddington, Milne, others, by a fellow scientist. 44pp. 5 3/8 x 8.
value
analytical
S308 Paperbound 600
A SHORT HISTORY OF ANATOMY AND PHYSIOLOGY FROM THE GREEKS TO HARVEY, Charles
Singer. Corrected edition of "The Evolution of Anatomy." Classic traces anatomy, physiology from prescientific times through Greek, Roman periods, dark ages, Renaissance, to
beginning of modern concepts. Centers on individuals, movements, that definitely advanced
anatomical knowledge. Plato, Diodes, Erasistratus, Galen, da Vinci, etc. Special section
on Vesalius. 20 plates. 270 extremely interesting illustrations of ancient, Medieval,
enaissance, Oriental origin, xii + 209pp. 53/8 x 8.
SPACE -TIME -MATTER,
Hermann Weyl. "The standard treatise on the general theory of

relativity," (Nature), by world renowned scientist. Deep, clear discussion of logical coherof general theory, introducing all needed tools: Maxwell, analytical geometry, nonEuclidean geometry, tensor calculus, etc. Basis is classical space-time, before absorption
of relativity. Contents: Euclidean space, mathematical form, metrical continuum, general
theory, etc. 15 diagrams, xviii
330pp. 5 3/a x 8.
ence
DOVER SCIENCE BOOKS

MATTER AND MOTION, James
Clerk Maxwell. Excellent exposition begins with simple parproceeds gradually to physical systems beyond complete analysis; motion, force
properties of centre of mass of material system; work, energy, gravitation
etc
Written
with all Maxwell's original insights and clarity. Notes by E. Larmor. 17 diagrams
I78pp
53/ x 8
S188 Paperbound
ticles,
$1.25
PRINCIPLES OF MECHANICS, Heinrich Hertz. Last work by the great 19th century physicist
is not only a classic, but of great interest in the logic of science. Creating
a new system
of mechanics based upon space, time, and mass, it returns to axiomatic analysis
under-
standing of the formal or structural aspects of science, taking into account logic, observation, a priori elements. Of great historical importance to Poincare, Carnap, Einstein
Milne
A 20 page introduction by R. S. Cohen, Wesleyan University, analyzes the implications of
Hertz's thought and the logic of science. 13 page introduction by Helmholtz. xlii
+ 274pp
53/s x 8
S316 Clothbound $3.50
-
FROM MAGIC TO SCIENCE,
Charles Singer. A great historian examines aspects of science

from Roman Empire through Renaissance. Includes perhaps best discussion of early herbals
penetrating physiological interpretation of "The Visions of Hildegarde of Bingen " Also
examines Arabian, Galenic influences;
Pythagoras'
sphere,
Paracelsus;
reawakening of
science under Leonardo da Vinci, Vesalius; Lorica of Gildas the Briton; etc
Frequent
quotations with translations from contemporary manuscripts. Unabridged
corrected edition. 158 unusual illustrations from Classical, Medieval sources, xxvii +
365pp. 53/8 x 8.
A HISTORY OF THE CALCULUS, AND ITS CONCEPTUAL DEVELOPMENT, Carl B. Boyer. Provides
laymen, mathematicians a detailed history of the development of the calculus, from beginnings in antiquity to final elaboration as mathematical abstraction. Gives a sense
of
mathematics not as technique, but as habit of mind, in progression of ideas of Zeno Plato
Pythagoras, Eudoxus, Arabic and Scholastic mathematicians, Newton, Leibniz, Taylor
Descartes, Euler, Lagrange, Cantor, Weierstrass,
and others. This first comprehensive critical
history of the calculus was originally entitled "The Concepts of the Calculus."
Foreword
by R. Courant. 22 figures. 25 page bibliography, v + 364pp. 53/8 x 8.
A
DIDEROT PICTORIAL
ENCYCLOPEDIA OF TRADES AND INDUSTRY,

Manufacturing and the
Technical Arts in Plates Selected from "L'Encyclopedie ou Dictionnaire
Raisonne des
Sciences, des Arts, et des Metiers" of Denis Diderot. Edited with text by C.
GillisDie
First
modern selection of plates from high-point of 18th century French engraving.
Storehouse
of technological information to historian of arts and science. Over
2,000 illustrations on
485 full page plates, most of them original size, show trades, industries of fascinating
era in such great detail that modern reconstructions might be made of them
Plates teem
with men, women, children performing thousands of operations; show
sequence general
operations, closeups, details of machinery. Illustrates such important, interesting
trades
industries as sowing, harvesting, beekeeping, tobacco processing, fishing,
arts
of
war'
mining, smelting
casting iron, extracting mercury,
making gunpowder, cannons, bells'
shoeing horses, tanning, papermaking, printing, dying, over 45 more categories.
Professor
Gillispie of Princeton supplies full commentary on all plates, identifies
operations
tools
processes, etc. Material is presented in lively, lucid fashion. Of great
interest' to all
studying history of science, technology. Heavy library cloth. 920pp. 9 x 12.
T421 2 volume set $18.50
DE MAGNETE, William Gilbert. Classic work on magnetism, founded new science
Gilbert
was first to use word "electricity," to recognize mass as distinct from weight, to
discover
effect of heat on magnetic bodies; invented an electroscope, differentiated
between static
electricity and magnetism, conceived of earth as magnet. This lively work
by first great
experimental scientist, is not only a valuable historical landmark, but a delightfully
easy
to follow record of a searching, ingenious mind. Translated by P. F.
Mottelay
25 page
biographical memoir. 90 figures, lix
368pp. 53/8 x 8.
S470 Paperbound
HISTORY OF MATHEMATICS,
$2.00
Most comprehensive, non-technical history of math

in English. Discusses lives and works of over a thousand major,
minor figures, with footnotes giving technical information outside book's scheme, and indicating disputed
matters
Vol.
I:
A chronological examination, from primitive concepts through Egypt
Babylonia
Greece, the Orient, Rome, the Middle Ages, The Renaissance, and to 1900
Vol 'II- The
development of ideas in specific fields and problems, up through elementary calculus
Marks an epoch
will modify the entire teaching of the history of science,"
George
Sarton. 2 volumes, total of 510 illustrations, 1355pp. 53/8 x 8. Set boxed
in
attractive
container.
T429, 430 Paperbound, the set $5.00
.
D.
E.
Smith.
THE PHILOSOPHY OF SPACE AND TIME, H. Reichenbach. An important landmark in development of empiricist conception of geometry, covering foundations of geometry, time theory
relativity, including: relations between theory and observationscoordinate definitions; relations between topological and metrical
properties of space
psychological problem of visual intuition of non-Euclidean structures; many more
topics
important to modern science and philosophy. Majority of ideas require only knowledge
of
consequences of Einstein's
intermediate math. "Still the best book in the

Carnap. 49 figures, xviii + 296pp. 53/8 x 8.
R.
field,"
Rudolf
Carnap.
Introduction
by
CATALOGUE OF
N. Campbell.
particularly physics. Examines why
critique of the most fundamental concepts of science,
philosophy,
from
science
demarcates
question,
certain propositions are accepted without
analyzes presuppositions of scientific thought: existence of material world,
Part
etc
of
applications
and
experiment
of
nature
covers
nature of laws, probability, etc; part 2
laws and theories,
mathematics: conditions for measurement, relations between numerical
space
motion,
force,
relativity
An appendix covers problems arising from
etc.
error
is," Higher Educational Journal
time. A classic in its field. "A real grasp of what science
55/8 x 83/s.
565pp.
xiii
FOUNDATIONS OF SCIENCE: THE PHILOSOPHY OF THEORY AND EXPERIMENT,

A
OF SCIENCE,
THE STUDY OF THE HISTORY OF MATHEMATICS and THE STUDY OF THE HISTORY
or mature worker. Describes

