Professional Documents
Culture Documents
IN SCIENCE IX
SYMMETRIES
IN SCIENCE IX
Edited by
Bruno Gruber
Southern Illinois University at Carhondale
Carhondale, Illinois
and
Michael Ramek
Technische Universitiit Graz
Graz, Austria
http://www.plenum.com
10987654321
Ali rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by
any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise, without written
permission from the Publisher
PREFACE
The international Symposium "Symmetries in Science IX" was held at the Col-
legium Mehrerau in Bregenz, Austria, during the period August 6-10, 1996. Some 40
invited scientists from 13 countries participated in the meeting.
Bruno Gruber
v
CONTENTS
vii
On Quadratic arid Nonquadratic Forms: Application to R 2m ----t R 2m-n
Nonbijective Transformations . . . . . . . . . . . . . . 153
M. Kibler
Quantum Mechanics on the Torus, Klein Bottle and Projective Sphere 313
C. Schulte
viii
q-Quantum Mechanics on Sl 331
R. Twarock
Contributors 349
Index . . . 351
ix
CLASSICAL VS. QUANTUM GROUPS AS
SYMMETRIES OF QUANTIZED SYSTEMS
INTRODUCTION
A q-deformation which seems to be the most relevant from the physical point of
view, concerns the operator algebra in Fock space. In this context, the standard al-
gebraic relations among the harmonic oscillator's creation and annihilation operators
are modified to include a dimensionless parameter q. One possible multidimensional
realization of such a generalization is the quantum group invariant q-oscillator. The
easiest way to understand these oscillators is to start with d commuting copies of the
q E R, (1)
to obtain
a) aia; - qa;ai = 1
b) = a}ai'
aia} i =I j (2)
c) aiaj = ajai, i =I j
Here the algebra is deformed in such a way that the different excitation directions
commute. In particular, this implies that the U (d) invariance of the standard oscillator
is broken. The number operators Ni of this q-deformed system are then given by
1- qNi
a;ai = (3)
1-q
Obtaining the unitary quantum group invariant q-oscillator from this simple system
requires the transformation [5]
Ci =q (N1 +N2++ N i
2
1)
ai (4)
It then follows that under the unitary quantum group U,,;q( d), the annihilation operators
Ci [6] transform as:
e --> Me (5)
where
(6)
Here the elements of the unitary quantum group matrix U are themselves operators
which satisfy the well known commutation relations of a quantum group matrix. They
are taken to commute with Ci. Or, equivalently, the elements of the matrix M act on
Ci by the standard tensor product. It turns out that these non-commuting elements of
the matrix M can also be parametrized in terms of d(d-1)/2 q-oscillators in (1) where
different oscillators commute [7].
The number of n-particle states are degenerate since different directions essentially
commute. Indeed there is a one to one correspondence between the states of the two
oscillators:
(7)
(
n + d -1 ) = (n + l)(n + 2) (n + d - 1)
(8)
d-l 12(d-l)
2
Having discussed quantum group invariant q-oscillators, we now turn to discuss
classical group invariant q-oscillators [8, 9J. A straightforward multidimensional gener-
alization of (1) is obtained by keeping (2.a) and modifying (2.b) such that these relations
can be combined into a single relation which is invariant under the group U( d):
It then follows that for q #- 1, the n-particle states created from the vacuum are
linearly independent. Hence the relation (2.c) is replaced by
c) if q #- 1
if q = 1
The key point is that, for a Fock space with a unique ground state the commutation
relation (2.c) or (9.c) are redundant. The commutation relations (9.a) and (9.b) are
invariant under the classical group U( d) acting on the annihilation operators by
where
(ll)
The number of linearly independent n-particle states for q #- 1 is given by dn . The Fock
space defined by (9) is a Hilbert space for -1 < q < 1, the scaJar product is positive
definite and the operators ai and a; are bounded [9J. For this oscillator the structure of
the Hilbert space is dramatically different for q = 1 and for q #- 1. For the former case,
the n-particle states consist of the totally symmetric representation occuring in the n-
fold product of the d-dimensional representations of U( d). Whereas in the latter case,
they consist of all representations occuring in the n-fold product of the d-dimensional
representation of U( d). In the q # 1 case some additional states iu the Hilbert space
appear, i.e. the states are split when q is introduced. These extra states created from
the vacuum become linearly dependent on the remaining ones in the q = 1 case.
In [2J a quadratic constrained model with a space-time metric having the signature
(+, -, -, -, -) or (+, -, -, -, +) was considered. We find that, in this model after
elimination of the first and fifth coordinates using the constraints and the gauge condi-
tion, we can obtain a proper Hilbert space with a positive definite inner product. The
lagrangian of this quadratic constrained system is given by
(12)
where f1 = 0,1,2,3,5, m and r are constants and N and .A are Lagrange multipliers
which will be used to generate primary constraints. We will mainly be interested in the
3
quantum properties of this system. For,\ = 0, (12) is the parametrization invariant
Polyakov action [10] for a free particle in 5 dimensions.
The Lagrange multiplier ,\ -1= 0 generates a constraint which describes the homoge-
nous space 80(4,1)/80(3,1) for 955 = -1 and 80(3,2)/80(2,2) for 955 = +1 if r2 < O.
For r2 > 0 the corresponding homogenous spaces are 80(4,1)/80(4) for 955 = -1 and
80(3,2)/80(3,1) for 955 = +1. Thus (12) can be interpreted as describing the motion
of a free particle in a De-Sitter universe for the cases sgn(r2) = sgn(955)' The standard
constraint analysis [3] yields
P"P,. - m 2 ~ 0, (13)
X"X,. _r2 ~ 0
For the classification of these constraints, we evaluate their Poisson brackets and obtain
(17)
the second class constraints as
Here a is a parameter which describes the 50(2) symmetry associated with the choice
of second class constraints. We choose
(20)
whereas
(21)
4
For this example, the dirackets including the constraints and the gauge condition
are given by
where Ckl and Dkl are the inverses of the matrices {Ih, Ol } and {Xk, xz} respectively.
The Poisson bracket of the second class constraints is
4r 3 m 3
{Ol,OZ} = -=--(2
sm a )
(23)
Since X2 is chosen to be a function of X s , Ps only, the last term in (22) does not introduce
any corrections to the commutators involving the componentsi,j = 1,2,3. Using the
constraints and the gauge condition to eliminate f1 = 5 and f1 = 0 from the phase space,
setting 01 ::::; O2 ::::; Xl ::::; Xz ::::; 0 and averaging over a to eliminate this extra parameter,
leaves us with the dirackets
{Xi,XiJD (24)
{Pi, PilD
{Pi,Xj}D
where [A, B]+ == AB + BA. These equations can be simplified by scaling Xi and Pi as
= (J Xi, Pi
Xi Pi = 7i (26)
5
we obtain that (25) is equivalent to
and obtain
(32)
where
l+k
q=-- (33)
1-k
This is precisely the q-oscillator algebra defined in (9). The Coon-Baker-Yu oscillator [8]
defined solely by (32) is a consistent algebra with a representation in a Hilbert space [9].
Hence we choose to keep (32) and discard (29). We call this procedure non-canonical
quantization. This is acceptable from the point of view of canonical quantization since
for q = 1 , (32) by itself implies (29). .
We can define the effective Hamiltonian of this system to be proportional to the
conserved Noether current associated with (pseudo) rotations in the Xo - Xs plane
(35)
which is idependent of the gauge condition chosen. This Hamiltonian is invariant under
the subgroup SO(3) x SO(1, 1) C SO(4, 1) or SO(3) x SO(2) C SO(3,2) depending
on the signature of the 5-dimensional metric.
CONCLUSIONS
Since the examination of our basic idea showed us that for constrained systems,
quantization in Fock space is a generalization of the canonical (Dirac) quantization,
we now want to show that this can also be applied to non-constrained systems. Our
starting point for quantization of non-constrained systems is
(36)
where the metric Dij is positive definite. In the case of a Hilbert space with a positive
definite norm, (36) gives the same result as the canonical quantization case. This can
be proven by constructing the Fock space representation of oscillator states. With the
assumption of the existence of the ground state, the action of any annihilation operator
on this ground state gives zero, i.e. ail) = 0 for any i. Then the oscillator states are
defined as
(37)
6
Using only (36) or (32), the matrix elements of ara} are given by
( k1 k 2 ... km latatli
Z J
i ... i n )
1 2
= 0m,n+2 N~.!hkm
't n
t)1112
=~ Ok! N k2 ":km.
~ 'tl 'tl_l
"1
.
"1+1 'tn '
(38)
1=1
(39)
which can again be calculated using only (36) or (32). For q 1 the N tensor and
hence the expression (38) will be completely symmetric under any permutation of the
lower or upper indices. This shows that
Note that, in the constrained case, the matrix element calculated above would be the
q generalized version of the above N tensor [11], which is not symmetric under the
interchange of i and j. Thus we cannot formulate any commutation relations between
ai and aj ; i.e. canonical and non-canonical cases are distinguished. Another point
which also is worth noting is that, if one starts with a d + 2 dimensional model, a d
dimensional oscillator is obtained by a similar analysis. This model would yield a q
oscillator system invariant under the classical group U( d) which in the q ---> 1 limit
reduces to the nonrelativistic U( d) invariant quantum oscillator.
The underlying classical symmetry group was not modified during the quantization
scheme considered in this work. Instead, the number of commutation relations were
reduced and the number of states were increased. However in the q ---> 1 limit the norms
of these extra states become zero and they are excluded from the Hilbert space [11]. An
alternative to this approach is to change the symmetry group itself and to require the
invariance of commutation equations under a quantum group. The quantum group con-
cept was first discovered through the quantization of nonlinear, completely integrable
models in two dimensions. In the language of q oscillators, a quantum group invariant
system is given by Pusz-Woronowicz q oscillators [6]. Since a quantum group is, in fact,
mathematically not a group, the concept of invariance under a classical group has to be
generalized. This depends on quantizing the Poisson brackets by using an exchange re-
lation. In this approach, the number of quantum commutators is equal to the number of
Poisson brackets leaving the number of states unchanged after the quantization. Exam-
ples of this kind of quantization scheme were given in [12]. If one has to discover which
quantization scheme nature has chosen, a possibility would be to set up an experiment
to count the number of quantum states provided that the measurements can be made
to be sensitive to the states that arise due to the deviation of the parameter q from
unity. The comparison of the number of states between the non-canonical quantum sys-
tem and its canonical analogue would directly give us the preferred quantization scheme.
ACKNOWLEDGEMENTS
7
REFERENCES
8
ALGEBRAIC MODEL OF AN OBLATE TOP
INTRODUCTION
The development of spectrum generating algebras and the study of exactly solvable
systems have played an important role in all fields of physics [1]. In particular, in
spectroscopic studies algebraic methods are very useful to study the symmetries and
selection rules, to classify the basis states, and to calculate matrix elements. An ex-
actly solvable model that is of special interest is the symmetric top. As an example,
we mention a study by Bohm and Teese [2] in which the rotational spectrum of a
prolate symmetric top was treated in terms of 50(3,2) with representation doubling
by parity.
The aim of this contribution is to study a realization of a symmetric top in which
all vibrational and rotational degrees of freedom are present from the outset. A good
candidate for such an approach is provided by the so-called algebraic method that was
developed for the collective vibrations and rotations of nuclei [3], and later extended
to rovibrational states of molecules [4, 5, 6] and collective excitations of the nucleon
[7J. This approach is based on the general criterion [8J to take U(k + 1) as a spectrum
generating algebra for a bound-state problem with k degrees of freedom and assigning
all states to the symmetric representation [N] of U( k + 1). For collective nuclei this led
to the introduction of the U(6) interacting boson model [3J and for diatomic molecules
to the U(4) vibron model [4J.
In this contribution we apply the algebraic method to a three-body system. The
dynamics of this system is determined by the six degrees of freedom of the two relative
coordinates, which in the algebraic approach leads to a U(7) spectrum generating al-
gebra. We develop the formalism for a system of three identical objects and discuss an
application to nonstrange baryon resonances which are interpreted as vibrations and
rotations of an oblate symmetric top.
(1)
Here rI, r2and r3are the coordinates of the three objects. Instead of using coordinates
and momenta we prefer to use a second quantized formalism, in which we introduce a
dipole boson for each independent relative coordinate and an auxiliary scalar boson
(3)
with (m,m' = -1,0,1) and (i,j = p,A). The introduction of the scalar boson is
just an elegant and efficient way by means of which the full dynamics of two vectors
can be investigated. This includes situations in which there is a strong mixing of the
oscillator basis (collective models). For a system of interacting bosons the model space
is spanned by the symmetric irreps [N] of U(7), which contains the oscillator shells with
n = np + n,\ = 0,1,2, ... ,N. The value of N determines the size of the model space.
For three identical objects (e.g. for X3 molecules or nonstrange qqq baryons) the
Hamiltonian (or mass operator) has to be invariant under the permuation group S3'
The permutation symmetry of three identical objects is determined by the transposition
P(12) and the cyclic permutation P(123) [9]. All other permutations can be expressed
in terms of these two elementary ones. Algebraically, these operators can be expressed
in terms of the generators Gij = :Em bj,mbj,m that act in index space (i,j = p, A). We
find
P(12)
P(123) (4)
The operators F2 and F3 are two components of the SU(2) pseudo-spin in index space,
which in turn is related to the vortex spin [10]
10
The operator it counts the total number of dipole bosons
Since the permutation group S3 is isomorphic to D3, one can either use the irreducible
representations of S3 or those of D3 to label the three symmetry classes. In this con-
tribution we use the D3 labeling: Al and A2 for the one-dimensional symmetric and
antisymmetric representations, and E for the two-dimensional representation.
All operators of interest are expressed in terms of the building blocks of eq. (2)
which transform under D3 as Ep, E>. and At, respectively. With the help of eq. (4)
one can construct physical operators with the appropriate symmetry properties. The
most general form of the Hamiltonian, that preserves angular momentum and parity,
transforms as a scalar (i.e. Ad under D3 and contains at most two-body interactions
is given by
PERMUTATION SYMMETRY
The permutation symmetry of a given wave function can be determined from the
transformation properties under P(12) and P(123) [9]. Since the Hamiltonian of eq. (7)
is invariant under the transposition P(12), basis states with np even and np odd do not
mix, and can therefore be treated separately. This allows one to distinguish wave
functions with t = Al or E;>, from wave functions with t = A2 or Ep
11
The cyclic permutation P(123) can be used to distinguish wave functions with t = Al
or Az from wave functions with t = E). or Ep
I1/;A,) 1 0 0 0 PA, )
11
I1/;AJ 0 1 0 0 I1/;A2)
P(123) (9)
I1/;EJ 0 0 cos(27r /3) - sin(27r /3) I1/;Ep)
I1/;E,J 0 0 sin(27r /3) cos(27r /3) \ I1/;EJ
In eqs. (8) and (9) we have used that the application of P(12) and P(123) on l1/;t)
means carrying out the inverse operation on the boson operators, bt,m and bl,m [9, 11).
In practice, the wave functions I1/;Ep) and I1/;EJ are obtained from separate diagonaliza-
tions, and hence are determined up to a sign. Eq. (9) can be used to determine their
relative sign, so that they transform as the two components of the mixed symmetry
doublet with t = E symmetry.
In addition to angular momentum, parity and permutation symmetry, Lf, the
Hamiltonian of eq. (7) has another symmetry. Since iI commutes with the opera-
tor Fz of eq. (5), its eigenstates can also be labeled by the eigenvalues of Fz: mF =
0, 1, ... ,N. We denote these eigenstates by lmF) . Since the transposition P( 12)
anticommutes with F2
(10)
the simultaneous eigenfunctions of P(12) and iI are given by the linear combinations
(11)
Here we have introduced the label v by mF = v (mod 3). These wave functions satisfy
(12)
The cyclic permutation P(123) acts in Fock space as a rotation generated by Fz (see
eq. (4
According to eq. (9) the wave functions 11/;1) (11/;2) transform for v = 0 as t = A2 (AI)'
and for v = 1,2 as t = Ep (E>,). The special connection between the label mF = v (mod
3) and the permutation symmetry only holds for Hamiltonians that commute with F2
All one- and two-body D3 invariant interactions indeed satisfy this property, and hence
their eigenstates can be labeled by MF = ImFI and Lf . However, at the three-body
level there are D3 invariant interactions that mix states with ilmF = 6 [6, 12). The
permutation symmetry can then still be determined by the general procedure outlined
in eqs. (8) and (9).
12
OBLATE SYMMETRIC TOP
An analysis of the equilibrium shape of the potential energy surface that corresponds
to eq. (7) yields that the only stable nonlinear configuration is that of an equilateral
triangle [13]. This equilibrium configuration is represented in U (7) by an intrinsic (or
coherent) state in the from of a condensate
_1_(bt )N 10)
v'Nf c ,
(14)
with
(15)
For the special case of R2 = 0, the intrinsic Hamiltonian has U(7) J U(6) symmetry
and corresponds to an anharmonic vibrator, whereas for R2 = 1 and 6 = 0 it has
U(7) J 50(7) symmetry and corresponds to a deformed oscillator.
In the more general case with R2 of. .0 and 6, 6 > 0, the intrinsic Hamiltonian of
eq. (16) describes the vibrational excitations of an oblate symmetric top. This can be
seen from a normal mode analysis. To leading order in N, Hint reduces to a harmonic
form [6,7]
(17)
with eigenfrequencies Cl = 4N6R2 and C2 = 4N~2R2 /(1 + R2) . The deformed bosons
are given by
The first term in eq. (17) represents the symmetric stretching mode (b u ) and the second
term a degenerate doublet of an anti symmetric stretching mode (b v ) and a bending
mode (b w ). This is in agreement with the point-group classification of the fundamental
vibrations of a symmetric X3 configuration [15]. Therefore, the deformed bosons of
eqs. (15) and (18) correspond to a geometry of an oblate symmetric top with the
threefold symmetry axis along the z-axis. In the large N limit the vibrational spectrum
is harmonic
(19)
13
In a geometric description, the excitations of an oblate top are labeled by
(Vb VDi K, Lf, M. Here VI denotes the number of quanta in the symmetric stretching
mode which has Al symmetry, and V2 the total number of quanta in the asymmetric
stretching and the bending modes, which form a degenerate doublet with E symme-
try. The label [ is associated with the degenerate vibration. It is proportional to the
vibrational angular momentum about the axis of symmetry and can have the values
[ = V2, V2 - 2, ... ,lor 0 for V2 odd or even, respectively. The rotational states, which
are characterized by the angular momentum L and its projection K on the three-fold
symmetry axis, are arranged in bands built on top of each vibration. The projection
K can take the values K = 0,1,2, ... , while the values of the angular momentum are
L = K, K + 1, K + 2, .... The parity is P = (_)K, t denotes the transformation char-
acter of the total wave function under D 3 , and M is the angular momentum projection.
For a given value of [ and K the degeneracy of a state with angular momentum L is
given by 4(2L + 1)/(1 + 8/,0)(1 + OK,O).
GEOMETRIC INTERPRETATION OF MF
Two sections ago we showed that the eigenstates of the algebraic Hamiltonian of
eq. (7) can be labeled by M F , Lf . The same holds for its intrinsic part, eq. (16),
which describes the vibrational excitations oran oblate top. In this section we wish to
elucidate the role of the label MF in the context of the oblate symmetric top, and, in
particular, its relation to the geometric labels K and [.
The connection between an algebraic and a geometric description of an oblate top
can be studied by means of intrinsic (or coherent) states [16]. In such an approach,
each vibrational band (VI, V~) is represented by an intrinsic state which can be obtained
from eq. (14) by replacing a condensate boson (bt) by one of the deformed bosons of
eq. (18). However, the deformed operators, bt and bt, do not have good projection
of the vibrational angular momentum on the symmetry axis. In order to construct
intrinsic states with well-defined projection on the symmetry axis, we transform the
Cartesian bosons of eq. (18) to spherical bosons, and introduce the linear combinations
[9]
(bl,m + i b~,m)/V2 ,
(bl. m - i b;,m)/V2 , (20)
(-bt - i b~)/V2 ,
(bt-ibU/h. (21)
In this representation, the operator F2 has the simple form of the difference between
two number operators
(22)
The intrinsic state for a vibration (VI, vD with projection l on the symmetry axis is
then given by (for N - t CXl)
14
which can be expressed in terms of a sum over (n'l/' nd configurations
For each of these configurations the projection K of the angular momentum along the
symmetry axis is
K = n( - n'l/ 21 = mF 21 . (25)
This shows that the algebraic label mF has a direct interpretation in terms of the geo-
metric labels, K and 1. For 1 = 0 we have mF = K, but for 1 > 0 there are two possible
values of mF for each K. Thus MF = ImFI = IK =F 2/1 provides an additional quantum
number which is needed for a complete classification oUhe rotational excitations of an
oblate top. For example, the (VI> v~=t) vibrational band has two LE levels which have
the same value of K = 3, but different values of MF . The rotational spectrum is given
by
X:t L(L + 1) - X:2 M'j;.
X:t L(L + 1) - X:2 (K2 T 4Kl + 412) . (26)
The last term contains the effects of the Coriolis force which gives rise to a 8X:2Kl
splitting of the rotational levels which increases linearly with K. The label MF plays
a role similar to that of the label G discussed by Watson [12] for X3 molecules, and
that of the label m by Bowler et al. [17] in the context of a harmonic oscillator quark
model in baryon spectroscopy. As an example of the assignments of K and MF we
show in Figures 1 and 2 the classification scheme for the levels with L ::; 3 belonging
to a (VI, v~) vibrational band with 1= 0 and 1 = 1, respectively.
With the exception of the levels with K = I = 0, all levels in Figures 1 and 2 are
doubly degenerate. According to eq.(26) the splitting of the levels with K = 0 is zero,
whereas for K > 0 there remains a double degeneracy because of the two projections
K on the symmetry axis. In particular, the rotational spectrum of Figure 2 does not
exhibit I-type doubling, and therefore for MF = 0 (mod 3) there is a degenerate doublet
consisting of At and A2 levels. This is a consequence of the fact that the one- and two-
body D3 invariant Hamiltonians of eqs. (7) and (16) commute with F2 The degeneracy
of the At and A2 states can be lifted by introducing higher order interactions that break
MF symmetry. For example, there exist three-body D3 invariant interactions that mix
states with f:::.MF = 6 [6]. A similar situation is encountered in Watson's effective
Hamiltonian [12], whose main terms are diagonal in the quantum number G '(which
plays a similar role as M F ), but which also contains small higher-order corrections with
f:::.G=6.
We note that U(7) can also accommodate other types of rotations (e.g. nonrigid)
and kinematics (e.g. relativistic). In such cases, the L(L + 1) and M'j;. terms in eq. (26)
will be replaced by a general function of the angular momentum Land MF [7].
WAVE FUNCTIONS
In the algebraic approach, the oblate top wave functions can be obtained by pro-
jection from an intrinsic state. In principle this is an exact procedure, but since the
expressions for the intrinsic states of eq. (23) are only valid for N ---+ 00, the same holds
for the results presented in this section. In the large N limit we find
..
IN, VI> V2, l, R, K, L, M} - Vfii+lJ
- ~ (L)* (0)
dO'DMK IN,. Vt, V2, I, R, 0) . (27)
15
__ 3~
__ 21. -2E __ 2~
__ It __ IE
__ 01.
K=O K=I K=2 K=3
Next we construct states with good D3 symmetry by considering the the action of the
transposition and the cyclic permutation on the projected wave function
P(12) IN, VI, V2, I; R; K, L, M) (-lt2+L IN, VI, V2, -1; R; -K, L, M) ,
P(123) IN, VI, V2, I; R; K, L, M) e27fi (K-21)/3IN , V b V 2" l R', K , L , M) (29)
Here we take without loss of generality K 2 0 and I 2 O. States with good D3 symmetry
are given by the linear combinations
I~I) = -z [IN, VI, V2, -1; R; K, L, M)-( -1)V2+LIN, VI, V2, l; R; -K, L, M)] ,
J2(1 +DK,oDl,O)
1~2) = (-It [IN, VI, V2, -1; R; K, L, M)+( _1)V2+LIN, Vb V2, I; R; -K, L, M)] .
J2(1 +DK,oDl,O)
(30)
16
_3~
_3~
__ 3 E
._lAlA2
_ot
J{=l J{=2
According to eq. (9) the wave functions 17)!1) (17)!2)) and 17)!3) (11fJ4)) transform for v = 0
as t = A2 (AI), and for v = 1,2 as t = Ep (EA)' These wave functions are consistent
with the choice of geometry in the deformed bosons of eqs. (15) and (18).
BARYON RESONANCES
In this section we discuss an application of the oblate top model in baryon spec-
troscopy. We consider baryons to be built of three constituent parts (quarks or oth-
erwise) with the string configuration of Figure 3. The full algebraic structure is ob-
tained by combining the geometric part, U(7), with the internal spin-flavor-color part,
SUs !(6) 0 SUc (3) (not considering heavy quarks),
(32)
17
1 " " ..... --3-.-.. . . . . .\
,,
.\
,,
1
,, \
,; ,
""
l \
\
"I " ' ... ...
" . ...
,1"" ...........,
I \
1\ 1 _...-...-------- 2 ")
' .......... _-_ ..... "..,... -................ _--_ ....
\
/
For the nucleon (isospin 1= 1/2) and delta (I = 3/2) families of resonances the three
strings of Figure 3 have equal lengths and equal relative angles. The three constituent
parts form an equilateral triangle with D3h :J D3 point group symmetry. Baryon reso-
nances are then interpreted in terms of vibrations and rotations of an oblate symmetric
top. In order to have total baryon wave functions that are antisymmetric, the permuta-
tion symmetry of the geometric (or spatial) part must be the same as the permutation
symmetry of the spin-flavor part (the color part is a color singlet, i.e. antisymmetric).
Therefore one can also use the dimension of the SU.f(6) representations to label the
states: Al f-4 [56], A2 f-4 [20] and E f-4 [70] . The nucleon itself is identified with the
oblate top ground state (Vl'V~); K,Lf = (0,0); 0,01" whereas the N(1440) Roper and
the N(1710) resonances are interpreted as one-phonon excitations with (1,0); 0, 01,
and (0,11); 0, O~, respectively (see Figure 4).
In such a collective model of baryons, the mass operator is written in terms of a
spatial and a spin-flavor part. The spatial part is given by eq. (16) plus a term linear
in the angular momentum L to reproduce the linear Regge trajectories, while the spin-
flavor part is given by the Giirsey-Radicati form [18]. With this mass operator we
obtained a good overall fit of the spectrum of the nucleon and delta resonances with a
r.m.s. deviation of 39 MeV [7, 13].
A far more sensitive test of models of baryon structure is provided by electromagnet-
ic, strong and weak couplings. In the next section we show how the general formalism
of the oblate top, which was developed in the previous sections, can be used to derive
closed expressions for helicity amplitudes that can be measured in photo- and electro-
production and strong decays of baryon resonances.
FORM FACTORS
(33)
18
N(1710)
3
oTH
N(1440)
2
1 1
_N_(9_39--...:
with (m = -1,0, 1). Here k = k2 is the photon (meson) momentum, ko is the photon
(meson) energy, m3 is the constituent mass and (3 represents the scale of the coordinate.
The operators Dp,m and D)",m are dipole operators in U(7)
x sP
(btp x b )(1)
- st m ,
The normalization factor Xv is given by the reduced matrix element between the Lf =
0t ground state and the first excited IE state (both belonging to the (vl,vD = (0,0 0 )
ground band)
(35)
Since D>. is a generator of the algebra of U(7), the matrix elements of fJ are repre-
sentation matrix elements of U(7), i. e. a generalization of the Wigner 'V-matrices for
SU(2). By making an appropriate basis transformation they can be obtained numer-
ically without having to make any further approximations. However, in the limit of
N -. 00 (infinitely large model space) the matrix elements of fJ, 1\,m and Tp,m can also
be derived in closed form.
We illustrate the method by evaluating the matrix elements of fJ that connect the
nucleon with its excited rotational states. In the collective model, the nucleon wave
function is that of the ground state of the oblate symmetric top
19
The vibrationally elastic matrix element connecting the ground state with a rotational
excitation is given by
(N, 0, 0, 0; R; K, L, MI U IN, 0, 0, 0; R; 0, 0, 0)
= v'2L 2+ 1
87r
J dndn , ('DMK (0) *( 0, ')- (
L ) ( 0, ) 'Doo '
N,O,O,O;R,n IU IN,O,O,O;R,n ') . (38 )
A
In the large N limit the matrix element appearing in the integrand becomes diagonal
in the orientation 0, of the condensate. The remaining angular integral can be obtained
in closed form in terms of a spherical Bessel function
With the wave functions of eq. (30) we find that the matrix elements connecting the
nucleon with its rotationally excited states are
(1/;11 U11/;0) = 0,
(01. I 101. ) K 2L + 1J(L+K)!(L-K)! [1 (-l)L- Kl' (k,8) (40)
2(l+OK,0) (L + K)!!(L _ K)!! +
<: ( )
0/2 U % = VM,O -1 Vi
A
JL
For the (Vb V~) = (0,0 0 ) ground band we have K = v (mod 3).
To summarize the results for the matrix elements of eq. (33) we;'ltroduce the
notation
Fi(k) (iIUI1/;o) ,
G;;A.m(k) (1/;,1 TA,m 11/;0) ,
Gi;p,m(k) (1/;11 Tp,m 11/;0) , (41)
for the elementary oblate top form factors. The nucleon wave function 11/;0) is given in
eq. (37), and the wave function of the resonance 11/;;) is given in eq. (30) for i = 1,2
and in eq. (31) for i = 3,4 . All matrix elements can be expressed in terms of spherical
Bessel functions. For the rotational transitions we find
,
OM,O Z(k,8) , (42)
where Z(k,8) can be obtained by comparing with eq. (40). The matrix elements of T)..,m
and Tp,m can be derived as
G1 ;)..,m(k) 0,
dZ(k,8)
G2 ;)..,z(k) -OM,om3ko,8~ ,
r-::c;-::--:-:- Z (k,8)
OM,l m3ko,8 VL(L + 1) ~ , (43)
20
Table 1. Analytic expressions of the (Vb vD = (0,0 0) (0,00) matrix elements of
---+
the transition operators of eq. (33) in the oblate top model for N ---+ 00. The final
states are labeled by [dim{SUsj (6)}, LP](Vl'V~);K . The initial state is [56,0+](0,00);0'
and
In eqs. (43) and (44) we have used G.;.,(k) = =fV2G.;.,1(k). We see that all relevant
matrix elements can be expressed in terms of F2(k) and G 1 ;p,(k) only. In Table 1 we
present the results for the elementary form factors that are relevant for the low lying
nucleon and delta resonances with L ::::: 2.
The matrix elements of eq. (33) for vibrational excitations are obtained as above
by projection from the corresponding intrinsic states. In Table 2 we present the results
for the elementary form factors for the vibrational excitations (0,0 0) ---+ (1,0 0 ), (0,11)
for nucleon and delta resonances with L ::::: 1. The oblate top form factors in Tables 1
and 2 correspond to the transitions indicated schematically in Figure 4.
21
Table 2. As Table 1, but for the (VI, V~) = (0,0) --; (1,0) and (0,11) transitions.
Xl = (1- R2)/2RVN and X2 = -VI + R2/RV2N.
-h k(3 j1 (k(3)
(70,0+](0,11);0 -X2
j2(k(3)
"f k(3
(70,1+](0,11);0
-X2
[56,1-](0,11);1 -i X2 {}- k(3 h( k(3)
[20,1-)(0,11);1
F4(k)
iX2 ~ [3h(k(3) - k(3jl(k(3)]
G3;p, (k )/ m3ko(3
[70,1-](0,11);1 -i x2vhk(3[2jo(k(3) - j2( k(3)) i x2 vk[h( k(3)+4jo( k(3) +3k(3 h (k(3))
In particular, we studied the case of three identical objects and showed how the
corresponding permutation symmetry can be taken into account exactly. Rather than
explicity constructing states of good permutation symmetry [9], we obtain these states
by diagonalizing a Hamiltonian that is invariant with respect to the point group sym-
metry. We developed a general procedure to determine the permutation symmetry of
any given wave function. It was shown that in the large N limit (large model space)
the algebraic Hamiltonian for the X3 system corresponds to an oblate symmetric top.
For the special case of one- and two-body interactions, the eigenstates can be labeled
by an additional quantum number Mp. This label plays a very interesting role. On
the one hand, it has a direct connection to the permutation symmetry. On the other
hand, in the large N limit it is directly related to the geometric labels /{ and 1, and
provides an extra label which is needed to classify the rovibrational states of the oblate
top uniquely.
It was shown that U(7) provides a unified treatment of both rotational and vi-
brational excitations of an oblate top. The ensuing algebraic treatment of the oblate
top has found useful applications both in molecular physics (X3 molecules [6, 22]) and
hadron physics (nonstrange qqq baryons [7, 20, 21)). As an example, we discussed
an application to baryon resonances which are interpreted as vibrations and rotations
of an oblate top. We derived closed expressions for a set of elementary form factors
for nonstrange resonances. These elementary form factors appear in all calculations
of electromagnetic, strong and weak couplings, and hence form the backbone of the
model. In a separate contribution to these proceedings [23] we discuss in more detail
the application to electromagnetic and strong couplings of baryon resonances.
Finally, we note that, although we discussed the specific case of three identical ob-
jects, the algebraic procedure is valid in general, both for three nonidentical objects and
for the many-body system. Applications of these techniques to polyatomic molecules
will be discussed in a separate contribution [24].
22
ACKNOWLEDGEMENTS
REFERENCES
[1] A. Bohm, Y. Ne'eman, A. O. Barut and others, Dynamical groups and spectrum
generating algebras, Volumes 1 and 2, World Scientific, Singapore, 1988.
[2] A. Bohm and R. B. Teese, J. Math. Phys. 17 (1976), 94.
[3] A. Arima and F. Iachello, Phys. Rev. Lett. 35(1975), 1069; for a rewiew see: F.
Iachello and A. Arima, The Interacting Boson Model, Cambridge University Press,
Cambridge, 1987.
[4] F. Iachello, Chern. Phys. Lett. 78 (1981), 581; for a review see: F. Iachello and
R. D. Levine, Algebraic Theory of Molecules, Oxford University Press, Oxford, 1993.
[5] A. Frank and P. van Is acker , Algebraic Methods in Molecular & Nuclear Structure
Physics, Wiley Interscience, 1994.
[6] R. Bijker, A. E. 1. Dieperink, A. Leviatan, Phys. Rev. A52 (1995), 2786; R.
Bijker, A. E. 1. Dieperink and A. Leviatan, in: Symmetries in Science VII: Spectrum
Generating Algebras and Dynamic Symmetries in Physics, B. Gruber and T. Otsuka
(eds.), Plenum Press, New York, 1993, 75.
[7] R. Bijker and A. Leviatan, Rev. Mex. Fis. 39, Suplemento 2 (1993), 7; R. Bijker,
F. Iachello and A. Leviatan, Ann. Phys. (N.Y.) 236 (1994), 69.
[8] F. Iachello, NucL Phys. A560 (1993), 23.
[9] P. Kramer and M. Moshinsky, Nucl. Phys. 82 (1966), 241; M. Moshinsky and Yu. F.
Smirnov, The harmonic oscillator in modern physics: from atoms to quarks, Gordon
and Breach, 1996.
[10] D. J. Rowe, Rep. Prog. Phys. 48 (1985), 1419.
[11] E. P. Wigner, Group theory, Academic Press, New York, 1959.
[12] J. K. G. Watson, J. Mol. Spectrosc. 103 (1984), 350; J. K. G. Watson, Can. J.
Phys. 72 (1994), 238.
[13] R. Bijker and A. Leviatan, in: Symmetries in Science VII: Spectrum Generating
Algebras and Dynamic Symmetries in Physics, B. Gruber and T. Otsuka (eds.),
Plenum Press, New York, 1993, 87.
[14] M. W. Kirson and A. Leviatan, Phys. Rev. Lett. 55 (1985), 2846.
[15] G. Herzberg, The spectra and structures of simple free radicals, Dover, 1971.
[16] See e.g. R. Gilmore and D. H. Feng, Nucl. Phys. A301 (1978), 189; A. M.
Perelomov, Generalized coherent states and their applications, Springer Verlag,
1986; W. M. Zhang, D. H. Feng and R. Gilmore, Rev. Mod. Phys. 62 (1990), 867.
[17] K. C. Bowler, P. J. Corvi, A. J. G. Hey, P. D. Jarvis and R. C. King, Phys. Rev.
D24 (1981),197; A. J. G. Hey and R. L. Kelly, Phys. Rep. 96 (1983), 71.
23
[18J F. Giirsey and L. A. Radicati, Phys. Rev. Lett. 13 (1964), 173.
[19) Particle Data Group, Phys. Rev. D54 (1996), 1.
[20) R. Bijker and A. Leviatan, Rev. Mex. Fis. 41, Suplemento 1 (1995), 62; R. Bijker,
F. Iachello and A. Leviatan, Phys. Rev. C, C54 (1996), 1935.
[21) R. Bijker and A. Leviatan, preprint nucl-th/96030l5, proceedings of the XIX Sym-
posium on Nuclear Physics, January 4-7,1996, Oaxtepec, Mexico, Rev. Mex. Fis.,
in press; R. Bijker, F. Iachello and A. Leviatan, preprint nucl-th/9608051, Phys.
Rev. D, in press.
[22) A. Frank, R. Lemus, R. Bijker, F. Perez-Bernal and J. M. Arias, preprint chem-
ph/9604005, Ann. Phys. (N.Y.), in press.
[23] A. Leviatan and R. Bijker, these proceedings.
[24J A. Frank, R. Lemus, R. Bijker, F. Perez-Bernal and J. M. Arias, these proceedings.
24
THE MASS-SQUARED OPERATOR AND THE
EINSTEIN-HILBERT ACTION FOR RESCALED
LORENTZ METRICS
1 Fakultiit
fiir Mathematik und Informatik
Universitiit Mannheim
D-68131 Mannheim, Germany
2Kopernikusstr. 52
D-68165 Mannheim, Germany
INTRODUCTION
The aim of these notes is to show a relation between Einstein equation and Klein-
Gordon equations with matter terms. This link is achieved by taking the trace of these
Einstein equations. Below we formulate this more precisely.
Basic to our notes is the following observation (Sniatycki (1980)): The geometric
quantization on a Lorentz manifold M with metric 9 yields the operator
ft (0 9 + -61 . R 9
)
Here pg denotes the Riemannian volume-form. Setting 9 := 1/;2 . g, for a real positive
state 1/;, a well known formula in differential geometry (d. Appendix) yields
"'
J,fg
(0 9 + -6
1 . R g, 01.)
'Y
= -61 . 1 D
R-9 . rg,
1/.-
T(g) being the energy-momentum tensor of o:(g). The g-trace trg of this equation is
(Og +1
6"' Rg)'Ij; = -'Ij; 3 . trgT (_g).
In fact Ho'lj; = _!'lj;3 . trgT(g) holds true. Therefore we may extend the remark on the
vacuum equations as follows:
The expectation value
is optimally observed, provided the above mentioned Einstein equation with a matter
term holds true.
The Einstein equation may be rewritten as
Clearly, o:(g) may chosen to be 'Ij; . Ho'lj;. In this case the equation above reads as
furthermore, in the special case of o:(g) = -!. m 2 . 'lj;2, for a constant m, this equation
turns into
+ 61 . Rg)'Ij; = m 2 . 'Ij;.
(Og
These lines of thaughts apply to 'Ij; = 1 + cpo If g satisfies Rg = 12 HoI, the g-trace
of the above mentioned Einstein equation for this sort of 'Ij; yields
The factor 2 on the right hand side is a consequence of the variation with respect to
the metric g.
All these results are easily obtained. To make these notes rather self contained we
collected in the Appendix the mathematical preliminaries and tools.
26
THE EXPECTATION VALUE OF THE MASS-SQUARED OPERATOR
Let (M,g) be a Lorentz manifold; here M is smooth and oriented and 9 is a smooth
Lorentz metric. Let D C M be a smooth, bounded, four-dimensional submanifold of
M. On the real vector space COO(D, JR.) of all smooth real-valued maps of D we have a
scalar product
'i y, 'ljJ E CCO(D, JR.) (1)
where flg is the metric volume-form. Given an operator H on COO(D, JR.) let us call
the expectation value of Hat 'ljJ, even though 'ljJ is not of norm one with respect to (1).
At first, we will investigate a very special operator. It arises from the mass-squared
function on TD associated with g; this function is given by g(v,v) for any v E TM. It
is twice the kinetic energy in Riemannian geometry.
In geometric quantization the operator associated with the mass-squared function
1S
1
'Ii . (0 9 + [; . Rg),
(3)
(4)
At first we treat the vacuum. Given a Lorentz metric 9 and any rescaling function
'ljJ we form
27
Setting g' := l/J-2 . g, we conclude that
(6)
1
(0';'-2. g + 6"' R';'-2.g)l/J = 0 (7)
holds b'ue. If on the other hand (1) holds fOT l/J = 1, then Rg = O.
Motivated by proposition 2 we consider a sixth of the Einstein- Hilbert action which
assigns (in our case) to each Lorentz metric 9 the value
(8)
F)g(H,,p) := ~. L R g /1iJ + L
6 a(g) . /1iJ (9)
28
with respect to g. Here a is a smooth density, differentiable with respect to g. If thus
T(g) denotes the energy-momentum tensor associated with oo(g) (d. Wald(1984)) we
find, as usual, Einstein's equation which reads in this case as
(13)
iftrgT(g) = 0 then Rg = 0 by (11) and hence (13) yields the Klein-Gordon equation
1
(Og + '6 . Rg)7j; = O.
( 14)
T(g)(v,w) = g(Tgv,w)
where
and
F(v,w) = g(Fgv,w) \/V,w E TqD and \/q ED.
Since Tg is trace free, equation (12) reads as
Ricq = 6 . T(g).
Hence the above equation shows Rg = 0 and (13) turns therefore into the Klein-Gordon
equation
29
which is the wave equation in case of Rg = O.
Next, the density a in (9) shall be of the form
with Ho as in (5). It is not g-dependent. The variation of (9) with a as in (15) with
respect to 9 yields Einstein's equation
1
Ricg - 2"' Rg . 9 = -3 ('Ij;-3. Ho'lj;) . g. (16)
for g. The energy-momentum tensor of a in (15) is T(g) = ('Ij;-3. Ho'lj;) g. In this case
only the variation of the volume-form /-lg contributes to the energy-momentum tensor.
We therefore have obtained the following:
Corollary 5
The expectation value -fg(Og + ~Rg + Ho,'Ij;) for a rescaling function 'Ij; E COO(D,IR) is
stationary at 9 if Einstein's equation
1
Ricg - - . Rg . 9 = -3 'Ij;-3 . Ho'lj; . 9 (17)
2
and hence
(18)
or equivalently
1 3 _ 1 1
Ho'lj; = -2"' 'Ij; . trgT(g) = 2"' (Og + 6"' Rg)'Ij;. (21)
The expectation value -fg (Og + ~ R g, 'Ij;) at 'Ij; is hence of the form
.
RZCg - 2"1 . Rg . 9_ = -2"3 . m 2 . 'Ij; - 2. -9 (22)
30
and hence
(23)
and in case of 9 =9
(25)
A GENERALIZATION
In this section we will extend (18) in corollary 5 to cp E Coo(D, 1R) yielding (27)
below. To this end we set
If hence Rg = 12 HoI (cf. (20)), vyhich is the case if JI,g(H, 1) is stationary at g, then
(26) yields the following:
Theorem 8:
Given H := Og + ~. Rg + Ho and let Rg = 12 HoI. The expectation value Jj)g(H, 1 + cp)
is stationary at 9 iff Einsteins equation
1
(Og + "6 . Rg - 2 . Ho)cp = O. (27)
for some symmetric smooth two-tensor Tl(jj). This holds true because T is defined on
!.
the whole conformal class of g. Since JI,( Og + Rg, 'lj;) is stationary at g, equations (20)
and (21) together yield
31
Corollary 9:
Let eg(H, 1) be stationary at g. If moreover eg(H, 1 + 'P) is stationary at 9 too, and if
in addition l1-g T(g) = 0 then
if hence Rg = 0 then
APPENDIX
To extend 9 to Ak(M), the collection of all smooth k-forms of M we need to define it only
on decomposable elements a = 0'1) 1\ .. . 1\ ak and B = f31, 1\ ... 1\ f3k with ai, f3i E Al (M)
for all i,j = 1"", k and then extend it linearly to all of Ak(M). We let
where the right hand side is the determinant of the matrix with entries g( ai, f3j).
The Hodge star operator *:Ak(M) ---+ A4-k(M) is given by
as well as
* * 0'= (_l)k(n-k)+lnd g . a
The divergence:
The divergence operator 15 : Ak(M) ---+ a k - 1 (M) associated with 9 is given by (cf.
Abraham, Marsden, Ratin 1983)
32
The d'Alembertian Dg is given by
Dg:= do + od.
The divergence of a vector field: Define the divergence divgX as
\IX E rTM
here iXPgC,,,J = /t(X,"',"'," .). All these equations are easily verified.
The last formula yields immediately
showing
diVg X = trV X.
For a orthonormal basis Eo,E1,E2,E3 with g(Eo, Eo) = -1 we deduce from above
3
k
rg(gmdcp,gmdcp'l/tq = kr cp' OgCp' flq + r cp' dcp'(n). /taD
kD
with flg the Riemannian volume-form and n denotes the outward directed unit normal.
Moreover flEW = inflg. The boundary term vanishes if <p and <p' have their support in
the interior of D.
Conformal change of the metric: Given f E Coo (M, ]Rn) we set
!J := e2J . g.
Here Hg (f) (X, Y) is the Hessian given by g(V xgradgf, V). Moreover Igradgfl2 =
g(gradgf, gradg!). The Ricci-tensor Ricg applied to any X, Y E rT M is hence given
by (d. Besse 1987)
Rici) (X, Y) = Ricg(X, Y) - 2Hg(f) (X, Y) + (Og(f) - 2IdfI 2 ). g(X, Y) + 2df(X) . dj(Y).
Here Idfl 2 = g(gradgf, gradgf}. Therefore, the scalar curvature Rg reads in terms
of Rg and f as (d. Besse 1987)
If 8 = e f the Ricci tensor Rici) and the scalar curvature Ri) of !J are expressed in terms
of 8 and Ricg resp. Rg as follows:
and
B3 . Rg = (Rg + 6 . Og)O.
If B = -lj; with lj; being everywhere negative the formula just above makes sense
for lj; as well. Therefore
(30)
holds for each nowhere vanishing ~), provided !J = lj;2 . g.
All these formulae are rather well known. However, they provide us with the basis
to verify the calculations in a convenient manner.
34
REFERENCES
35
MULTICHANNEL DYNAMIC SYMMETRY
J. Cseh
INTRODUCTION
In such a case the eigenvalue problem of the energy still has an analytical solution, the
labels of irreducible representations are still good quantum numbers, but the degeneracy
corresponding to an exact G dynamic symmetry splits up. The Hamiltonian with this
kind of dynamic symmetry contains more general interactions than the one with an
exact G symmetry, however, the symmetry-breaking terms have to be very special,
expressed in terms of the invariant operators of the subalgebras.
The dynamic symmetry (in the more general sense) turned out to be applicable in
hadron, nuclear, atomic and molecular physics. Its usefullness is manyfold. i) It can
(easily) be applied for the description of experimental data, and even if it does not
hold exactly, it can be a useful reference for comparison and classification. ii) It can
connect descriptions of different methods, like algebraic and geometric, or microscopic
and phenomenologic. iii) It helps to reveal the relations of different models, associated
with different physical pictures.
Both the collective and the cluster models were connected to the shell model in
1958.
Elliott showed that a rotational band of a nucleus consists of a set of shell model
states with the same SU(3) symmetry character [4]. The SU(3) appears as a surviver
of the oscillator symmetry group due to the fact that the shell model potential for light
nuclei can be approximated with an oscillator potential.
At the same time Wildermuth and Kanellopoulos found an important transforma-
tion of the Hamiltonian of the many-body harmonic oscillator problem [5]. In this
approximation the Hamiltonians of the shell and cluster models can be rewritten into
each other exactly. As a consequence, the energy spectra of the two models are the
same, and their basis states are also connected to each other. Kg. a basis state of
the cluster model can be expressed as a linear combination of those shell model basis
states, that belong to the same energy eigenvalue. Thus the cluster states have well-
defined relation to the shell model states, and through this connection to the quadrupole
collective states as well.
There is, however, an essential difference between the two ways, how the collective
and cluster models are connected to the shell model. In the first case the relation
is established for a realistic interaction, while for the second one only for harmonic
oscillator forces.
The main part of the Hamiltonian of the Elliott model is:
H = HHO + xQ . Q, (2)
i.e. in addition to the harmonic oscillator potential quadrupole-quadrupole effective
two-nucleon-forces are applied. Recalling that the (algebraic) Q operator is related to
38
the second-order Casimir operator of SU(3) C~~3 by
3
2Q Q + 4L. L,
(2) _
CSU3 - (3)
(4)
1 C(2) 3
H-
_ C(l)
U3 + 2"X SU3 - SX
C(2)
S03' (5)
This is a general dynamic symmetry in the sense of (1); without the residual interaction
the Hamiltonian would be unrealistic, e.g. one would get degenerate states within one
rotational band. But the SU(3) quantum numbers are not destroyed by the additional
interaction, the relation between the wavefunctions of the shell and collective models is
valid for this realistic Hamiltonian as well.
On the other hand, in the case of the cluster model no realistic interactions were
known that can do the same job.
Recently we have applied interactions corresponding to the U(3) dynamic symme-
try in the framework of the Semi microscopic Algebraic Cluster Model (SACM). These
interactions leave the harmonic oscillator connection$ between the cluster model and
the other two models unchanged: their energy spectra are the same, and the relations
of their wavefunctions are simple. The remarkable thing is that this kind of interaction
proved to be realistic in several applications.
ug (4) 181 Uc, (3) 181 ug (4) 181 UC 2 (3) 181 ... 181 ug~( 4) 181 UCK (3)
I8iUR,(4) 181 UR2(4) 181 ... 181 UR K _, (4)
:J Ug T (4) 181 Uc (3) 181 UR (3) :J Ug(2) 181 U(3)
:J Ug(2) 181 0(3) :J U(2) :J 0(2). (6)
The basis states of the model are characterised by the labels of irreducible representa-
tions of these groups.
Model space
A major point in constructing the model space is related to the exclusion of the
Pauli-forbidden states and the spurious excitations of the center of mass. One can follow
39
different routes in order to exclude the forbidden states. A simple procedure is to make
an intersection between the cluster model basis corresponding to group (6) and that of
the (fully antisymmetric) 5U(3) shell model of the whole nucleus. This procedure is
based on the equivalence of the Hamiltonians of the shell model and the cluster model,
as discussed in more details in Subsection 3.3. The SU(3) basis obtained in this way is
free from the Pauli-forbidden states (as well as from the spurious excitations), and it
describes the cluster configuration, due to its construction.
The fully microscopic construction of the model space provides us not only with the
allowed set of basis states, but also with the cluster spectroscopic factors [11]. However,
these calculations can be very complicated, for several cluster problems prohibitively
involved.
An intermediate level of complexity is represented by the techniques similar to that
of the usual coefficients oHractional parantages [12]. This method is able to produce, in
addition to the allowed basis states, the relative values of the (microscopically obtained)
cluster spectroscopic factors.
Physical operators
The physical operators of the model are expressed in terms of the generators of
group chain (6) with parameters to be fitted to the experimental data. Depending
on the number of clusters, and on the open or closed-shell nature of them the group
structure can undergo considerable simplification. The simplest case is that of a system
consisting of two closed-shell clusters. For such a problem the physical operators are
obtained in terms of a single UR( 4) group of the relative motion (simple Vibron Model).
In this case the dynamic symmet.ry is defined by the group-chain:
The relations of the labels of irreducible representations are: n" = N, N - 1, ... 0 ; and
L = n", n" - 2, ... 1 or 0 , while their physical contents are as follows: L is the angular
momentum of a state, n" is the number of oscillator quata in the relative motion, and
N is its upper limit. When considering only linear and quadratic operators in the
Hamiltonian it reads:
(8)
(9)
This simple energy formula is able to reproduce the spectrum of the well-established
160 + a cluster bands of 20 N e [13, 14]. The "YR value is determined from the oscillator
parameter systematics: 45A-t - 25A-~ = 13.185 MeV. The other parameters are
obtained from a least square fit (also in MeV): E = -63.998, DR = -0.4641, (3 =
0.1563.
When there are more than two clusters,or they have open sheH structures, then
other degrees of freedom couple to the relative motion. Hamiltonians having a U(3)
dynamic symmetry can always be constructed in a simple way. For several cases they
are discussed in detail in ref. [7].
40
Relation to other models and their interactions
The comparison of the (5) and (8) Hamiltonians shows that the shell model inter-
action and the one corresponding to the dynamic symmetry of the cluster model are
very similar. The only difference is that while the Q . Q interaction fixes the relative
strength of the Casimir operators of the U(3) and 0(3) groups, they are independent in
the Hamiltonian of the dynamic symmetry. In other words we can say that in addition
to the quadrupole force of the shell model, a further L . L interaction is involved in
the algebraic cluster model. The relation between the basis states of the two models is
valid for the more general interactions of Eq. (8), too. In this respect the major point
is that the basis states belonging to different irreducible representations of SU(3) are
orthogonal to each other. Therefore, in the expansion of a basis state of the cluster
model, only those basis states of the shell model can occur, that have the same SU(3)
symmetry.
In the fully microscopic description of the cluster configurations specific effective
two-nucleon forces are applied. Our phenomenologic algebraic Hamiltonians can be
compared to them in a simple way. The set of SU(3) basis state is the same in the two
approaches, therefore one can calculate the matrix elements of different interactions
between the same basis states, and equate them. This has been done for the 16 0 + 0'
system in ref. [13]. It turns out that the general features of the effective two-nucleon
forces are similar to those of the phenomenologic one that gives the best fit to the exper-
imental data, however, the corresponding E, '/R, 8R, /3 values (of Eq. (9)) are somewhat
different.
A further interesting question is: how the intercluster algebraic interaction of Eq.
(9) is related to potentials. This kind of mapping can be carried out, like in case of
other algebraic models. with the help of the coherent states techniques [15]. It results
in a shifted anharmonic potential for the 16 0 + 0' interaction:
For the E2 and El transitions of the 18 0 nucleus a detailed analysis was carried
out in ref. [16] in terms of core-plus-alpha-particle configuration. .
A large number of states of 38Ar have been described in terms of core-plus-alpha-
particle model [17] too, including new experimental data as well.
Low-lying states 24Mg and 56 Ni are described together with high-lying molecular
resonances of the 12C + 12C and 28Si + 28Si systems in refs. [18, 19].
41
Based on the pseudo 5U(3) scheme [20] the question was investigated, whether the
210 Pb+ 14C cluster configuration is present in the ground-state wave function of the
224Ra nucleus [21]. The answer turns out to be affirmative, and in addition the low-lying
bands could be described in terms of this cluster configuration.
When harmonic oscillator interactions are applied, not only the cluster Hamiltonian
can be obtained from that of the shell model
Realistic interactions
The U (3) dynamic symmetry of the harmonic oscillator is a simple one (or an exact
one), as discussed in the Introduction. The question is if the requirements i)-iv) are
fulfilled also for more general interactions. The U(3) dynamic symmetry of the SACM
provides us with more general interactions, but, as it has been treated so far, it has
a single-channel nature. From this viewpoint the question is if it can be extended to
a multichannel one in the sense of i)-iv). The answer is affirmative: Hamiltonians of
this kind can be constructed [22]. It is not a priori evident that the (phenomenological
42
intercluster and coupling) interactions of different configurations are uniquely related
to one another, but one can put this extra constraint (most easily in terms of the
eigenvalue-equations) in order to determine an interaction with multichannel character.
Let us consider two different cluster configurations of the same kind: those consisting
of one open and one closed-shell clusters. An example is the 24Mg+ 4H e and 12C + 16 0
configurations of the 28 Si nucleus. They can be described by a model with the group
structure of:
Uc(3) (9 UR(4) :::> Uc(3) (9 UR(3) :::> U(3):::> SU(3) :::> 0(3) :::> 0(2)
l[nf,nf,n],[N,O,O,O], [n".,O,O,], [nI,n2,n3],(A,Il),K ,L ,M). (14)
where t, ,",(, ;3 and B are parameters to be fitted to the experimental data, while F( A, Il, L)
is a function of the indicated quantum numbers. The n" quantum number depends
on the fragmentation: n"il i = 1,2, while the A, Il and L quantum numbers do not.
E and its parameters also carry the channel index i. If, however, several states are
common in the two configurations, then their energies have to be the same, according
to the requirement of the multichannel symmetry. This circumstance establishes a
straightforward transformation between the phenomenological interactions in Eq. (15).
Let the relation of the relative motion quantum numbers be n", = n"2 + n"o. Then the
The consequences of this symmetry can be interesting both with respect of studying
special nuclear states, and from the viewpoint of treating complex spectra.
For specific nuclear states we obtain a selection rule for the various cluster configu-
rations, that are directly related to the channels of cold fission and fusion.
In ref. [23) the link between the superdeformed and cluster states of alpha-like
nuclei (N = Z = even) was explored. It turns out, that in most cases several different
clusterizations have the same wavefunction and in addition, the same cluster structure
appears at different excitation energies of the same nucleus, depending on the relative
orientations of the deformed clusters.
In ref. [24) the allowed and forbidden binary fission modes of ground-state-like con-
figurations in sd shell nuclei were determined.
For the studies of complex spectra the consequence of the multichannel dynamic
symmetry is that it gives a strong correlation between the distributions of different
43
cluster configurations at low and high energies[22]. The high-lying cluster states are
usually populated as resonances in heavy-ion reactions. Their analysis in terms of mul-
tichannel dynamical symmetry involves much more constraints, than in terms of the
phenomenological models (algebraic or geometric), therefore we have much less ambi-
guities in this description than usual. E.g. for some cases parameter-free predictions
can be given for the energy spectrum in an interval determined by the experimental
circums t ances.
The 24Mg+ 4He and 12C+ 16 0 configurations of the 28Si nucleus have been studied
in ref. [22].
The relation of different clusterisations of the 24 M 9 nucleus is investigated in ref. [25].
The [26] work deals with the effects of the microscopic structure on the cold binary
fission channels of the 2 52 C f nucleus. some of which have recently been observed [27].
The logical structure of the multichannel dynamic symmetry described in the pre-
vious section resembles somewhat to that of the dynamic super symmetry (SUSY) in
nuclear physics [28]. Both symmetries have a composite nature in the following sense.
One applies a model consisting of two sectors, each of them having a (usual) dynam-
ic symmetry, and some symmetry transformation connects them with each other. In
the SUSY case there are the bosonic and the fermionic parts of the model, equipped
with dynamic symmetries, and the supertransformations take these degrees of free-
dom into each other. Here we have two different cluster-configurations, and the trans-
formations between the two sets of Jacobi coordinates take these configurations into
each other.
The mathematical description of the multichannel dynamic symmetry has not yet
been completely explored. In this respect the way followed in [29, 30] may be useful.
CONCLUSIONS
My intention with this contribution was to show that the concept of dynamic sym-
metry can be fruitfully applied in nuclear cluster studies, both for the description of
experimental data and in building up the relation between the cluster, shell and collec-
tive models.
Furthermore, the possibility of a new type of symmetry, called multichannel dynamic
symmetry, seems to exist, which is related to different clusterisations of an atomic
nucleus. The multichannel dynamic symmetry puts a very strong constraints on the
relation of different cluster-cluster (and coupling) interactions. Therefore, a unified
treatment can be given for phenomena related to different cluster configurations, e.g.
low-energy clusterization, and high-lying molecular resonances. This may decrease the
ambiguity of the description considerably.
ACKNOWLEDGEMENT
This work was supported by the OTKA Grant No. T14321 and by the US-Hungarian
Joint Fund (NO. 345/93).
44
REFERENCES
45
[29] P. Kramer and M. Moshinsky, in: Group Theory and Its Appilication, E. M. Loebl
(ed.), Academic Press, New York, 1968, 339.
[30] P. Kramer, G. John and D. Schenzle, Group Theory and the Interaction of Com-
posite Systems, Vieweg, Braunschweig, 1981.
46
KAZHDAN-LUSZTIG POLYNOMIALS,
SUBSINGULAR VECTORS AND CONDITIONALLY
INVARIANT (q-DEFORMED) EQUATIONS
V. K. Dobrev
INTRODUCTION
Recently, there was a lot of interest in the study and applications in mathemat-
ics and physics of the singular [1 J and subsingular vectors [2J of Verma modules. In
particular, were considered the singular vectors of Verma modules of: the Virasoro
algebra [3-7J the super-Virasoro algebras with N = 1 [8], N = 2 [9,10J Kac-Moody
algebras [11-13J quantum groups [14,15] W-algebras [16, 17J of Fock modules of the
Virasoro algebra [18J. The physical applications are mostly in two-dimensional (su-
per) copformal field theory, topological field theory, Calogero-Sutherland model, etc.
Subsingular vectors figured prominently (though without explicit formulae) in the
BRST analysis of the Fock modules of the Virasoro [19J and 8l(2) Kac-Moody [20J
algebras. Our interest in (sub )singular vectors is motivated by their relation to (con-
ditionally) invariant equations. This relation stated in condensed form is: to every
singular, resp., subsingular vector of a Verma module over a semi-simple (also reduc-
tive) Lie algebra 9 there corresponds a differential operator and equation invariant,
resp., conditionally invariant with respect to g, cf. [21,22J. Both statements are valid
for the corresponding Drinfel'd-Jimbo quantum group Uq (9), cf. [22,23J and also for
the corresponding Lie group (with some additional subtleties [21 J).
Thus, it is important to have explicit formulae for the (sub )singular vectors.
While there is a lot of such formulae for singular vectors in the literature cited above,
there are very few results for subsingular vectors. Furthermore, these are usually
treated via the occurrence of non-trivial Kazhdan-Lusztig polynomials [24J in the
character formulae for semisimple, affine Kac-Moody and W algebras (d. e.g., [25]),
though this relation is not well understood as we shall see in the present article.
SUBSINGULAR VECTORS
Preliminaries
Let 9 be a semisimple Lie algebra with triangular decomposition 9 = g+ EEl H EEl
g-, where H is a Cartan sub algebra of g, g+, resp., g-, are the positive, resp.,
negative root vector spaces of the root system .6. = .6.(9, H), corresponding to the
decomposition.6. = .6.+ u.6. - into positive and negative roots. Thus, g = EEl PEt. 9 P ,
(dim 9P = 1), and we will denote by Xi the element spanning gp, f3 E .6.+. The
elements Xi belong to the Cart an- Weyl basis of g. Let .6. s = {O'i t i = 1, ... ,r =
rank g} be the system of simple roots of .6.. We use the standard deformation Uq (9)
[26,27] given in terms of the Chevalley generators X; == X;;., Hi E H, i = L ... ,r of
9 by the relations:
[Hj, Hd = 0, [Hj, xtJ = ajkXt, [xj, X;J = bjk [HjJqj, (la)
where (ajk) = ((0' j, O'k)) is the Cartan matrix of g, 0' V == 20'/( 0',0'), (-, .) is the scalar
product of the roots normalized so that for the short roots 0' we have (0',0') = 2,
48
qj = q(OIj,OIj)/2, (;;')q = [n]q!/[m]q![n - m]q!, [m]q! = [m]q[m - l]q ... [l]q, [m]q =
(qm/2 _ q-m/2)/(ql/2 _ q-l/2). The elements Hi span the Cartan sub algebra H of g,
while the elements xt generate the sub algebras Uq(Q).
A lowest weight module (LWM) MA over Uq(sl(Q)) is given by the lowest weight
A E H* (H* is the dual of H) and a lowest weight vectorvo so that X Vo = if
X E g-, Hvo = A(H)vo if HE H. In particular, we use the Verma modules VA over
Uq(sl(Q)) which are the lowest weight modules such that VA ~ Uq(g+)Q9vo. Further,
for brevity we shall omit the sign 129, i.e., we shall write instead of Q9Vo just Vo.
Note that the numbers:
(p(Hk) = (p,ak) = 1), completely determine the lowest weight A and shall be used
also for the characterization of the LWM. The collection of these numbers shall be
called the signature of A and denoted X(A) or just x:
(3)
Further we shall use also the following notions. The signature X = (ml' ... , mr)
is called dominant, semi-dominant, anti-dominant,anti-semi-dominant, resp.,
if mk EN, mk E 1+ , mk E -N, mk E 1_ , respectively.
Analogously, we shall also use numbers corresponding to arbitrary positive roots:
(4)
where H 0' E H corresponds to the root a by the isomorphism H ~ 1i*. Certainly, each
m(X is a fixed linear combination of mi, however, these numbers have independent
importance as we shall see just below. Naturally, ma; = mi.
In the present paper we restrict to the case when the deformation parameter q is
not a non-trivial root of 1. In this case a Verma module VA is reducible [2] (q = 1),
iff at least one of the numbers ma is a positive integer [28]:
(5)
Whenever (5) is fulfilled there exists a singular vector v::~ in VA such that v::~ tj
CVo, Xv~n~ = 0, V X E g- and Hv';'~ = (A + maa)(H)v';'~, VHE 1i. The space
Ima = Uq(9+) v~na is a proper submodule of VA isomorphic to the Verma module
VA+maO' with a shifted lowest weight A+maiX [28]. Clearly, this implies that VA and
VA+ma a have the same values of the Casimir operators.
Remark 1: Note that if we were considering highest instead of lowest weights, the
analog of the numbers mk, (m a ), would be defined as m~w. = 1 + A(Hi), (m~w. =
(p + A)(Hcx )). However, the statement about the reducibility is unchanged; only the
shifted weight is A - m~wiX [28]. Moreover, in terms of mk (anti- )dominant weights
have (anti- )dominant signatures.
It is important that one can find explicit formulae for the singular vectors. The
singular vector introduced above is given by [21,28]:
(6)
49
POI,m a is unique up to a non-zero multiplicative constant. References 21 and 28
contain all explicit singular vectors needed in the present paper. Note that we refer
to both, since one gives formulae for q = 1 [21], while the other gives such formulae
for general q [28]. (More general explicit formulae for singular vectors, including all
singular vectors for Uq(sl(n)), are contained in [14]. Note that the modules considered
in refs. 14 and 28 are highest weight modules and the singular vectors are polynomials
in X i-; the translation of those formulae to the lowest weight module setting is
straightforward in view of the above Remark.)
Certainly, (5) may be fulfilled for several positive roots (even for all of them).
Let us denote: fA ~=. Ulma , where the union is over all positive roots fOE which (5) is
fulfilled. Clearly, IA is a proper submodule of VA. Let us also denote FA =. VA / fA.
The Verma module VA contains a unique proper maximal sub module IA (2 fA)
[2,29]. Among the lowest weight modules with lowest weight A there is a unique
irreducible one, denoted by LA, i.e., LA = VA / IA. (If VA is irreducible then LA =
VA.)
It may happen that the maximal submodule [A coincides with the submodule
fA generated by all singular vectors. This is, e.g., the case for all Verma modules if
rank 9 :s: 2, or when (5) is fulfilled for all simple roots (and, as a consequence for all
postitive roots). Here we are interested in the cases when fA is a proper submodule
of IA. We need the following notion.
x Vsu E VX E g- (7)
where QiOl are homogeneous polynomials such that the r.h.s. is a homogeneous poly-
nomial, and 6.i is a subset of liA C 6.+, such that et E 6.i iff QiOl is a non-zero
polynomial. Let us denote by 6. su the union of 6.i: 6.su =. Ui=l6. i. We shall call 6.su
the set of roots associated to the subsingular vector Vsu. The corresponding set of
singular vectors {vOl,rn a let E 6.su } will be called singular vectors associated with the
subsingular vector Vsu. Clearly t:.su is a subset of 6.A and in general a proper subset.
Let Isu =. UOiELlsuIOI(~ fA), Fsu =. VA / Isu ; then Vsu becomes a singular vector in Fsu ,
i.e., when we factorize all singular vectors associated with it.
Clearly, Vsu and fA generate a sub module It;,
so that:
(9)
where all embeddings are proper. Of course, there is no claim that I/:u coincides with
IA.
50
Further we shall use also the following notions. The singular vector VI is called
descendent of the singular vector Vz ~ CVl if there exists a homogeneous polynomial
P12 in xt so that VI = P 1Z V2. Clearly, in this case we have: [1 C [Z , where [k is the
submodule generated by Vk. If a singular vector v~ is descendant of another singular
vector then v~ is called a composite singular vector. Clearly, if two singular vectors
VI and V2 belong to .6. A (.6.;, .6. su ) and VI is descendent of V2 , then we can omit VI
from the set .6.A (.6.; , .6. su ).
We restrict now to g = sl(4). The simple roots are 0:1,0:2,0:3 with non-zero
scalar products: (O:j,O:j) = 2, j = 1,2,3, (0:1,0:2) = (0:2,0:3) = -1. The Chevalley
generators are denoted by X; , i = 1,2,3. The non-simple roots are: 0:12 = 0:1 + 0:2,
0:23 = 0:2 + 0:3, 0:13 = 0:1 + 0:2 + 0:3 , and the corresponding Cartan-Weyl generators
are given by [27,28,30]:
Xik =q":f1 / 4(ql/4XjX;_q-l/4X;Xf), (jk)=(12),(23), (lOa)
X~ = q":fl/4 (l/4Xt X~ - q-l/4 x~xt) =
(lOb)
= q":fl/4(ql/4X~Xi' _q-l/4Xi'X~) .
All commutation relations between the generators which we shall use follow from
these definitions and (1).
For the six positive roots of the root system of sl(4) one has from (2) and (4)
(see ref. 21):
ml=l-A(Hl) (lla)
mz=1-A(H2) (llb)
m3 = 1- A(H3) (llc)
m12 = 2 - A(HIZ) = m} + m2 (lld)
mZ3 = 2 - A(Iiz3) = mz + m3 (lle)
m13 = 3 - A(H13) = ml + mz + m3 (llf)
Thus the signature here is: X = (m},mZ,m3).
For further reference we give the value of the sl( 4) second order Casimir operator
[31] in terms of the above notation:
Cz = Xr{mi3 + m~ + h(mi - ma)2) - 5 (12)
which is normalized to take zero value on the trivial irrep (mk = 1) (and thus on all
representations partially equivalent to it).
(13a)
51
However, only the singular vector V2 is relevant, the others are descendants of V2 .