G Sarton. Excellent introductions, orientation, for beginning
of previous generations. Exduty of mathematical historian, incessant efforts and genius
methods.
200 item bibliography with
plains how today's discipline differs from previous
best treatises
mathematicians,
modern
of
biographies
available
best
evaluations,
critical
volumes bound as one
on historical methods is especially valuable. 10 illustrations. 2
Paperbound $1.25
T240
8.
53/8
x
75pp.
113pp.
MATHEMATICAL PUZZLES
selected and edited by Martin Gardner. 117 choice
from his famous "Cyclopedia
puzzles by greatest American puzzle creator and innovator,
" All unique style, historical flavor of originals. Based on arithmetic, algebra,
of Puzzles
sliding block, operations research, geotopology,
tracing,
route
theory,
game
probability,'
Horse of
"14-15" puzzle which was national craze,
metrical dissection. Includes famous
120 line drawings diagrams Solutions
a Different Color" which sold millions of copies.
Paperbound
$1.00
T498
167pp. 53/8 x 8.
xx
MATHEMATICAL PUZZLES OF SAM LOYD,
Carroll. "Symbolic Logic" is not concerned

but is instead a collection of over 380 problems posed with
fascinating diagrammatic method of
and
a
charm and imagination, using the syllogism,
imagination devises a
drawing conclusions. In "The Game of Logic" Carroll's whimsical
to manipulate hundreds of
(included)
counters
logical game played with 2 diagrams and
is a lagniappe of 101 additional puzzles
Miss"
or
"Hit
section,
final
The
syllogisms.
tricky
this reprint edition, both of these books were rarities
in the delightful Carroll manner. Until
Game of Logic: 96pp.
costing upSo $15 each. Symbolic Logic: Index, xxxi + 199pp.
T49Z Paperoouna i.ou
2 vols, bound as one. 53/8 x 8.
SYMBOLIC LOGIC and THE GAME OF LOGIC, Lewis

with
modern symbolic
logic,
ê
rarest of all Carroll's

PILLOW PROBLEMS and A TANGLED TALE, Lewis Carroll. One of the
"Pillow Problems" contains 72 original math puzzles, all ^P"* W**:***thought them out,
he
remain exactly as
Carroll's answers which
larly fascinating are
in "A Tangled Ta e" are in story form
reflecting his actual mental process. The problems
works
gives the solutions, but

a monthly magazine serial. Carroll not only
approaches and misleading paths, and
uses answers sent in by readers to discuss wrong
Pillow
rarities until this edition
erades them for insight. Both of these books were
Problems: Preface and
Problems" costing up to $25, and "A Tangled Tale" $15. Pillow
illustrations. I52pp -Two vols,
mtroduction by Levis' Carroll, xx + 109pp. A Tangled Tale: 6
T493 Paperbouna $i.au
bound as one. 53/8 x 8.
original"? appearing as
comPUZZLES, G. L. Kaufman. 100 brand new challenging puzzles on words,

beginners
never before published. Most are new types invented by author for
symmetrical
words;
build
to
and experts both. Squares of letters follow chess moves
syllable puzzles where
designs made of synonyms; rhymed crostics; double word squares;
other types, all new. Solutions.
you fill in missing syllables instead of missing letter; many
X 8
5
vi
figures,
122
PP
puzzles.
196
+
"Excellent," Recreation. 100
44 paperbound $1
NEW WORD
binations,
%3
probMATHEMATICAL EXCURSIONS, H. A. Merrill. Fun, recreation, insights into elementary math

travelled in elementary
lem solving. Math expert guides you on by-paths not generally
systems for pi; odd,
courses divide by inspection, Russian peasant multiplication; memory
Tchebichev s machine;
even magic squares; dyadic systems; square roots by geometry;
G
a c as s
, ?
dozens more. Solutions to more difficult ones. "Brain stirring stuff
h 'f$1.00
T350 Paperbound
50 illustrations. 145pp. 53/8 x 8.
'
&
new mateall
THE BOOK OF MODERN PUZZLES, G. L. Kaufman. Over 150 puzzles, absolutely
puzzles, but with different principles,
rial based on same appeal as crosswords, deduction
observation puzzles,
techniques. 2-minute teasers, word labyrinths, design, pattern, logic,
many'others.
situations
puzzles testing ability to apply general knowledge to peculiar
Paperbound $1.00
T143
Solutions. 116 illustrations. 192pp. 53/8 x 8.
R. V. Heath. Over 60 puzzles,

mentally
numbers. Easy techniques for. multiplying large numbers
any
in
V^- 'ncludes Th e Lost Digit,
identifying unknown numbers, finding date of any day
others not easily found else
cubes,
trjangles,
squares,
magic
Bridge,
Psychic
3 Acrobats
53/8 * 8.
whe?e Edited by J. S Mly4r. 76 illustrations. 128pp.
MATHEMAGIC
on
MAGIC PUZZLES, AND GAMES WITH NUMBERS,
parties
of
DOVER SCIENCE BOOKS

QUIZ AND STUNT FUN, J. Meyer. 238 high-priority puzzles, stunts, tricks math
Please Help Alice; mysteries, deductions
like The Clever Carpenter, Atom Bomb,
The Bridge of Sighs, Secret Code; observation puzzlers like The American Flag, Playing
squares, tongue twisters, puns, anawith
magic
Cards, Telephone Dial; over 200 others
grams. Solutions. Revised, enlarged edition of "Fun-To-Do." Over 100 illustrations. 238
puzzles, stunts, tricks. 256pp. 53/a x 8.
PUZZLE
puzzles
like
101 PUZZLES IN THOUGHT AND LOGIC, C. R. Wylie, Jr. For readers who enjoy challenge,
stimulation of logical puzzles without specialized math or scientific knowledge. Problems
entirely new, range from relatively easy to brainteasers for hours of subtle entertainment.
Detective puzzles, find the lying fisherman, how a blind man identifies color by logic, many
more. Easy-to-understand introduction to logic of puzzle solving and general scientific
method. 128pp. 53/8 x 8.
CRYPTANALYSIS, H. F. Gaines. Standard elementary, intermediate text for serious students.

Not just old material, but much not generally known, except to experts. Concealment,
Transposition, Substitution ciphers; Vigenere, Kasiski, Playfair, multafid, dozens of other
techniques. Formerly "Elementary Cryptanalysis." Appendix with sequence charts, letter
frequencies in English, 5 other languages, English word frequencies. Bibliography. 167
x 83/8
codes. New to this edition: solutions to codes, vi -f 230pp.
5%
CRYPTOGRAPY, L. D. Smith. Excellent elementary introduction to enciphering, deciphering

secret writing. Explains transposition, substitution ciphers; codes; solutions; geometrical
patterns, route transcription, columnar transposition, other methods. Mixed cipher systems;
single, polyalphabetical substitutions; mechanical devices; Vigenere; etc. Enciphering Japanese; explanation of Baconian biliteral cipher; frequency tables. Over 150 problems. BibT247 Paperbound $1.00
liography. Index. 164pp. 53/s x 8.
MATHEMATICS, MAGIC AND MYSTERY, M. Gardner. Card tricks, metal mathematics, stage
mind-reading, other "magic" explained as applications of probability, sets, number theory,
etc. Creative examination of laws, applications. Scores of new tricks, insights. 115 sections
math guarantees
on cards, dice, coins; vanishing tricks, many others. No sleight of hand
easy to follow," Mathematics Teacher.
success. "Could hardly get more entertainment
174pp. 53/8 x 8.
115 illustrations, xii
IN MATHEMATICS, H. E. Dudeney. Foremost British originator of math puzzles,

intriguing, paradoxical in this classic. One of largest collections. More than
problems on number manipulations,
Mazes, games,
problems, paradoxes.
puzzles,
unicursal, other route problems, puzzles on measuring, weighing, packing, age, kinship,
chessboards, joiners', crossing river, plane figure dissection, many others. Solutions. More
258pp. 53/8 x 8.
than 450 illustrations, viii
AMUSEMENTS
always witty,
430
H. E. Dudeney. Chaucer's pilgrims set one another problems in