Indeed, we have the following embed dings:
n n (14)
iA C IA
where h, Ii j , are the submodules generated by V2 , v:j' resp. All embeddings are
proper, including the last one since there is the following subsingular vector:
It is easy to see:
X[Vbgg V~3
X;Vbgg 0
X;Vbgg -V12 I
thus, indeed, (7) is fulfilled. Formula (15a) is in the unordered Chevalley basis. An
expression in the cordered PBW basis is:
(15b)
which for q = 1 is exactly equal to (15a) and for q -I 1 differs from (15a) by the
inessential term (ql/2 - q-l/2)XtX: x;tvo E lA, For q = 1 a third expression
coinciding with (15a, b) is:
Note that we have translated the result of ref. 2 into our lowest weight module setting
and that the actual expression for Vbgg in ref. 2, (given naturally for q = 1), is not
correct. (Also formulae (13b) and (14) are not given in ref. 2.) Clearly, Vbgg becomes
a singular vector in the factor-module F2 = VA / h. Let 12) denote the lowest weight
vector of F 2 . Then the singular vectors in (13) become null-conditions, the relevant
one (13a) giving:
(16)
If we factor out also Vbgg we have the following null-conditions in the resulting irre-
ducible module with lowest weight vector I 2 ):
x;tl 2) = 0 (17a)
(xtx;tX: -X:x;tXt)12) = 0 (17b)
52
The Other Archetypal 81(4) Example
Vf = P Va , (18)
P -- X+X+
13 2 - q-1/2X+X+
12 23 (19)
P = [q-1/~
a-I q
(Xi xi - [2]qxi xi) ([a - l]qXt xi- [aJqXi xi) +
+ [a -1]
I q
xi ([a - l]qxi xi - [a]qXi xi) xt (22b)
The need for the introduction of these special forms will become clear below.
Consider now a Verma module VA with lowest weight A satisfying the condi-
tions:
A(H3) = 0 { = } m3 = 1 (23a)
A(H1 +H2) = 1 { = } m12 = 1 (23b)
X1(a)=x(A)=(a,1-a,1), aEC (23c)
cf. (l1c, d). From these conditions follow that there are two singular vectors which
are explicitly given by [28]:
V3 = xi Va, m3 = 1 (24a)
V12 = ([a -l]qXt xi - [a]qXi xi) Va, m12 = 1 (24b)
53
There is also a singular vector corresponding to m13 = 2 [21,28], which, however, is
a composite one:
xi 11) = 0, m3 = 1 (26a)
([a -1]qXt xi - [a]qXixt) 11) = 0, ml2 = 1 (26b)
For the Verma module under consideration the vector (18) is a subsingular vector
iff a = 1. Indeed, one can check (using also (20a)) that for any a:
X;-Vj = 0
XzVj = q-l ([a + l]qXi xt - [a]qXt Xi) V3 + q-l xivl2 (27)
X3'Vj = 0
Vj = [q-l/~
a-I q
(xi xi - [2]qX; xi) V12 +
(28)
+ -[a-I
_1_] xi
q
([a - l]qXt xi - [a]qX; xi) V3
If a rJ. Z there are no other non-descendent singular vectors, besides (24) and the
maximal invariant submodule is: IA = I~ = [C'3 U Iet12. We denote by L; = V A/ I~
the corresponding irreducible factor-module, and by I I' ) the lowest weight vector of
L;. Then the expressions in (24) become null-conditions, namely:
xiiI' ) = 0 (29a)
([a -lJqXt xi - [aJqXi xt) II' ) = 0 (29b)
(30)
54
and two descendents corresponding to m23 = 2 - a, mI3 = 2. Thus the maximal
invariant submodule is: II. = I~' = r a u r l2 U l Ci 2 , L~ = VA / I~' is the irreducible
factor-module, I I" ) is the lowest weight vector of L~. Then the null-conditions are:
Xill")=O (31a)
([a -1]qXt Xi - [a]qXi xt) II" ) = 0 (31b)
(33)
Thus the maximal invariant submodule is: II. = I[V = lCia U l Ci 12 U l Ci 1 , Lfv =
VA/I[V is the irreducible factor-module, 11 lV ) is the lowest weight vector of LF.
Then the null-conditions are:
xill lV ) = 0 (34a)
([a -1]qXt xi - [a]qXi Xt)II lV ) = 0 (34b)
(xtr 11 lV ) = 0 (34c)
(35)
From the above proof it remained to show that vf can not be represented as a linear
combination of descendents of VI and V3, and thus does not belong to iI., which is
easy to see also by inspecting (21).
We denote by iA = IO'l U lCiS the submodule generated by these singular vectors,
by FI = VA / iA the factor-module, by II) the lowest weight vector of Fl. We have
the following null-conditions in Ft:
xiiI) = 0 (36a)
xtlI) = 0 (36b)
vj,=Pll) (37)
55
Factoring out the sub module built on Vh we obtain the the irreducible factor-module
L1 = VA /I{\. We denote by I 1 ) the lowest weight vector of L 1. Then the null-
conditions are:
=0
xi 11) (38a)
=0
xt 11) (38b)
(Xi xt - [2]qXt Xi) xt xt I 1 ) = 0 (38c)
where for (38c) we have used (3Sa) and (21d). An equivalent condition to (3Sc) is:
VI = xt Vo, m1 =1 (40a)
V23 = ([a - l]qXi xt - [a]qxt xi) Vo, m23 = 1 (40b)
xt 13) = 0, m1 = 1 (41a)
(ra -1]qXtXt - [a]qXt Xi) 13) = 0, m23 =1 (41b)
For the Verma module under consideration the vector (18) is a subsingular vector iff
a = 1. Indeed, one can check (using also (20b)) that for any a:
X;-vf = 0
X2-vf = q-l ([a + l]qXt xt - [a]qXt Xi) VI + q-1 Xtv23 (42)
X;;vf = 0
However, if a =f. 1 then vf itself belongs to the maximal invariant submodule fA
generated by the singular vectors, and thus is not a subsingular vector by definition.
Explicitly we note that if a =f. 1 then using (22b) we have:
-1/2
Vf = [q ] (xt xt - [2]qXt Xi) V23
a-I q
+
(43)
+ -[a-I
_1_] xt ([a -1]qXi xt -
q
[a]qXt Xi) VI
56
Note that a lowest weight can satisfy both (23) and (39) which happens exactly
for the special case a = A(H2) = 1 and A(H1) = A(H3) = 0. In this case there are
four singular vector given in (24) and (40). However, V12, (d. (24b)), becomes a
descendent of the singular vector V1, (d. (40a)), while V23, (d. (40b)), becomes a
descendent of the singular vector V3, (d. (24a)). Then in the Verma module will
have both intersecting submodules I1 and I3 :
Proposition 1: Let the Verma moduie VA satisfy at least one of the two sets of
conditions (23), (39), i.e., it has signature X(A) = X1(a) = (a,l - a, 1) or X(A) =
X3(a) = (1,1 - a, a). Then there exists a subsingular vector given by (18) iff a =
A(H2) = 1, i.e., X(A) = X1(1) = X3(1) = (1,0,1). It becomes a singular vector for
the factor-module FA = VA /(I1 U I3). <)
Conditions (34) and (38) ((38c, c' ) in a different, but equivalent form) were given
first in [32]. The corresponding irreps (for a E N) were shown [32] to be a construction
of the irreducible massless representations of a q-conformal algebra (with jqj = 1)
characterized by the helicity h = (a -1)/2 E (1/2)Z+.
KAZHDAN-LUSZTIG POLYNOMIALS
57
where 1!( x), x E W is the length function on W, Ex == (-1 )f(x) is the sign function on
W.
Suppose now that the Py,w' are known for all y S; w' with 1!(w') < 1!0. Let
1!(w) = 1!0 and write w = sv with s E S, (thus 1!(v) = 1!0). Then Py,w may be
determined recursively from the following relation ((2.2.c) of ref. 24):
if sy <y (48)
c= {~ if sy >y
ux == uf(x) , x EW
Py,w(O) = I, Vy S; w (52)
For this note that for u = 0 only one of the first two terms in the r.h.s. of (48) survives
(depending on the value of c), while the summation term vanishes since the leading
power is positive (as Z < w). Thus (48) is reduced to a simple recursive relation which
is used repeatedly until (52) is achieved.
An important property of the KLP which may be proved also using recursively
(48) is [24]:
deg Py,w ~ H1!(w) - 1!(y) - 1), y < w (53)
58
Thus -< is related to the polynomials of maximal possible degree. (Note that in the
exposition of ref. 24 this, together with (49), is one of the defining conditions.)
For illustration let us consider also the example w = sys', (as reduced expres-
sions) with s,s' E S, thus, few) = fey) + 2. Then in (48) c = 0 since sy > yand
the summation is empty, since the only possible z = y does not fulfil sz < z. Then
the first term in (48) Psy,ys' is zero since sy is not comparable with ys'. Thus, only
the second term contributes and one has Py,sys' = Py,ys" which is equal to 1 because
of (50). Analogously one may consider the other possibilities for lengths differing by
two and obtain [24]:
Py,w = 1, if C(w)=C(y)+2 (54)
Let us mention also the following relation (cf. (2.3.g) of ref. 24):
From this follows (cf. Lemma 2.6.vi of ref. 24) if W is finite with longest element Wo:
There is a conjecture [24] that all coefficients in the KLP are non-negative inte-
gers. This conjecture is already proved in many important cases, e.g., when W is a
Weyl and affine Weyl group [33]; for more information cf. [34].
Note that if W is a dihedral group with S = (8,t) then Pv,w = 1 for all y ::::; w
[24,34].
The first non-trivial case is the Weyl group W of A3 = sl(4). Denote S =
(SI,82,83) with: 8i = 8~ = 8~ = e, (SI82)3 = e, (82S3)3 = e, 81S3 = 83S1. There are
only two cases in which y -< wand degPy,w > 0, namely [24J:
PS2,S281S3S2 = 1+ U (57a)
PSlS3,81S382"1"3 = 1+u (57b)
There are also several cases in which y < wand deg Py,w > 0, however, y -f, w,
namely:
Pe ,S2 S 1 S 3 S 2 = 1 + u (58a)
Pt ,S,S3 S ,S,S3 = 1 + u, t = e, 81,.53 (58b)
Note that (58a) may be obtained from (57a) using (55) with y = e, 8 = 82, W =
In the same way (58b) may be obtained from (57b) using (55) first with
82.51.53.52'
y = e, .5 = .51,.53, W = .518382.51.53, and then with y = 81,83, 8 = .53,81, resp.,
w = 81.53.528183.
59
(Note that Vo = CVo.) Following [1] let E(1{*) be the associative abelian algebra
consisting of the series L:p,E1{* c"e({t) ,where cp, E C,cp, = 0 for {t outside the union
of a finite number of sets of the form D(A) = {{t E 1{*\{t :::: A}, using some ordering
of 1{*, e.g., the lexicographic one; the formal exponentR e( {t) have the properties:
e(O) = 1, e({t)e(v) = e({t + 1/).
Then one defines the character of V by:
(61)
(62)
The vVeyl group W is generated by the simple reflections Si == SO'i' ai E t:J.. s . Thus
every element W E W can be written as the product of simple reflections. It is said
that w is written in a reduced form if it is written with the minimal possible number
of simple reflections; the number of reflections of a reduced form of w is called the
length of w, denoted by( w ).
The Weyl character formula for the finite-dimensional irreducible LWM over 9,
i.e., when A E - f +, has the form [1]:
ch LA = L (_l)fCw) ch v w .A , A E - f+ (63)
wEW
(64)
where WA is a unique element of W with minimal length such that the signature of
A = W A1 . A is anti-semi-dominant:
1 - A(ET() E Z_ (66)
60
Note that Py,w(l) E N for y :::: w. Formula (65) is valid both for q = 1 [24], and for
generic q (not a root of 1) [36]. (If q is a root of 1 then the formula involves the affine
Weyl group which. is the untwisted affinization of W [36].)
Note that (63) follows simply from (65). Indeed, the signature X(A) of the finite-
dimensionalirreducibleLWM LA is dominant: mk = l-A(Hk) E N, (since A E -r +),
and hence w A = Wo , where Wo is the maximal length element of W, since the signature
of w- 1 . A is anti-dominant only for w = woo Then one notes Pw,wo = 1 and w :::: Wo
Vw E W. Thus the summation range coincides and it remains to make the change of
summation w = w'wo and to note that f(wow'wo) = f(w').
where 51,52,53 arc the simple reflections of W = W(51(4)) with 5183 = 5351, and e
is the unit clement of W.
Thus, there are only two kinds of LWM LA whose character formula contain these
non-trivial KLP. We shall parametrize these LWM using the parametrization of the
LWM Ao with semi-dominant signatures XO = (ml,m2,m3), mk E Z+. Namely, for
each such signature there would be at most one of those two kinds of LWM. Before
making the precise statement we denote by Ao the lowest weight anti-semi-dominant
signature Xo = (-m}, -m2, -m3), which for fixed mk is equal to Wo . Ao , where Wo is
the maximal length clement of VV with a reduced expression, e.g., Wo = 8153825}5352.
Then we introduce notation for the two special elements of W, which appear in
(67):
,
Wo = 52 5 1 5 3 8 2 (68a)
Wo" = 51 5 3 5 2 5 15 3 (68b)
N ow we can state:
n n (70a)
61
VAo C VA'
n (70b)
VAil VAo
C
VAo C VA'
n n (70c)
VAil VAo
ii) The elements W AI, WAil of minimal length which correspond to A', A" in (65) are
r
given as follows:
Proof: i) Consider first the case of dominant signature XO = (m1' m2, m3), mk E N.
The embedding picture (70a) is shown as follows. We recall that whenever (5) is
fulfilled for some G' E ~ + the Verma module VA is reducible and contains as a
submodule the Verma module VA+maO! = Va A , cf. also (64). Now we can check
that:
This shows the embeddings w.r.t. Ao , (though omitting the intermediate embeddings
evident in (72. For the embeddings w.r.t. Ao it remains to note:
To show the embeddings in (70) in the cases when some of the parameters mk are
zero one has to observe when A' or A" coincides with Ao. The coincidences are clearly
seen from (73) and so we easily obtain the rest of (70). Finally, we note that Ao may
coincide with either of A' or A" only if all mk = 0 (cf. the signatures), which we have
excluded.
ii) Consider again the case of dominant signature XO = (m!, m2, m3), mk E N. We
need to know the signatures of W to which we denote by X(w Ao) for the elements
W which occur in the statement of the Lemma. We have:
62
X(SZSlS3 1\0) = (-m12' m13, -mZ3) (74a)
X(SZS3 1\0) = (-m13' m12, -mz) (74b)
X(SZSl 1\0) = (-mZ, m23, -m13) (74c)
X(SlS3SZS3 1\0) = (m13, -mZ3, mz) (75a)
X(SlS3S2S1 1\0) = (mz, -m12, m13) (75b)
X(SlS3SZ 1\0) = (mZ3, -m13, m12) (75c)
X(SlS2S3 1\0) = (m13' -m3, -m2) (75d)
X(S3S2S1' 1\0) = (-mZ,-ml,m13) (75e)
Clearly, X' and X" do not coincide with any of these signatures when all mk > 0.
Next we consider the cases when some of the parameters mk are zero. From (74),
(75) we obtain first that X', X", resp., does not coincide with any of these signatures
when m2 > 0, ml, m3 > 0, respectively. Further from (74) and (75) we obtain:
= X' ,
1\0) m3 > 0, ml = mz = (76c)
X(SZSl
= X",
1\0) ml = 0, m2, m3 > (77a)
X(SlS3 S ZS 3'
From (76) and (77) statement ii) of the Lemma follows by just noting the lengths of
the elements, which are 3,2,2 in (76a,b,c), resp., while (wb) = 4, and 4,4,3,3,3 in
(77a, b,c,d,e), resp., while (w~) = 5 .
Using the above result we shall give explicitly all character formulae in the case
Uq (sl(4)) which involve non-trivial KLP, i.e., ch LA', ch LA" , resp., when mz > 0,
ml, m3 > 0, respectively. For the simplification of the character formulae we use
notation for the formal exponents corresponding to the simple roots: tj = e(aj),
j = 1,2,3; then for the three non-simple roots an = al + az, a23 = a2 + a3, a13 =
al + az + a3 we have: t12 = e(a12) = t 1t 2, tZ3 = e(a23) = t2h, t 13 = e(a13) = t 1t 2i 3.
In terms of these the character formula for the Verma module is:
Proposition 2: The character formulae of the irreducible LWM LA', LA" with
signatures in (69) are explicitly given as follows:
63
+ t~'t~3 + t;"'3t~3 + t~'t-:"3 + t;"12t-:'23 +
+ t;"12t'.{" + t'.{'3t';'23 - 2 t;"12t'.{"3{~'23 _
(SO)
_ t~'t'.{'3t;"3 _ t;"12t'.{'lt;"3 - t;"'8t'.{'8t;'28 _ t;"13t'.{"t;;3 +
+ 2 tm,2tm,3
1 23
+ 2. tm,3tm28
12 3
+ .tm13tm,+m3tm13
1 2 3
m,3 )
_ 2 t 13
ml, m3 EN, m2 E Z+ .
Proof: Let first mk > 0, k = 1,2,3. The proof amounts to the correct identification
of the Verma modules which are submodules of VA' , VAil. In particular, the Ver-
ma modules which contribute two times because of (67) have the following weights
(expressed in terms of Ao and t from (67)):
A~ = t . AD , t = e, 82 (Sla)
A~' = t Ao, t = e, 81, 83, 8183 (SIb)
i.e., A~ = A~ = Ao. Their weight differences (which enter the character formulae)
are:
A~2 - A' = m23al + m13a2 + m12 a 3 (S2a)
Ao - A' = m23a12 + m12a23 (S2b)
= m12 a l + m13 a 2 + m23 a 3
A~'S3 - A" (S3a)
= m12al + m13a23
A~, - A" (S3b)
A~3 - A" = m13 a 12. + m23 a 3 (S3c)
Ao - A" = m13a13 (S3d)
where a12 = al + a2, a23 = a2 + a3, al3 = al + a2 + a3. The extension to the cases
when some mk are zero is done as in the Lemma.
For future reference we record the signatures ofthe lowest weights in (Sl) besides
Ao (in evident notation):
X~2 = (-m23, m2, -mI2) (S4a)
X~, = (m3, -m2a, -ml) (S4b)
X~3 = (-m3, -mI2, ml) (S4c)
X~'S3 = (ma,-mI3,ml) (S4d)
Note that the character formulae (79) and (SO) are simplified when some mk are
zero (as specified). In particular, (79) has six terms when mlm3 = O,.ml +ma > 0,
and four terms when ml = ma = 0, while (SO) has eight terms when m2 = 0:
ch LA' -- ch VA' .( 1 - t 2m23 _ t Hm2 + tm23tm2
2 a -
(S5a)
(S5b)
- tm2tm,2
1 23 + t 12m2 t 23m'2) , m3 =0
ch LA' = ch VA' ( 1 - t'.{'2) ( 1 - t;';2), ml = m3 = 0 (S5c)
ch LA"= ch VAil ( (1 - t;"') (1- t;'3) -
(S5d)
- t 12 m3 (1 - t 1m3) (1 - tml)
m't 23 3)' m2 = 0
64
We record also the factorization of (85d) for m1 = m3:
(85e)
m2=0, m1=m3=m, xll=(m,O,m)
to the same embedding schemes (70). We also stress that unless all mk > there
VA" for fixed mk have the same values of their Casimir operators, since they belong
Corollary 1: The Verma modules VA', VA" for fixed mk have the following maximal
common submodule generated by singular vectors if m2 > and m1 + m3 > 0:
where 5jk == 5 ajk , and we have also recalled that A~, A~ have appeared already in
(75c), (74a), respectively_
Proof: From the signature of VA- it follows that its maximal submodule generated
by singular vectors is:
(87)
It remains to notice that none of the four Verma modules VA~ is a submodule of the
other three, and that all of them are proper submodules of both VA' and VA" _ This
is ensured by the restrictions on the parameters mk _
65
Remark 4: The contributions of the Verma modules VA~ in the character formulae
of LA' and LA" are the terms (with minus sign) on the first line of (79) and the terms
on the second line of (80).
One may wonder about the character formulae of LA' and LA" m the excluded
cases m2 = 0 and mlm3 = 0, respectively. Certainly, formulae (79) and (80) are not
valid in these cases, and the formal substitution of m2 = 0 in (79) and of ml = a or
m3 = 0 in (80) gives zero. We shall give these formulae only for LA" since we shall
need them below. We first introduce notation for the lowest weights with signatures
which appeared already in (75a, b):
n n (89a)
where A~ was given in (75c) and (86). The character formulae are:
(90a)
(90b)
66
ch LA; = ch VA; ( 1 - t';'23 - t~2 - t';'12 +
+ tm23tm12
1 3
+ tm23tm2
1 23
+ tm2tm12
12 3
_ tm23tm2tm,2
1 2 3
) (90c)
(9Ia)
(9Ib)
(9Ic)
valid when m2 E N, m1 = m3 = O. <>
The folklore says that a subsingular vector of the Verma module VA is expected
whenever Pw,wA(I) > 1 for some w. It is clear from (65) that such a subsingular
vector should have the following weight:
(92)
Here we shall demonstrate that subsingular vectors may not appear in the above
situation. We begin by describing the standard situation in detail. We use again the
Uq (sl(4 case.
From the character formulae (79) and,,(80) follows that the subsingular vectors
are vectors in the Verma modules VA ,VA with weights equal to the weight differ-
ences given in the proof of Proposition 2, (82), (o3), and which are valid under the
restrictions there, namely, m2 > 0 for (82), m1m3 > 0 for (83). Thus the simplest
subsingular vectors are expected to have the weights:
67
m2 = 1, m1 = m3 = , I
X = XO = (0,1,0) (93c)
A"81 8 3 -A" = 0'1 + 20'2 + 0'3 (94a)
A"s, -A"= 0'1 + 20'2 + 20'3 (94b)
A" -A" = 0'1 + 20'2 + 20'3
83
(94c)
Ao -A" = 20'13 (94d)
m2 = 0, m1 = m3 = 1, = (1,0,1)
/I
X = XO (94e)
I
Vbgg = X+
2 Vbgg (95)
A' A'
which is a composition using the singular vector V8 '2 - xivo '2 of the Verma
VA
A'
'2 (with lowest weight vector Vo '2); the latter singular vector gives the
I
module
embedding V A :2 ::) via. Indeed, substituting (15a) we have:
In the same way one may check that there are no subsingular vectors of weights
(94b, c, d). In particular, there are three descendants corresponding to the embedding
scheme:
VAa A"
C V"
n n (97)
VTA~'3 C V A"'1'3
68
Conjecture: Consider a Verma module VA with A E f. Let w be such that
Pw,wA(l) E N + 1, where WA is the unique element of W with minimal length such
that the signature of A = W;;:1 . A is anti-semi-dominant, cf. (65) and (66). A
subsingular vector exists iff W -< WA . (;
where A E H*, and X is the right action of the generators of the algebra 9
, d
(X'P )(g) == dt 'P (g exp( tX)) It=o (101)
Right covariance is used also to pass from functions on the group G to the so-called
reduced functions <p on the coset space G/B, where B = exp(H)exp(9-) is a Borel
subgroup of G. Note that G/B is a completion of G+ = exp(g+) and in practical
calculations one is usually using the local coordinates of G+ .
The weight A completely characterizes these representations, which we denote by
C A , each of which is then in correspondence with the lowest weight representations
with the same lowest weight, in particular, with the Verma module VA.
Now the main ingredient of the procedure [21] is that to every singular vector
there corresponds an intertwining differential operator. Namely, to the singular vector
Vs = v""m Q (d. (6)) of the Verma module VA there corresponds an intertwining
differential operator
Dc,m O
A' = A + m",O: (102a)
\f X E 9 (102b)
where Xi denotes the infinitesimal left action (99) acting in the space CA. The
operator is given explicitly by:
D ""m a = -n",,1n(X'
r
+
1 j""
X' r+) (103)
69
where POl,nt is the same polynomial as in (6), and jet is the right action (101). This
operator gives rise to the y-invariant equation:
(104)
In the same way a subsingular vector Vsu produces a differential operator Dsu and
equation which are conditionally invariant. The latter means that this invariance hold
only on the intersection of the kernels of all intertwining operators DOl,rn a such that
a and the singular vectors vOl,rna are associated with the singular vector in question
(d. "Preliminaries" in section 2), i.e., on the space:
(105)
where (:A is such that CSli c (:A C C A holds, and we use that C su is y-invariant:
(107)
A conditionally invariant equation has non-trivial r.h.s. if we take the situation cor-
responding to the reducible factor-module FA = VAliA; the latter is realized when
we do not impose in FA the null-condition corresponding to the subsingular vector
which in FA is a singular vector. A conditionally invariant equation has trivial r.h.s.
if we take the situation corresponding to the irreducible factor-module LA = VA I [A,
i.e., if we impose in FA the null-condition corresponding to the subsingular vector.
Below we consider both situations, for which we are prepared by the detailed analysis
of section 2.
Remark 5: Note that one may exchange the left and right actions in the above con-
siderations, i.e .. consider the representations acting as right regular representations
with properties of left covariance. Independently, if one uses highest weight represen-
tations (d. Remark 1) one then uses the coset CIB', where B' = exp(H)exp(9+) is
the Borel subgroup of C conjugate to B. <>
Referring further the general case to ref. 21, we restrict here to y = 31(4), C =
SL( 4). We pass to functions on the flag manifold Y = SL( 4)1 B, where B is the Borel
subgroup of S L( 4) consisting of all upper diagonal matrices. (Equally well one may
take the flag manifold S L( 4) I B', where B' is the Borel subgroup of lower diagonal
matrices.) We denote the six local coordinates on Y by x, v, ii, z, z. From the explicit
form of the singular vectors it is clear that we need only the right action of the three
simple root generators. Denoting this right action of xi
by R k , we have [21]:
a
RI = az == az
R2 = zza+ + zav + zav + a_ (108)
a
R3 = az == az
where
(109)
70
Things are arranged so that in the conformal setting we can use the same co-
ordinates [37]. In this case the coordinates x, v, v are related to the Minkowski
space-time coordinates XO,Xl,X2,X3:
(llO)
We start with the equations arising from the BGG example of subsingular vec-
tor. Substituting (108) in (16) we obtain the following sl( 4) and su(2,2) invariant
equation:
(ll3a)
71
while the subsingular vector Vbgg gives rise to the following conditionally invariant
equation:
where 1j; E C A and satisfies (113a), 1j;'ECA ', A' = A - a13, the corresponding
signatures being X = (0,1,0), X' = (-1,1, -1). (Note that the second Casimir
operator has the same value in the two representations: C2 (X) = C2 (X') = -4, cf.
(12).) If we consider the irreducible factor-module LA, which means that we should
use (17) instead of (16), we have instead of (113b):
(113c)
(114)
which means that our functions do not depend on the variable z - this is valid for
the signature Xl (a) and arbitrary a. (In the conjugate situation with signature X3( a)
our functions do not depend on the variable z.)
Further, we have the equations arising from the singular vector V2, when a E Z_ ,
(null-conditions (31c) and (32b:
(115)
Next, we have the equations arising from the singular vector VI, when a E N,
(null-conditions (34c), (36b), and (38b:
which means that our functions are polynomials in the variable z of degree a - 1.
Thus for a = lour functions do not depend also on z.
Next we write down the equation arising from the singular vector V12 (null-
conditions (29b), (31 b), (34b :
It is also valid in all cases, however, for a = 0 it follows from (115) and for a = 1 it
follows from (116). Now, since (117) is a first degree polynomial in z, on which our
72
functions do not depend, it actually consists of two equations, though not invariant
by themselves, i.e., we have:
((a-1-zaz)av-a_az) 0 =0 (118a)
((a-1-zaz)a+-avaz) 0 =0 (118b)
where we used (114) in passing from (119b) to (119c). Thus, we have recovered the
d' Alembert operator. Note that (119) is valid for arbitrary a since we have used only
condition (114) which is valid for all of our representations.
Now for a = 1 if we take only invariant equations arising from the conditions (36),
(i.e., we work with the counterpart of the factor-module Fd, we have the following
system of differential equations:
Oz 0 =0 (120a)
az 0 =0 (120b)
0 0 =tp" (120c)
where 0 E C A and satisfies (120a, b), 0' E CAl, A' = A - 0'13 - 0'2, the corresponding
signatures being X = (1,0,1), X' = (1, -2, 1). (Note that the second Casimir operator
has the same value in the two representations: C 2 (X) = C 2 (X') = -3, cf. (12).) If
we consider the irreducible factor-module L I , which means that we should use (38)
instead of (36), we have instead of (120c):
o 0 = 0 (120d)
Thus, from the subsinguiar vector vf we have obtained the d'Aiembert equations
(120c, d) as conditionally si( 4) and su(2, 2) invariant equations.
Now we pass to the cases when a # 1. In these cases the vector vf is a linear
combination of the singular vectors VI and V12 and it becomes zero when these singular
vectors are factorized. Since vf gives rise to the d'Alembert operators for all a we
expect that the d'Alembert equation (120d) will hold automatically if the invariant
equations (114), (117) (arising from VI, V12) hold. This is indeed so. We use the two
equations (118) which are the two components of (117). First we take OJ; derivative
from (118a) and a_
derivative from (118b) and subtracting the two we get:
73
This still follows from (120d). Analogously, taking 0+ derivative from (l1Sa) and ov
derivative from (l1Sb) and subtracting the two we get:
This also follows from (120d). Now, clearly from (121a, b) follows:
(a-1)D<p=0 (121c)
instead of (117). Furthermore the analogs of (l1Sa, b), (114), (116), resp., are:
We summarize now the results of this section. In the sl( 4) setting we have that
the d' Alembert equation (120d) holds in the representation spaces with signatures
Xl (a) = (a, 1- a, 1), resp., X3(a) = (1,1- a, a), if our functions do not depend on the
variable z, resp., Z and in addition satisfy (120a, b), resp., (123a, b). For a = 1 the
d'Alembert equations (120c,d) are conditionally sl(4) and su(2,2) invariant, while
for a i= 1 the d'Alembert equation (120d) just follows from (120a, b), resp., (123a, b).
If a E N then our functions are polynomials in z, resp., z, of degree a - l.
Summarizing our results in the su(2, 2) setting we again recall that the variables
z, Z are representing the spin dependence coming from the Lorentz representation
[37-39J. The above result then is restated so in the case a ~ N: the d'Alembert equa-
tion holds if the fields carry holomorphic (depending only on z) or antiholomorphic
(depending only on z) infinite-dimensional representations of the Lorentz algebra;
in addition they satisfy (120a, b), resp., (123a, b). In the case a E N we restrict to
Lorentz representations which are finite-dimensional; in fact, of dimension a.
The case a = 1 is remarkable in one more respect, namely, in this case one may
have a non-trivial r.h.s., cf. (120c). It is easy to check that there are no other cases
with non-trivial r.h.s .. In fact, for a -I- 1 (120d) follows from (120a, b) or (123a, b).
74
This can be shown also independently. Indeed, in the first case the candidate signa-
tures would be: Xl (a) = (a, I-a, 1), X~ (a) = (a, -I-a, 1). We know that a necessary
condition to have-an invariant equation is that the two representations would have the
same Casimir operators, in particular, one should have C2(XI(a - C2(x~(a = 0,
where C 2 is given in (12). Calculating this difference we obtain:
(124)
Remark 7: Actually, in ref. 38 the general Maxwell equations were considered with
non-zero current, i.e.,
(126)
which are then equivalent to a modification of (117), (122) with non-zero right-hand-
sides which are given explicitly in formulae (5a,b) of ref. 38. More than that, an
hierarchy of Maxwell equations involving two conjugated families of representations
is discussed in ref. 38: X~ = (n + 3, -n - 2, n + 1), X;;- = (n + 1, -n - 2, n + 3),
n E Z+ , from which the Maxwell case is obtained for n = O. Note that there is no
other intersection of this Maxwell hierarchy with the two families Xl (a) and X3( a)
(d. (23c) and (39c)) which we consider in the present paper. <>
We may write out many other equations with indices, however, one of the main
points here is that in this form equations (120) and (123) are valid for different rep-
resentation spaces, the different representations manifesting themselves only through
the parameter a.
Remark 8.1: It is interesting to note that there are other conditionally invariant
equations involving the d' Alembert operator, from which (120c) is a partial case
(m = 1), namely:
(127)
where cp E C A, cp' E CA', A' = A-m(aI3+a2), the corresponding signatures
being X = (m, 0, m), X' = (m, -2m, m). These are produced by subsingular vectors of
weights m (0013 + (02), d. (82b). The functions cp, cp' carry irreducible Lorentz repre-
sentations which are symmetric traceless tensors of rank m - 1. (For early examples,
namely, (127) with m = 2, obtained from other considerations, d. [40-42].)
Remark 8.2: We should note that there are conditionally invariant equations in-
volving the d'Alembert operator, which do not arise from subsingular vectors but
from reduction of integral intertwining operators. These equations are also given by
(127), however, the corresponding signatures are: X = (m, n, m), X' = (m, n-2m, m),
m, n E N, cf., e.g., [43].
Remark 8.3: We should note that in (most of) the physical applications (127) is
75
not considered conditionally invariant. The reason is that there only representations
induced from finite-dimensional Lorentz representations are considered. The fact that
these representations are also subspaces of reducible representations is ignored and
thus the restriction to these subspaces is not considered to be a condition (cf. [40-
44]). <)
cp(Y)
i,j,k,R.,m,nE'iL+ (128)
CPijklmn = zi v j x~ x~ vm zn
MK cp(Y) L
i,j,k,R.,m,nEZ+
{lijklmn JVIK CPijklmn , (129a)
TK cp(Y) L
i,j,k,R.,m,nEZ+
{lijkmn T" CPijklmn , (129b)
76
Note that although M;: 1 is not defined if the corresponding variable is of zero degree,
the operator VI< is well defined on such terms, and the result is zero (given by the
action of (T" - T,,-1 )). Of course, for q -+ 1 we have V" -+ 0".
Using the above operators the representation (left) action was given for general
n [23] and for n = 4 r39].
The q-difference analogues ofthe operators Rk, i.e., the right action of Uq (sl(4))
on our functions is also known from ref. 23. Adapting this to our notation we have:
(133a)
A A A -1 -1 A -
+ Mz Mz V+ (TvT-) Tf; +q Mz Vf;- (133b)
- A Mv Mi V_ V+ Tv) Tv T z- 1
R~ = Vz Tz (133c)
Substituting (133) in (16) we obtain the following Uq ( sl( 4)) and Uq ( su(2, 2))
invariant equation:
R 2q <pA = (qA A
Mz Vv Tv2 T_2 + V_
-
T_ +
The subsingular vector Vbgg gives rise to the following conditionally invariant equation:
+ q>-.. lVIv V_ V+ Vz Tz Tz Tv +
+ q>-.. q Mz Vv Tv T_
(
A A 2 2
+ V_
-
T_ +
A A - -1 -1 - A
+Mz Mz V+ (TvT_) Tv +q Mz V v -
where 0, E C A and satisfies (134a), 0' E CAl, A' = A - (Xl3 , the corresponding signa-
tures being as in (113). Clearly (134a, b) go into (113a, b) for q = 1. If we consider
the irreducible factor-module LA, which means that we should use (17) instead of
(16), then we have a zero r.h.s. in (l34b) as in (113c).
77
The q-Difference Equations from the Other Archetypal sl(4) Example
Finally, we write down the q-difference analogues of the d' Alembert equation and
of the equations ensuring its Uq ( sl( 4)) and Uq ( su(2, 2)) invariance and from which it
follows (except for a = 1). Substituting (133) in (38) we obtain:
(135a)
- qI-a (
q Mz Vv Tv2 T_2
A A
+ V_ T_ +
A
+ Mz M .. V+ (TvT_)-1 Tv + q-l M .. Vv -
As in the q = 1 case we use (135a) to split (135b) in two equations and to simplify
(135c). Finally, we have:
(136a)
3 -1 -2) -1
'Dv 'Dv Tv Tv - q 'D_ 'D+ T_ Tz Tv -
A A A A
{ (
o (136d)
In the scalar case a = 1 the relevant equations are (136a, c, d), in particular,
using (136c) and adding a non-trivial r.h.s. we obtain the conditionally Uq (sl(4)) and
Uq ( su(2, 2)) invariant q-d' Alembert equation:
3 -1 -2)-1
'Dv 'Dv Tv Tv -q'D_'D+T_ Tv -
A A A A
{ (
78
Analogously one may write down explicitly the conjugated invariant equations.
Clearly, for q = 1 (136c, e) go into the d'Alembert equations (120d, c), respec-
tively.
ACKNOWLEDGMENT
The author was supported in part by BNFR under contracts Ph-401 & Ph-643.
REFERENCES
79
[21] V. K. Dobrev, Rep. Math. Phys. 25 (1988), 159.
[22] V. K. Dobrev, J. Phys. A28 (1995),7135.
[23] V. K. Dobrev, .T. Phys. A27 (1994), 4841 & 6633.
[24] D. Kazhdan and G. Lusztig, Inv. Math. 53 (1979), 165.
[25] K. De Vos and P. Van Driel, Lett. Math. Phys. 35 (1995), 333; BONN-TH-95-14,
Aug 1995, hep-th/9508020.
[26] V. G. Drinfeld, Soviet. Math. Dokl. 32 (1985), 2548; in: Proceedings of
the International Congress of Mathematicians, Berkeley (1986), Vol. 1, The
American Mathematical Society, Providence, RI, 1987, 798.
[27] M. Jimbo, Lett. Math. Phys. 10 (1985), 63; Lett. Math. Phys. 11 (1986), 247.
[28] V. K. Dobrev, in: Proc. of the International Group Theory Conference (St.
Andrews, 1989), C. M. Campbell and E. F. Robertson (eds.), Vol. 1, London
Math. Soc. Lect. Note Ser. 159 (1991) 87.
[29] C. De Concini and V. G. Kac, Progress in Math. 92. ilirkhiiuser, Boston, 1990,
471.
[30] V. K. Dobrev, J. Phys. A26 (1993), 1317.
[31] T. Yao, J. Math. Phys. 8 (1967),1931; 9 (1968), 1615; 12 (1971), 315.
[32] V. K. Dobrev and R. Floreanini, J. Phys. A27 (1994), 483l.
[33] D. Kazhdan and G. Lusztig, Proc. Symp. Pure Math. Vol. 34, The American
Mathematical Society, Provicence, RI, 1980, 185.
[34] .T. E. Humphreys, Reflection Groups and Coxeter Groups, Cambridge Studies on
Advanced Math. Vol. 29, Cambridge Univ. Press, 1990.
[35] M. Rosso, Comm. Math. Phys. 117 (1987), 58I.
[36] G. Lusztig, Adv. Math. 70 (1:988), 237.
[37] V. K. Dobrev, J. Math. Phys. 26 (1985), 235.
[38] V. K. Dobrev, Phys. Lett. 341B (1994), 133 & 346B (1995), 427.
[39] V. K. Dobrev, in: Symmetries in Science VIII, B. Gruber (ed.), Plenum, New
York, 1995, 55.
[40] G. M. Sotkov and D. Ts. Stoyanov, J. Phys. A13 (1980), 2807; J. Phys. A16
(1983), 2817.
[41] B. Binegar, C. Fronsdal and W. Heidenreich, J. Math. Phys. 24 (1983), 2828.
[42] P. Furlan, V. B. Petkova, G. M. Sotkov and 1. T. To dorov , Riv. del Nuovo Cim.
8 (3) (1985), l.
[43] P. Furlan, V. B. Petkova and G. M. Sotkov, J. Math. Phys. 28 (1987), 25l.
[44J A. Salam and J. Strathdee, Phys. Rev. 184 (1969), 1760.
80
ON A PATH TO NONLINEAR QUANTUM MECHANICS
(1) Observables - for fixed t - are given by elements A of the set SA (H) of (linear)
self-adjoint operators. The spectral representation allows a probability interpre-
tation of the observables.
(2) States - for fixed t - are described through (linear) positive trace class operators
W t (TrWt = 1) on H called density matrices. The result of a measurement of
the observable A in the state W t is determined by
This decomposition of W t is in general not unique. The set {Aj, 0j,tl jEI is denoted
as a mixed state; those mixed states corresponding to the same W t are physically
equivalent, and a density matrix (2) defines an equivalence class Ct in the set of
mixed states.
(3)
on a dense set {}t"t2 C H (depending on Ut2 ,t,), with Ut"t, = 1 and Uts,t, oUt2 ,t, =
Uts,t, on
{}t1,t2'. Probability conservation implies (non-relativistic case)
(4)
The propagator is norm-preserving. In the time translation invariant case we
have Ut := Ut,o and irtfJt1/;t = H( 1/;t), where iH = -n8tUt is in general a nonlinear
operator or "Hamiltonian" and not an observable in the sense of (1). Using our
principle one can show that Ut (1/;j), j E I, yield a unique evolution of W t if and
only if the equivalence class GtI evolve into an equivalence class Gt2 for all t2 > t l .
This is only possible if Ut is linear. Hence we get a unitary U t = U(t) and
(6)
The corner stones of this successful and consistent framework show that self-adjoint
operators and density matrices as mathematical models for observables and states imply
together with the principle that pure states in iJ> rule the system, a linear' and unitary
time evolution. Only evolution equations of von Neumann type are allowed, those
of, e. g., Lindblad type, characteristic for dissipative quantum system are excluded.
Furthermore, also nonlinear evolution equations for pure states, i. e. nonlinear Ut or H,
are forbidden.
The linear creed was questioned already in the beginning of quantum mechanics.
E. P. Wigner mentioned in a footnote to the first sentence of his famous paper "On
unitary representations of the inhomogeneous Lorentz group" [1] "The possibility of
a future nonlinear cilaracter of the quantum mechanics must be admitted, of course."
In the history of attempts to go beyond the linear creed various nonlinear Schrodinger
equations were proposed with more or less physically motivated arguments or with ad
hoc additional nonlinear terms.
In recent years quantum mechanical precision experiments became possible and
hence nonlinear Schrodinger equations are presently of wider interest.
S. Weinberg proposed in [2] "a framework for introducing nonlinear corrections . .. ,
that can serve as a guide to experiments that would be sensitive to such corrections"
and he considered a theory with observables given by homogeneous functionals f[1/;,~]
82
of the wave functions 'Ij;. In particular, Weinberg used a Hamiltonian functional h['Ij;, 'Ij;]
generating the time evolution'
(7)
1. Bialynicki-Birula and J. Mycielski observed in [3] that "All linear equations de-
scribing the evolution of physical systems are known to be approximations to some
nonlinear theories, with the notable exception of the Schrodinger equation". They
proposed a nonlinear version of the form
(8)
which fulfills a separation property for multiparticle wave functions. The nonlinearity
is proportional to 0:1 which can be measured through scattering experiments. Such
experiments were made [4] with an upper bound for 0:1 as a result.
In connection with Weinberg's nonlinear quantum mechanics N. Gisin [5] and J. Pol-
chinski [6] showed that the inconsistency of a nonlinear evolution of pure states with
a well defined evolution of mixtures, explained in the first section, leads to acausal
effects, i. e. supraluminal communications with is a "no-go" argument for any nonlinear
evolution in the framework based on the linear creed.
A convincing strategy to go beyond this creed needs two parts:
1. A physical and also mathematical respectable nonlinear evolution equation should
be a consequence of physically motivated first principles. We describe in the fol-
lowing section possible candidates: The DG family of nonlinear evolution equa-
tions is derived
The DC family can be constructed also from so called nonlinear gauge transforma-
tions [11-13]. We explain Borel quantization, the generalized Ehrenfest relation
and the DG family in the next section. Nonlinear gauge transformations, shortly
mentioned later, were described in detail in the lecture of G. A. Goldin in this
conference.
83
A PATH TO NONLINEAR EVOLUTIONS: BOREL QUANTIZATION
with a3 three dimensional commutative and with the Euclidean Lie-algebra e(3).
To generalize this idea to arbitrary configuration manifolds without metric, note
that there are no preferred vectorfields, like Killing vectorfields, on M to construct the
kinematics. We use as a generalization of S the semi-direct product of the function space
C=(M), viewed as a commutative Lie algebra, modeling general position operators, and
the complete vectorfields X E Vectc(M) on M, modeling generalized momenta. Hence
the kinematical algebra on M is
where [ denotes the Lie derivative. S(M) is the kinematical algebra of a gener'alized
elementary system localized and moving on M. Note that we get for S(R3) more general
results as for S. Time evolutions based on S(M) are discussed in the section "Quantum
Dynamics" below.
To quantize S(M) we construct a mapping
84
(I) The "partial" Lie-algebra structure of S(M) should survive, i. e., Q is a partial
Lie-algebra homomorphism.
(III) The elements in L2(M,/1) are C-valued functions over M. With the motivation
that in quantum mechanics differential operators play an essential role we want
to have differential operators on L2(M, /1) or a differentiable kinematics. For
this we need a differential structure V on the set M X C compatible with the
differential structures on M and C. L2(M, /1) is defined in the measurable category
and not in the differentiable one. A possibility to define V in thiR setting is to
consider sections in a line bundle over M, TJ = (E, pr, C, (., .)), with total space E,
projection pI', fiber C and inner product (., .). TJ is a smooth manifold and admits
connections and related differential operators.
(IV) We view now L 2 (M,/1) as the Hilbert space L 2 (TJ,/1), i.e. the completion of the
space of square integrable sections ljJ of TJ. On these sections we need a locality
condition for the operators P(X), i. e.
(16)
= x + TiW(X) ,
~
'i7 X (18)
85
The above result yield representations Q(S(M)) on L2( 7], fJ), which are characterized
by a number C E R and an element 00 E 7rl(M)*. One can show that the Q(S(M))
are irreducible for any c and 00. Hence these numbers can be considered as quantum
numbers of a generalized elementary system with kinematics SCM).
In the linear framework unitary equivalent representation of ge(3) are physically
equivalent description of the kinematics of free particles (because of (1)).
The notion of unitary equivalence can be formulated also for Q(S(M)). Two rep-
resentations on L2(M,fJ) with (7]i,Wi) and Ci, i = 1,2, are equivalent if there is an
isometric isomorphism between (7]I,Wl) and (7]2,W2) and if Cl = C2. This implies in the
standard case that closed one-forms w which differ by a logarithmic exact one form give
equivalent Q(S(M)). Unitary equivalence leads here also to physical equivalence; this
is indicated below in connection with a discussion of nonlinear gauge transformations.
Summarizing the discussion we have:
(19)
The observables arc given as:
generalized position operators: Q(J)1/J = J1/J
generalized momentum operators: P( X)1/J = ~ \l x1/J + (c + fi) (div I'X) 1/J .
Quantum Dynamics
The kinematical structures discussed in section "Quantum Borel Kinetics" are the
basics for the introduction of dynamics in our quantization program. As a first step
we assume the existence of a distinguished Riemannian metric g on the configuration
manifold M. We will demand in a second step that the implicit time dependence of the
classical observables is given through a second order differential equation - the typical
form of classical dynamics. This second order Riemannian dynamics yields a relation -
(23) - on T* M. In a third step we "lift" SCM) and Q to T* M to get a quantized version
of the dynamics in terms of Q(S(M)), called generalized (1.) EhrenJest relation.
For the second step take J E COO(M) and a solution curve c(t) of a second order
differential equation on M. Then we get
(20)
(21 )
86
with Y = g:Y E Vect(T*M); J is the lift of f to T*M; B = L-Pidqi is the Liouville
form and X denotes the canonical lift of X E Vect(M) to T* M. Hence on T* M our
characterization of the dynamics looks
J= B(gradgf) . (23)
Q:Fe(T*M)EflsFz(T*M) -+ 5A(L2(M,/1))
(24)
J + B(X) f-+ Q(f) + P(X) .
Fe and Fz are th~ C=-functions on T* M, whic~ are constant resp. linear on the fibres.
Because of (23) J is in Fz(T* M) and Q sends J to P(gradgf).
The time evolution of classical position observables in (23) implies for the expecta-
tion values (1) of the quantized classical observables in 5(M) in a state Wt,
(25)
Now we assume a natural compatibility condition between classical and quantum time
evolutions in the form
as a relation which characterizes the quantum version of a classical second order Rie-
mannian dynamics.
Equation (26) is called the generalized (1.) Ehrenfest relation because on M = R~
with metric gij = m8ij and with W t = P,pt it implies
This generalized Ehrenfest relation is the last corner stone in Borel quantization.
The result can be considered as a description of a generalized elementary system in the
sense of Wigner localized and moving non-relativistically on M and with kinematical
algebra 5(M) and a time evolution ruled by the generalized Ehrenfest relation.
A version of the Borel kinematics on 51 and the N-point discretization of 5Jir in
which the generalized momenta were modeled not through differential operators but
through q-difference operators, is discussed by R. Twarock [20J in these proceedings.
87
Nonlinear Evolution
The Ehrenfest relation restricts evolution equations for states, i. e. the equation
for the time dependence of density matrices Wt. To evaluate this condition note that
the decomposition of density matrices as a convex sum of pure states P ,pj,t is the most
general case.
Hence we discuss first pure states which evolve by construction into pure states and
below we give the results for the general case.
If we insert for P(gradgJ) equation (17) we get (for a given wand c),
~
dt
r f(m)pt(m)dfl(m) = 1Mr f(m) (-divj;"(m) + c!;;'gpt(m)) dfl(m) ,
1M
(27)
pt(m) .- '1/yt(m);j,t(m),
j;"(m) .- H;j,t(m )(gradgt)(m) - t(m)(gradg;j,t)(m)) + pt(m)gU w .
(28)
Because (28) is valid for all f E C=(M) we get a Fokker-Planck like equation
(29)
(30)
(31 )
The DG family. The equations (31) are a result of our quantized versions of
a classical second order Riemannian dynamics. They give a nonlinear Schrodinger
equation with a nonlinear term enforced through the quantization of the kinematical
algebra. The usual linear Schrodinger equation is included for c = 0 and R = O.
Note, that we derived first a Fokker-Planck like equation (29) for p = 8t (;j,?jJ) with an
additional quantum mechanical diffusion term -c!;;'gPt. Equation (29) is nonlinear in
?j;t,?jJt. Hence 8t inherits in general a nonlinearity from (29) which is only real for c = O.
It was known already in the beginning of quantum mechanics that a nonlinear real
88
term in the Schrodinger equation is consistent with the quantum mechanical continuity
equation (see, e. g., [23]).
It is interesting that the Borel quantization yields probability conservation
! fMPtdP, =0 (32)
(b) Such a nonlinear choice of N will transform a system governed by the ususal,
linear Schrodinger eqaution to a physically equivalent system obeying a NLSE -
one that is, of course, linearizable (by construction).
In [13] T\{ was assumed to be local in the sense that N[?ji] at x, t should depend only on
x, t and on the value 1jJ(x, t), i. e. (N = NF)
NF[?ji](X, t)= exp[i(F( ?ji(x, t)j x, t))] ?ji(x, t) .
Usual gauge transformations Ug = exp(i8(x,t)) are special cases of N. With the as-
sumption that Ug intertwines with N,
89
and that N seperates (see above) one gets a real 2-parameter (((i), A(i)) group of
nonlinear transformations {N{-y,A)}:
(N{-Y,A)[I,bJ) (x,i) = IVJ(x,iWexp[i({(t)lnll,b(x,t)1 + A(i)Argl,b(x,i))).
These are nonlinear gauge transformations. Their physical significance and use were
described in [12]. Another method [26] to derive N{-y,A) observes that the actual
Hilbertspace is projective and assures that N respects its projective structure.
With these nonlinear gauge transformations we get the DG familiy as follows: Con-
sider the set VL of solutions of the Schrodinger equation
Construct the set VNL = {N{-y,.\)VL}. One can prove that VNL is a set of solutions of
nonlinear Schrodinger equations contained in a subset of the DG family and character-
ized through relations between the family parameters C1, . , C5' Break these relations.
On the resulting subset apply again N("(,.\). Get a larger subset. Repeat this procedure.
It ends after a finite number of steps and the DG family is the result.
This derivation is completely independent of the quantization procedure explained
the above sections "Quantum Borel Kinematics" and "Quantum Dynamics". It reflects
the simple physical statement that all measurements are "in principle" position and
time measurements.
Mixed states. In the previous paragraphs we discussed the consequence of the
generalized Ehrenfest relation (26) for the evolution of pure states. For mixed states,
i. e. for arbitrary Wt, we proceed exactly along the same lines. We give the result for
M = R 3 ,gik = mOik:
3
where
1i?
H = - 2m b. + V(x)
F t is a functional of Wt, OJ W t , ... , x, t such that Tr(Fd) = 0 for all f E C=(R3, R).
For F t == 0 the resulting equation:
. 3
is a generalization of the von Neumann equation (6) and a master equation of Lindblad
type [27] which appear in the description of, e. g., open quantum system.
The properties of this master equation for mixed states and of the DG family for
pure states are different. The DG family is linear only for c = 0, the master equation
(35) is linear for all c. The DG family evolve pure states into pure states. Because
of the remarks in the first section the DG family is incompatible with the notion of a
density matrix. One can show that in (35) pure states at t = 0 evolve into mixed states
at t > O.
Concerning the structure of the generalized von Neumann equation (35) we refer to
a recent result of A. S. Holevo [28J. Its physical interpretation in connection with our
approach is an open question.
90
THE PRESENT SITUATION
ACKNOWLEDGEMENTS
The paper incorporates the results of a number of common papers with B. Anger-
mann, G. A. Goldin, W. Liicke, U. Miiller, P. Nattermann and J. Tolar. We benefitted
from many discussions and remarks by P. Nattermann.
REFERENCES
91
[9J G. A. Goldin and D. H. Sharp, in: 1969 Battelle Rencontres: Group Represen-
tations, V. Bargmann (ed.), Lecture Notes in Physics 6, Springer, Berlin, 1970,
300; G. A. Goldin, J. Math. Phys. 12 (1971), 462; G. A. Goldin, R. Menikoff, and
D. H. Sharp, in: Measure Theory and its Applications, G. A. Goldin and R. F. Wheel-
er (eds.), Northern Illinois Univ. Dept. of Mathematical Sciences, DeKalb, IL, 1981,
207; G. A. Goldin, R. Menikoff, and D. H. Sharp, Phys. Rev. Lett. 51 (1983), 2246.
[10] H.-D. Doebner and G. A. Goldin, Phys. Lett. A 162 (1992), 397; J. Phys. A:
Math. Gen. 27 (1994), 1771.
[11J H.-D. Doebner, G. A. Goldin, and P. Nattermann, in: Quantization, Coherent
States, and Complex Structures, pages J.-P. Antoine, S. Ali, W. Lisiecki, 1. Mlade-
nov, and A. Odzijewicz (eds.), Plenum, New York, 1996, 273l.
[12J H.-D. Doebner and G. A. Goldin, Phys. Rev. A 54 (1996), 3764.
[13J H.-D. Doebner, G. A. Goldin, and P. Nattermann, Clausthal-preprint ASI-TPA/
21/96.
[14J B. Mielnik, Comm. Math. Phys. 37 (1971), 22l.
[15J R. Haag and U. Bannier, Comm. Math. Phys. 60 (1978), 1.
[16J P. Nattermann, these proceedings.
[17J W. Lucke, in: Nonlinear, Deformed and Irreversible Quantum Systems, H.-D. Doeb-
ner, V. K. Dobrev, and P. Nattermann (eds.), World Scientific, Singapore, 1995,
140~154.
[18J J.-D. Hennig, in: Nonlinear, Deformed and Irreversible Quantum Systems, H.-
D. Doebner, V. K. Dobrev, and P. Nattermann (eds.), World Scientific, Singapore,
1995, 155~ 165.
[19J R. Abraham and J. E. Marsden: Foundations of Mechanics, 2. edition, Benjamin/
Cummings Publishing Company, Reading, 1978.
[20] R. Twarock, these proceedings.
[21J H.-D. Doebncr and P. Nattermann, Acta Phys. Pol. B 27 (1996), 2327.
[22J C. Schulte, these proceedings.
[23J E. Madelung, Zeit. Phys. 11 (1926), 322; D. Schuch, K.-M. Chung, and H. Hart-
mann, J. Math. Phys. 24 (1993), 1652.
[24J see contributions in: Nonlinear, Deformed and Irreversible Quantum Systems, H.-
D. Doebner, V. K. Dobrev, and P. Nattermann (cds.), World Scientific, Singapore
1995; Group21 - Physica.l Applications and Mathematical Aspects of Geometry,
Groups and Algebras. H.-D. Doebner, P. Nattermann, W. Scherer, and C. Schulte
(eds.), World Scientific, Singapore 1997, to appear.
[25J R. Feynman and A. Hibbs, Quantum Mechanics and Path Integrals, McGraw-Hill
Book Company, New York, 1965.
[26J H.-D. Doebner, G. A. Goldin, and P. Nattermann, in preparation.
[27J G. Lindblad, Comm. Math. Phys. 48 (1976), 110.
[28J A. S. Holevo, in: Lecture Notes in Physics, A. Bohm, H.-D. Doebner, and P. Kiela-
now ski (eds.), Springer, Berlin 1997.
92
QUANTUM MECHANICAL PROBLEMS WITH
q-DEFORMATIONS AND OVER THE
p-ADIC NUMBER FIELDS
1. H. Duru
University of Trakya,
Mathematics Department
Edirne, Turkey
and
TUBITAK, Marmara Research Centre
Physics Department,
P.K. 21, 41470 Gebze, Turkey
INTRODUCTION
When we are interested in the quantum mechanical motion of a particle over a con-
figuration space which is not equivalent either globally or. locally to Iftn , the Schrodinger
equation approach is not the most convenient method. First of all, when we do not
have a differentiable manifold, it is difficult to find the correct operators replacing the
derivatives. On the other hand since it is always possible to construct an integration
theory on any given set or number field, the path integration may in principle be the
suitable method of quantization.
The techniques for solving the path integral in quantum mechanics are quite ad-
vanced. In fact almost all of the non-relativistic potential problems, which are exactly
solvable by the traditional Schrodinger approach, are solvable by the path integrals too.
A short list of these problems is:
(i) Harmonic oscillator and the linear potential (1].
(ii) Inverse square potential (2].
(iii) H-atom (3].
(iv) Morse potential (4].
(v) Poschl-Teller, Hulthen, Wood-Saxon and Rosen-Morse potentials, which are
related to the particle motion over the SU(2) manifold [5].
(vi) Moving potential barrier V = cosh- 2 (x -wi) [6]; and potentials in motion with
constant accelerations or velocities [7].
The techniques employed in solving the above problems may also be adopted to
a 1
V=---- (1)
x 321lX2
where a is the coupling constant and Il is the mass.
In its usual form the above potential can exactly be solved by path integrations:
one first employs a point canonical transformation with the variable change
u=vx (2)
then passes to the new "time" parameter s given by [3J
(3)
Then the probability amplitude for a particle with mass Il, to travel from the space-time
point x, t = 0 to the point x', t = T can be expressed as
Here Koss( U, U'; s) is the probability amplitude for a particle with mass Il moving in
(u,s)-"space-time" under the influence of the oscillator potential ~IlW2U2 with w =
2J-2EIIl. Since the exact closed form of Koss is well known [1], the above equation
provides the exact Green function for the potential (1).
94
To obtain the energy spectrum and the wave functions, we decompose the Green
function Koss in terms of the oscillator wave functions Wn(u). Eq.(4) then becomes
K(x, x'; T) =
00
1 o
00 ds 100 _e-
dE ET e- 4"'!S(xx')1/4
-00 271"
. x
1
or
dE 00 -iET
L (
00
e I 1/4 --.-,
K(x,x,T)--~
I. _ .
-2 /2)(xx) 4>n(X)4>n(x) (6)
-00 71" n=O 4a + w n + 1
with 4>n( x) = Wn(u) + Wn( -u). The pole at w = -4a( n + 1/2)-1 in (6) gives the correct
energy spectrum for the potential (1)
(7)
corresponding to the wave functions x 1/ 4 4>n(x).
The procedure summarized above can easily be adopted to the q-deformed version
of the potential (1): in place of (5) we have
Kq(x, x'; T) = 1 ~! J
-00
00
w- 1 dqa-1e-iET f e-
n=O
i n+1(2)+4 ajw )
(8)
where
~=eWs and 4>~(x)=W~(u)+W~(-u). (9)
If, for example, for the q-deformed oscillator states we employ the ones given in terms
of the q-l-hermite polynomials [10]
W~( u) = (27rkj32)1/4(2/3tq(n+l/2)2 /4( q; q );,1/2 e-k(3l"wu 2
Kq(x, x'; T) = ~
00 1
00 dE
-
e- iET
n=O -00 271" wq-4' w([i(n + 1/2)]- [4ia/w])
'Q .
Since the p-adic oscillator is well known [10], if we adopt the relation (4) or (5),
we obtain the p-adic version of the potential (1). The Green function for the p-adic
oscillator is [11 J:
u2 + u12
K%ss( u, u'; s) = Ap( s/ fl) I fl/S 1!/2 X(flW( - tanws
+ 2-.-.
uu'
smws
(12)
95
The e-value
(13)
takes the values
(14)
or
a p = p-V(ao + alP + ... + a v _ 2pV-2). (15)
Here 0'0 =I 0 and O'i = 0,1, .... ,p - 1. The values of v are
If we decompose the Green function(12) in terms ofthe p-adic oscillator wave functions
I]! "p/Jp (with the number ;3p distinguishing the degenerate states [11, 12]) as
L X(wO'ps)!"p/Jp(x)fap/Jp(x'). (18)
"p/Jp
Here! is the p-adic wave function of the potential (1)
(19)
and X's are the character functions which correspond to the exponential functions.
Using the formula [13]
r dsX(;s) = 8(;) ,
lop
(20)
(22)
or
(23)
EP __2
- _2f-ta (24)
Kep - 0'2
P
96
Note that the minimum value of the above energy spectrum is rather pathological for
IYp starting from zero. Finally it must be stressed that the square root of every p-adic
number is not necessarily a p-adic number. Thus according to the Hensels' lemma [11],
the method we presented is valid for the set of numbers given by
(25)
DISCUSSION
In the preceding sections we have sketched a method for studying the q-deformed and
p-adic versions of a toy potential, which is related to the harmonic oscillator by a point
canonical transformation. Some other potentials which can classically be converted to
the harmonic oscillators (for example the Morse potential) may also be studied by the
similar procedUFe. However the class of such potentials is rather limited. On the other
hand, for the large class of potentials, the most suitable and unambiguous approach
is the adoptions of Green functions (which are originally developed over the related
classical group manifold) to the given non-commutative geometry or the non-trivial
number field.
REFERENCES
[lJ R. P. Feynman and A. R. Hibbs, Quantum Mechanics and Path Integrals, Mc Graw-
Hill, New York, 1965.
[2J D. Peak and A. Inomata, J. Math. Phys. 10 (1969), 1422; 1. H. Duru, Phys. Lett.
112A (1985), 42l.
[3] 1. H. DUFu and H. Kleinert, Phys. Lett. 84B (1979), 185; Fortschr. Phys. 30
(1982), 40l.
[4J 1. H. DUFu, Phys. Rev. D28 (1983), 2689.
[5J 1. H. Duru, Phys. Rev. D30 (1984), 2121; Phys. Lett. 119A (1986), 163.
[6] H. Ahmedov and I. H. Duru, J. Phys. A: Math. Gen. (in press).
[7] 1. H. Duru, J. Phys. A: Math. Gen. 22 (1989), 4827.
[8] H. Ahmedov and 1. H. Duru, Green function over the q-symmetric space
SUq (2)jU(I), TUBITAK preprint(1996).
[9] 1. H. Duru and N. Unal, Phys. Rev. D34 (1986), 959.
[10J M. M. Atakishiev, A. Frank and F. B. Wolf, J. Math. Phys. 25 (1994),3253.
[l1J B. Dragovich, Int. J. Mod. Phys. 10 (1995), 2349.
[12J P. Ruelle, E. Thiran, D. Verstegen, J. Weyers, J. Math. Phys. 30 (1989), 2854.
[13J P. G. O. Freund and M. Olson, Nucl. Phys. B297 (1988), 86.
97
A SYMMETRY ADAPTED ALGEBRAIC APPROACH
TO MOLECULAR SPECTROSCOPY
INTRODUCTION
The study of molecular vibrational spectra [1] requires theoretical models in order to
analyze and interpret the measurements. These models range from simple parametriza-
tions of the energy levels, such as the Dunham expansion [2], to ab initio calculations,
where solutions of the Schrodinger equation in different approximations are sought
[3,4, 5, 6]. In general, the latter involve the use of internal coordinates and the evalua-
tion of force field constants associated to derivatives at the potential minima. While this
method can be reliably applied to small molecules [7], it quickly becomes a formidable
problem in the case of larger molecules, due to the size of their configuration spaces.
New calculational tools to describe complex molecules are thus needed.
In 1981 an algebraic approach was proposed to describe the roto-vibrational struc-
ture of diatomic molecules [8], subsequently extended to linear tri- and four- atomic
molecules [9] and certain non-linear triatomic molecules [10]. Although these were en-
couraging results, the model could not be extended to polyatomic molecules, due to
the impossibility of incorporating the underlying discrete symmetries. This difficulty
could be surmounted by treating the vibrational degrees of freedom separately from
the rotations. In 1984 Van Roosmalen et al. proposed a U(2) based model to describe
the stretching vibrational modes in ABA molecules [11] which was later extended to
describe the stretching vibrations of polyatomic molecules such as octahedral and ben-
zene like molecules [12]. Recently the bending modes have also been included in the
framework, which was subsequently applied to describe C2v -triatomic molecules [13] and
the lower excitations of tetrahedral molecules [14], using a scheme which combines Lie-
algebraic and point group methods. In a different approach, it has also been suggested
The model is based on the isomorphism of the U(2) Lie algebra and the one dimen-
sional Morse oscillator
whose eigenstates can be associated with U(2) :J SO(2) states [22]. In order to see
how this isomorphism comes about, consider the radial equation
-1 ( - -
1- - + -u
2
d rd + r2 ) cf>(r) = (N + 1)cf>(r) , (2)
2 r dr dr r2
r2 =(N + l)e- P ,
100
eq. (2) transforms into
(3)
This can be identified with eq. (1) after defining x = pd and multiplying by h Z /2f1d z ,
provided that
D (4)
= (5)
where we have defined m = cr/2. In the framework of the U(2) algebra, the operator N
corresponds to the total number of bosons and is fixed by the potential shape according
to eq. (4), while m, the eigenvalue of the 50(2) generator J z , takes the values m =
N/2, (N - 2)/2, .... The Morse spectrum is reproduced twice and consequently
for these applications the m-values must be restricted to be positive. In terms of the
U(2) algebra, it is clear from Eqs. (3-5) that the Morse Hamiltonian has the algebraic
realization
(6)
In addition, the U(2) algebra includes the raising and lowering operators J+ and L,
which connect different energy states, while the angular momentum operator is given
by JZ = N(N + 2)/4, as can be readily shown.
The Morse Hamiltonian of eq. (6) can be rewritten in the more convenient form
(7)
where we have used the relation J; = JZ - (J+L + LJ+)/2 and added a constant
term AN2 /4 in order to place the ground state at zero energy. The parameters Nand
A are related to the usual harmonic and anharmonic constants We and XeWe used in
spectroscopy. To obtain this relation it is convenient to introduce the quantum number
(8)
which corresponds to the number of quanta in the oscillator. In terms of v, the corre-
sponding energy expression takes the form
N2 A
E' = -A(m 2 - T) = -2(N + 1/2) + A(N + 1)( v + 1/2) - A( v + 1/2)Z, (9)
We A(N + 1)
A. (10)
Thus, in a diatomic molecule the parameters A and N can be determined by the spec-
troscopic constants We and XeWe.
101
Wf} now c<?nsi~er t~e Ui(2) J 8Ui(2) J 80i(2) algebra, which is generated by the
set {G i } == {Ni , J+,i, L,i, Jo,;}, satisfying the commutation relations
As a specific example, we consider the Be4 cluster, which has a tetrahedral shape.
'D3h molecules can be similarly treated. In the Be4 case there are six U;(2) algebras
involved (i = 1, ... ,6). In the present approach each relevant interatomic interaction
is associated with a U;(2) algebra. The operators in the model are expressed in terms
of the generators of these algebras, and the symmetry requirements of the tetrahedral
group To. can be readily imposed [14, 24]. The local operators {O;} acting on bond i
can be projectedto any of the fundamental irreps r = AI, E and F2 Using the lit,;
generators we obtain the To. tensors
(12)
where It = ,O and I denotes the component of r. The explicit expressions are given
by
(13)
102
v = 2:i Vi is the total number of quanta. The local basis. states for each oscillator
are usually written as INi , Vi), where Vi = (Ni - 2mi)/2 = 0,1, ... [Ni/2] denotes the
number of oscillator quanta in the i-th oscillator. The states with one quantum V = 1
are denoted by Ii) with Vi = 1 and Vj#i = O. Using the same projection technique as
for the generators (13), we find the six fundamental modes
6
The expansion coefficients are the same as in eq. (13). The states with a higher number
of quanta IV r:P~) can be constructed using the Clebsch-Gordan coefficients of ~ [14,24].
Since all operators are expressed in terms of powers of the Ui (2) generators, their matrix
elements can be easily evaluated in closed form. The symmetry-adapted operators of
eq. (13) and symmetry-adapted basis states are the building blocks of the model.
We now proceed to expicitly construct the Be4 Hamiltonian. For interactions that
are at most quadratic in the generators the procedure yields
(16)
with
(17)
Note that we have not included VA, in Ho, since the combination
1 6 , ,
= 4N L:Ni(N; + 2) , (18)
i:::::l
is a constant 3(N + 2)/2 . The five interaction terms in eq. (16) correspond to linear
combinations of the Casimir operators of [14]. However, for a good description of the,
vibrational energies of Be4 it is necessary to include interactions which are related to the
vibrational angular momenta associated with the degenerate modes E and F2 These
kind of terms is absent in the former versions of the model [12, 14]. We now proceed
to show how they can be obtained in the present formalism. In configuration space the
vibrational angular momentum operator for the E mode is given by [25]
(19)
where qf and q!f are the normal coordinates associated to theE mode. This relation
can be transformed to the algebraic space by means of the harmonic oscillator operators
(20)
to obtain
(21 )
103
Here b~ = 2:i (Y.~i bi , with a similar form for b~t, while the (Y.~i can be read from
Eqs. (12,13). In order to find the algebraic expression for fA2 we first introduce a scale
transformation
(22)
where
Ni
2 (24)
A A
Vi = ~ JO,i.
The other two commutation relations of eq. (11) are not modified by the scale transfor-
mation of eq. (22). In the harmonic limit, which is defined by Ni -+ 00, eq. (23) reduces
to the standard boson commutator [bi, bJ] = 1. This limit corresponds to a contraction
of SU(2) to the Weylalgebra and can be used to obtain a geometric interpretation
of algebraic operators in terms of those in configuration space. In the opposite sense,
eq. (22) provides a procedure to construct the anharmonic representation of harmonic
operators through the correspondence b; -+ b; = L,;/ VJ1; and bi -+ bi = j +,;/ VJ1;.