Also Adventures of the Puzzle Club, the Strange Escape of the King's Jester,
Riddlewell, the Squire's Christmas Puzzle Party, others. All puzzles are
original, based on dissecting plane figures, arithmetic, algebra, elementary calculus, other
branches of mathematics, and purely logical ingenuity. "The limit of ingenuity and in225 pp.
tricacy," The Observer. Over 110 puzzles, full solutions. 150 illustrations, viii
53/8 x 8.
THE CANTERBURY PUZZLES

story form.
the Monks
of
MATHEMATICAL PUZZLES FOR BEGINNERS AND ENTHUSIASTS,
G. Mott-Smith. 188 puzzles to

interpretation, algebra, dissection of plane figures, geometry,
decimation, permutations, probability, all are in these delightful
problems. Includes the Odic Force, How to Draw an Ellipse, Spider's Cousin, more than 180
others. Detailed solutions. Appendix with square roots, triangular numbers, primes, etc.
135 illustrations. 2nd revised edition. 248pp. 53/8 x 8.
test mental
properties
agility.
of
Inference,
numbers,
MATHEMATICAL RECREATIONS, M.
Kraitchik.
Some 250
puzzles,
problems,
demonstrations of
mathematics on relatively advanced level. Unusual historical problems from

recreation
Greek, Medieval, Arabic, Hindu sources; modern problems on "mathematics without numbers," geometry, topology, arithmetic, etc. Pastimes derived from figurative, Mersenne,
Fermat numbers: fairy chess; latruncles: reversi; etc. Full solutions. Excellent insights
"Strongly recommended to all who are interested in the
fields of math.
into special
lighter side of mathematics," Mathematical Gaz. 181 illustrations. 330pp. 53/8 x 8.
FICTION
FLATLAND, E. A. Abbott. A perennially popular science-fiction classic about life in a 2dimensional world, and the impingement of higher dimensions. Political, satiric, humorous,
moral overtones. This land where women are straight lines and the lowest and most dangerous classes are isosceles triangles with 3 vertices conveys brilliantly a feeling for
many concepts of modern science. 7th edition. New introduction by Banesh Hoffmann. 128pp.
Tl Paperbound $1.00
53/8 x 8.
CATALOGUE OF
SEVEN SCIENCE FICTION NOVELS OF H. G. WELLS. Complete texts, unabridged, of seven of
Wells' greatest novels: The War of the Worlds, The Invisible Man, The Island of Dr. Moreau,
The Food of the Gods, First Men in the Moon, In the Days of the Comet, The Time Machine.
Still considered by many experts to be the best science-fiction ever written, they will offer
amusements and instruction to the scientific minded reader. "The great master," Sky and
Telescope.
1051pp. 53/8 x
T264 Clothbound $3.95
8.
WELLS. Unabridged! This enormous omnibus contains

Men Like Gods, Star Begotten plus 26 short stories of space, time,
2 full length novels
invention, biology, etc. The Crystal Egg, The Country of the Blind, Empire of the Ants,
The Man Who Could Work Miracles, Aepyornis Island, A Story of the Days to Come, and
writers of today," The English Journal.
20 others "A master ... not surpassed by
915pp. 53/s x 8.
28 SCIENCE FICTION STORIES OF H.
G.
FIVE ADVENTURE NOVELS OF H. RIDER HAGGARD. All the mystery and adventure of darkest
captured accurately by a man who lived among Zulus for years, who knew African
ethnology, folkways as did few of his contemporaries. They have been regarded as examples
of the very best high adventure by such critics as Orwell, Andrew Lang, Kipling. Contents:
She, King Solomon's Mines, Allan Quatermain, Allan's Wife, Maiwa's Revenge. "Could spin
a yarn so full of suspense and color that you couldn't put the story down," Sat. Review.
821pp. 53/s x 8.
Africa
CHESS AND CHECKERS

LEARN CHESS FROM THE MASTERS, Fred Reinfeld. Easiest, most instructive way to implay 10 games against such masters as Marshall, Znosko-Borovsky, Bronprove your game
stein, Najdorf, etc., with each move graded by easy system. Includes ratings for alternate
moves possible. Games selected for interest, clarity, easily isolated principles. Covers
Ruy Lopez, Dutch Defense, Vienna Game openings; subtle, intricate middle game variations;
all-important
144pp.
end
53/s
game.
Full
annotations.
Formerly
x 8.
"Chess by Yourself." 91 diagrams, viii

REINFELD ON THE END GAME IN CHESS, Fred Reinfeld. Analyzes 62 end games by Alekhine,
Rubinstein, Lasker, Reshevsky, other masters. Only
Flohr, Tarrasch, Morphy, Capablanca,
tell exactly what is wrong with each move
1st rate book with extensive coverage of error
you might have made. Centers around transitions from middle play to end play. King and
pawn, minor pieces, queen endings; blockage, weak, passed pawns, etc. "Excellent ... a
boon," Chess Life. Formerly "Practical End Play." 62 figures, vi + 177pp. 53/8 x 8.
HYPERMODERN CHESS as developed in the games of its greatest exponent, ARON NIMZOVICH, edited by Fred Reinfeld. An intensely original player, analyst, Nimzovich's approaches
startled, often angered the chess world. This volume, designed for the average player,
shows how his iconoclastic methods won him victories over Alekhine, Lasker, Marshall,
Rubinstein, Spielmann, others, and infused new life into the game. Use his methods to
startle opponents, invigorate play. "Annotations and introductions to each game ... are
excellent," Times (London). 180 diagrams, viii + 220pp. 5 3/8 x 8.
Lively reader, by one of America's finest chess
masters, including: history of chess, from ancient Indian 4-handed game of Chaturanga
great players of today; such delights and oddities as Maelzel's chess-playing automaton
that beat Napoleon 3 times; etc. One of most valuable features is author's personal recollecNimzovich, Emanuel Lasker, Capablanca, Alekhine,
tions of men he has played against
etc. Discussion of chess-playing machines (newly revised). 5 page chess primer. 11 illus3
x
8.
diagrams.
5
53
296pp.
trations.
/8
THE ADVENTURE OF CHESS, Edward Lasker.

to
THE ART OF CHESS, James Mason. Unabridged reprinting of latest revised edition of most
famous general study ever written. Mason, early 20th century master, teaches beginning,
intermediate player over 90 openings; middle game, end game, to see more moves ahead,
plan purposefully, attack, sacrifice, defend, exchange, govern general strategy. "Classic
... one of the clearest and best developed studies," Publishers Weekly. Also included, a
complete supplement by F. Reinfeld, "How Do You Play Chess?", invaluable to beginners
question-and-answer method. 448 diagrams. 1947 Reinfeld-Bernstein text.
lively
for
its
Bibliography, xvi + 340pp. 53/8 x 8.
to
MORPHY'S GAMES OF CHESS, edited by P. W. Sergeant. Put boldness into your game by
flowing brilliant, forceful moves of the greatest chess player of all time. 300 of Morphy's
best games, carefully annotated to reveal principles. 54 classics against masters like
Anderssen, Harrwitz, Bird, Paulsen, and others. 52 games at odds; 54 blindfold games; plus
over 100 others. Follow his interpretation of Dutch Defense, Evans Gambit, Giuoco Piano,
Ruy Lopez, many more. Unabridged reissue of latest revised edition. New introduction by
x 8.
352pp.
Reinfeld. Annotations, introduction by Sergeant. 235 diagrams, x
F.
5%
DOVER SCIENCE BOOKS

Former World's Unrestricted Checker
game, expert's shots, traps, problems for beginner, standmoves to draw when
ard openings, locating best move, end game, opening "blitzkrieg"
Appendix. 75 probbehind etc Over 100 detailed questions, answers anticipate problems.
Paperbound $1.00
T363
53/
8 x 8.
lems with solutions, diagrams. 79 figures, xi + 107pp.
WIN AT CHECKERS, M. Hopper.