Applying this method to the vibrational angular momentum we find
(25)
For the vibrational angular momentum f~! associated with the F2 mode we find a similar
expreSSlOn. The corresponding interactions are
where in the last step we used eq. (24). The spectrum of eq. (27) has an anharmonic
correction, analogous to the quadratic term in the Morse potential spectrum. We are
thus substituting harmonic oscillators by Morse oscillators.
A more interesting application is to use our model to fit. the spectroscopic data of
several polyatomic molecules. In the case of Be4 the energy spectrum was analyzed by
ab initio methods in [17], where force-field constants corresponding to an expansion of
the potential up to fourth order in the normal coordinates and momenta were evaluated.
We have generated the ab initio spectrum up to three quanta using the analysis in [25].
For the algebraic Hamiltonian we take [16]
104
Table 1. Fit to ab initio [26] calculations for Be4. The values of the parameters
are WI = 636, W2 = '153, W3 = 532, X33 = 44.276, X 12 = 4.546, X 13 = -2.539,
g33 = -15.031, t33 = -1.679 and i 23 = -1.175. The parameters and energies are given
in cm- 1.
The terms 6 33 and 6 23 represent the algebraic form of the corresponding interactions
in [25] which are responsible for the splitting of the vibrational levels in the (VI, V;', V~) =
(0,0,22) and the (0,11,11) overtones [16].
In Table 1 we show the the results of a least-square fit to the vibrational energies
of Be4 with the Hamiltonian of eq. (28). The r.m.s. deviation obtained is 2.6 cm-I,
which can be considered of spectroscopic quality. We point out that in [25, 26] several
higher order interactions are present which we have neglected. Since our model can be
put into a one to one correspondence with the configuration space calculations, it is in
fact possible to improve the accuracy of the fit considerably, but we have used a simpler
Hamiltonian than the one of [25, 26]. When no ab initio calculations are available (or
feasible) the present approach can be used empirically, achieving increasingly good fits
by the inclusion of higher order interactions [16].
We note that the Be4 Hamiltonian of eq. (28) preserves the total number of quanta
V. This is a good approximation for this case according to the analysis of [25, 26], but it
is known that Fermi resonances can occur for certain molecules when the fundamental
mode frequencies are such that (V, V') states with V i= V' are close in energy. These
105
Table 2. Least-square energy fit for the vibrational excitations of Ht, Be3 and Nat.
The energy differences fiE = Eth - Eexp are given in cm- I .
H+
3 Be3 Nat
V (VI, V~) r fiE fiE fiE
Parameters 8 4 4
interactions can be introduced in the Hamiltonian but the size of the energy matrices
grows very rapidly, so the best way to deal with this problem is through perturbation
theory.
V 3 h TRIATOMIC MOLECULES
106
THE METHANE MOLECULE
We now turn our attention to the CH 4 molecule, for which we shall make a detailed
description. For methane we have four U(2) algebras corresponding to the C-H inter-
actions and six more representing the H-H couplings. The assignments and the choice
of the Cartesian coordinate system are the same as in [14]. The molecular dynamical
group is then given by the product
(29)
The labeling is such that i = 1, ... ,4 correspond to the C-H couplings while the other
values of i are associated with H-H interactions. Consequently there are two different
boson numbers, Ns for the C-H couplings and Nb for the H-H couplings, which corre-
spond to the stretching and bending modes, respectively. The tetrahedral symmetry
of methane is taken into account by projecting the local operators {G';}, which act on
bond i, on the irreducible representations r of the tetrahedral group 'LJ. The explicit
expressions for the 'LJ tensors for the stretching modes are
1
2L
4 ,
T:~" JIL,i ,
.=1
(30)
(31)
107
tensor in the construction of the Hamiltonian, and (ii) diagonalizing this Hamiltonian
in a symmetry-adapted basis from which the spurious mode has been removed following
the procedure of [14]. We note that the condition on the Hamiltonian that was used in
[14] to exclude the spurious contributions, does not automatically hold for states with
higher number of quanta.
According to the above procedure, we now construct the Ta. invariant interactions
that are at most quadratic in the generators and conserve the total number of quanta
(32)
Here r =:: At, F2 for the stretching vibrations x = s and r = E, F2 for the bending
vibrations x =:: b. In addition there are two stretching-bending interactions
(33)
The interaction VAI., has not been included since, in analogy to eq. (18), the combination
1 4 , ,
= -
4N "
L N(N+2)
,. , (35)
s i=l
.
11m J+,i
-- b; ,
Ni-+ OO ,;N;
.
11m Li
-'-
Ni-+ OO ,;N;
2Jo,I'
lim __ = 1. (36)
Ni-+ OO Ni
108
where the operators bi and b} satisfy the standard boson commutation relation rbi, b}J =
Applying the harmonic limit to the interactions of Eqs. (32) and (33) we obtain
8 ij .
lim
lVx-+OO
Hrx ==
0,
o. (37)
Here the operators b~ x t are given in terms of the local boson operators b! through the
coefficients o~~ given in Eqs. (30,31)
10
with a similar relation for the annihilation operators. From eq. (37) the physical inter-
pretation of the interactions is immediate. The Hrx terms represent the anharmonic
counterpart of the harmonic interactions, while the Vrx terms are purely anharmon-
ic contributions which vanish in the harmonic limit. In an application to the ozone
molecule it was found that these terms can account for the strong anharmonicities and
incorporate the effect of Darling-Dennison type couplings [20J.
The zeroth order Hamiltonian of eq. (34) is not sufficient to obtain a high-quality
fit of the vibrations of methane. Several physically meaningful interaction terms that
are essential for such a fit are not present in eq. (34). They arise in our model as
higher order interactions. It is an advantage of the symmetry-adapted model that the
various interaction terms have a direct physical interpretation and a specific action on
the various modes [16J. Hence the addition of higher order terms and anharmonicities
can be done in a systematic way. For the study of the vibrational excitations of methane
we use the 'Ti, invariant Hamiltonian [21]
,)2
+Xll (?iA
" ,, + X 22
(,)2
?iEb + X33 (?iF2"
,)2 + X (,)2
?iF2,b 44
The interpretation of the Wi and 03 terms follows from eq. (37). The Xij terms are
quadratic in the operators Hrx and hence represent anharmonic vibrational interac-
tions. The gij terms are related to the vibrational angular momenta associated with
the degenerate vibrations. As mentioned before, these interactions, which are funda-
mental to describe molecular systems with a high degree of symmetry, are absent in
previous versions of the vibron model in which the interaction terms are expressed in
109
terms of Casimir operators and products thereof [12,14]. They give rise to a splitting of
vibrational levels with the same values of (VI, V2, V3, V4) but with different [2, [3 and/or
[4' Their algebraic realization is given by
-i '2~[tEo
VL- Nb
E
- x t + o]A2 ,
The square brackets in eq. (40) denote the tensor coupling under the point group 'Tci
(41)
where the expansion coefficients are the Clebsch-Gordan coefficients for 'Tci [14, 24l In
the harmonic limit the expectation value of the diagonal terms in eq. (39) leads to the
familiar Dunham expansion [27]
(42)
Here d; is the degeneracy of the vibration. The t;j terms in eq. (39) give rise to further
splittings of the vibrational levels (Vb V2, V3, V4) into its possible sublevels. They can be
expressed in terms of the tensor operators of Eqs. (30) and (31) as
(43)
In the harmonic limit the t;j termS have the same interpretation as in [25]. The 0 ss,
Obb and Osb terms give rise to a splitting of the E and F2 vibrations belonging to the
(VI, V~\ V~3J V~4) = (0,0 0 ,2 2 ,0 0 ), (0,0 0 ,0 0 ,2 2 ) and (0,0 0 ,11, 11) levels, respectively. Simi-
larly, the 02s and 02b terms split the Fl and F2 vibrations belonging to the (0, 11 ,1 1 ,00 )
and (0,1 1 ,00 ,1 1 ) overtones, respectively.
The Hamiltonian of eq. (39) involves 23 interaction strengths and the two boson
numbers, Ns and N b The vibron number associated with the stretching vibrations
is determined from the spectroscopic constants We and XeWe for the CH molecule to
be Ns = 43 [23]. The vibron number for the bending vibrations, which are far more
harmonic than the stretching vibrations, is taken to be Nb = 150. We have carried
out a least-square fit to the vibrational spectru:m of methane including 44 experimental
energies from [28]. The values of the fitted parameters are presented in the second
column of Table 3 (Fit 1). In Table 4 we compare the results of our calculation with
the experimentally observed energies. All predicted levels up to V = 3 quanta are
included. We find an overall fit to the observed levels with a r.m.S. deviation which is
110
Table 3. Parameters in cm- I obtained in the fits to the vibrational energies of CH 4 .
Ns 43 43 43 00 00 00
111
an order of magnitude better than in previous studies. Whereas the Lm.s. deviations of
standard vibron model studies in [14] and [29] are 12.16 and 11.61 cm- 1 for 19 energies,
we find a value of 1.16 cm- l for 44 energies.
In order to address the importance of the (\;3 term, which is completely anharmonic
in origin and vanishes in the harmonic limit (see eq. (37)), we have carried out a
calculation without this term (Fit 2 of Table 3). With one less interaction term the
r.m.s. deviation increases from l.16 to 4.49 em-I. This shows the importance of the (\;3
term to obtain an accurate description of the anharmonicities that are present in the
data. Similarly, the third calculation (Fit 3) shows that, in a fit without the tij terms,
the r.m.s. deviation increases from 1.16 to 7.S1 em-I.
In addition, we have carried out a fit in the harmonic limit (N., Nb -+ 00). In this
limit the (\;3 term vanishes and the algebraic Hamiltonian of eq. (39) reduces to the
vibrational Hamiltonian of [25], the harmonic frequencies Wi and anharmonic constants
Xij, gij and tij having the same meaning: The r.m.s. deviation increases to 20.42 cm- l
(Fit 4). When we use the vibrational Hamiltonian of [25], which contains one additional
interaction (called 334 in Table 3) the quality of the fit does not improve. In fact, the
Lm.s. deviation increases to 20.90 cm- l (Fit 5). The importance of the nondiagonal
elements is demonstrated in a calculation (Fit 6), in which the data are fitted by the
(diagonal) Dunham expansion of eq. (42). For this case the r.m.s. deviation is 21.00
em-I, almost the same value as for the other two calculations in the harmonic limit.
The small differences in the parameter values and the Lm.s. deviation of Fits 4 and
6 show that, unlike for finite Ns and Nb (Fits land 3), in the harmonic limit the
nondiagonal contributions from the iij terms are not very important.
A comparison between the parameter values and the r.m.s. deviations of Fits 1-6
in Table 3 shows that the (\;3 term and the anharmonic effects in the interaction terms
of eq. (39) can only be compensated for in part by the anharmonicity constants X ij .
The r.m.s. deviation increases from 1.16 to 4.49 and 20.42 cm- 1 for Fits 1, 2 and 4,
respectively.
112
Table 4. Fit to vibrational excitations of CH 4 The values of the parameters are given
in the second column of Table 3. Here b.E = Eca/ - Eexp. The experimental energies
are taken from [28]. The wave numbers are given in em-I.
r (Vl, 1.12, V3, V4) Eca/ Eexp tlE r (VI, V2, Va, V4) Eoa/ Eexp tlE
113
be eliminated exactly. It will be important to further explore the scope and applica-
bility of the present approach. A more extensive study of methane including rotation-
vibration couplings, states with a higher number of quanta and transition intensities is
in progress.
ACKNOWLEDGEMENTS
We thank Prof. J. C. Hilico for his interest and for making available to us his
compilation of observed level energies. This work was supported in part by the Euro"
pean Community under contract nr. CI1*-CT94-0072, DGAPA-UNAM under project
INI05194, CONACyT-Mexico under project 400340-5-340lE and Spanish DGCYT un-
der project PB95-0533.
REFERENCES
[IJ J. M. Hollas, Modern Spectroscopy, John Wiley 1992; Kazuo Nakamoto, Infrared
and Raman Spectra of Inorganic and Coordination Compounds. Wiley - Interscience
publication, 1978.
[2] G. Herzberg, Infrared and Raman Spectra of Polyatomic Molecules, van Nostrand,
New York, 1950.
[3J E. B. Wilson, Jr., J. C. Decius and P. Cross, Molecular Vibrations, Dover, New
York,1980.
[4] W. T. Raynes, P. Lazzeretti, R. Zanesi, A. J. Sadly and P. W. Fowler, Mol. Phys.
60 (1987), 509; G.D. Carney and R.N. Porter, J. Chern. Phys. 65 (1976), 3547.
[5] D. L. Gray and A. G. Robiette, Mol. Phys. 37 (1979), 1901.
[6] T. J. Lee, J. M. 1. Martin and P. R. Taylor, J. Chern. Phys. 102 (1995), 254.
[7] J. F. Ogilvie, J. Mol. Spectrosc. 69 (1978), 169; W. Meyer and P. Botschwina, J.
Chern. Phys. 84 (1986), 89l.
[8] F. Iachello, Chern. Phys. Lett. 78 (1981), 581; F. Iachello and R. D. Levine, J.
Chern. Phys. 77 (1982), 3046.
[9] F. Iachello, S. Oss and R. Lemus, J. Mol. Spectrosc. 146 (1991), 56; Ibidem 149
(1991), 132.
[10] F. Iachello and S. Oss, J. Mol. Spectrosc. 142 (1990),85.
[11] O. S. van Roosmalen, 1. Benjamin and R. D. Levine, J. Chern. Phys. 81 (1984),
5986.
[12J F. Iachello and S. Oss, Phys. Rev. Lett. 66 (1991), 2976; F. Iachello and S. Oss,
Chern. Phys. Lett. 187 (1991), 500; F. Iachello and S. Oss, Chern. Phys. Lett.
205 (1993), 285; F. Iachello and S. Oss, J. Chern. Phys. 99 (1993), 7337.
[13] J. M. Arias, A. Frank, R. Lemus and F. Perez-Bernal, Rev. Mex. Fis. 41 (1995),
703.
[14] R. Lemus and A. Frank, J. Chern. Phys. 101 (1994), 8321; A. Frankand R. Lemus,
Phys. Rev. Lett. 68 (1992), 413.
[15] R. Bijker, A. E. 1. Dieperink and A. Leviatan, Phys. Rev. A52 (1995), 2786.
114
[16] F. Perez-Bernal, R. Bijker, A. Frank, R. Lemus and J.M. Arias, Chern. Phys. Lett.
258 (1996), 30l.
[17] A. P. Rendell, T. J. Lee and P. R. Taylor, J. Chern. Phys. 92 (1990), 7050.
[18] S. Carter and W. Meyer, J. Chern. Phys. 93 (1990), 8902.
[19] A. Frank, R. Lemus, R. Bijker, F. Perez-Bernal and J. M. Arias, preprint chem-
ph/9604005, Ann. Phys. (N.Y.) 252 (1996), in press.
[20] F. Perez-Bernal, J. M. Arias, A. Frank, R. Lemus and R. Bijker, preprint chem-
ph/9605001, submitted to Journal of Molecular Spectroscopy.
[21] R. Lemus, F. Perez-Bernal, A. Frank, R. Bijker and J. M. Arias, preprint chem-
ph/9606002, submitted to Chemical Physics Letters.
[22] Y. Alhassid, F. Gursey and F. Iachello, Ann. of Phys. 148 (1983), 346.
[23] F. Iachello and R. D. Levine, Algebraic Theory of Molecules, Oxford University
Press, 1995; A. Frank and P. Van Isacker, Algebraic Methods in Molecular and
Nuclear Structure Physics, Wiley, New York, 1994.
[24] S. L. Altmann and P. Herzig, Point Group Theory Tables, Clarendon Press, Oxford,
1994.
[25] K. T. Hecht, J. Mol. Spectrosc. 5 (1960), 355.
[26] A. P. Rendell, T. J. Lee and P. R. Taylor, J. Chern. Phys. 92 (1990), 7050.
[27] G. Herzberg, The Spectra and Structures of Simple Radicals, Dover, 1971.
[28] J. P. Champion, J. C. Hilico, C. Wenger and L.R. Brown, J. Mol. Spectrosc. 133
(1989), 256; J. C. Hilico, J. P. Champion, S. Toumi, V. G. Tyuterev and S. A.
Tashkun, J. Mol. Spectrosc. 168 (1994), 455; J. C. Hilico, S. Toumi and L. R.
Brown, in: Laboratory and Astrophysical High Resolution Spectra, A. S. P. Conf.
Ser. 81 (1995), 322; J. C. Hilico, private communication (March 26, 1996); M.
Chevalier, A. de Martino and F. Michelot, J. Mol. Spectrosc. 131 (1988), 382; D.
Bermejo and S. Montero, J. Chern. Phys. 81 (1984), 3835; J. S. Margolis, Appl.
Opt. 29 (1990), 2295.
[29] Zhong-Qi Ma, Xi-Wen Hou and Mi Xie, Phys. Rev. A53 (1996), 2173.
115
DYSON BOSON-FERMION REALIZATION
OF LIE (SUPER)ALGEBRAS
INTRODUCTION
The aim of the present paper is to construct Dyson type [1, 2J boson-fermion realiza-
tions (BFR) of the arbitrary finite-dimensional contragredient Lie (super)algebra [3, 4J
and to discuss how these realizations can be used to determine the matrix representa-
tions of (super)algebra. (Here we use the name "(super)algebra" in the following sense:
It includes all semisimple finite-dimensional Lie algebras and all finite-dimensional COll-
On leave of absence from Institute for Nuclear Physics at Moscow State University, Moscow
119899, Russia
PRELIMINARY
(1)
(2)
(3)
(4)
for i,j = 1,2, ... ,1'; a E fl.+; (3,,,/ E fl., where coefficients Nf3'Y are called the structure
constants. Here the bracket [e, e], called a supercommutator, has the properties
(-1 )0(a)0(c) [a, [b, ell + (-1 )0(c)0(b) [e, [a, bll + (-1 )0(b)0(a l [b, [e, a]] = 0 (6)
118
for all the homogeneous elements a, b, c E g.t)
Let Ug be the universal enveloping algebra of g. This algebra U(g) may be consid-
ered as an unital associative (super)algebra generated by the Cartan-Weyl basis hi, e"
with the defining relations (1)-(4), where the supercommutator is defined by the fol-
lowing explicit formula
[a, bJ = ab - (_1)8(a)8(b)ba (7)
for all homogeneous elements a,b E g. By using eqs. (1) and (2), we can easily obtain
the more general relations in U(g).
(8)
(11)
and
(12)
(13)
where E = 0 or 1 and we use the notation 0(a) := 0(e a ). The mapping (11) and (13)
is called the hermitian (grade) conjugation.
Let pj(g) be an elementary representation or a Verma module of 9 [19J. The rep-
resentation pj (g) is a highest weight cyclic one realized in U (n_), where n_ is a linear
span of basic vectors e_ a , a E 6+. Moreover, the extremal or highest weight vector of
this representation coincides with the unite of U(n_), i.e.
(14)
for i = 1,2, ... ,r and any a E 6+. The operators pj(h;) and pj(eh) act on arbitrary
elements u E U( n_) by the formulae
(16)
for any a E 6+. In the last equation the element u is proposed to be of the form
u = I:,6E~+ e_,6u~,u~ E U(n_). For the sake of convenience we put
(18)
t The elements of go and gl are called even and odd, respectively. Elements that are either even
or odd are called homogeneous.
119
Let us choose a basis in the space U( n_) of the elementary representation Pi> consisting
of the following vectors
(19)
where the order of factors in the product coincides with a normal (or convex) ordering
in .6.+ (see [IS]). Such a basis will be called an ordered one. We introduce also the
(super )symmetric basis:
In ) >(s)-In
, .- 0'1'
n 0'2"'" n erN'.)" >(8)._
.-
~s'(
1,
n.
II Ena) I)
-O!
> (20)
"Ei>+
(21 )
According to Perelomov [7] the GeSs for arbitrary complex simple Lie group G are
defined by the relation
Ie) = T(a)'lIo , (22)
where T(a) is the representation operator of the element a of the group G, and \jio is
some vector (highest or lowest vector as a rule) of the representation. Using Gauss
decomposition of the algebra 9
(23)
(24)
where XO! are complex parameters, and 'lIo is highest vector of the representation.
Eqs. (22)-(24) can be generalized to the case of the superalgebras. Since the generators
eO! are graded (even or odd), the parameters XO! should be also graded and belong to a
Grassmann algebra. Let us introduce this algebra.
Let 9 be the complex Grassmann algebra [3] with generators 6, e2, e3, ... ,eN and
relations
(25)
The element
(26)
m2: 0 i1 <12 <"""<jm
120
is called an even one if the terms with odd min (26) are equal to zero, and it is called
an odd one if the terms with even m are equal to zero .. The set of even elements will be
denoted by 90 and the set of odd elements by 91(9 = 90 E9 91)' Both of odd and even
elements are called homogeneous. The parity function (}(1]) is defined on homogeneous
elements by the formula
(27)
(28)
The Grassmann envelope with the permutation relation (28) is called the Grassmann
envelope of the second type [3].
Two types of a semilinear mapping 1] --+ 1]* and 1] --+ 1]0 that conserve the graduation
can be defined in the Grassmann algebra 9, with the properties
(30)
and
(34)
and
(35)
121
where Ij} is a highest weight vector in the U(n_) algebra and E-cxeC/ E g(9),a E !ltv'
This definition implies that only even U(g(9)) elements 'are present in the powers of
the exponents in (34) and (35), because the Grassmann envelope g(9) consists of even
elements by definition.
Note that we will drop out indices 0 and s in the designations of the GCS, bases,
and other quantities, provided the corresponding expressions are valid both for ordered
and symmetrical forms.
According to the permutation relations (28) for the superalgebra generators and the
Grassmann algebra elements, the following formula can be written down
(36)
Individual factors in the r.h.s. of (34) are normally ordered. The different normal
orders generate the different GCSs (34). The connection between the sets of the GCS
parameters corresponding as to the different normal orders as to the symmetric form
of the GCS can be obtained using the Backer-Campbell-Hausdorff (BCH) formula [21],
In the case under consideration it is not necessary to generalize the BCH formula to
allow for some peculiarities connected with superalgebras, because the elements of the
Grassmann envelope g(9) are even according to the definition. Also note that the GCS
(34) and (35) do not formally belong to U(g), because only finite series belong to U(g),
whereas the exponents in the definition of GCS result in an infinite series.
~
Now we define the functional image f (A) of some operator A (Ec. or Hi) as
the functional image f (A) includes only first-order partial derivatives and may be
found using two formulae
Aexp(C) = exp(C)(A
(-It
+ L -,-['" [C,A] ']n), (38)
n.
n=l
exp(C) 8a= exp(C)(C 8c. + L / _I)n), [... [C, C 8c.] "]n), (39)
n=1 n +1 .
where 8,,=8 /01]" denotes the right derivative. Since B( C) = 0, ordinary commutators
[A,C] = AC - CA (40)
are only used in (38) and (39).
It is easy to see that the following relation can be obtained from formulae (38) and
(39):
(41 )
In the case [E_ a , E-c.] = N a ,,,E_ 2 c. = 0 this formula leads to the ordinary expression
(42)
Now we can write down the explicit formulae of the functional realization of the gener-
ators of 9 superalgebra.
The factorized (ordered) form of the GCS (34) leads to the following results [6]:
p-l
122
N
N
fo (h;) = ji - 2:=(al,ai) 8al (aI' (45)
1=1
~ ~ Ba,(I")
~ Oa n
(
an ~C> + a.,) ~'"
')1'
-
(as, as)
--2- OCt,
~ 1'2
~a,
n=l
N N
- 2:= (a., at) 8al ("I(Ct, + 2:= [Jar B~:((Ct) , (46)
where ginv is an invariance algebra of the highest weight vector U), a E tl + are a simple
roots and
I'kll'k~ I'kL,
B- Ct p(1' ) - '"' ~l ~2 ", ~l-l .Nkl Nk~ , , ,NkL , (47)
Ctl ~Ct - ~ kl kl '" kl , -Ctp,-CX, -cxp-kla,,-Ct2 -cxp-ZI.:2, klcx;,-c'! '
{kl} 1 2 I-I' .-
k-l
N!,(3 = II N,,+i(3,(3 , (48)
i:;:;:O
I-I
provided that there exist all intermediate roots a p + aI, a p + 20'1, , , , , a p + ki al, a p +
ki al + a2 . .. and so on. The summation in formula (47) is extended on all possible
sets of {kl} satisfying (49).
The symmetric form (35) of the GCS leads to the result [5]
We have also for the generators e" belonging to the invariance algebra ginv the
following expression
Is (e,,) = 2:= N_(3,cx 8-(3+0' e(3 . (51)
(3EL1;nv
For the lowering generators e_" (a E tlf;,v) we obtain
dk = 1,-1/2,1/12,0" .. , (53)
123
and the sums over aI, ... , ak are performed over all roots for which (31, ... ,(3k are roots
and ai E .6.7;. .
As for raising generators e,,( a E .6.;;'.) the explicit expression will be given only for
the functional realization of generators corresponding to simple roots
It is interesting to note that the coefficients dk in (53) and d~ in (55) are connected
with the Bernouilli numbers Bk in the following manner
(57)
The property
B 2k + 1 = 0, k = 1,2,3, ... (58)
of the Bernouilli numbers simplifies the formulae (52) and (54).
It is clear that the bases (19) and (20) are non-orthonormal, but they may be used
for the construction of the matrix representation of the generators of the (super ) algebra.
The explicit expression of the matrix representation of the Hi and EOI operators in
the bases (19) and (20) could be obtained by direct calculation. For the basis (19) it
was done in ref. [6], however for the symmetrical basis (20) such a procedure is rather
cumbersome.
It appears very convenient to use the GCS as a generating functions for the vectors
of symmetric and ordered bases and then to find the matrix representation of the
superalgebra generators (A = Eh or Hi) in an ordered and symmetric bases, using the
functional representations found above.
124
To do it, we expand exponents in the formulae (34) and (35)
Ie; 0) = II
0:
f
no::::O
~(E_"e"ta Ij) = L L In; o)e:~N ... e:~'
nC(- n n
, (60)
Ie,s) = L ~(LK"(,,tlj)
n,
= LL In,s)(:~N .. . (;~, , (61)
. n
is used. Here the sum is extended on all possible non-negative integers n = (nl' n2, ... ,
nN) satisfying the condition nl + n2 + ... + nN = n.
Let us introduce the monom of the form
(63)
which corresponds to the normal ordered basis ({ ry,,} = {e,,} ) or to the symmetric basis
({ry,,} = {(ex}). In the latter case the order of factors is assumed to be the same as in
the definition of symmetric basis (20). Hence the GCS can be written as
Now we substitute this expression in the definition (26) of the functional realization
The action of the functional image f (A) on the monom (63) can be realized in direct
manner and the result may be represented in a form
where k, A~_k 11 are the same as in eq. (66). In other. words, to find the matrix rep-
resentation of ~ome operator A it is necessary to identify the coefficient A n - k 11 in the
expansion (67) with the coefficient A~_k n from eq. (66). '
Using the evident relations '
(nl7)ap = (n + 1a p l( _l)B("p) Lf:,' B(",;) , (68)
125
USUI OPERATOR TECHNIQUE
For Lie algebras the transition from functional realization to the boson realization
is fulfilled by means of Bargmann representation [22] for the boson operators
b,
<- +
b .
x ~ a~ (70)
As a result, a so-called elementary BR (EBR) is obtained that is connected with the
DBR through an adjoint operation
(71)
In order to construct the BFR of Lie super,algebras, this procedure should be general-
ized.
Let COt, C;;( a E ~ +) be boson and fermion creation and annihilation operators
satisfying the ordinary commutation relations
The second type is a generalization of the usual adjoint operation, i.e. the grade adjoint
operation
(C.,)# = (-1)I'O(Ot)C;;, (C:)# = (_l)(I'+l)O(a)ca ,
Ca)#)# = (-l)O(a)ca, (C",C{3)# = (-1)8(a)8(,s)(C,s)#(Ca)#, (74)
where J.L = 0 or 1.
In analogy with the ordinary harmonic oscillator CS, the GCS for HW superalgebra
is of the form
Iz) = exp (L ctz",)IO), (75)
aElI+
(76)
After substituting the factors z'" by the operators C'" and the derivatives 80t
by the
operators C;; in the formulae of the functional realization of operators Ea and Hi, the
EBFR pE(ea),pE(hi) can be obtained. It is clear that, in principle, the intermediate
functional realization step can be omitted and the EBFR can be directly found. To do
it, the Usui operator concept may be used [23, 2].
126
The vectors of the Fock boson-fermion space (BF) will be denoted as
The bra-vectors
(nl = (01 II C~Q (78)
CteL1tnv
and
Us = L Lin; s)(nl = (01 exp( L E_"Ca) Ii). (82)
n n CtEA.tnv
(83)
with A = Ea or Hi gives the representations realized in terms of BF operators and
is equivalent to the representation A described in section 2. The BFR given by the
formulae (83) will be refered below as EBFR. It can be derived in analogy with the
functional realization.
Similarly to the permutation relations (28) it is possible to introduce the permuta-
tion relation for boson-fermion operators Ca, ct
and some operator A (Ep or Hi)
C"A = AC,,( -1 )8(a)8(A) , (84)
which means that the following supercommutators vanish
In analogy with the adjoint operation (30) and (31) let us define the corresponding
adjoint operation
(86)
and
(CaEa + C(JE(J)+ = (Ea)+(Ca)# + (Ep)+(Cp)# . (87)
It can be readily verified that
(88)
127
Using the adjoint operation "+", the transition from EBFR to DBFR can be done.
Let us find the relations connecting EBFR and DBFR. The application of the adjoint
operations for both sides of (83) leads to two pairs of relations corresponding to the
adjoint and the grade adjoint operations in Lie and HW superalgebras, respectively:
(89)
and
(pE(A))#(U)+ = (U)+(A)+ . (90)
From (89), a rather natural formula connecting EBFR and DBFR follows
(91)
(92)
However, in fact, it provides a set of DBFR because the grade adgoint operation "+"
in the Lie superalgebra and the operation "#" in the HW algebra can be chosen in a
different manner (i.e., f1t <:). Nevertheless all these versions are compatible with the
adjoint operation "+,, for combinations EaCa.
Formally speaking, the DBFR can be obtained using the adjoint Usui operator U+
instead of the operator U
(94)
Then the explicit form of the DBFR should be found from the equations
(95)
which are similar to eq. (83).
In the definitions (19) and (20) (for the U(n_) bases), (34) and (35) (for the GCS),
and (81) and (82) (for the Usui operators), the generators forming the highest weight
invariance subalgebra gin. (see (44)) are assumed to be omitted. Therefore for the
highest weight with a non-trivial gin. algebra the number of active bosons and fermions
operators used in the BFR is smaller than in a general case. Such shortened BFR were
called non-maximal BFR [24].
Note that in the ordered cases the boson realization of generators forming ginv
algebra can be obtained from the ones corresponding the full realization by omitting all
boson-fermion operators connected with the ginv, provided that in the definition of the
Usui operator Uo the generators included in ginv stan9. just before the highest weight
vector Ij).
It is appropriate to formulate the formal receipt to find the matrix representation
of some operator, provided its DBFR is known: each term (C,:t)kC", containing the
operator C", should be substituted by the factor (n", + 1 - k), and the substitution
n --+ n + la - k" should be done in the vector In); each term (C:)k leads to the
substitution n --+ n - k" in the vector In).
128
THE MIXED BOSON-FERMION REALIZATION
It was pointed out in ref. [24J that one can introduce a so-called mixed boson real-
ization, where only a part of the generators is mapped on their boson images. Recently
such boson realizations became very popular in connection with their fruitful applica-
tion in the microscopical nuclear collective model Sp(2d, R) [12]. Besides, mixed boson
realizations appear to be very useful in pure representation theory, for example, to
find the explicit form of IR in the bases corresponding to the non-canonical reductions
SU(3) :) SO(3) [25], SO(5) :) SU(2)x U(l) [26], SO(8) :) SU(4) [27], Sp(6) :) SU(3)
[28J etc.
The mixed BFR can be obtained in the framework of the method developed by us in
the previous sections, however some modification should be done, namely the formalism
of the vector coherent states [9J should be used instead of usual GCS.
Let K be a subalgebra embedded regularly in 9 algebra and containing the Cartan
subalgebra. The generalized Gauss decomposition can be fulfilled
(96)
(97)
Let us take some basis vector IjAA) belonging to some definite IR DA of the K sub-
algebra and construct the expressions
(100)
In the case of the regular embedding of K into 9 (only this case is under considera-
tion), the highest weight vector Ij) of IR Dj of algebra 9 should belong to the "highest"
representation DA.
For VCS the full theory can be developed similar to the theory of usual GCS (see
refs. [9,29,30,31]), but we do not intend to discuss all arising problems (the complete-
ness of VCS,the identity resolution, the uncertainty relations), because we use the VCS
theory only as an instrument to C(])l1struct a mixed BFR.
129
The consideration is carried out in the same scheme as in sect. 3, namely, we act
by the generators K" E_ a and M{J( M{J E J<) on the yeS (98) and (99). Then, us-
ing formulae (37) and (38), these generators should be permuted with corresponding
exponents. A single difference is that the generators M{J of the J{ sub algebra should
be consequently permutated with exponents in order to stand just before the vectors
of "highest" representation IjAA). At this position they should be substituted by the
so-called "intrinsic" generators M{J [9] acting in a standard manner on the basis vectors
Ij AA) of the "highest" irreducible representation DA of the J< subalgebra.
The permutation relations of the "intrinsic" J{ sub algebra generators and the gen-
erators of Grassmann algebra should be taken in the form
(101)
similarly to eq. (28). It assumes also that, similarly to (84) and (85), the supercom-
mutator of the "intrinsic" generators M{J with the boson-fermion operators Ca , ct
vanish
(102)
As a result the mixed BFR Pmix(E-a), Pmix(Ea), Pmix(M{J) of algebra 9 generators in a
form of polynomials of operators ct, ca , and M{J can be obtained.
Of course the operators of sub algebra J{ can be also represented by means of the
boson-fermion operators and substituting these expressions in the mixed BFR gives the
full BFR.
EXAMPLES
(106)
where 10) is the boson vacuum, Ij) is the highest weight of U(3) IR (Aiilj) = jilj)), and
B kl , B~t (v = 1,2 or sand k, I = 1,2,3) are the boson creation and annihilation
operators
i,
[Bk'l' Bft] = 818f + 8i 8 Bk'l = Elk , Bk't = B'l/ . (107)
Following the general scheme developed above, we can construct the DBFR using the
Usui operator U:. The boson images of the generators A 23 , A13 can be obtained
immediately
(108)
130
(109)
In order to find the boson image of the generator A 12 , one should commute it with
exponents exp(A23Bit) and exp(A13Bit) to stand just after exp(A12Bit)
(110)
The generators A 21 , A31, A32 are annihilating on the highest weight bra-vedor UI.
Hence, substituting them with exponents in order to stand them just after the vedor
UI, we can obtain the following boson images of these generators:
1+( . - )2. - N12
PID(A 21 ) -- B 12)1 1 ) - B1+B1
13 23, (111)
PI 31 - 23 12]I - )2 - N 12
D(A ) - B1+B1+( . . 1 ) + B1+(
13)1
1 - N 13
. - )3. - N 1Z 1 -
1 ) (112)
N 23'
pf(A32) = Bit(j2 - j3 + N{2 - N{3 - Ni3) + Bit Bi2 , (113)
where Nt = Bt+ Bi~(k = 1,2, s). Similarly we can find the boson images of the Cart an
subalgebra generators
pf(All) = jl - N{z - N{3 , (114)
pf(A22) = j2 + Ni2 - Ni3 , (115)
pf(A33) = h + Ni3 + Ni3 . (116)
The boson images generated by the Usui operators U: and ut can be found in a
similar way. The choice of the Usui operator in a form ut leads to the following boson
realization
p?(AI3 )=Bi3' (117)
p?(An) = Bi2 , (lIS)
P?(A23) = B~3 + B;t B;3 , (119)
p?(A32 ) = B~t(j2 - j3 - N~3) + Bit Bi2 , (120)
pf(A31) = B~t Bit(j2 - h - Ni3) + Bit(jl - h - N122 - Ni3 + Ni3)' (121)
pf(A21) = Bit(jl - j2 - N;2 - N;3 - Ni3) - Bit Bi3 , (122)
P?(All) = j1 - N;2 - N;3 , (123)
= j2 + Ni2 - Ni3 ,
p?(A 22 ) (124)
pf(A33) = j3 + Ni.3 + N;3 . (125)
The symmetric form of the Usui operator U: results in the expressions:
(126)
131
. )Bs+ S+B S+B S
B 23 1 B S+BS+)Bs+ BS
(J2. - J3 23 - 23 23 -"21 (Bs+
13 -"2 12 23 23 13
S+ . 1 ( .
. )B13 2' )B s+BS+ (Bs+ s+BS
(11. - 13 + "2 11 + J3. - J2 12 23 - 13 -"21 BS+
12
BS+)B
23 23 23
-
s+
(B 13 + "21 BS+Bs+)Bs+Bs
12 23 12 12 -
(Bs+Bs+
13 13
+ 4"1 (Bs+Bs+)2)BS
12 23 13 ,
(130)
S
-
(8 8+
. 13 -"21 BS+Bs+)B
12 23 23' (131)
The comparison of the Lh.s. of the last expression with the symmetric form of the Usui
operator Us+ leads to the relations
s+ 1+ B 8 + B1+
B 12 = B1+
12'
B 23
S+
= B 23, 13 = 13 +"21 B1+B1+
12 23'
B 12 S
1 -- B 12 + 2"I Bs+Bs
23 13'
B123 -- B 23 + 2"I Bs+Bs
S
12 13'
S
B113 -- B 13' (135)
The connection between pf(Aij) and p?(Aij) can be found using the permutation
of two exponents exp(A12Bin and exp(A23B5t):
132
The non-maximal boson realizations that corresponds to the case .il = j2 or j2 = h
:an be obtained from pf(Aij) and p(Aj) by simple cancellation of terms containing
boson operators Bi2' Bit or B~3' B;t, respectively.
There are two forms of mixed boson realizations for U(3) algebra. For example, if
the subalgebra f{ is chosen as the sum of the Cartan subalgebra h( HI = All - An, H2 =
AZ2 - A 33 ) and the algebra SU(2) is formed by the generators A 2b A 12 , HI = An- A 22 ,
the Usui operator U;;;ix can be written as follows
(138)
where IH R) is the "highest" representation of f{, i.e. the representation containing the
highest weight vector Ij) of the U(3) algebra. The application of the general technique
described above gives the result
p~ix(AI3) = Bn , (139)
p~ix(A23) = B 23 , (140)
p~ix(A31) = Bi;Azl + Bt3(HI + Hz - N12 - N 13 - N 23 ), (141)
p~ix(A32) = Bi;(H2 - N 13 - N 23 ) + BtA12 , (142)
p~ix(A12) = A12 - BtB23 , (143)
p~ix(A21) = AZI - Bi;B13 , (144.)
p~ix(Hd = HI + N 13 + N 23 , (145)
p~ix(H2) = H2 + 2N23 + N13 . (146)
If the intrinsic operators A 12 , A 2l , HI, H2 in (139)-(146) will be substituted by their
Dyson boson images
pD(A2l) = B12 , (147)
n D(.A'0) = BtUl - j2 - Nn), (148)
pD(Hl) = jl - j2 - 2N12 , (149)
pD (H 2 ) = j2 - j3 + N12 , (150)
the total DBR (108)-(116) are obt.ained.
Let us consider the osp(2/1) superalgebra. Its generators satisfy the commutation
relations:
1 1 1
[10, VJ = 2"V , [V, VJ = 2"h , [V+, V-l = -2"Jo. (151 )
The operators J, Jo are even, but the generators V are odd. There are two lowering
operators J_ and V_. Therefore the Usui operator is of the form (because of the
commutativity oJ Land V_ the Usui operators Uo and Us are the same)
U = (O/exp(Lb)exp(V_a)/j), (152)
(153)
133
pE(JO ) =j - b+b - ~a+a, (154)
pE(J+) = (2j - b+b - a+a)b, (155)
1
pE(V_) = a+ - 4"b+a, (156)
(158)
If the grade adjoint operation for the fermion operator is taken in the form
pD(J+) = b, (160)
pD(V- ) = (-I)'*(~b+ba
4
+ b+a - ~ja+)
2' (164)
ACKNOWLEDGMENTS
134
For simple roots (as E ~t..) the following expression can be found from (46):
s-2 ( )
s s -1
p (E", )1 n,o ) -"""'
- ~B"n",,(1)1
k",n n-k,o ) + [( as,].) - -a.,a
- 2 - n- 2 -'
n=1
N N
-L (as,al)n",]ln -1"",0) + ~B~:(m~,n)ln - ml,o), (168)
l=s+1 r=s+1
where
(169)
For lowering generators E_",(a E ~t.v) we obtain from eq. (52) the result
k k
135
k k
- L 8(,Bk+l - a;) - 8(,Bk+l - 8) - 8(,Bk+1 - a)) /n - L la, + l{Jk+l - 1" - 1d, (173)
i=l i=l
where
(174)
As an example we consider the matrix representation of the U(3) algebra in sym-
metric basis. Making use of the DBR given by the formulae (126)-(134) and the known
rules of the derivation of the matrix representation, we obtain the following results
REFERENCES
136
V. N. Tolstoy, in: Symmetries and Nuclear Structure, R. A. Meyer and V. Paar
(eds.), Harwood Acad. Publ., 1987, 24.
[7] A. M. Perelomov, Generalized Coherent States and Their Applications, Springer,
Berlin, Heidelberg, New York, 1986.
[8] J. Wei and E. Norman, J. Math. Phys. 4 (1963), 575.
[9] D. J. Rowe, J. Math. Phys. 27 (1984), 2662; K. T. Hecht, Lecture Notes in Physics
290, Springer, Berlin, Heidelberg, 1987.
[10] F. Iachello, Phys. Rev. Lett. 44 (1980), 772; F. Iachello and O. Scholten, ibid. 43
(1979), 679.
[11] A. B. Balantekin, 1. Bars and F. Iachello, Phys. Rev. Lett. 47 (1981), 19; Nucl.
Phys. A 370 (1981), 284.
[12] G. Rosensteel and D. J. Rowe, Ann. Phys., NY 126 (1980), 343.
[13] K. Ito and S. Komata, Mod. Phys. Lett. A 6 (1991), 581; B. L. Feigin,
E. V. Frenkel, Commun. Math. Phys. 128 (1990), 16I.
[14] B. Gruber and A. U. Klimyk, J. Math. Phys. 25 (1984), 755.
[15J E. Chacon and M. Moshinsky, J. Phys. A: Math. Gen. 20 (1987), 513, 4595.
[16J O. Castanos, E. Chacon, M. Moshinsky and C. Quesne, J. Math. Phys. 26 (1985),
2107.
[17] M. Scheunert, W. Nahm and V. Rittenberg, J. Math. Phys. 18 (1977),146.
[18] R. M. Asherova, Yu. F. Smirnov and V. N. Tolstoy, Mat. Zametki 26 (1979), 15.
[19J J. Dixmier, Algebres enveloppantes, Gauthiers-Villars, Paris, 1974.
[20] F. A. Berezin and V. N. Tolstoy, Commun. Math. Phys. 78 (1981), 409.
[21J see for example R. Gilmore, Lie Groups, Lie Algebras, and Some of Their Appli-
cations, Wiley, New York, 1974.
[22J V. Bargmann, Comm. Pure Appl. Math. 14 (1961), 187; Comm. Pure Appl.
Math. 20 (1967), I.
[23] T. Usui, Prog. Theor. Phys. 23 (1960), 787.
[24J M. Havlicek and W. Lassner, Rept. Math. Phys. 12 (1977), 1; P. Exner,
M. Havlicek and W. Lassner, Czech. J. Phys. B 26 (1976), 1213.
[25J R. Le Blanc and D. J. Rowe, J. Phys. A: Math. Gen. 18 (1985), 189I.
[26J Yu. F. Smirnov and V. N. Tolstoy, Rept. Math. Phys. 4 (1973), 97.
[27J K. T. Hecht, Nucl. Phys. A 444 (1985), 189.
[28J K. T. Hecht, J. Phys. A: Math. Gen. 18 (1985), L1003.
[29J D. J. Rowe, J. Math. Phys. 25 (1984), 2662.
[30J D. J. Rowe, G. Rosensteel and R. Gilmore, J. Math. Phys. 26 (1985), II.
[31J J. Deenen and C. Quesne, J. Math. Phys. 25 (1984), 8.
137
FORMAL LANGUAGES FOR QUASICRYSTALS
Departamento de Fisica
Universidad de Oviedo
E-33007 Oviedo, Spain
and
Departamento de Fis-ica
Universidad Europea-CEES
Campus Universitario
E-28670 V. de Odon, Madrid, Spain
INTRODUCTION
Rotational symmetries of order five or higher than six are not compatible with
translations in the plane or in 3D space so they can not be found in ordinary crystals
which are periodic. This fact is known as crystallographic restriction.
A crystal has been defined recently, as a solid with an essentially discrete diffraction
diagram. The discrete part is called the Bragg spectrum. When we speak about its
symmetry we refer to that appearing in the diffraction diagram. A quasi crystal is a
crystal with a diffraction diagram showing symmetries forbidden by the crystallographic
restriction [1].
A structure has long range order, if its Fourier transform has discrete component i.e.
delta functions supported by a module over the integers. The existence of quasicrystals
contradicts the idea that periodicity and long range order are equivalent.
Tiling theory is one of the methods more accepted for the study of quasi crystal
structur~s. Although already in 1974, R. Penrose proposed a tiling of the plane with
fivefold symmetry[2J, it was the discovery of quasicrystals in 1984 [3] that stimulated
more the development of the theory of aperiodic tilings. The reason is that some of
them have long range order.
There are several methods for the generation of aperiodic strutures. Symmetries of
order five or higher than six can be compatible with traslations, if the space dimension is
increased. Projection methods allows to obtain tilings in a space of arbitrary dimension
p,by projection of p-facets ftom lattices in dimension n > p [4,5,6]. The number of basic
tiles could be very high if the orientation of the p-plane in the n-space is arbitrary. But
if the p-plane is invariant under a subgroup of the group of symmetry of the lattice in n
Since the time of the ancient world the golden section has played an important role
in both science and art [9]. In geometry it appears in relation with the regular pentagon
and two of the five regular solids that Plato adscribed to the shapes of the four elements
and to the cosmos. In number theory the golden section is related with the Fibonacci
sequences. It is also found in plant growth and in the proportions of the human body.
In art, although sometimes not calculated and present only in the subconscious of the
artist as a natural principle of form, it has been used in architecture (Partenon, Firenze
cathedral), painting (Leonardo in La Gioconda), poetry (Holderlin in Die Aussicht) and
music (Bela Bartok in the Music for strings, percussion and celest or in the Sonate for
two pianos and percussion).
The best known examples of aperiodic structures in 1D are the Fibonacci chains.
They can be obtained [7] by projection on a straight line from a parallel band in a two
dimensional square lattice, if the slope of the line is the inverse of the golden section
r= 2cos7r /5= (1 + v'5) /2.
Fibonacci chains can be generated also by substitution. Consider the alphabet
140
E = {L, S}, the set of production rules
L J----+ L S, S J----+ L (1)
and the aXIOm w = L. The language consists in the words L, LS, LSL, LSLLS,
LSLLSLSL, .... Every letter represents a segment such that (length of L)/(length
of S)= T. The infinite chain is selfsimilar: a change of scale by a factor T, produces
another sequence represented by the same word. The geometric interpretation in 1D is
clear: if two letters follow one another, the corresponding segments will appear together
in the structure.
Another example of a selfsimilar structure can be obtained with the alphabet E =
{U, V, W}, axiom w = U and production rules
If lu Ilv = {3 and Iv Ilw = {32 - (3 - 1 = 2 cos(27r 17) then the scale factor is (3 =
1 + 2 cos(27r 17) and satisfies the equation {33 = 2{32 + (3 - 1.
PENROSE PATTERNS
The Robinson decomposition of the Penrose patterns uses two triangles as basic tiles.
A colour is associated with every edge X and this is denoted by Xi with i = 1,2,3,4.
T(A, B, C) represents a triangular tile with edges A, B, C.
It is possible to describe the patterns in a similar way than in previous section for
the one dimensional case. The alphabet E = {ai, bi, ai, b;, (,)} contains two brackets
and letters of type ii and 1:; which represent tiles with the same shape but they are
distinguished by a colour: white for ii, black for 1:;, where i E ZlO denotes relative
orientation.
The tiles T(LI S4 L4) T(L 3 S3 2) T(S4 S2 U) and T(SI S3 L 3) are repre-
, " '" " , ,
sented by letters of type a, a, band b respectively. The orientation is chosen such that
the tiles T(A, B, C) represented by letters with subindex 1 have the edges A and C
oriented in the same way. The oriented tile with subindex i is obtained by rotation of
7r(i - 1)/5 through the vertex An C.
Every element belonging to E representing a tile can be used as axiom. We take for
instance w = al' The set of production rules for Penrose Patterns is:
R1 (ai) = a9-i R1 (b i ) = b3 - i
R 2 (ai) = a7-i R2 (b i ) = b1- i
141
it can be seen that the inflation represented by cp p satisfies Rkcp'j'; = CP'j'; Rk for k = 1,2
and m ~ l.
The BKSZ patterns with five-fold symmetry were obtained in [11] by projection from
the root lattice A4 and they have the same basic tiles as the Robinson decomposition of
the Penrose patterns. BKSZ patterns and the class of patterns with eigth-fold symmetry
due to Ammann [12] can be described also in terms of DOL systems [13]. A different type
of grammar can also be used: context free grammars (see [14], [15] and the Appendix).
A class of patterns with seven-fold symmetry and matching rules have been in-
troduced in [16]. For a given prototile, different arrows are placed on the edges of
length U, V, W which are labeled with 0 or 1 depending whether the arrow orientation
is anticlockwise or clockwise respectively. If we choose anticlockwise orientation, the
substitution rules for the edges are
cpn[b] = [r 3 , ((3n-l - (3n+l )U6]cpn-l [b] U [r, _(3n U6 ]cpn-l [d] U [1, 0] cpn [C]
(5)
cpn[c] = [1, o]cpn-l [a] U [r8, ((3n+l - (3n-l )uo]cpn-l [b]
<l>n[d] = [r8, (3n+1uo]cpn[b] U [r9, (3n+l uo ]cpn-l [b] U [r3, _((3n+1 + (3n)us]cpn-l [aj
U[rlO, (3n-l uo - ((3n+1 - (3n-l )U6]cpn-l [d]
There is also an interpre~ation of the patterns in terms of DOL systems. The alphabet
K,
is ~ = {ai, bi, bi, Ci, di, ai, bi , '2;, d;, (,)} with i E Z14 and ti = [ri-t, O]t. The production
rules can be obtained from eq(5). For instance bi f---+ (cp[biD = (Ci+6bi-6d;+6di-S)
142
The generators of the Coxeter group H(7) acts on the symbols representing tiles as
follows
Rl(Xi) = X11-i R 2 (Xi) = X9-i
(6)
Rl(Yi) = fh-i R 2 (Yi) = fh-i
where x = a, b, band Y c, d. In this way the inflation represented by iji satisfies
Rkijim = ijimRk for k = 1,2 and m 2:: l.
Now I prove that the tilings are facet to facet. If the mirror image of the word a
is denoted by Mir(a), and the projection of Lj into L by P(Lj) we have the following
properties 1) Mir( a) = f3 if and only if Mir(f3) = a
2) Mir( af3 ... 'Y) = Mirb) ... Mir(f3) Mir( a),
3) P(af3 ... 'Y) = P(a)P(f3) ... Pb)
Only LO and Ll are common edges of two adjacent tiles. Therefore the fact that the
tilings are facet to facet is equivalent to
(7)
This can be proved by induction on n for every Li. For instance, if LO = Uo, we
have for n = 1:
This tiling was obtained by inflation. On the other hand, Nischke and Danzer have
given a construction of an inflation rule for d-fold symmetry based on the system of
the tangents of the deltoid (17). It turns out that the information about the inflation
rules necessary to derive the heptagonal patterns is contained in the ND cgnstruction
if the arrowing is changed as I will explain below. Based on this observation, tilings of
the plane with nine-fold (which is not considered in the original ND construction) and
eleven-fold symmetry have been derived in (18). The construction contains a system of
lines G"" Jl = 0,1, ... d - 1. The angle between G", and the positive x-axis is equal to
27r III d. If d = 2m + 1, 5 v = sin( V7r Id), and [[y II denotes the integer part of y, the pairs
of lines
{G[["'t"-)) , Gd-[["'t"-))}, ... , {G2 , Gd- 2 }, {Gm - l , Gd-m+1}, { Gl , Gd-d, {Gm, Gd- m}
(10)
intersect Go on consecutive points separated by distances 5d-2,' .. ,55,53' The figure
obtained is a triangle pattern with edges 51,52, ... ,5 m .
In order to get the arrowing we construct, for every I} = 1,2, ... , m, a star Sv with
d segments of length 5 v inclined according to the system of lines G w For the star
corresponding to the longest segment 5 m we put arrows pointing outwards its center.
The arrows point towards the center, for the star Sm-l. The arrowing is then changed
alternatively, so that Sv and Sv-2 have the same.
If we do not consider the arrows, the substitution matrix M for the edges is a
triangular matrix with Mij = 0 for i + j < m and Mij = 1 otherwise. The eigenvalues
143
of Mare Ap = 8(2P-l)m/82p-l for p = 1,2, ... ,m. The largest eigenvalue is AI, which is
also the scaling factor of the tilings.
In the nine-fold symmetry case the edges of the basic tiles are H = 81, [ = 82,
J = 83,.K = 84. By taking into account the arrows as in the case d = 7, the following
substitution rules are obtained:
HO 1--+ q,[HO] = KO , HI I - - t q,[Hl] = Kl ,
[0 I--t q,[ID] = J OKl , Jll--t q,[PJ = KOJl ,
(11)
J OI - - t q,[JO] = [0 KO Jl , Jll--t q,[Jl] = J OKl Jl ,
KO ~ q,[KO] = ~JOKl[l, Kl I - - t q,[Kl] = [0 KO Jl HI .
If instead of arrows we put colours on the tiles, only two are necessary for d = 9
(four colours are necessary for d = 11 as in d = 7). Up to colour (or mirror reflection),
the basic triangles are in this case T([O,Ho,J l ), T(J\Hl,KO) and T(J,I,KO)
A species of 3D tilings with icosahedral symmetry have been proposed by Danzer [19,
20]. The prototiles are the tetrahedra T = A, B, C, K. They have the property that
every plane containing a facet t' (t =a, b, c, k and i = 1,2,3,4) of T is parQllel to one of
the fifteen mirror planes of a fixed regular icosahedron. The tetrahedra can be defined
by giving, for {t l n t 2, t 2 n t 3, t 3 n tI, t 2 nt4, t l n t4, t 3 n t4}, the pairs Ca, I) where a is
the angle between two facets and 1 the length of their common edge :
(12)
with p= ~ 4
and (J'= fl.
2
If we choose a cubic coordinate system based on a set of three orthogonal 2-fold
axes of the icosahedron, the Miller indices of a plane and .the direction perpendicular
to it are the same. I choose the following mirror planes:
PI = (T, 1,1 + T), P2 = (-2T,0,0), P3 = (-1- T,T, 1).
Eight basic tetrahedra represented by the letters 'Z = A, B, C, K, 1, ii, C, K are
used. The tetrahedra T and T are mirror images withT == T. The notation T;;: with
p, q, T, 8 = 1, 2, ... , 15 means that the tile T has his facets tl, t2, t 3 and t 4 lying on planes
parallel to the mirror planes Pp , Pq , Pr and p. respectively.
Consider the Coxeter group H3 = < RI, R2, R3/(Rl R2)3 = (R2R3Y' = (R3Rl)2 =
R~ = .m = R~ = e >. The reflections on the planes PI, P2 and P3 correspond to
the generators R l , R2 and R3 respectively. The mirror planes which are not fixed are
transformed under the action of the generators of H3 in the following way:
RI[PS ] = P2 ,RI [P7] = PU ,RtlP9 ] = PIS, RI[PlO] = P13 ,RI [Pd = P6 ,RI [P14] = P4
R2[Ps] = PI, R2[P7 ] = P6 , R2 [P9 ] = Ps, R2 [PIO] = Pu, R2 [P12] = P13 , R2 [P14 ] = P3
R3 [PS ] = P13 , R3 [P7 ] = P4 , R3[P9 ] = P6 , R3 [PlO ] = P2 , R3 [PuI = P15 , R3[P14] = Pu
(13)
The action of the generator Rk in T;;: is Rk[T;;:] = T;': if RdPy] = Pfi with
Y = p,q,T,S.
144
In [21] the tilings are interpreted in terms of languages generated by grammars of
type DOL. The alphabet contains two brackets and letters T;:;
representing tiles. The set
Jf production rules is of the type T;:; f - + (iIl[T;::]). For the tiles K = K~:;3, C = c;:io,
B = B1 I0,15 A A 41,3,9 we h ave
,3' =
iIl[K] = r,,[K]r;3[B]
iIl[C] = R1[K]R3[K]Rr[C]r6[C]A
(14)
iIl[B] = r-y1[K]R 81 [K]r,[B]R<1[B]Rr[K]ro[K]C
iIl[A] = iIl[rdBllR(3[C]R1,,[K]R3,,[KhdB]
145
Let u w( k) be the number of occurences of w in the first k letters of Wn and lw the
length of the segment associated with w for w = a, b.
A mass distribution T(x) is placed inside every tile. Its Fourier transform is defined
as usual
F(T(x)) = T( q) = exp( -iqx)T(x)dx .J (16)
The density function for atoms of unit mass placed on the extremes of the segments is
T(x) = L:k o(x -Xk) with Xk = laua(k) +hUb(k), k = 0, I, 2, .... The Fourier transform
tv
of a chain with N atoms is iN = L:k exp(iqxk)' It is possible to compute recursion
relations for the Fourier amplitudes along the lines of [22]:
f n+l- f n -- ",ttn+l-
l
L...k=ttn exp ('zqXk ) --
exp(iqxttn) + exp(iq(xttn + Xl)) + ... exp(iq(xttn + Xttn-d) +
exp(iq2xttn) + exp(iq(2xttn + Xl)) + ... exp(iq(2xttnxttn_l-I)) =
(17)
exp(iqxttn) ( exp(iqxo) + exp(iqxl) + ... + exp(iq(xttn-1)) +
exp(iq2Xttn) ( exp(iqxo) + exp(iqxt) + ... + exp(iqXttn_l-1)) =
exp( iqxttn)in + exp( iq2xttn)fn-1
Following the same method one can obtain recursion relations for other quasiperiodic
structures ( in = Xttn):
In what follows the Fourier transform for tilings in dimension higher than one is
studied. Under a linear isometry I or a translation v, the Fourier amplitudes changes
into
F(T(lx)) = T(l-lq)
(IS)
F(l(x + v)) = exp( -iqv)T(q)
The geometric description given in eq. (5) can be used in order to derive recursion re-
lations between Fourier amplitudes for the heptagonal tiling (the remaining amplitudes
can be obtained in a similar way):
146
where An(q) is the Fourier amplitude associated with <J)n[a], etc.
Recursion relations between Fourier amplitudes for the Danzer tiling can be obtained
by using eq. (15). If T,,(q) denotes the Fourier amplitude associated with the object
<J)n[T] we have:
Kn( q) = exp( _iqr n- 1pk 14 )Kn- 1(r;:;-lq) + Bn- 1(r~lq)
+ exp (-zq
. (1 T-1p)) + exp (-zq
r- pa23 + -2-a13 . (pa13 + r- 1pa14 + -Z-a24)))
T- 1 "
(21)
In [24] the atoms are placed at each vertex of the prototiles. One vertex of the
tetrahedron I< is not considered as an atomic site for the quasi crystal, and the prototile
is an octahedron consisting of eight tetrahedra I<. In our example, only one facet of
I< contributes to the total mass, if the octahedra is taken as prototile with atomic
decorations on its facets. In that case Ka(q) = !exp(-iq(r- 1pk I4 + 9k34)) (see [25]
for other decorations).
ON BRAGG POSITIONS
Recently, necessary and sufficent conditions under which the Fourier transform of a
selfsimilar tiling has Bragg spectrum have been given [26], extending previous results
in 1D and 2D [22,27]. In [26] it is shown that the scaling factor of the tiling should be
147
a Pi sot-Vijayaraghavan number, which is an algebraic integer greater than one with all
its conjugates (the remaining roots of its minimal polynomial) strictly less than one in
absolute value.
We now discuss the derivation of Bragg positions. The results holds for any mass
distribution on the elementary tiles.
An algebraic derivation of the Fourier module associated with a substitution in 1D
has been given in [28]. If the largest eigenvalue of the substitutional matrix is denoted by
Al and v is the right eigenvector associated with AI, and normalized by the condition
that the sum of all its components is unity, then the Bragg peaks take place (up to
possible extinctions) at reduced wave vectors q/27r given by integer linear combinations
of the components of v. In this way, the Fourier modules :F for the Fibonacci and
UVW systems (see section 2) are
(22)
(23)
(25)
where mk are integers. For a wave vector of the form given by eq. (25) it can be
checked that all the phases in eq. (19) are integer multiples of 27r in the n ---+ 00 limit
by taking into account cos(7r/7) = ((32 - (3 -1)/2, cos(27r/7) = ((3 -1)/2, cos(37r/7) =
(_(32 + 2(3 + 1)/2 and eqs. (23) and (24).
Danzer tilings are selfsimilar tilings with scaling factor T. Therefore if the diffraction
pattern is studied numerically with the help of eq. (20) and (21), a set of Bragg peaks
will be found, although no closed expressions for the amplitudes can be obtained with
this method.
148
CONCLUSION
An alternative way to describe the tilings is through context free languages [14, 15J.
A context free grammar is a quadruple
(26)
149
start symbol S. The production rules for the Penrose pattern are
(27)
The vertical bars separate two alternative possible images under the map given by
the arrows. An allowed word contains terminal elements only.
There exists a class of automata which are specific for the context free languages:
the push-down automata [31]. The generating automaton contains an input tape, a
memory tape and a computing unit with two operating heads: one to write on the
input tape and the other to read, erase and write on the memory tape. The sym-
bols written on the input tape belong to the vocabulary ET , those written on the
memory tape belong to ET U EN. In addition there is one special symbol (j written
on the memory tape. The computing unit can be in one of the following internal
states:
[0], [1], [Ai], [Bi] , [Ai], [Bi] , i E ZlO . (28)
if c = (j then it moves the tape so that the head is scanning the symbol to the
left of v and it erases v;
if c is a string of symbols then it writes the string to the right of the last symbol
and moves the tape so that the head is scanning the right most symbol written.
At the end the automaton will be in a state that we label So with the symbol (j written
on the extreme left of the memory tape.
150
The rules for the automaton generating the Penrose pattern are:
REFERENCES
151
[16] J. Garda Escudero, J. Phys. A: Math. Gen. 29 (1996), 6877.
[17] K. P. Nischke and 1. Danzer, A construction of inflation rules based on n-fold
symmetry, to appear in Discrete and Computational Geometry (1996).
[18] J. Garda Escudero, On inflation rules for odd symmetries, preprint (1996).
[19] L. Danzer, Discrete Math. 76 (1989), 1.
[20] L. Danzer, in: Lecture Notes in Physics 382, Springer, Berlin, 1991, 561.
[21] J. Garda Escudero, J. Phys. A: Math. Gen. 28 (1995),5207.
[22] E. Bombieri and J. E. Taylor, J. Phys. FranceColloq. C3 (1986), 19.
[23] J. Garda Escudero, to appear in J. Phys. Soc. Jpn. 65 (1996).
[24] G. Kasner and H. Bottger, Int. J. Mod. Phys. B 7 (1993),1487.
[25] L. Danzer and A. Talis, in: Symposia Gaussiana, de Gruyter, Berlin and New York,
1995.
[26] F. Gahler and R. Klitzing, to appear in Proc. NATO ASI "The Mathematical
Foundations of Long Range: Aperiodic Order" (1995).
[27] C. Godreche and J. M. Luck, J. Stat. Phys. 55 (1989), 1.
[28] J. M. Luck, C. Godreche, A. Janner and T. Janssen, J. Phys. A: Math. Gen. 26
(1993), 1951.
[29] S. E. Burkov, Phys. Rev. B. 47 (1993), 12325.
[30] A. Connes, Geometrie non commutative, InterEditions, Paris, 1990.
[31] M. Gross and A. Lentin, Introduction to Fornlal Grammars, Springer, Berlin, 1970.
152
ON QUADRATIC AND NONQUADRATIC FORMS:
APPLICATION TO R 2Ill-+R 2Ill -n NONBIJECTIVE
TRANSFORMATIONS
Maurice Kibler
Institut de Physique Nucleaire de Lyon
IN2P3-CNRS et Universite Claude Bernard
43 Boulevard du 11 Novembre 1918
F -69622 Villeurbanne Cedex
France
INTRODUCTION
U = Uo + 2:= Uk ek (1)
k=l
be an element of A(c) where the real numbers Uo, Ul,, UZm-1 are the components
of U and the set {el' e2, ... , e2m-l} is a system of generators of A( c). The product
W = uv of the hypercomplex numbers U and v of A( c) is defined once the multiplication
rule
2m-1
(3)
w = H(u;c)v (4)
tw = (Wo, WI,, W2m-l) tu = (uo, Ul,, Uzm-J) tv = (Vo, VI,, V2m-J) (5)
The matrix H(u; c) depends on c and the components of u. For 2m = 8, the matrix
154
H(u; c) is the following one
U7 Us -U5 U4 U3 Uz -Ul Uo
The matrix H(Uic) for 2m = 4 can be obtained from (7) by omitting the 5-, 6-, 7-
and 8-th lines and columns. Similarly, the matrix H(UiC) for 2m = 2 correponds
to the two first lines and two first columns of the matrix (7). It can be shown that
the matrix H(u; c) for the algebra A(c) can be developed as a linear combination of
Clifford matrices. Indeed, we have
L
2m-l
(11)
We shall use the qualification compact or noncom pact according to as the metric
'f/ is Euclidean or pseudoEuclidean. The C-D algebra A(c) may be normed in the
compact case or pseudonormed (corresponding to a singular division algebra with a
cone of zero divisors) in the noncompact case. We close this introductory section with
a lemma that is of central importance for what follows.
155
The latter anti-involutions (i.e. involutive anti-automorphisms) j : A( e) -+ A( e)
can be described in the following way. In the case m -I- 1, one anti-involution jo
corresponds to the complex conjugation in the C-D algebra A(e) while the 2m - 1
remaining anti-involutions jl,h, ,hm-l correspond to the complex conjugation in
the various m-dimensional C-D sub algebras of A( e). (The complex conjugate u of the
element U of eq. (1) is defined in A(e) by u = Uo - z::i:;-l
Ukek.) For m = 1, the
two-dimensional C-D algebras A( el) have only one anti-involution since jo coincides
with jl.
HURWITZ TRANSFORMATIONS
We associate to the element U of A( e), see eq. (1), the hypercomplex number
2m-l
U= Uo + L ck Uk ek (12)
k=l
QUADRATIC TRANSFORMATIONS
The column vector x contains n components that are equal to o. The integer n, which
can take several values (0 :::; n :::; 2m -1), depends on c. In other words, the Hurwitz
transformation T[I; c; c] generates, via eq. (15), a R 2m -+ R 2m - n mapping: To each
156
vector, of components u'" (a = 0,1"", 2m - 1), in R 2 m is associated a vector in
R 2 m-n the components of which are quadratic functions of the variables Un. Let us
emphasize the following property which directly follows from eq. (11).
(16)
X2 = 2UOU2
(17)
with the property
(18)
Eq. (17) can be written as
(k = 1, 2, ... , 7) (19)
in condensed form.
From the geometrical point of view, the R 2m --+ R 2m transformations (2m = 8, 4
and 2) may be arranged in two classes. The compact case corresponds to the fibration
S2m-I --+ s2m-I /Z2 ~ Rp2m-I of discrete fiber Z2 while the noncompact case
corresponds to the fibration H 2m - I (m, m) --+ H2m-I(m, m)/Z2 of discrete fiber Z2.
(We use H 2m-I (m, m) to denote the hyperboloid of equation 1 U; - 2::
u; 2:7:m+l
in R2m.)
Note that the transformation T[I; -1; 1] corresponds to the so-called Levi-Civita
(conformal) transformation used in the restricted three-body problem.
Type B 1 . The mapping T[I; Cl, C2, C3; f] where
157
Table 1. Fibrations (up to homeomorphisms) and Lie algebras under con-
straints for the R 2m -+ R 2 m-n quadratic transformations with n = m - 1 +
,s(m,l)
S3 -+ S2 SI so(2) so(4,2)
R4 -+ R3 R2 X SI -+ R2 SI sp(8,R) so(2) so(4,2)
R2 x S1 -+ R X SI R so(l, 1) so(3,3)
S7 -+ S4 S3 so(3) so(6,2)
R8 -+ RS R4 X S3 -+ R4 S3 sp(16, R) so(3) so(6,2)
R4 x S3 -+ R2 X S2 R2 X SI so(2,1) so(4,4)
158
classes. In the compact case, we have the fibration on spheres s2m-l -T sm of com-
pact fiber sm-l. The noncompact case corresponds to two kinds of fibrations on
hyperboloids, viz. one with a compact fiber and the other one with a noncompact
fiber [10]. In addition, the R2m -T R transformations (2m = 8, 4 and 2) associated
to the anti-involution)o of ACc) may be classified in two classes. The compact case
corresponds to the fibration s2m-l -T {1} of compact fiber s2m-l and the noncom-
pact case to the fibration H 2m - l (m,m) -T {1} of noncompact fiber R m x sm-l (up
to homeomorphisms). The fibrations corresponding to the anti-involution h (= ]0)
for 2m = 2 and to the anti-involutions of type 11,]2,'" ,12m-l for 2m = 4 and 8 are
reported in Table 1.
The case 2m = 4 deserves three remarks. First, is to be noted that the transfor-
mation T[l;-l,-l;EJ with E = diag(l,-l,l,l) can be identified to the so-called
Kustaanheimo-Stiefel [3J transformation (associated to the celebrated Hopf fibra-
tion on spheres S3 -T S2 rv S3 / S1 rv CPl) used for the regularization of the Ke-
pler problem in classical mechanics. Furthermore, the transformation T[l; -1, 1; cJ
with c = diag(l, -1.1, 1) corresponds to the transformation introduced by Iwai [7J
for reducing an Hamiltonian system by an SI action. Finally, the transformation
T[I; Cl, C2;cJ with c = diag(l, -1, 1, 1) and Cl = -C2 = 1 (or Cl = Cz = 1) corresponds
to a transformation, introduced by Lambert and Kibler [lOJ, inequivalent to the two
previous ones.