Champion discusses principles
(Formerly "Checkers.")
of
Fred Reinfeld. If you have trouble finishing off your opponent,

strokes and combinations from actual tournament play.
Starts with 1-move checkmates, works up to 3-move mates. Develops ability to lock ahead,
gain new insights into combinations, complex or deceptive positions; ways to estimate weaknesses, strengths of you and your opponent. "A good deal of amusement and instruction,"
Times, (London). 300 diagrams. Solutions to all positions. Formerly "Challenge to Chess
Players." 111pp. 53/8 x 8.
HOW TO FORCE CHECKMATE,

here
is
collection
of
lightning
Delightful collection of anecdotes,

short stories, aphorisms by, about masters; poems, accounts of games, tournaments, photoepisodes, comments, word
historical
graphs; hundreds of humorous, pithy, satirical, wise,
portraits. Fascinating "must" for chess players; revealing and perhaps seductive to those
who wonder what their friends see in game. 49 photographs (14 full page plates). 12
306pp. 53/8 x 8.
diagrams, xi
A TREASURY OF CHESS LORE, edited by Fred Reinfeld.
Fred Reinfeld. 300 practical chess situations, to sharpen your eye, test skill
against masters. Start with simple examples, progress at own pace to complexities. This
selected series of crucial moments in chess will stimulate imagination, develop stronger,
more versatile game. Simple grading system enables you to judge progress. "Extensive use
of diagrams is a great attraction," Chess. 300 diagrams. Notes, solutions to every situation.
120pp. 53/8 x 8.
Formerly "Chess Quiz." vi
WIN AT CHESS,
MATHEMATICS:
ELEMENTARY TO INTERMEDIATE
QUICKLY, H. Sticker. Tried and true method to help mathematics of
Awakens "number sense" ability to see relationships between numbers as
serious course of over 9000 problems and their sblutions through
A
whole
techniques not taught in schools: left-to-right multiplications, new fast division, etc. 10
HOW TO CALCULATE
everyday
life.
quantities.
minutes a day will double or triple calculation speed. Excellent for scientist at home in
higher math, but dissatisfied with speed and accuracy in lower math. 256pp. 5 x 7Va.
Paperbound $1.00
PROBLEMS OF ELEMENTARY GEOMETRY, Felix Klein. Expanded version of 1894

Easter lectures at Gbttingen. 3 problems of classical geometry: squaring the circle, trisecting angle, doubling cube, considered with full modern implications: transcendental numbers, pi, etc. "A modern classic ... no knowledge of higher mathematics is required,"
92pp. 53/8 x 8.
Scientia. Notes by R. Archibald. 16 figures, xi
FAMOUS
HIGHER MATHEMATICS FOR STUDENTS OF CHEMISTRY AND PHYSICS,
Mellor. Practical,
J. W.
not abstract, building problems out of familiar laboratory material. Covers differential calculus, coordinate, analytical geometry, functions, integral calculus, infinite series, numerical
equations, differential equations, Fourier's theorem probability, theory of errors, calculus
of variations, determinants. "If the reader is not familiar with this book, it will repay
him to examine
it,"
Chem. and Engineering News. 800 problems. 189
figures,
xxi
641pp.
53/8 x 8.
A. A. Klaf. 913 detailed questions, answers

particularly useful in clearing
cover most important aspects of plane, spherical trigonometry
quadrants, functions, graphical repreangles,
plane
trig,
areas.
Part
I:
special
in
difficulties
up
sentation, interpolation, equations, logs, solution of triangle, use of slide rule, etc. Next
188 pages discuss applications to navigation, surveying, elasticity, architecture, other
problems.
special fields. Part 3: spherical trig, applications to terrestrial, astronomical
Methods of time-saving, simplification of principal angles, make book most useful. 913
questions answered. 1738 problems, answers to odd numbers. 494 figures. 24 pages of forT371 Paperbound $2.00
629pp. 53/8 x 8.
mulas, functions, x
TRIGONOMETRY REFRESHER FOR TECHNICAL MEN,
A. A. Klaf. 756 questions examine most imcalculus. Part I: simple differential calculus, constants, variables, functions, increments, logs, curves, etc. Part 2: fundamental ideas of
integrations, inspection, substitution, areas, volumes, mean value, double, triple integration,
etc. Practical aspects stressed. 50 pages illustrate applications to specific problems of civil,
engineering, electricity, stress, strain, elasticity, similar fields. 756 questions
nautical
431pp.
answered. 566 problems, mostly answered. 36pp. of useful constants, formulas, v
CALCULUS REFRESHER FOR TECHNICAL MEN,

portant aspects
of
integral,
differential
53/8 x 8.
CATALOGUE OF
MONOGRAPHS ON TOPICS OF MODERN MATHEMATICS,
edited by J. W. A. Young. Advanced

mathematics for persons who have forgotten,
or not gone beyond, high school algebra.
monographs on foundation of geometry, modern pure geometry, non-Euclidean geometry,
fundamental propositions of algebra, algebraic equations, functions, calculus, theory of
numbers, etc. Each monograph gives proofs of important results, and descriptions of leading methods, to provide wide coverage. "Of high merit," Scientific American. New introduction by Prof. M. Kline, N.Y. Univ. 100 diagrams, xvi + 416pp. 6Vs x 9V4.
9
MATHEMATICS IN ACTION, 0. G. Sutton. Excellent middle level application of mathematics

study of universe, demonstrates how math is applied to ballistics, theory of computing
machines, waves, wave-like phenomena, theory of fluid flow, meteorological
problems,
statistics, flight, similar phenomena. No knowledge of advanced math required. Differential
equations, Fourier series, group concepts, Eigenfunctions, Planck's constant, airfoil theory,
and similar topics explained so clearly in everyday language that almost anyone can derive
benefit from reading this even if much of high-school math is forgotten. 2nd edition. 88
figures, viii
236pp. 53/s x 8.
to
MATHEMATICS
ELEMENTARY
FROM
AN
ADVANCED
STANDPOINT,
Felix
Klein.
Classic
text,
an outgrowth of Klein's famous integration and survey course at Gbttingen. Using one field
to interpret, adjust another, it covers basic topics in each area, with extensive analysis.
Especially valuable in areas of modern mathematics. "A great mathematician, inspiring
teacher,
deep insight,"
Bui.,
Amer. Math Soc.
Vol. I. ARITHMETIC, ALGEBRA, ANALYSIS. Introduces concept of function immediately, enlivens discussion with graphical, geometric methods.
Partial
contents: natural numbers,
special properties, complex numbers. Real equations with real unknowns, complex quantities. Logarithmic, exponential functions, infinitesimal calculus. Transcendence of e and pi,
theory of assemblages. Index. 125 figures, ix
x 8.
274pp.
Vol.
etry
II.
5%
GEOMETRY. Comprehensive view, accompanies space perception inherent
in
geom-
analytic formulas which facilitate precise formulation. Partial contents: Simplest

manifold; line segments, Grassman determinant principles, classication of configurations of space. Geometric transformations: affine, projective, higher point transformations, theory of the imaginary. Systematic discussion of geometry and its foundations. 141
x 8.
illustrations, ix
214pp.
with
geometric
5%
A TREATISE ON PLANE AND ADVANCED TRIGONOMETRY, E. W. Hobson. Extraordinarily wide

coverage, going beyond usual college level, one of few works covering advanced trig in
By a great expositor with unerring anticipation of potentially difficult points.
circular functions; expansion of functions of multiple angle; trig tables; relations
between sides, angles of triangles; complex numbers; etc. Many problems fully solved.
"The best work on the subject," Nature. Formerly entitled "A Treatise on Plane Trigonomx 8.
etry." 689 examples. 66 figures, xvi
383pp.
full
detail.
Includes
5%
NON-EUCLIDEAN GEOMETRY, Roberto Bonola. The standard coverage
of non-Euclidean geomExamines from both a historical and mathematical point of view geometries which
have arisen from a study of Euclid's 5th postulate on parallel lines. Also included are
complete texts, translated, of Bolyai's "Theory of Absolute Space," Lobachevsky's "Theory
of Parallels." 180 diagrams. 431pp. 5% x 8.
etry.
GEOMETRY OF FOUR DIMENSIONS, H. P. Manning. Unique in English as a clear, concise introduction. Treatment is synthetic, mostly Euclidean, though in hyperplanes and hyperspheres
at infinity, non-Euclidean geometry is used. Historical introduction. Foundations of 4-dimensional geometry. Perpendicularity, simple angles. Angles of planes, higher order. Symmetry,
hyperpyramids, hypercones, hyperspheres; figures with parallel elements;
order, motion;
volume, hypervolume in space; regular polyhedroids. Glossary. 78 figures, ix
348pp.
5%
x 8.
MATHEMATICS: INTERMEDIATE TO ADVANCED

GEOMETRY (EUCLIDEAN AND NON-EUCLIDEAN)
THE GEOMETRY OF RENE DESCARTES. With this book, Descartes founded analytical geometry.
French text, with Descartes's own diagrams, and excellent Smith-Latham translation. Contains: Problems the Construction of Which Requires only Straight Lines and Circles;
On the Nature of Curved Lines; On the Construction of Solid or Supersolid Problems. DiaS68 Paperbound $1.50
grams. 258pp. 5% x 8.
Original
10
DOVER SCIENCE BOOKS

T. L. Heath. All the known works of the great Greek
Archimedes. Contains: On
of
mathematician, including the recently discovered Method
Definitive
Circle, Spirals, Conoids, Spheroids, etc
of
a
Measurement
Cylinder,
SDhere and
of ancient world. 186 page study by Heath disedition of greatest mathematical intellect
S9 Paperbound $2.00
53/8 x 8.
563pp.
mathematics.
Greek
of
history
and
cusses Archimedes
the WORKS OF ARCHIMEDES, edited by
first to contain complete

monographs, 3 complete
enormous importance in
of
been
have
which
papers
miscellaneous
lecture courses, 15
other areas of mathematics. Edited by
relativity, topology, theory of complex variables,
English introduction by
Dedekind, H. Weber, M. Noether, W. Wirtinger. German text;
R
Hans Lewy. 690pp. 53/8 x 8.
COLLECTED WORKS OF BERNARD RIEMANN. Important sourcebook,

unabridged. 31
text of 1892 "Werke" and the 1902 supplement,
Sir Thomas Heath. Definitive edition

world. Complete translation of Heiberg text,
of one of very greatest classics of Western
Euclid,
Greek, Medieval mathematics
on
introduction
page
150
XIV.
Book
spurious
Dlus
each definicommentators, etc. Elaborate critical apparatus parallels text, analyzing
texts
refutations, supports, extrapolations
matters,
textual
covering
proposition,
postulate,
tion,
of Cambridge U. 2nd edition. 3 voletc This is the full Euclid. Unabridged reproduction
S88, 89, 90, 3 volume set, paperbound $6.00
umes. 995 figures. 1426pp. 53/8 x 8.
THE THIRTEEN BOOKS OF EUCLID'S ELEMENTS, edited by
'
D. M. Y. Sommerville. Presupposes no
to higher dimensional
previous knowledge of field. Only book in English devoted exclusively
angles
parallelism, perpendicularity
incidence,
of
ideas
fundamental
Discusses
geometry
and metric
between linear space, enumerative geometry, analytical geometry from projective
figures
60
hyperspacial
of
content
situs,
analysis
ideas
in
elementary
views, polytopes,
diagrams. 196pp. 53/8 x 8.
AN INTRODUCTION TO GEOMETRY OF N DIMENSIONS,
D. M. Y. Sommerville. Unique in proceeding steponly good knowledge of high-school geometry and algebra, to grasp elecurvature and its
mentary hyperbolic, elliptic, analytic non-Euclidean Geometries; space
parataxy and parallelimplications; radical axes; homopethic centres and systems of circles;
exceptional clarity. 126 probism- Gauss' proof of defect area theorem; much more, with
274pp. 53/8 x 8.
lems at chapter ends. 133 figures, xvi
ELEMENTS OF NON-EUCLIDEAN GEOMETRY,

by-step
Requires
THE FOUNDATIONS OF EUCLIDEAN GEOMETRY, H. G. Forder. First connected, rigorous acrecourse to empiricism,
count in light of modern analysis, establishing propositions without
hypotheses. Based on tools of 19th and 20th century mathematicians,

possible to remedy gaps and complexities, recognize problems not earlier
Begins with important relationship of number systems in geometrical figures.
discerned
for magnitudes, groups,
Considers' classes, relations, linear order, natural numbers, axioms
(at length), particular axioms
ouasi-fields fields, non-Archimedian systems, the axiom system
Lists:
similarity, etc.
congruence,
constructions,
Axioms),
Parallel
the
on
chapters
(two
x
axioms employed, constructions, symbols in frequent use. 295pp. 53/8
without multiplying
who made
it
CALCULUS, FUNCTION THEORY (REAL AND COMPLEX),
FOURIER THEORY
Provides complete, readily
FIVE VOLUME "THEORY OF FUNCTIONS" SET BY KONRAD KNOPP.
yet without sacrifice of
followed account of theory of functions. Proofs given concisely,
M.I T., Chicago,
completeness or rigor. These volumes used as texts by such universities as
readable, concise,
remarkably
introduction
"Excellent
NY City College, many others.
.
clear,
rigorous,"
J.
of
the
American
Statistical
Association.
further
ELEMENTS OF THE THEORY OF FUNCTIONS, Konrad Knopp. Provides background for of comFoundations, system
volumes in this set, or texts on similar level. Partial contents:
mapping by
sphere of numbers
plex numbers and Gaussian plane of numbers, Riemann
linear functions, normal forms, the logarithm, cyclometric
only for the young student, but also for the student who
Mathematical Journal. 140pp. 53/8 x 8.
functions,
knows
all
binomial
Not
series.
about what is in it.