Type C 1. The mapping T[I; Cl, cz, C3; cJ, where c neither is the unit matrix 1 nor
corresponds to an anti-involution] of A( Cl, C2, C3), yields new quadratic transforma-
tions only when I:i=1 Ek = -3 or 5. For example, the case
X3 = 2( -C3U4U7 + C3U5U6)
X4 = 2( CZUZU6 - Cl CZU3U7) X5 = 2(CZU2U7 - C2U3U6)
(27)
159
DIFFERENTIAL ASPECTS OF QUADRATIC TRANSFORMATIONS
n = 2H(u;c)cdu (28)
(29)
(30)
(31)
when Lo is semisimple. The Lie algebra Ll is called a Lie algebra under constraints.
Theorem 2. The constraint Lie algebra Lo and the Lie algebra under constraints
Ll are characterized by the compact or noncompact nature of the fiber of the fi-
bration associated to the corresponding R2m ---> R m+1-8(m,1) quadratic transforma-
tion.
The triples (L,Lo,L 1 ), Lo eLl c L, are reported in Table 1 for the quadratic
transformations R 2 ---> R, R 4 ---> R 3 and R 8 ---> R 5 .
160
NONQU ADRATIC TRANSFORMATIONS
Eqs. (15) and (32) define R 2m - t R 2 m-n quadratic and nonquadratic transformations
(0:::; n :::; 2m - 1) in a unified way. The property (16) may be generalized as
(33)
H(u; ct)N u. General properties of these transformations of type AN (N =f. -1) can
be easily derived for 2m = 2. Indeed, we have the five following properties:
dx = (N+ 1) H(u; cd N du
dx~ - Cldxi = (N + 1? (u~ - CIUi)N (du~ - Cldui)
\7 x = (N + 1)-1 (u~ - clui)-N '1/ H(u; ct)N '1/ \7 u
Example 2. The mapping T[-1; Cl, cZ; 5], where 5 = diag(1, -1, 1, 1), yields the
generic nonquadratic formulae
(35)
161
APPLICATIONS
N umber Theory
First, the Hurwitz theorem shows that eq. (6) admits solutions only for n = 2,
4 and 8. (The case n = 1 is trivial.) The noncompact extension of this theorem
concerns the expressions
A == u~ + ui - u~ - ui
F == u~ + ui + u~ + ui ( 42)
Similarly, for C2 == C3 == 0 we get from eq. (38) the Diophantine equation
A Z _ clBz == [z ( 43)
162
with the solutions
(44)
Note that the introduction of specific relations between the U oc in eqs. (17), (21) and
(26) may lead to new Diophantine equations. For instance, by putting UI = 0 and
U2 = U3 in eq. (42), we obtain that
are solutions of
( 46)
an equation which comes from eq. (41).
Equation Solution
163
Third, the Hurwitz transformations of type AN, B N and eN with N = 2,3,4, ...
lead to Diophantine equations where the variables A, B, C, . are homogeneous poly-
nomials of degree N + 1 in U", (0 = 0,1,, 2m -1). As a typical example, from the
Hurwitz transformation T[N; el; 1] we see that
N = 2,3,4, (47)
(48)
Particular solutions of more complicated equations generalizing eq. (47) can be ob-
tained from transformations of type AN, B N and eN.
Dynamical Systems
(49)
where p = JU5 + ui and ;j == ;j(u) is the transform of IjJ == ljJ(x) under the considered
transformation. Furthermore, let us impose that 2N - o(N + 1) = o. Then, eq. (50)
reduces to
(51)
164
field) into a Schrodinger equation for a four-dimensional isotropic harmonic oscillator
(with quartic and sextic anharmonic terms) subject to a constraint [4,5,8].
ACKNOWLEDGMENTS
Part of the present work was done in collaboration with several people. In this
respect, the author is grateful to his co-authors of refs. 5 and 8-13 for fruitful col-
laborations. The applications to Number Theory were not published before. In this
connection, the author acknowledges Profs. J.-P. Nicolas, T. Ono, and A. Schinzel for
interesting comments. A preliminary version of this paper appears in ref. 16. Finally,
the author is very indebted to Prof. Bruno Gruber to have given him the opportunity
to present this work to the beautiful symposium "Symmetries in Science IX" .
REFERENCES
165
QUANTIZATION OF SYSTEMS WITH CONSTRAINTS
John R. Klauder
INTRODUCTION
The resultant equations of motion that arise from the action read
8 q) + Aa 8<Pa(P,
= 8H(p, j H} + Aa{ j J, }
8 q) =
- {q,
.j
q q ,'I'a ,
Pj Pj
. = _
PJ
oH(p,q) _ Aao;a(p,q)
8 qJ. 8
= { .
- PJ'
H} + \"{ PJ'. 'l'a
A
J, }
,
qJ
<Pa(p,q) = 0, (2)
where L} denotes the Poisson bracket. The set of conditions {;a(p,q) = O} define
the constraint hypersurface. If the constraints satisfy
where <Pa and H denote self-adjoint constraint and Hamiltonian operators, respectively.
Following Dirac [2], we adopt the quantization prescription given by
where W denotes any function of the kinematical operators {Qj} and {Pj } which are
taken as a self-adjoint, irreducible representation of the commutation rules [QJ, Pk ] =
i5n with all other commutators vanishing. The equations of motion hold for all time
t, say 0 < t < T. On the other hand, the conditions
that determine the physical Hilbert space are imposed only at time t = 0 as the analog
of the initial value equation; the quantum equations of motion ensure that the constraint
conditions are fulfilled for all time.
The procedure of Dirac has potential difficulties if zero lies in the continuous spec-
trum of the constraint operators for in that case there are no normalizable solutions of
the constraint condition (8). We face the same problem, of course, and our resolution
is discussed below; see also ref. [1].
168
equivalent point per gauge orbit, and in addition a factor (in the form of a determinant)
is introduced to formally preserve canonical covariance. The result is the path integral
J exp{ i J~[pjqj - H(p, q)] di}8{X(p, q)} det( {X a , <Pb} )5{ <p(p, q)} Dp Dq . (10)
namely, as a path integral over a reduced phase space in which the 5-functionals have
been used to eliminate 2A integration variables. '
The final expression generally involves curvilinear phase-space coordinates for which
the definition of the path integral is typically ill defined. Additionally, in the form
(10), the Faddeev-Popov determinant often suffers from ambiguities connected with
inadmissible gauge fixing conditions [4]. Thus this widely used prescription is not
without its difficulties.
BRST-BFV formulation
By extending the phase space to include Grassmann variables, it is possible to develop
alternative and more powerful methods to discuss systems with constraints. These
methods are well documented, e.g., [5], and will not be discussed here. Instead, our
interest focuses on what can be said without enlarging the number of variables beyond
those that make up the original phase space augmented by the necessary Lagrange
multipliers.
(12)
where 10) traditionally denotes a normalized, unit frequency, harmonic oscillator ground
state. Here { Qj} and {PJ, 1 ::; j ::; J, denote an irreducible set of self-adjoint operators
satisfying the Heisenberg commutation relations. The coherent states admit a resolution
of unity in the form
where the integration is over IR 2J and this integration domain and the form of the
measure are unique. For a general operator H(P, Q) we introduce the upper symbol
169
Using these quantities, the coherent state path integral for the time-dependent
Hamiltonian H(P, Q) + Aa(t)<pa(P, Q) is readily given by
( p,q
II lilT e -i JT0 [H(p,Q)+,\a(t)i!>a(P,QlJ dt I'
p,q')
= ~~ JII N
(PI+1, ql+1le- iE (H+'\l'i!>a) Ipl, ql)
N
II d/1(PI, qz)
J
~o ~1
= exp{ i f[i(p, ql( djdt)lp, q) - (p, qlH + Aa<pa Ip, q) J dt} 'D/1(p, q)
In the second line we have set PN+l,qN+l = p",q" and Po,qo = p',q', and repeatedly
inserted the resolution of unity; in the third and fourth lines we have formally inter-
changed the continuum limit and the integrations, and written for the integrand the
form it assumes for continuous and differential paths (M denotes a formal normal-
ization constant). The result evidently depends on the chosen form of the functions
pa(t)}.
(16)
defines a projection operator onto the subspace for which <Pa 0 provided that {j~
denotes the normalized, J(j~ = 1, group invariant measure. Based on (5), (6), and (16)
it follows that
e-iTa\!>aE =E , (17)
e-iHTE = E e-iHTE = E e-i(E HE )TE (18)
We now project the propagator (15) onto the quantum constraint subspace which leads
to the following set of relations
lU' I'
- (" lilT e -iJ[H+,\a(t)\!>a]dt ""p,q ')
-p,q
where T a incorporates the functions Aa as well as the structure parameters Cab C and ha b
Alternatively, this expression has the formal path integral representation
170
On comparing (19) and (20) we observe that after projection onto the quantum con-
straint subspace the propagator is entirely independent of the choice of the Lagrange
multiplier functions. In other words, the projected propagator is gauge invariant; see
refs. [I, 5, 6].
We may also express the physical (projected) propagator in a more general form,
namely,
J exp{ i J[PA j - H(p, q) - ).,a cPa (p, q)] dt} 'Djj(p, q) 'DC( >..)
= (p", q"le- iT1i IE Ip', q') (21 )
provided that J'DC(>") = 1 and that such an average over the functions {Aa} introduces
(at least) one factor IE .
which is a bounded, continuous function for any projection operator IE, especially in-
cluding the unit operator. It follows that K(p", q"; p', q')* = K(p', q'; p", q") as well
as
K
L IYkcqK(Pk,qk;PI,ql):::: 0 (23)
k,l=l
for all sets {IYk}, {Pk, qd, and all J{ < 00. The last relation is an automatic consequence
of the complex conjugate property and the fact that
K(p", q"; p', q') = J K(p", q"; p, q) K(p, q;p', q') dJ-l(p, q) (24)
holds in virtue of the coherent state resolution of unity and the properties of IE. The
function K is called the reproducing kernel and the Hilbert space it engenders is termed
a reproducing kernel Hilbert space. A dense set of elements in the Hilbert space is given
by functions of the form
K
and the inner product of this function with itself has two equivalent forms given by
K
As usual, the inner product of two distinct functions may be determined by polarization.
Clearly the entire Hilbert space is characterized by the reproducing kernel K. Change
the kernel K and one changes the representation of the Hilbert space. Following a
suitable limit, it is even possible to change the dimension of the Hilbert space, as we
discuss in the next section.
171
Reduction of the reproducing kernel
Suppose the reproducing kernel depends on a number of variables and additional pa-
rameters. We can generate new reproducing kernels from a given one by a variety of
means. For example, the expressions
each generate a new reproducing kernel provided the resultant function remains con-
tinuous. Sometimes, however, the inner product in the Hilbert space generated by the
new reproducing kernel is only given by an analog of (26) and not by (27), although
frequently some sort of local integral representation for the inner product may also
exist.
Let us present an example of the reduction of a reproducing kernel. Consider the
example
which defines a reproducing kernel for any 5 > 0 that corresponds to an infinite dimen-
sional Hilbert space. Let us multiply this expression by 7['1(2/(25) and take the limit
5 -+ O. The result is the expression
1("2
K(p";p') = e-'2 p +.p'2) , (32)
(33)
APPLICATIONS
Example 1
The following example is based on Problem 5.1 in ref. [5]. Consider the two-degree of
freedom system with vanishing Hamiltonian described by the classical action
(34)
For notational convenience all indices have been placed as subscripts. The equations of
motion become
pj = 0, Pi = 0, j = 1,2. (35)
172
Evidently the Poisson bracket {P}'P2} = o.
As a second version of the same dynamics, consider the classical action
(36)
Since eC~' P2 = 0 implies that P2 = 0, it follows that the two formulations are equivalent
despite the fact that in the second case {PI,e cQ'P2} = -ce cQ'P2, which has a completely
different algebraic structure when c =I- 0 as compared to c = o.
Let us discuss these two examples from the point of view of a coherent state quan-
tization. For the first version we consider
(38)
(p,q p,q{,
" "IIEI' '\ (39)
where
= 7r- I ;g 15
2 5
exp[-t(kl - PI')2 + ik1(ql' - ql) - t(k/ - pl)2J dk/ . (41)
Let us reduce this reproducing kernel by multiplying this expression by 7r /(28)2 and
passing to the limit 8 - t O. The result is the reduced reproducing kernel given by
(42)
173
where in the present case the fully reduced form of this expression is proportional to
(46)
which once again represents a one-dimensional Hilbert space.
Thus we have obtained a c-dependent family of distinct but equivalent quantum
representations for the same Hilbert space, reflecting the c-dependent'family of equiv-
alent classical solutions. Observe that in the quantum theory, just as in the classical
theory, all observable effects are independent of c.
Example 2
In discussing constraints one often pays considerable attention to the regularity of the
expressions involved; see [5], Sec. 1.1.2. Consider, once again, the simple example of a
single constraint p = as illustrated by the classical action
The equations of motion read q = >., p = 0, and p = 0. On the other hand, one may ask
about imposing the constraint p3 = or possibly pl/3 = 0, etc., instea~ of p = 0. Let
us incorporate several such odd (function) examples by studying the classical action
Je_i~PIPI~Sin(!~+10 de
= E(_p+l < PIPI"! < 8"!+1)
= E( -8 < P < 8) . (49)
Thus, from the operator point of view, it is possible to consider the constraint operator
PIPI" justas easily as P itself. In particular, it follows that
174
where we have appended / to the measure for the Lagrange multiplier). to emphasize
the dependence of that measure on /. The reduction of the reproducing kernel proceeds
exactly like the cases discussed earlier, and we determine for all / > -1 that
"(p"
lim _V_
Ii q"IIElp' q') = e-2"1( p"2+ p'2) (51 )
0--->0 (28) " ,
representative of a one-dimensional Hilbert space. Just like the classical theory, note
that the ultimate form of the quantum theory is independent of /.
It is natural to ask how one is to understand this acceptable behavior for the quan-
tum theory for irregular constraints while there are difficulties that seem to be present
in the classical theory. In the first section we discussed the definition of the classical
generator as derived from the quantum generator. Just like the classical and quantum
Hamiltonians, the connection between the classical and quantum constraints is given
by
With this rule we typically find that r/J(p, q) -I- <l>(p, q) due to the fact that li -I- 0,
but the difference between these expressions is generally qualitatively unimportant. In
certain circumstances, however, that difference is qualitatively significant even though
it is quantitatively very small. Since that difference is O(li) let us explicitly exhibit the
appropriate li-dependence hereafter. First consider the case of / = 2. In that case
(53)
where we have introduced the shorthand ((-)) == (01()10). Since (P 2) = li/2 it follows
that for the quantum constraint p 3 , the corresponding classical constraint function is
given by p3+(3li/2)p. For Ipi ~ v'n, this constraint is adequately given by p3. However,
when Ipl ~ v'n-as must eventually be the case in order to actually satisfy the classical
constraint-then the functional form of the constraint is effectively (3li/2)p. In short, if
the quantum constraint operator is p 3 , then the classical constraint function is in fact
regular when the constraint vanishes.
A similar discussion holds for a general value of,. The classical constraint is given
by
For Ipi ~ v'n the first line of this expression effectively yields r/J-y(p) plpl-Y. On
the other hand, for p :::::J 0, and more especially for Ipl ~ v'n, the second line of
this expression shows that this constraint function vanishes linearly, speci fically as
r/J-y(p) ~ kp, where
(56)
175
Thus, from the present point of view, irregular constraints do not arise from consis-
tent quantum constraints; instead, irregular constraints arise as limiting expressions of
certain consistent, regular classical constraints as 11, ---+ 0.
There is one category of irregular constraints that is not covered by the foregoing
discussion, namely the even constraints, e.g., a classical constraint given by p2 = 0
rather than p = O. This case differs from those treated above because the classical
constraint is strictly non-negative. However, from the operator point of view, the case
of p2 is not qualitatively different from the other cases because we still have
Je-i~p2 sin~~20 de
= E(p2 < 82 )
= E( -8 < P < 8) . (57)
From the classical point of view, however, it follows that
(58)
which evidently never vanishes so long as 11, > O. This result is not surprising; as the
expectation value of a non-negative operator with continuous spectrum, it cannot van-
ish. If we adopt our previous interpretation that there are no irregular odd constraints
thanks to a nonvanishing 11" then we are forced to admit that for even constraints there
is no suitable classical analog. Of course, such examples are in no way restricted to
terms involving n. One need only consider the classical constraint p2 + q2 + 1 = 0 which
evidently has no solution for real phase space variables.
However, there is another way to look at even classical constraints [7]. For the sake
of illustration, let us initially focus on the simple case p2 = as an even representative
of the odd classical constraint p = O. From a simple physical point of view, a system
will follow a constraint at least approximately provided it costs a great deal of energy
to violate it. onsider the simple system described by the action functional
(59)
where A is a large positive constant. The equations of motion for this case read q = Ap,
and p = 0, and the solution to these equations of motion is p( t) = pi and q( t) = Ap't +q'.
We insist that the solution should be independent of the large parameter A, and so it
follows that pi = O(A-l). As a consequence, the "energy" Ap'2/2 = O(A-l) as well. In
the limit that A ---+ 00, it follows that pi = 0, the "energy" vanishes, while q(t) = ct + ql
for some c. This solution is seen to be an example of one from the usual formulation
using a Lagrange multiplier. Indeed, we can also make q(t) into a rather general function
of time by allowing for A to be time dependent as well as large; however, for convenience,
we ignore this r.ather evident generalization. An alternative example of the same kind
is given by
(60)
The analysis in this case proceeds as before with the final result that
+ pi sin(ct) .
q(t) = ql cos(ct) p(t) = _ql sin(ct) + pi cos(ct) , (61)
subject to the requirement that pl2 + ql2= 1. Note well, in these cases, that we deal
with even constraints and a large positive parameter A in order that no cancellation
among large terms of opposite sign is possible.
176
Following this brief introduction to an alternative classical formulation of a theory
with a constraint, let us turn our attention to the quantum mechanics of such a system.
Let W denote the basic constraint operator we wish to maintain; for instance, W = P2/2
for the first case and W = (: p 2 + Q2 : -1)2/2 for the second case. Maintaining tt = 1
temporarily, interest centers on the expression
(62)
= lim JiAT/21l'
A~oo
jex [-t(k- pll )2
p
(65)
Thus, from the point of view of deriving the classical action from the quantum theory,
we are not led to the original starting action (59) but instead to
(67)
177
The equations of motion, q = Ap 3 and p= 0, lead to an A-independent form of
the classical solution given by q(t) = ct + q', for some c, and p(t) = O. Quantum
mechanically, we are led to conclude that
(6S)
for a suitable constant k. The coherent state path integral formulation leads to
(69)
(70)
If we take this form seriously, we are led to conclude that as A -7 00 and ipi ~ -Iii, only
the quadratic term becomes important in the equations of motion as compared with
the quartic term. In short, the p4 term is just as "regular" as the p2 term. Thus, once
again, we conclude that from the coherent state point of view there are no irregular
constraints so long as 1i > 0; it is only in the limit where 1i -7 0 that certain regular
constraints turn into irregular constraints.
ACKNOWLEDGEMENTS
Thanks are expressed to Bruno Gruber for his efforts which resulted in a very
pleasant conference. Communications with J. Govaerts, M. Henneaux, and S. Shabanov,
and discussions with B. Whiting are gratefully acknowledged.
REFERENCES
178
AUTOMORPHISMS AND DISCRETE FIBER
BUNDLES
INTRODUCTION
We follow Magnus, Karrass and Solitar [9] and denote the free group with n gener-
ators by
(1)
This notation implies that there are no relations between the generators except for the
ones which ensure the unique inverse,
(2)
Without these inverses one would have the weaker algebraic structure of a monoid,
compare Lothaire [10]. The general elements of Fn are words w obtained by associative
concatenation,
(3)
Here the integer aij marks the power of the jth occurence of the generator Xi. We shall
also need the commutative group with n generators which we write with a multiplicative
composition,
(4)
Its elements are words z which, due to commutativity, may be written up to order as
The powers of the generator bi of zn are obtained by summing in w over all the powers
of Xi.
Example 1: For the word w = X1IX~XIX21 we get homl(w) = h
Vie emphasize some general subgroups of the free group. The homomorphism homi
has a kernel
kel(homl):= (u E Fnl2:.:::aij = o'Vi). (7)
j
With these relations it is easy to show that the commutator group is normal in G. Our
first important claim for the free group and its subgroups is
Prop: The normal subgroup ker(homl) and the commutator subgroup K of Fn are
identical.
Proof: We refer to [9J for a general proof but sketch a constructive procedure which
illuminates the analysis in later sections. Clearly any element of the commutator sub-
group by its power sums eq. (6) is in ker(homl) and thus K < ker(homr). Conversely,
180
given a general word w E F n , we can reorder it with the help of commutators so that
the reordered word reads
(9)
(10)
181
We particularize the discrete construction to a pair G &> H with H a normal subgroup.
Now the basis B of the bundle may be identified with the factor group of G with respect
to H, the projection Jr becomes the homomorphism Jr : G -+ G / H, and the fiber is the
kernel ker( hom) = H. When does the left action yield a closed path in the basis? It
is easy to check that this is the case if and only if one acts with an element it E H, for
then
(12)
The transitive action of H on the fiber from the right guarantees that the shift in the
fiber becomes independent of the starting point.
We summarize:
Prop: A principal fiber bundle can be obtained from a group/subgroup pair G &> H.
The total space is E = G, the basis points are the left cosets B = G / H, in one-to-
one correspondence with the factor group, and the homomorphism hom : G -+ G / H
has as its kernel the fiber F = H. Any left action with an element from the normal
subgroup H produces in the bundle basis a closed path. In the fiber H at the coset
generated by c;, this action yields an algebraic torsion given by the left action with
( Ci-1- )
uCi .
As an example for the construction of a discrete bundle let us consider the free group
F2 (Xl, X2)' As the normal subgroup we take the commutator subgroup K. We shall
=
use an explicit set of left coset generators. Applying the reordering discussed in the
second section we can write a general element w E F2 in the form
(13)
(E,B,Jr,F) (14)
= (F21 PI, homl, K)
The numbers (ml' m2) determine uniquely an element of the commutative translation
group PI with two generators as the image under hom!. Similarly we decompose a
group element acting from the left as
(15)
Then the left action yields
(16)
The changes (rih, m2) in the coset generators determine a unique action in the com-
mutative translation group. In the particular case g = v E K we find for it the expres-
sIOn
(17)
According to our definition in the previous section this is the algebraic torsion.
182
GEOMETRIC INTERPRETATION ON A GRAPH
The algebraic construction of the last section has a simple interpretation in terms of
a graph r . Define the graph r as the set of vertices and edges of the square lattice in
the plane, although metrical properties play no role at the moment. We associate the
generators (x!, X2) of F2 with the two types of edges pointing in the 1,2 direction of this
lattice respectively. Upon choosing an origin on r, any point with coordinates (mI' m2)
is in one-to-one correspondence to the coset decomposition eq. (13) and moreover to a
unique element of the factor group
(18)
(19)
The word u provides a closed loop around two squares, and the conjugation with the
coset generator links the initial and final vertex of this loop to the vertex which cor-
responds to the unit element e of Pl. On the right-hand side of Figure 1 we give the
decomposition of the loop into two elementary loops around single squares. The corre-
sponding deformation of the path is allowed by the relation XIX;:-I = e, eq. (2) on the
group F2
We have stated in section 2 that any word from the commutator group K may be
decomposed into a product of conjugates of elementary commutators. In terms of the
present geometric view this means that the path and the torsion may be decomposed
into a sequence of elementary loops with torsion. We summarize our findings:
Prop: The fiber bundle constructed from F2 t> PI with basis B = PI and fiher
F = K can be represented on the graph r(Pl). The torsion along a closed path
183
can be interpreted as the complete memory of the path followed on f(P1). It may be
decomposed into elementary square loops with torsion.
We wish to extend the fiber bundle construction to all the 17 space groups on the
plane. The bundle basis in each case should be a space group G, but moreover it should
admit a simple geometric representation. This representation can be obtained from the
Cayley graphs of the planar space groups which were derived and discussed by Coxeter
and Moser [14]. In the following we summarize the main properties of a Cayley graph:
By the symbol R we denote a relator, that is a word in the generators which in the
group reduces to the identity. In our geometric interpretation we take advantage of the
fact that relators are nontrivially represented by loops in the Cayley graph.
In [14] one finds the Cayley graphs f( G) for anyone of the 17 space groups G on
the plane. These groups are finitely generated but infinite periodic, due to the presence
of an infinite periodic commutative subgroup. Hence all Cayley graphs are infinite
periodic. It therefore suffices to give a part of the infinite periodic graph. Although
the vertices of each graph are in one-to-one correspondence to the group elements, the
edges depend on the choice of generators a.nd so there are many different Cayley graphs
for a single group. Similarly, the relators which determine closed loops depend on the
choice of the generators. A convention on a Cayley graph is that a relation of the type
R = (Xj)2 = e is represented by an undirected edge. In the following section we give
the Cayley graphs for various space groups.
For the construction of a fiber bundle whose base is a space group G we need as the
total space a group g, a homomorphism 9 -+ G which serves as the projection, and
the kernel of the homomorphism which provides the fiber. We wish to maintain the
free group as the preimage of the commutative translation subgroup of G. At the same
time we wish to gain the flexibility for introducing the notion of point groups.
In order to achieve this construction we turn to the group of automorphisms Aut(Fn)
of the free group Fn. The group Aut(Fn) was studied by Nielsen [9] in great detail.
This group can be specified by giving its action on the free group. Both groups are
finitely generated, and so it suffices to give the action of generating automorphisms on
184
the generators of Fn. An automorphism may then be specified as an invertible map
Here the Yi) expressed as words in Fn, are the image of the generators Xi under the
automorphism <p. For the role played by the free group, the crucial observation is that
the normal subgroup
(21)
is isomorphic to Fn. It means that the free group can be embedded as a normal subgroup
into its group of automorphisms. We shall use the embedding
(22)
as the preimage of the commutative translation group. The preimages of point trans-
formations must then be built from automorphisms not belonging to Inn(Fn). In par-
ticular there are subgroups of finite order in Aut(Fn). A systematic construction of all
automorphisms of finite order is given by McCool [15].
Next we turn to homomorphisms. We have already described the homomorphism
homl: Fn -+ zn called abelianization. For Aut( Fn) there is a natural homomorphism [9]
to the integral linear group G L( n, Z) defined as follows: A fixed automorphism say <p can
be specified by giving the n images Yi eq. (20) of the generators Xl ... Xn. If we abelianize
the n words Yi, we obtain from the sums of the exponents an n X n array of integer
entries called the substitution matrix of <p. Multiplication of automorphisms yields
multiplication of substitution matrices. The inverse of <p must exist, and so the inverse
substitution matrix must also have integer entries. Hence the substitution matrix is
an element of GL(n, Z). The group GL(n, Z) is the group of automorphisms of the
commutative translation group. It forms a basis for all notions of crystallography [16].
The point groups of commutative Euclidean crystallography are subgroups of finite
order of G L( n, Z) [17].
The homomorphism homz unfortunately fails for the preimages of translations: The
elements of the free group appear in Aut(Fn) as inner automorphisms, which means
that the corresponding images are all of the form
185
Example 4: As examples we choose a 2fold rotation S2 and the two automorphisms
Tt, T2 which are the images of the generators (Xl, X2) under the isomorphism F2 rv
Inn(F2). We give in columns the generators (Xl, X2, X3), their images (YI, Y2, Y3) under
the automorphism from Aut(F3)' and the substitution matrix hom2. Note the block
triangular form of the matrices hom2(TI ), hom2(T2).
e S2 hom2 TI hom2 T2 hom2
Xl (XI)-l -1 0 0 Xl 1 0 0 (X2tlxlX2 1 0 0
(25)
X2 (X2)-1 0 -1 0 (X1t 1x 2XI 0 1 0 X2 0 1 0
X3 X~X3 0 2 1 Xl-1 X2
-1 -1 -2
Xl X2 X3 -2 0 1 X2 X3 0 -2 1
Now we turn to the 17 non-commutative space groups g. These groups were con-
structed in [1] by the embedding Aut(F2) < Aut(F3) which we just described. They
are characterized as subgroups of Aut(F2} in two ways:
(1) By generators from Aut(F2)' by sets of relators RI equal to unity in Aut(F2)'
and by sets of relators R2 which in Aut(F2) belong to the commutator subgroup of F2
and when put equal to unity enforce a homomorphism 9 ----; G.
(2) By the embedding of these groups 9 into Aut(F3)' which under hom2(Aut(F3))
maps them into G.
We give examples from the 17 space groups G in the order and notation of the
International Tables [18]. The none commutative space group 9 is denoted by the same
symbol but with script letters. For each group we first give alternative sets of generators
(a), (b), .. , as in [14], with some changes of notation to avoid symbols used for Aut(F2).
We use alternative sets of generators to display representative elements for each space
group. Under (hom )-1 we give in the next row a preimage in Aut( F2) for each generator.
For the details of the generators we refer to [1]. Under R we give the relators R =
(R1' R 2). In the row that follows we give under ker for the relators R2 their expressions
in Aut(F2). They determine elements of infinite order in g.
P2
(a) x Y T
(hom)-l T1 T 2- 1 0"10"2
R T2 TXT X TYTY XY X-I y- l
ker IT1,T2- 1)
(b) U1 = YT U2 = T U3 = XU2
(hom)-l 53 52 51
R (Ur)2 (U2)2 (U3)2 (U1U2U3 )2
ker IT2-r, Tr)
(26)
Pg
(a) X P
(homtl T2TI T1Cl3
R P- 1XPX
ker IT2' T1- 1)
(b) P Q=PX
(hom)-l T1Cl3 C13Tl
R P2Q-2
ker ITI ,T2- 1 )
(27)
186
Pg
(a) X P
(hom)-l T2T1 T1C13
R P-1XPX
ker K(T2,T1- 1)
(b) P Q=PX
(homt1 T1C13 C13T1
R P 2 Q-2
ker K(T1,T2- 1)
(28)
em
(a) P Q R
(hom)-l T1C13 C13T1 C13
R R2 RPRQ-1 P 2Q-2
ker K(T1, TZ- 1)
(b) R S=PR
(hom)-l C13 T1
R R2 (SR)Z(RS)-Z
ker K(T1, TZ- 1)
(29)
P2mm
(a) Y R R' R2
(hom)-l T- 1 (12T2 (12 (11
1
R R2 (R')2 (R2)2 R'YR'y- 1
ker
R R2RRzR R zYR 2Y RYRy- 1 R 2R'R2R'
ker K(r,-:-l
2 , T-
1
1 )
P3
(a) U1 U2 U3
(hom)-l Cl2 C23 Cl2 C23 Tl T1- 1C12 C23
R (UJ)3 (U2)3 (U3 ? U1 U2U3
ker K(T- 1 2 1)
1 , T-
(b) U1 U2
(hom)-l Cl2 C23 C12 C23 T 1
R (U1)3 (U2)3 (U1U2)3
ker K(T-2 , T-
1 1
1 )
(31)
187
T
Figure 2. Cayley graph of P2 generated by two translations X, Y and one involution T. The
torsion is carried by two quadrangles.
From both characterizations we get individual pairs 9 t> G and can construct in-
dividual fiber bundles whose base is the space group G. It remains to construct the
kernels of the homomorphism, from it the fibers, the closed paths and the torsions.
Then we can use the Cayley graphs f( G) for a geometric interpretation.
The kernels of the homomorphisms were analysed in detail by U. Beck [19] with the
general result:
Prop: The kernel of the 17 homorphisms 9 ~ G is in all 17 cases isomorphic to
the commutator group K, of F 2 .
It is clear that the commutator group lies in the kernel, but it has to be shown that
no other elements of 9 have this property.
This proposition allows us to state that the fiber of the bundle to be constructed is
always the commutator group of F 2 The proposition applies even to non-symmorphic
pairs like Pg, Pg in which the preimage of the translation group is not isomorphic to the
free group F2 but only to its subgroup generated by even words in the generators. The
commutator group K, is contained in this subgroup. We now have all the ingredients
for the construction and the analysis of the fiber bundles.
In section 4 we constructed the principal fiber bundle eq. (14). The Cayley graph
is the lattice f(Pl). We rewrite this bundle in the form
(E,B,7r,F) (32)
= (F2 , Pl,homl,Q
= (PI, PI, homl, K,)
The elementary torsion arises from minimal square loops which are commutators of
the form K(Xl1,X21), the torsion for a closed path is the product of conjugates of
elementary contributions.
188
... . . . . . . ............?~.............................
o o
............ ...............
.......... ....
o
.... .... ....... .
. . ... .... .....
...... . . ......
189
,,
p p ,, Q Q
,, ,
,, ,,
0 :I
n ,, p R
,
,,
Figure 5. Cayley graph of em !',ema<w,!d by two glide reflections P, Q and one involution R.
The torsion is carried by each rhombus.
,...; .... [ -- J
. .--
....... " ...
.
...............~
.- - . ~
190
In Figure 3 we generate P2 by three involutions. The relators which yield the
involutive property are free of torsion, but the hexagons formed by three pairs of these
involutions carry elementary torsion.
From these examples we can infer the torsion that would be obtained by a general
closed path on f(P2): First of all we may decompose the path into parts without double
points. Each such part contains in its interior a certain amount of elementary loops
with torsion. The torsion on this path is a product of conjugates to the elementary
commutators which arise from these elementary loops. Comparing with the case of
PI we see that the total torsion will represent a partial memory of the closed path on
f(P2), since a deformation to a new path which differs by loops free of torsion will give
no contribution.
Clearly we can give similar constructions for all the 17 space groups G. We sum-
manze:
Prop: For any pair 9, G of the 17 non-commutative and commutative space groups,
there is a principal fiber bundle with total space g, basis G, and fiber the commutator
group JC of the free group F 2 The basis can be represented as the Cayley graph r(G).
The torsion can be interpreted as a partial memory of the closed path followed on f( G).
It can be decomposed into contributions from elementary loops with torsion which are
specific for the pair 9, G.
CONCLUSION
We present the 17 space groups G of the plane as bases of discrete principal fiber
bundles. A closed path in the basis then becomes a product of space group elements
which represents the identity or a circuit in the terminology of Pond and Hirth [8].
We show that this can be given a meaning by use of the Cayley graph f( G). As
the fiber we choose in each case the commutator group of the free group F 2 We
define a specific torsion from the left action of the groups. This torsion is built up
from particular elementary loops in the graph. By this construction, the noncom-
mutativity is cast into a partial path memory on the Cayley graph. In future stud-
Ies we wish to explore in more detail the relations to a discrete theory of disloca-
tions.
REFERENCES
191
[7] S. 1. Ben-Abraham, in: Fundamental Aspects of Dislocation Theory, J. A. Simmons,
R. de Wit, and R. Bullough (eds.), Nat. Bur. Stand. (U.S.) Spec. Publ. 317 II (1970),
999.
[8JR. C. Pond and J. P. Hirth, Solid State Physics 47 (1994), 288.
[9] W. Magnus, A. Karras, and D. Solitar, Combinatorial Group Theory, Dover, New
York,1976.
[10J M. Lothaire, Combinatorics on Words, Addison-Wesley, Reading, MA, 1983.
[11] J. Dieudonne, A History of Algebraic and Differential Topology 1900-1960., Birk-
hauser, Basel, 1989, 385-419.