Konrad Knopp. With volume II, provides coverage of basic

contents: numbers and points, functions of a complex
Cauchy s integral
integral of a continuous function, Cauchy's intergral theorem,
variable
expansion and analytic function in a power series,
terms,
variable
with
formulae, series
Laurent expansion types
analytic continuation and complete definition of analytic '-nctions,
vii
+ 146pp. 53/8 x 8.
THEORY OF FUNCTIONS, PART

concepts
of
and
theorems.
I,
Partial
singularities,
Konrad Knopp. Application and further development of

II,
Weierstrass. Meromorphic
general theory, special topics. Single valued functions, entire,
Multiple valued functions. Riemann surfaces.
functions.
Periodic
functions: Mittag-Leffler.
Ôpp. 3/8 * 8
Algebraic functions. Analytical configurations, Riemann surface, x
|r ou d $1 35
THEORY OF FUNCTIONS, PART
11
CATALOGUE OF
PROBLEM BOOK
THE THEORY OF FUNCTIONS, VOLUME
IN
theory, for
mentary
Cr
com Dlex V a
C
?es
Sirs *T+*
PROBLEM BOOK
Konrad Knopp. Problems in eleuse with Knopp's "Theory of Functions," or any other
text
Arranged
d iff,;C * y
Fu n damental c nC pts se ^nces of numbers' and
iSte
h remS
a b,e
m ?te gral tth6
deV6l? Pment
SX.p
r
l26;p
'
53/s
^Ti^S&JSF^
'
THE THEORY OF FUNCTIONS, VOLUME
IN
I,
Konrad Knopp. Advanced theorv

of functions, to be used with Knopp's "Theory of
Functions," or compaVabl
text Singular
t,r
and mer morphic f unctions
Periodic,
analytic,
continuation
multiple -willed
fSS{.II psurfac es
conf, rma
capping. Includes section of elementary problems
\.
ThP riiffirni'fTI!
CU t task f S ele ct
P roblems Just within the reach of the ueKm,,er
beginner ls
is
hI ro masterfully
ml ! .. accomplished,"
u r,? .;
here
AM. MATH. SOC. Answers. 138pp. 53/8 x 8.
II,
'
ADVANCED CALCULUS,
E
B Wilson. Still recognized as one of most
comprehensive useful
amount of well-represented, fundamental material, including
chapters on
functions, ordinary differential equations, special functions
calculus of variations
etc., which are excellent introductions to these
areas. Requires only one yea
of calculut
Over 1300 exercises cover both pure math and applications to
engineering and phvscai
problems. Ideal reference, refresher. 54 page introductory review,
ix
566pp
x 8
Immense
texts.
vector
5%
S504 Paperbound
LECTURES ON THE THEORY OF ELLIPTIC FUNCTIONS,
H.
$2.45
Hancock.
Reissue of only book in

English with so extensive a coverage, especially of Abel,
Jacobi, Legendre
Weierstrass
Hermite, Liouville, and Riemann. Unusual fullness of treatment, plus
applications as well as
n
dlscuss n g universe of elliptic integrals, originating in
y
works of Abel and
\ a ?nl
c
e
e
,n
r0 " i6e m St general the0ry
40 " page table of formulas.
76 fiiurSJ
Jill + AQnn
figures, xxm
76
53/8 x 8.
498pp. ?g
'
,'.
RY
F
,
,
Volterra.
"
UN CT 0NALS AND 0F INTEGRAL AND

k ,
L
u,
INTEGRO-DIFFERENTIAL
EQUATIONS
Vito
Unabridged republication of only English translation, General theory

of functions
depending on continuous set of values of another function. Based
on author's concept of
transition from finite number of variables to a continually
infinite number
Includes much
material on calculus of variations. Begins with fundamentals,
examines generalization of
analytic
functions
in
0d C
taker xxxx
taker,
functional
derivative
n
C
?
? fi nn 53/8 !"
x 8.
226pp.
EV3nS
"
equations,
Biogra P h y-
applications, other directions if theory

jticism of Volterra's work by E
Whit
-
C|
RO DUCTI N T F0URIER METH0DS AND THE LAPLACE TRANSFORMATION,

Philip
!lJ.' !y r
? +
Franklin.
Concentrates
on essentials, gives broad view, suitable for most
applications
Requires only knowledge of calculus. Covers complex qualities with
methods of computing elementary functions for complex values of argument and finding
approximations by chartsFourier series; harmonic anaylsis; much more. Methods are
related to physical problems
of heat flow, vibrations, electrical transmission, electromagnetic
radiation, etc
828 emblems, answers. Formerly entitled "Fourier Methods." x
289pp. 53/8 x 8.
A
THE ANALYTICAL THEORY OF HEAT, Joseph
Fourier.
This
which revolutionized mathematical physics, has been used by generations of mathematicians

and physicists interested
he at 0l a PPl'catipn of Fourier integral. Covers cause and reflection
'"
\.
of rays of heat
u
radiant heating
Fourier integral,
book,
heating of closed spaces, use of trigonometric series

Translated by Alexander Freeman. 20 figures, xxii
in
theory
of
heat'
+ 466pp 53/8 x 8
etc.
ELLIPTIC INTEGRALS, H. Hancock. Invaluable in work involving differential

equations with
cubics, quatnes under root sign, where elementary calculus methods
are inadequate
Practical solutions to problems in mathematics, engineering,
physics; differential equations requiring integration of Lame's, Briot's, or Bouquet's equations;
determination of arc of
ellipse, hyperbola, lemiscate; solutions of problems in elastics;
motion of a projectile under
resistance varying as the cube of the velocity; pendulums; more.
Exposition in accordance
with Legendre-Jacobi theory. Rigorous discussion of Legendre
transformations. 20 figures
5 place table. 104pp. 53/8 x 8.
S484 Paperbound
$1.25
Y L0R S ERIES
A N INTRODUCTION TO THE THEORY OF FUNCTIONS OF A COMPLEX
n
n
P. Dienes. Uses Taylor series to approach theory of
functions, using ordinary
calculus only, except in last 2 chapters. Starts with introduction to
real variable and complex algebra, derives properties of infinite series, complex differentiation,
integration, etc
Lovers b:umform mapping, overconvergence and gap theorems, Taylor series
on its circle
of convergence
etc. Unabridged
corrected reissue of first edition. 186 examples, many
fully worked out. 67 f.gures. xii
_+ 555pp. 53/8 x 8.
'
!"/D,m
r
VARIABLE,
LINEAR
INTEGRAL EQUATIONS, W. V. Lovitt. Systematic surwey of general theory, with

some
application to differential equations, calculus of variations, problems of
math, physics
Includes: integral equation of 2nd kind by successive substitutions;
Fredholm's equation
as ratio of 2 integral series in lambda, applications of the Fredholm
theory, Hilbert-Schmidt
theory of symmetric kernels, application, etc. Neumann, Dirichlet, vibratory
problems
ix
253pp.
53/s
8.

12
DOVER SCIENCE BOOKS

DICTIONARY OF CONFORMAL REPRESENTATIONS,
H. Kober. Developed by British Admiralty to

Laplace's equation in 2 dimensions. Scores of geometrical forms and transformations
engineers, Joukowski aerofoil for aerodynamics, Schwartz-Christoffel transelectrical
formations for hydro-dynamics, transcendental functions. Contents classified according to
analytical functions describing transformations with corresponding regions. Glossary. TopoS160 Paperbound $2.00
logical index. 447 diagrams. 6Vs x 91/4.
solve
for
ELEMENTS OF THE THEORY OF REAL FUNCTIONS, J. E. Littlewood. Based on lectures at

Trinity College, Cambridge, this book has proved extremely successful in introducing graduate
students to modern theory of functions. Offers full and concise coverage of classes and
cardinal numbers, well ordered series, other types of series, and elements of the theory
71pp. 53/8 x 8.
of sets of points. 3rd revised edition, vii

INFINITE SEQUENCES AND SERIES, Konrad Knopp. 1st publication in any language. Excellent
introduction to 2 topics of modern mathematics, designed to give student background to
penetrate further alone. Sequences and sets, real and complex numbers, etc. Functions of
a real and
Expansion
complex variable. Sequences and series. Infinite series. Convergent power series.
elementary functions. Numerical evaluation of series, v + 186pp. 5% x 8.
5152 Clothbound $3.50
5153 Paperbound $1.75
of
THE THEORY AND FUNCTIONS OF A REAL VARIABLE AND THE THEORY OF FOURIER'S SERIES,
E. W .Hobson. One of the best introductions to set theory and various aspects of functions
and Fourier's series. Requires only a good background in calculus. Exhaustive coverage of:
metric and descriptive properties of sets of points; transfinite numbers and order types;
functions of a real variable; the Riemann and Lebesgue integrals; sequences and series
of numbers; power-series; functions representable by series sequences of continuous funcrepresentation of functions by Fourier's series; and much
tions; trigonometrical series;
more. "The best possible guide," Nature. Vol. I: 88 detailed examples, 10 figures. Index,
xv + 736pp. Vol. II: 117 detailed examples, 13 figures, x + 780pp. 6Vs x 9V4
Vol. I: S387 Paperbound $3.00
Vol. II: S388 Paperbound $3.00
.
ALMOST PERIODIC FUNCTIONS,

known
mathematician
A.
S. Besicovitch. Unique
detail the two stages
as a generalization of
and important summary by

Bohr's
of development in
well
theory
pure periodicity, with results and
Stepanof, Wiener, Weyl, and Bohr in generalizing the theory.
covers
in
of almost periodic functions:

proofs; (2) the work done by
xi
180pp. 53/8 x 8.
(1)
INTRODUCTION TO THE THEORY OF FOURIER'S SERIES AND INTEGRALS, H. S. Carslaw. 3rd

introduction,
edition, an outgrowth of author's courses at Cambridge. Historical
rational, irrational numbers, infinite sequences and series, functions of a single variable,
definite integral, Fourier series, and similar topics. Appendices discuss practical harmonic
x 8.
368pp.
analysis, periodogram analysis, Lebesgue's theory. 84 examples, xiii
revised
S48
5%
Paperbound
$2.00
SYMBOLIC LOGIC
ELEMENTS OF MATHEMATICAL LOGIC, Paul Rosenbloom. First publication in any language. For mathematically mature readers with no training in symbolic logic. Development
of lectures given at Lund Univ., Sweden, 1948. Partial contents: Logic of classes, fundamental theorems, Boolean algebra, logic of propositions, of propositional functions, expressive languages, combinatory logics, development of math within an object language, paradoxes, theorems of Post, Goedel, Church, and similar topics, iv
214pp. 5% x 8.
THE
INTRODUCTION TO SYMBOLIC LOGIC AND ITS APPLICATION, R. Carnap. Clear, comprehensive,

rigorous, by perhaps greatest living master. Symbolic languages analyzed, one constructed.
Applications to math (axiom systems for set theory, real, natural
numbers), topology
(Dedekind, Cantor continuity explanations), physics (general analysis of determination, causality, space-time topology), biology (axiom system for basic concepts). "A masterpiece,"
Zentralblatt fur Mathematik und Ihre Grenzgebiete. Over 300 exercises. 5 figures, xvi
241pp.
53/8
x 8.
AN INTRODUCTION TO SYMBOLIC LOGIC, Susanne
K. Langer. Probably clearest book for the

scientist, layman
no special knowledge of math required. Starts with simplest
to give remarkable grasp of Boole-Schroeder, Russell-Whitehead systems,
clearly,
quickly.
Partial
Contents: Forms, Generalization, Classes,
Deductive System of
Classes, Algebra of Logic, Assumptions of Principia
Mathematica, Logistics, Proofs of
simplest introduction
Theorems, etc. "Clearest
the intelligent non-mathematician
should have no difficulty," MATHEMATICS GAZETTE. Revised, expanded 2nd edition. Truthvalue tables. 368pp. 5 3/8
8.
philosopher,
symbols, goes on
...
13
CATALOGUE OF
TRIGONOMETRICAL SERIES, Antoni Zygmund. On modern advanced
level. Contains carefully

analyses of trigonometric, orthogonal, Fourier systems of functions, with clear
theory, conjugate series,
series,
proximation
of
Fourier
summability
of
descriptions
adequate
convergence, divergence of Fourier series. Especially valuable for Russian, Eastern EuroS290 Paperbound $1.50
53/
8.
x
8
pean coverage. 329pp.
organized
logic some 100 years

the 1st significant attempt to apply logic to all aspects of human endeavour.
interpretations,
eliminations, condilaws,
signs
laws,
of
and
derivation
Partial contents:
tions of a perfect method, analysis, Aristotelian logic, probability, and similar topics,
THE LAWS OF THOUGHT, George Boole. This book founded symbolic

ago.
xvii
It
is
424pp.
5%
x 8.
edition of probably most cited

2nd
SYMBOLIC LOGIC, C. I. Lewis, C. H.
book in symbolic logic. Wide coverage of entire field; one of fullest treatments of paradoxes;
plus much material not available elsewhere. Basic to volume is distinction between logic
of extensions and intensions. Considerable emphasis on converse substitution, while matrix
system presents supposition of variety of non-Aristotelian logics. Especially valuable sections on strict limitations, existence theorems. Partial contents: Boole-Schroeder algebra;
truth value systems, the matrix method; implication and deductibility; general theory of
x 8.
propositions; etc. "Most valuable," Times, London. 506pp.
Langford.
revised
5%
GROUP THEORY AND LINEAR ALGEBRA,
SETS, ETC.
LECTURES ON THE ICOSAHEORON AND THE SOLUTION OF EQUATIONS OF THE FIFTH DEGREE,
Felix Klein. Solution of quintics in terms of rotations of regular icosahedron around its
axes of symmetry. A classic, indispensable source for those interested in higher algebra,
geometry, crystallography. Considerable explanatory material included. 230 footnotes, mostly
detailed, readable book," Math. Gazette. 2nd edibibliography. "Classical monograph
tion, xvi + 289pp. 53/s x 8.
.
INTRODUCTION TO THE THEORY OF GROUPS OF FINITE ORDER, R. Carmichael. Examines

fundamental theorems and their applications. Beginning with sets, systems, permutations,
etc., progresses in easy stages through important types of groups: Abelian, prime power,
permutation, etc. Except 1 chapter where matrices are desirable, no higher math is needed.
783 exercises, problems, xvi + 447pp. 53/8 x 8.
First published some 40 years ago,
clearest introductions. Partial contents: permutations, groups independent of
of a group with itself, Abelian
isomorphism
representation, composition series of a group,
groups, prime power groups, permutation groups, invariants of groups of linear substitunumerous problems
tion, graphical representation, etc. "Clear and detailed discussion
512pp. 53/8 x 8.
which are instructive," Design News, xxiv
THEORY OF GROUPS OF FINITE ORDER, W. Burnside.

still
one of
V. N. Faddeeva, translated by C. D. Benster.

1st English translation of unique, valuable work, only one in English presenting systematic
exposition of most important methods of linear algebra classical, contemporary. Details
of deriving numerical solutions of problems in mathematical physics. Theory and practice.
Includes survey of necessary background, most important methods of solution, ior exact,
iterative groups. One of most valuable features is 23 tables, triple checked for accuracy,
252pp. 53/8 x 8.
unavailable elsewhere. Translator's note, x
COMPUTATIONAL METHODS OF LINEAR ALGEBRA,
E. V. Huntington. This famous book

elementary account of the modern theory of the continuum as a type
which requires no technical
theory,
ordinal
Cantor-Dedekind
of serial order. Based on the
knowledge of higher mathematics, it offers an easily followed analysis of ordered classes,
numbers. "Admirable
traYisfinite
Cantor's
discrete and dense series, continuous series,
reading easy," Science Progress.
introduction to the rigorous theory of the continuum
Clothbound $2.75
S129
2nd edition, viii + 82pp. 53/8 x 8.
THE CONTINUUM AND OTHER TYPES OF SERIAL ORDER,

gives a
systematic
E. Kamke. Clearest, amplest introduction in English, well suited for independent study. Subdivisions of main theory, such as theory of sets of points, are discussed,
is on general theory. Partial contents: rudiments of set theory, arbitrary sets,
their cardinal numbers, ordered sets, their order types, well-ordered sets, their cardinal
numbers, vii + 144pp. 53/8 x 8.
THEORY OF SETS,
but emphasis
CONTRIBUTIONS TO THE FOUNDING OF THE THEORY OF TRANSFINITE NUMBERS, Georg Cantor.

These papers founded a new branch of mathematics. The famous articles of 1895-7 are
translated,
background
with
of
an
his
82-page introduction by P. E. B. Jourdain dealing

discoveries, their results, future possibilities, ix
with
Cantor, the
53/8 x 8.
211pp.
14
DOVER SCIENCE BOOKS

NUMERICAL AND GRAPHICAL METHODS, TABLES
JACOBIAN ELLIPTIC FUNCTION TABLES,
L.
M. Milne-Thomson. Easy-to-follow, practical, not
only useful numerical tables, but complete elementary sketch of application of elliptic
properties; complete elliptic integrals; Fourier
principle
of
description
functions. Covers
series, expansions; periods, zeros, poles, residues, formulas for special values of argument;
graphs form body of book:
problem;
etc. Tables,
cubic, quartic polynomials; pendulum
Graph, 5 figure table of elliptic function sn (u m); en (u m); dn (u m). 8 figure table of
figure
table
of Jacobian zeta-function
complete elliptic integrals K, K', E, E', nome q. 7
123pp. 53/8 x 8.
Z(u). 3 figures, xi
TABLES OF FUNCTIONS WITH FORMULAE AND CURVES,
E. Jahnke, F. Emde. Most comprehensive

1-volume English text collection of tables, formulae, curves of transcendent functions. 4th
corrected edition, new 76-page section giving tables, formulae for elementary functions not
in other English editions. Partial contents: sine, cosine, logarithmic integral; error integral;
Legendre, Bessel, Riemann, Mathieu, hypergeometric
theta functions;
integrals;
elliptic
functions; etc. "Out-of-the-way functions for which we know no other source." Scientific
Computing Service,
Ltd.
212 figures. 400pp.
5%
8%.
Dwight. Covers in one volume almost every function of importance in applied mathematics, engineering, physical sciences. Three extremely fine
tables of the three trig functions, inverses, to 1000th of radian; natural, common logs;
b 2 ) exp
V2a; complete elliptical insquares, cubes; hyperbolic functions, inverses; (a 2
x);
tegrals of 1st, 2nd kind; sine, cosine integrals; exponential integrals; Ei(x) and Ei(
binomial coefficients; factorials to 250; surface zonal harmonics, first derivatives; Bernoulli,
function;
normal
probability
integral;
Gamma
over
logs
to
base
of
10;
Euler numbers, their
60pp. Bessel functions; Riemann zeta function. Each table with formulae generally used,
sources of more extensive tables, interpolation data, etc. Over half have columns of
x 8.
231pp.
differences, to facilitate interpolation, viii
MATHEMATICAL TABLES,
H.
B.
5%
PRACTICAL ANALYSIS, GRAPHICAL AND NUMERICAL METHODS,
F. A. Willers. Immensely practical hand-book for engineers. How to interpolate, use various methods of numerical differentiation and integration, determine roots of a single algebraic equation, system of linear
equations, use empirical formulas, integrate differential equations, etc. Hundreds of shortcuts for arriving at numerical solutions. Special section on American calculating machines,
by T. W. Simpson. Translation by R. T. Beyer. 132 illustrations. 422pp.
x 8.
5%
NUMERICAL SOLUTIONS OF DIFFERENTIAL EQUATIONS, H. Levy, E. A. Baggott. Comprehensive

collection of methods for solving ordinary differential equations of first and higher order.
2 requirements: practical, easy to grasp; more rapid than school methods. Partial contents:
graphical
integration of differential equations, graphical
methods for detailed solution.
Numerical solution. Simultaneous equations and equations of 2nd and higher orders.
"Should be in the hands of all in research and applied mathematics, teaching," Nature.
21 figures, viii
238pp. 53/8 x 8.
NUMERICAL INTEGRATION OF DIFFERENTIAL EQUATIONS, Bennet,
Milne, Bateman. Unabridged

republication of original prepared for National Research Council. New methods of integration
by 3 leading mathematicians: "The Interpolational Polynomial," "Successive Approximation,"
A. A. Bennett, "Step-by-step Methods of Integration," W. W. Milne. "Methods for Partial
Equations," H. Bateman. Methods for partial differential equations, solution
Differential
of differential equations to non-integral values of a parameter will interest mathematicians,
physicists. 288 footnotes, mostly bibliographical. 235 item classified bibliography. 108pp.
53/s x 8.
Write for free catalogs!

Indicate your field of interest. Dover publishes books on physics, earth
sciences, mathematics, engineering, chemistry, astronomy, anthropology, biology, psychology, philosophy, religion, history, literature, mathematical recreations, languages, crafts, art, graphic arts, etc.
Write
to
Dept. catr
Dover Publications,
Scien
180 Varick
15
St.,
Inc.
N. Y. U, N. Y.
,..,;.
'>- ,,
--
\m
Si
MARSTON SCIENCE
Date Due
^Returned
Due
.01
II
Ikl
fiP
7 2088
LIBRARY
UNIVERSITY OF FLORIDA
M
3
1262
05704
ll
4652
SC/-HSSL
/-
HECKMAN
BINDERY
INC.
1X1
|H|
APR 97
W-To-P,^
N.MANCHE|rER,

Introduction To T 00 Cars

Uploaded by

Document Information

Original Title

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

Introduction To T 00 Cars

Uploaded by

Copyright:

Available Formats

=

FOURIER'S SERIES AND INTEGRALS

Sc.D., LL.D., F.R.S.E.

PROFESSOR OF MATHEMATICS IN THE UNIVERSITY OF SYDNEY

THIRD EDITION, REVISED AND ENLARGED

DOVER PUBLICATIONS, INC

Second Edition, 1921.

PREFACE TO THE THIRD EDITION

and considerable new matter has been introduced without

upper and lower limits of indetermination

series are included.

and the Second Theorem

Additional tests for

and Integrals have been

enlarged by the insertion of questions drawn from recent Cambridge

papers set in the Mathematical Tripos.

bounded variation extends the

the elementary discussion of Fourier's Series given in the text

In the chapters dealing particularly with Fourier's Series space

has been found for the Riemann-Lebesgue Theorem and

quences, and for Parse val's

ordinary purposes the discussion of the properties of

and Fourier's Constants given

in the text will,

hoped, be found both sufficient and satisfactory.

wishes to go further a treatment of the

Lebesgue Definite Integral

new Appendix, which

the place of the former Appendix containing a detailed bibliography

tried to show, in as simple a

what the Lebesgue

respects the rules to which

and are superior

those for the classical

PREFACE TO THE THIRD EDITION

Trigonometrical Series, and

on Fourier's Series, Fourier's Constants,

their titles, brought

would cover many

disposal other works from

PEEFACE TO THE SECOND EDITION

now for some time out of

Integrals, as well as in the

Heat Conduction, that

Mathematical Theory of the

Conduction of Heat, published in 1906, and

has seemed advisable to write

new work, and to issue the same in two volumes. The

Integrals, with special reference to Fourier's

The second volume

Mathematical Theory of the Conduction of Heat.

one can properly understand Fourier's Series and Integrals

without a knowledge of what

involved in the convergence of

development of the idea of a limit and a function, and both are

In Chapter IV the Definite

treated from Kiemann's point of view, and special atten-

given to the question of the convergence of infinite integrals.

whose terms are functions

and the theory

are discussed in Chapters,

theories are closely related,

and can be developed on similar