[12J W. D. Curtis and F. R. Miller, Differential Manifolds and Theoretical Physics,
Academic Press, New York, 1985.
[13] S. 1. Goldberg, Curvature and Homology, Dover, New York, 1982, 132-144.
[14J H. S. M. Coxeter and W. O. Moser, Generators and Relations for Discrete Groups,
Springer, New York, 1965.
[15] J. McCool, Trans. Am. Math. Soc. 260 (1980), 309.
[16] H. Brown, R. Biilow, J. Neubiiser, H. Wondratschek, and H. Zassenhaus, Crystal-
lographic Groups of 4-Dimensional Space, Wiley, New York, 1978.
[17J M. Lorente and P. Kramer ,in: Symmetries in Science VIII, B. Gruber (ed.), Plenum,
New York, 1995, 315.
[18] T. Hahn, Ed. International Tables for Crystallography, 3rd revised ed., Kluwer,
Dordrecht, Boston, London, 1992.
[19J U. Beck, Diplomarbeit, Universitiit Tiibingen, 1996.
192
ALGEBRAIC APPROACH TO BARYON STRUCTURE
INTRODUCTION
u u
on the electromagnetic helicity amplitudes [7] and calculate strong decay widths [8].
The predictions of the collective model are compared with those of the single-particle
description and with experimental data on these observables. The reader is encouraged
to consult first [6] and to find more details in [5, 7, 8, 9].
MASS SPECTRUM
We consider the baryon as an object composed of. three identical constituents car-
rying internal quantum numbers: flavor = triplet = u, d, S; spin = doublet = 1/2; and
color = triplet. The geometric arrangement of this three-body system can be phrased
in terms of two relative Jacobi vectors. It is possible [5, 6] to cast the dynamics of
these spatial degrees of freedom in algebraic form, in terms of a U(7) spectrum gen-
erating algebra. The full algebraic structure is obtained by combining the geometric
part, with the usual spin-flavor-color part, resulting in total wave functions which are
representations of U(7) Q9 SUsj (6) Q9 SUc (3),
(1)
The classification under SUsj (6) and its flavor-spin subgroups SUj (3) 10 SUs (2), are
denoted by the dimension of the corresponding irreducible representations. We consider
a collective model in which non strange baryons are interpreted as excitations of the
string configuration in Fig. 1. In this case the spatial part in eq. (1) consists of an oblate-
top wave function [6] characterized by the labels: (Vl,V2); K, Lf, where (Vb V2) denotes
the vibrations (stretching and bending) of the string; K denotes the projection of the
rotational angular momentum L on the body-fixed symmetry-axis; P the parity and t
the S3 symmetry type of the state under permutations. The permutation symmetry of
the spatial part must be the same as that of the spin-flavor part to ensure anti symmetry
of the full space-spin-flavor-color wave-function (SUsj (6) species: [56] symmetric, [20]
anti symmetric, [70] mixed symmetry). The spin Sand L are coupled to total angular
momentum]. In this notation the nucleon (S = 1/2, flavor octet) and delta (S =3/2
flavor decuplet) ground state wave functions are given by
and (2)
For harmonic oscillator dynamics, the spin-flavor part in is the same, while in the spatial
part the (Vl' V2); K labels are replaced by the harmonic oscillator quantum number n.
An S3-invariant operator can now be constructed along the lines outlined in [6], and
used to describe properties of nucleon and delta resonances. In the collective model the
resulting mass spectrum exhibits rotational states (Lf, K) arranged in bands built on
194
top of each vibration (VI, V2)' For a large model.space the spectrum follows closely a
mass formula of the form
(3)
The spatial U(7) contribution to the mass consists of a vibrational part (M;ih =
Al VI +A2V2) and a rotational part (M;ot = aL). The spin-flavor contribution is expressed
in a Cursey Radicati form [2] associated with the chain SUsj (6) :::> SUj(3) SUs(2).
The resulting fit (r.m.s deviation of 39 MeV) for 3* and M nucleon and delta res-
onances is shown in Fig. 2, along with a comparison to the non-relativistic [3] and
relativized [4J quark models. As can be seen, the quality of the fits are compara-
ble although the underlying dynamics and corresponding wave functions are differ-
ent. A typical oblate-top collective wave-function for the ground state is shown in
Fig. 3, expanded in an harmonic oscillator basis. It exhibits strong mixing of many
oscillator shells and reflects a correlated motion of the constituents. This signifi-
To consider decay processes of baryon resonances we need two ingredients: (i) the
wave function of the initial and final states and (ii) the form of the transition operator. It
is assumed that the operators inducing the electromagnetic (strong) transitions involve
an absorption or emission of a photon (elementary meson) from a single constituent.
In such circumstances, the couplings discussed below can be expressed in terms of the
operators
(4)
(5)
195
3.0 I-
N(I=1/2)
2.5
~
->
Q,I
Q
CJ
~
0
~ 2 .0
:E
~ Q IE!::l m 0
E::J .+0
1 .5 ~ ~ 'i'D
0
1 .0
......,A
I I
c:J
2.5 I-
6 (1=3/2 )
:> 2.0 I- +
0
@
Q,I
~
CIJ
D G;:!::;I t!!I ~
:E q m 0
"
~
0
1.5 I- -
1 .0 r- -
3/2
Figure 2 . Nucleon and delta mass spectrum (M vs. JP) . Collective model [5] (+ ), nonrela
tivistic quark model [3] (x ), relativized quark model [4] (0) .
196
1.0
0.9
0.8
~ 0.7
:cIII
..c 0.6
e
c.. 0.5
0.4
0.3
0.2
0.1
o 2 4 6 8 10 12 14 16 18 20
n
Figure 3. Probability distribution of a typical ground state collective oblate-top ground state
wave function in an harmonic oscillator basis.
where (3 is a radial coordinate and a is a scale parameter. The collective form factors are
obtained by folding the matrix elements of U and Tm with this probability distribution
Here if' denotes the spatial part of the baryon wave function. The ansatz of eq. (5)
for the probability distribution is made to obtain the dipole form for the elastic form
factor. With the same distribution we can now derive closed expressions for inelastic
form factors connecting other final states. A sample of collective transition form factors
of the distributed string are given in Table 1, which also lists the corresponding harmonic
oscillator form factors. The single-particle and collective form factors differ in their k
dependence. In particular, all collective form factors drop as powers of k. This property
is well-known experimentally and is in contrast with harmonic oscillator quark models
in which all form factors which fall off exponentially.
ELECTROMAGNETIC COUPLINGS
In electromagnetic processes such as photo- and electroproduction we encounter
both elastic transitions involving diagonal matrix elements and inelastic transitions
involving non-diagonal matrix elements of the transition operator. We consider below
the corresponding observables (elastic electric and magnetic form factors of the nucleon
and helicity amplitudes) which can be measured.
197
Table 1. Collecti ve and harmonic oscillator form factors
[56,0+]{0,0);0 1 4ka 0
{1+k'a 2)2 {1+k2a 2)3
[70,1-]{0,0);1 -z
.V3 {1+k2a2
ka
)2
.V3 {1+k2a
Z
1-3k2a2
2)3 =t=i v'6 {1+kt2)2
Final stateb) (flU Ii) (fIt Ii) /m3 k of3 (f1'hli)/m3 k of3
a) Collective form factors (for large model space): the final states are labeled by
[dim{SUs J(6)}, LP](Vl,V,);K, the initial state is [56, O+]{O,O);O; H(x) = arctan x - x/(l +x 2 ).
b) Harmonic oscillator form factors: the final states are labeled by [dim{SU s j(6)} , LP]n,
the initial state is [56,0+]0'
where III denotes the nucleon wave function 2802[56,0+](0,0);0 with N = p (n) for proton
(neutron). Further e3, /13 = eg3/2m3, m3, g3, 83 = 0"3/2 are the charge (in units of e:
en = 2/3, ed = -1/3), scale magnetic moment, mass, g-factor and spin, respectively, of
the third constituent. Using the results of Table 1 we obtain
1 G
+ Pa2)2;
p
GE = (1 E= 0, (8)
for the charge form factors. The scale parameter in the distribution (5) is related to the
proton charge radius (r 2 )j, = 12a 2 Similarly, for the magnetic form factors we obtain
(9)
respectively. Here we have assumed spin-flavor symmetry so that the mass and the
g- factor of the up(u) and down (d) constituents are identical, mu = md= mq and
198
gu = gd = g. Accordingly /1u = /1d = /1 and /1 = eg/2m q in eq. (10). The corresponding
harmonic oscillator elastic form factors are obtained by replacing the dipole function
(1 + Pa 2 )-2 by the exponential function e- k2 (32/6 with (3 related to the harmonic os-
cillator size parameter. The form factors in eqs. (8)-(9) (and their harmonic oscillator
analogues) satisfy G~ = flGZ;; and the relations G'E = 0 and GM/G~ = -2/3 for all
values of the momentum transfer. These relations are due to spin-flavor symmetry, but
are not obeyed by the experimental data. Within a truncated three-constituents con-
figuration space, to have a nonvanishing neutron electric form factor, one must break
SU8j (6) [11]. This breaking can be achieved in various ways, e.g. by including in the
mass operator a hyperfine interaction [12], or by distorting the oblate-top geometry,
allowing for a quark-diquark structure [13]. Within the model discussed here we study
the breaking of the SUsj (6) symmetry by assuming a flavor-dependent distribution of
the charge and the magnetization along the strings of Fig. 1,
With this dependence and for small symmetry-breaking the electric nucleon form factors
become
GnE ~
~ (12)
to leading order in D.. Here D. == (a~ - a~)/O,2 with 0,2 = (a~ + a~)/2.
If the length of the string in Fig. 1 is slightly different for u and d, so is their mass
and thus in principle, their magnetic moment. Applying the same procedure to the
magnetic form factors gives (to leading order in D.)
Here /1ueu and flded are the magnetic moments of the u and d constituents which
determine the proton and neutron magnetic moments
(14)
In the calculations reported below we have used the following procedure to determine
the parameters. In all cases we take gu = gd = l. We assume that the constituent masses
mu and md are determined from the measured magnetic moments. This fixes the scale
magnetic moments flu andfld. The scale parameters au and ad in the distributions (11)
are determined from a simultaneous fit to the proton and neutron charge radii, and
to the proton and neutron electric and magnetic form factors. For the calculations in
which the SUs!( 6) symmetry is satisfied (/1u = fld = fl and au = ad = a) this procedure
yields a = 0.232 fm and fl = /1p = 2.793 flN (= 0.126 Gey-l), which corresponds to a
constituent mass of mu = md = 0.336 GeY. For the calculations in which the SUs](6)
symmetry is broken we find au = 0.230 fm and ad = 0.257 fm (implying D. = 0.2 in
eqs. (12)-(13) ), flu = 2.777flN (= 0.126 Gey- 1 ), /1d = 2.915flN, (= 0.133 Gey-l)
corresponding to mu = 0.338 GeY and md = 0.322 GeY, respectively.
199
Fig. 4 shows the elastic electric form factors of the proton and the neutron divided
by the dipole form, Fv = 1/(1 + Q2/0.71)2. The division by Fv emphasizes the effect of
the breaking of spin-flavor symmetry. Fig. 5 shows the results for the elastic magnetic
form factors. We see that while the breaking of spin-flavor symmetry can account for
the non-zero value of G E and gives a good description of the data, it worsens the fit
to the proton electric and neutron magnetic form factors. This implies either that
the simple mechanism for spin-flavor breaking of eq. (11) does not produce the right
phenomenology and other contributions, such as polarization of the neutron into p+ 7r- ,
play an important role in the neutron electricform factor [14]. (A coupling to the meson
cloud through p, wand cP mesons is indeed expected to contribute in this range of Q2
[15].) This conclusion (i. e. worsening the proton form factors) applies also to the other
mechanisms of spin-flavor symmetry breaking mentioned above, such as that induced
by a hyperfine interaction [12] which gives au < ad ('moves the up quark to the center
and the down quark to the periphery'). This pattern is a consequence of the fact that
within the framework of constituent models Gr;;, GE, G~ and G'M are intertwined as is
clearly evident from the expressions in eqs. (12)-(13).
Helicity Amplitudes
Other (observable) quantities of interest are the helicity amplitudes in photo- and
electroproduction. The transverse helicity amplitudes between the initial (ground) state
of the nucleon and the final (excited) state of a baryon resonance are expressed as [5]
where v = 1/2, 3/2 indicates the helicity. The orbit- and spin-flip amplitudes (A and
B, respectively) are given by
(17)
Here M is the nucleon mass, W is the mass of the resonance, and _Q2 can be interpret-
ed as the mass squared of the virtual photon (for elastic scattering we have k 2 = Q2). In
200
2.0
1.5
c
u..
~ 1 .0
't.?
0.5
N(939)P 11
0.0
0 .0 1 .0 2.0 3 .0 4.0 5.0 6.0 7.0 8 .0 9 .0 10.0
Q2 [ (GeV/c)2 )
0.20
0.1 5
c
u..
~ 0.10
'G
0.05
0.00
Figure 4. Comparison between the experimental proton (G~;) and neutron (G'EJ electric
form factors with the corresponding collective form factors. Dashed (solid) lines correspond
to a calculation with (without) flavor breaking. The form factors are divided by the dipole
form factor, FD = 1/(1 + Q2/0.71)2.
201
1.5
N(939)P II
u.."
::l
..... 1.0
~=+
' t +' t
CI. ::!
t
<.?
0.5
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 B.O 9.0 10.0
02 [ (GeV/c)2)
N(939)P1I
1.5
Cl
u..
::::l
'b::! 1.0
0.5
Figure 5. Comparison between the experimental proton (G~/f.1p) and neutron (G M/f.1n)
magnetic form factors with the corresponding collective form factors. Dashed (solid) lines
correspond to a calculation with (without) flavor breaking. The form factors are divided by
the dipole form factor, FD = 1/(1 + Q2/0.71)2.
202
assessing the quality of the fits it should be noted that there are no adjustable param-
eters involved in the calculation of helicity amplitudes. All the parameters appearing
in eq. (16) and in the distributions (ll) are those extracted previously from the elastic
form factors.
Representative results for transverse helicity amplitudes A1 / 2 and A3 / 2 are shown in
Fig. 6 for the nucleon resonances N(1520)D 13 and N(16S0)F15' From this figure it is
seen that the effect of spin-flavor breaking is rather small. Only in those cases in which
the amplitude with SUs j(6) symmetry is zero, the effect is of some relevance. Such is
the case with proton helicity amplitudes for the 4S J [70, L P ) multiplet (e.g. the L P = 1-
resonances N(1675)D 15 and N(1700)D13) and with neutron helicity-3/2 amplitudes for
the 2S J [56, L P ) multiplet (e.g. the L P = 2+ resonance N(16S0)FlS)' The conclusion
that one can draw from this analysis is that, for all purposes, with the exception of the
electric form factor of the neutron, the breaking of spin-flavor symmetry according to
the mechanism of eq. (11) is of little importance.
In a string-like model of hadrons one expects on the basis of QCD [16) that strings
will elongate (hadrons swell) as their energy increases. This effect can be easily included
in the present analysis by making the scale parameters of the strings energy dependent.
We use here the simple ansatz
a=aol+~ ( W~M)
M ' (IS)
where M is the nucleon mass and W the resonance mass. This ansatz introduces a new
parameter (0, the stretchability of the string.
Fig. 7 shows the effect of stretching on the helicity amplitudes for N(1520)D13 and
N(16S0)FlS. It is seen that the effect of stretching, especially if one takes the value
~ ~ 1 suggested by the arguments of [16) and the Regge behavior of nucleon resonances
(see e.g. Fig. 5 of [5)), is rather large. In particular, the data for N(16S0)FJ.5 and
N(1520)D13 show a clear indication that the form factors are dropping faster than
expected on the basis of the dipole form. (Of course for the elastic form actors there is
no stretching.) We suggest that future data at CEBAF and MAMI be used to analyze
the effects of stretching on the helicity amplitudes.
STRONG COUPLINGS
We consider in this section strong decays of baryons of the form B -+ B' + M [S).
The process involves an emission (by one of the constituents in B) of an elementary
pseudoscalar meson meson (M = 7r or 1]) with energy ko = EM = EB ~ EB' and
momentum k = PM = P ~ P' = kz. Here P = Pzz and pI (= P:z) are the momenta of
the initial (B) and final baryon (B ' ). The calculations are performed in the rest frame
of B (Pz = 0).
In the collective model the strong couplings (similar to the electromagnetic cou-
plings) are obtained by folding the matrix elements of the transition operator inducing
the decay with the distribution function g((3) of eq. (5). These collective matrix ele-
ments can be expressed in terms of helicity amplitudes. For decays in which the initial
baryon has angular momentum j = L + and in which the final baryon is either the
nucleon or the delta with wave functions (2), the (strong) helicity amplitudes are
+~(L,~I,S'V+IIJ,v)(_Z+(k)] . (19)
203
200
tt N(1520)D'3
+
v =3/2
O L~f.--~==~~===l~
+ T
100
~~t~f+~====~~_=_~_==
v = 1/2
200
N(1 680)F'5
v = 312
'?
o
o ~~------------------------------------~~
100
v = 1/2
204
Figure 7. Effect of hadron swelling for excitation of N(1520)D 13 (a factor of +i is suppressed)
and N(1680)F15. The curves are labelled by the value ofthe stretching parameter ~ of eq. (18).
205
The coefficients (m are spin-flavor matrix elements [8] and Zm(k) (m = 0, ) are the
radial matrix elements
Zo(k) = 6[gk-~hk]F(k)*-6h9z(k)*,
Z(k) = -6h9'f(k)*. (20)
The coefficients 9 and h denote the strength of two terms in the transition operator.
The radial matrix elements Zm (k) involve the same collective form factors F( k), 9'f (k)
discussed in section 3. The reason for the complex conjugation in eq. (20) is that here we
consider an emission process, whereas Table 1 shows the form factors for an absorption
process.
The decay widths for a specific channel are given by [17]
Wo(k)
(23)
Partial widths for other models of the nucleon and its resonances can be obtained
by introducing the corresponding expressions for the elementary amplitudes Wi( k). For
example, the relevant expressions in the harmonic oscillator quark model are
Wo(k) PIP} e-
3i {[gk - "61 hk1k(3 + hm3ko(3(3 - -3-) k 2 f32/6
,
e =
47r _EB=--'-,---(k-,--)E-.-::.:M,,--,(---,k)_k
Pi (25)
206
with EB,(k) = Jm1, + k 2 and EM(k) = Jm1r + Po
We consider here decays with emission of 7r and T). The calculated values depend on
the two parameters 9 and h in eq. (20) and on the scale parameter a of eq. (5). In the
present analysis we determine these parameters from a least square fit to the N 7r partial
widths (which are relatively well known) with the exclusion of the S11 resonances. For
the latter the situation is not clear due to possible mixing of N(1535)Sll and N(1650)Sll
and the possible existence of a third S11 resonance [18]. As a result we find g = 1.164
GeV- 1 and h = -0.094 GeV- 1 . The relative sign is consistent with a previous analysis
of the strong decay of mesons [19] and with a derivation from the axial-vector coupling
(see e.g. [17]). The scale parameter, a = 0.232 fm, extracted in the present fit is found
to be equal to the value extracted in the calculation of electromagnetic couplings [7].
We keep g, h and a equal for all resonances and all decay channels (N7r, NT), tl7r, tlTJ).
In comparing with previous calculations, it should be noted that in the calculation
in the nonrelativistic quark model of [20] the decay widths are parametrized by four
reduced partial wave amplitudes instead of the two elementary amplitudes 9 and h.
Furthermore, the momentum dependence of these reduced amplitudes are represented
by constants. The calculation in the relativized quark model of [21] was done using a
pair-creation model for the decay and involved a different assumption on the phase space
factor. Both the nonrelativistic and relativized quark model calculations include the
effects of mixing induced by the hyperfine interaction, which in the present calculation
are not taken into account.
The calculations of decay widths of 3* and 4* resonances into the N 7r and tl7r
channels are in fair agreement with experiment. Representative results are shown in
Fig. 8. The results are to a large extent a consequence of spin-flavor symmetry. The use
of 'collective' form factors improves somewhat the results when compared with older
(harmonic oscillator) calculations. This is shown in Table 2 where the decay of a tl
Regge trajectory into N 7r is analyzed and compared with the calculations of [17], which
are based on the harmonic oscillator model discussed in [22]. We also include the results
of more recent calculations in the nonrelativistic quark model [20] and in the relativized
quark model [21]. There does not seem to be anything unusual in the decays into 7r
and our analysis confirms the results of previous analyses.
Contrary to the decays into 7r, the decay widths into T) have some unusual properties.
The calculation gives systematically small values for these widths. This is due to a
combination of phase space factors and the structure of the transition operator. Both
depend on the momentum transfer k, however, due to the difference between the 7r
and T) mass, the momentum carried by the T) is smaller than that carried by the 7r.
Therefore, the T) decay widths are suppressed relative to the 7r decays. The spin-flavor
part is approximately the same for N7r and NTJ, since 7r and T) are in the same SUj (3)
multiplet. We emphasize here, that the transition operator was determined by fitting
the coefficients g and h to the N 7r decays of the 3* and 4* resonances. Hence the TJ
decays are calculated without introducing any further parameters.
The experimental situation is unclear. The 1992 Particle Data Group (PDG) compi-
lation gave systematically small widths (rv 1 MeV) for all resonances except N(1535)Sl1'
The 1994 PDG compilation deleted all T) widths with the exception of N(1535)Sl1' This
situation persists in the latest PDG compilation [23], where N(1650)Sl1 is now assigned
a small but non-zero T) width. The results of our analysis suggest that the large "7 width
for the N(1535)Sll is not due to a conventional q3 state. One possible explanation is the
presence of another state in the same mass region, e.g. a quasi-bound meson-baryon
S wave resonance just below or above threshold, for example NT), KE or K A [24].
Another possibility is an exotic configuration of four quarks and one antiquark (q4ij).
207
JP L S
1950
1920
1910
7/']
3/2+
1/2+ 2 3/2
1905 5/2+
208
Table 2. Strong decay widths for.6.* -+ N +7r and N* -+ N+7r in MeV. Experimental
values are from [23].
N(1520)D 13 1 85 74 115 67 9
N(2190)G 17 3 48 34 67 27
N(2600)hn 5 11 9 49 20
ACKNOWLEDGEMENTS
The results reported in these proceedings are based on work done in collaboration
with F. Iachello (Yale). The work is supported in part by grant No. 94-00059 from the
United States-Israel Binational Science Foundation (BSF), Jerusalem, Israel (A. L.)
and by CONACyT, Mexico under project 400340-5-3401E and DGAPA-UNAM under
project IN105194 (R. B.).
REFERENCES
[1] M. Gell-Mann, Phys. Rev. 125 (1962), 1067; Y. Ne'eman, Nucl. Phys. 26 (1961),
222.
[2] F. Gursey and L. A. Radicati, Phys. Rev. Lett. 13 (1964), 173.
[3] N. Isgur and G. Karl, Phys. Rev. D18 (1978),4187; D19 (1979), 2653; D20 (1979),
1191.
[4] S. Capstick and N. Isgur, Phys. Rev. D34 (1986), 2809.
209
[5] R. Bijker, F. Iachello and A. Leviatan, Ann. Phys. (N.Y.) 236 (1994), 69.
[6] R. Bijker and A. Leviatan, these proceedings.
[7] R. Bijker, F. Iachello and A. Leviatan, Phys. Rev. C54 (1996), 1935.
[8] R. Bijker, F. Iachello and A. Leviatan, preprint nucl-th/9608057, Phys. Rev. D, in
press.
[9] R. Bijker and A. Leviatan, in: Symmetries in Science VII: Spectrum Generating
Algebras and Dynamic Symmetries in Physics, B. Gruber and T. Otsuka (eds.),
Plenum Press, New York, 1993, 87; R. Bijker and A. Leviatan, Rev. Mex. Fis.
39, Suplemento 2, (1993) 7; ibid. 42 Suplemento 2 (1995), 62; A. Leviatan and R.
Bijker, 7r N Newsletter 11 (1995), 142.
[10] N. Isgur, G. Karl and R. Koniuk, Phys. Rev. Lett. 41 (1978), 1269.
[11] J. L. Friar, Part. and Nucl. 4 (1972),153; R. D. Carlitz, S. D. Ellis and R. Savit,
Phys. Lett. B68 (1977), 443.
[12] N. Isgur, G. Karl and D. W. 1. Sprung, Phys. Rev. D23 (1981),163.
[13] Y. Tzeng and S. S. Hsiao, 11 Nuovo Cimento 106 (1993), 573.
[14] F. E. Close and R. R. Hogan, Nucl. Phys. BI85 (1981), 333; O. V. Mawell and
V. Vento. Nucl. Phys. A407 (1983), 366; S. Theberge, A. W. Thomas and G. A.
Miller, Phys. Rev. D22 (1980), 2838.
[15] F. Iachello, A. D. Jackson and A, Lande, Phys. Lett. B43 (1973), 191. G. Hohler,
E. Pietarinen, 1. Sabba-Stefanescu, F. Borkowski, G. G. Simon, V. H. Walther and
R. D. Wendling, Nucl. Phys. B114 (1976), 505.
[16] K. Johnson and C. B. Thorn, Phys. Rev. DI3, (1974) 1934; 1. Bars and H. J.
Hanson, Phys. Rev. DI3, (1974) 1744.
[17] A. Le Yaouanc, L. Oliver, O. Pene and J.-C. Raynal, Hadron transitions in the
quark model, Gordon and Breach, 1988.
[18] Z. Li and R. Workman, Phys. Rev. C53 (1996), R549.
[19] C. Gobbi, F. Iachello and D. Kusnezov, Phys. Rev. D50 (1994), 2048.
[20] R. Koniuk and N. Isgur, Phys. Rev. Lett. 44 (1980), 845; Phys. Rev. D2I (1980),
1868.
[21] S. Capstick and W. Roberts, Phys. Rev. D47 (1993), 1994; ibid. D49 (1994),
4570.
[22] R. R. Dalitz, in: Quarks and Radronic Structure, M. Morpurgo (ed.), Plenum, New
York, 1977.
[23] Particle Data Group, Phys. Rev. D54 (1996), 1.
[24] N. Kaiser, P. B. Siegel and W. Weise, Phys. Lett. B362 (1995), 23.
210
DISCRETE REFLECTION GROUPS AND INDUCED
REPRESENTATIONS OF POINCARE GROUP ON
THE LATTICE
Miguel Lorente
Departamento de Fisica
Facultad de Ciencias
Universidad de Oviedo
E-33007 Oviedo, Spain
INTRODUCTION
The assumption of a physical model on discrete space and time leads to some changes
of the mathematical equations. In particular, the symmetries of the model has to be
restricted to integral transformations acting on some vector space defined over the inte-
gers. The space-time groups are subgroups of the set of non-singular integral matrices,
and the classical and quantum laws are written with the help of difference operators.
In [1] we have developed some properties of non-compact groups acting on lattice with
non-Euclidean metric. In [2] we have modified some of the postulates of quantum me-
chanics in order to incorporate the hypothesis of discrete space and time. In [3] we have
proposed a new scheme for the Klein-Gordon and Dirac wave equation. The Quantum
fields on the lattice is the subject of very extensive literature [4] and the theoretical
difficulties of it have not yet been solved.
In section 2, we review some results of discrete reflection groups that give raise to
a complete classification of n-simplex reflection groups of the spherical, Euclidean or
hyperbolic type, and of the generalized hyperbolic reflection groups whose connection
with integral space-time groups is made via Coxeter groups.
In section 3 and 4 we introduce the definitions and main properties of a discrete
differential geometry in order to construct covariant expression for the Maxwell, Dirac
and Klein-Gordon difference equations.
In section 5 we apply the standard formulation of induced representations to calcu-
late the transformation properties of the Mackey, Wigner and covariant state functions
that belong to the carrier space for the unitary representations of Poincare group on
discrete momentum space.
REFLECTION GROUPS
ii) the exponent kij is either a positive integer or 00 for each i,j;
The Coxeter graph of G is the labeled graph with vertices J and edges
{(i,j) : k ij > 2} .
Each edge (i, j) is labeled by kij . For simplicity, the edges with kij = 3 are usually not
labeled in a Coxeter graph.
Only some particular types of Coxeter groups have been classified. We present here
some of the most fundamental and useful Coxeter groups: the simple reflection groups
and the generalized simplex reflection groups.
Let ~ be an n-simplex in X (= sn, En or Hn) all of whose diedral angles are
submultiple of Jr (by n-simplex we define a convex n-dimensional polyhedron in X with
212
exactly n + 1 vertices). The group r generated by the reflections of X in the sides of
lI. is an n-simplex reflection group.
If n = 1 then lI. is a geodesic segment in X, and r is the dihedral group of order 2k
with k > 1, where 1r / k is the angle of 6. The Coxeter graph is
k
or
For X = 51, k is finite; for X = El or HI, k = 00.
Assume that n = 2. Then lI. is a triangle in X whose angles 1r / a, 1r /b, 1r / care
submultiple of 1r. If we call T( a, b, c) the triangle determined by the integer numbers
(a, b, c), then the group r generated by the reflections in the sides of T( a, b, c) is denoted
by G(a, b, c) and it is called a triangle reflection group. If X = 52 the only spherical
triangle reflection groups have the following Coxeter graph:
v
v
G(2,2,2) G(2,2,c) G(2,3,3)
4
v
v
G(2,3,4) G(2,3,5)
L"--v--'
G(3,3,3)
4
v
G(2,4,4)
4
v
G(2,3,6)
6
If X = H3 we have the hyperbolic triangle reflection groups with Coxeter diagram:
~--v,,",---'
'-.-'
G(2,b,c) G(a,b,c)
Let r be the group generated by the reflections of X in the sides of an n-simplex lI.
all of whose dihedral angles are submultiples of 1r. The group r said to be irreducible
if and only if its Coxeter graph is connected.
The classification of all the ireducible n-simplex (spherical, Euclidean and hyperbol-
ic) reflection groups is complete. Spherical and Euclidean n-simplex reflections groups
exist in all dimensions; however, hyperbolic n-simplex reflections groups exist only for
dimension n ::; 5 [5].
Another type of Coxeter groups that have been classified are the generalized simplex
reflection groups, which are defined only in Hn. A generalized n-simplex in Hn is a
n-dimensional polyhedron with n + 1 generalized vertices (a generalized vertex of a
convex polyhedron P is either a vertex of P or an ideal vertex of P).
The generalized hyperbolic triangle reflection groups have the following Coxeter
graphs:
213
00
00 00
b>2
The generalized hyperbolic n-simplex reflections groups exist only for n ::; 10 [6].
Some particular very interesting cases of the last hyperbolic generalized reflection
groups are the following:
r 2,1 : _ _ 4---'.0--_
00_ . r 3 ,1 : _ __ 4 __ 4 _...
r 4,1 : ----<1 . - - -. . 1- 4
--.. r 5 ,1 - r 9 ,1 : ----I--.. ------~
If we take a Cartesian basis {ei}, i = 0,1, ... ,9, a realization of the reflections
c00)
corresponding to these Coxeter groups can be given in matrix form:
r 2,1 : 51 = o
o
0 1
1 0
, 52 = C0: ),
0 1
0 0 -1
53 = ( !,
-2
2
-1
-2 -1
~, )
f 3 ,1 :
S,~ U
0
0
1
0
0
1
0
0
D' S,~ U
0
1
0
0
0
0 o
0 1
1 o
0) ' 53 = C )
0
0
0
1
0 0
0 0 0
0 1 0
0 0 -1
('
1 1
54 =
-1 0 -1
-1 -1 0
~1
-1
)
-1 -1 -1 0
f n ,l ; 5i = Xi ...... Xi+1 i = 1,2, ... , n - 1; n = 4,5, ... ,9.
5n = Xn ...... -Xn
2 1 1 1
S"+, ~ (
-1 0 -1 -1
-1 -1 0 -1
-1 -1 -1 0
0 1:_' )
It can be checked that these reflections satisfy the presentation of the Coxetergroups
The set of all proper Lorentz transformations that leave the Lorentian lattice An in-
variant is generated by the elements of the Coxeter group r n,l, (n = 2, ... ,9), the proof
214
is given by Kac [7]. Similarly Coxeter has proved [8] that all integral Lorentz transfor-
mations (including reflections) are obtained by combining the operation of permuting
the spacial coordinates x!, X2, X3 and changing the signs of the coordinates Xo, X!, X2, X3,
together with the transformation
( 2-1 -1 -1)
-1 0 -1 -1
-1 -1 0 -1 .
-1 -1 -1 0
We have arrived at the same result in the case of r 2 ,1 using the isomorphism between
SO (2, 1) n GL (3, Z) and GL (2, Z): formulae (77) and (78) of ref. [1] correspond exactly
to the reflections given by the Coxeter groups r 2,1.
Given a function of one independent variable the forward and backward differences
are defined as
tl,cf(x,y) f(x+tlx,y)-f(x,y) ,
tlyf(x,y) f(x,y + tly) - f(x,y) ,
1
Li.J(x,y) - 2"{f(x+tlx,Y)+f(x,y)}
1
Liyf(x,y) 2"{f(x,y+tlY)+f(x,y)} ,
tlf(x,y) f(x + tlx, y + tly) - f(x, y) ,
1
Lif(x,y) 2"{f(x+tlx,y+tlY)+f(x,y)} .
These definitons can be easily generalized to more independent variables but for the sake
of brevity we restrict ourselves to two independent variables. From the last definitions
the following identities can be proved:
215
Given a vectorial space vn over Z we can define a real-valued linear function over Z
The forms w constitute a vectorial linear space (dual space) *V n , and can be expanded
in terms of a basis w Ot
The basis e{3 of vn and w Oi of *vn can be contracted in the following way
(6)
hence
(7)
with u = u"e".
If we take w{3 = D.x{3 as coordinate basis for the linear forms, we can construct
discrete differential forms (a discrete versIon of the continuous differential forms) [9). A
particular example of this discrete form is the total difference operator (3) of a function
of several discrete variables written in the following way:
For these discrete forms we can define the exterior product of two form (j and p
p 1\ (j = -(j 1\ p ,
which is linear in both arguments.
For the coordinate basis we also have
where 5 is the n-dimensional Levy-Civitta totally antisymmetric tensor (5123 ... == 1).
EXTERIOR CALCULUS AND LORENTZ TRANSFORMATIONS
flw
where in the last expression we have used the properties of the exterior product.
This definition of exterior difference can be easily written for I-form in n-dimensional
space.
Given a 2-form in a 3-dimensional space,
flw
(12)
The exterior derivative applied to the product of a O-form (scalar function f) and a
I-form (w = aflx + bfly) is
(14)
The exterior difference of the product of a p-form p and a q-form (J is
(15)
(16)
Some examples:
From the Faraday 2-form F.= !Fl'vflxl' /\ flxv we write down one set of Maxwell
difference equations
217
(17)
from the Maxwell 2-dual form *F we get the other set of Maxwell equations
t.*F = 4?r*J ,
( LiXLS)izt.tP
t.t +
t.xLiyLizLitlx Lixt.yLizLitly LixLiyt.zt.tlz)
t.x + t.y + t.z .
t.t 1\ t.x 1\ t.y 1\ t.z = 0 . (18)
Note that the coefficient of the difference form is the discrete version of the continuity
equation.
From a scalar function we get the wave equations in terms of difference operators,
namely
- * t. * t. == D , (19)
where D is the discrete d'Alambertian operator:
(21)
where the t.xl' are the space-time intervals in Minskowski space-time. From
(22)
where A~' is a global Lorentz transformation, and from the coordinate-free expression
for w we get
(23)
recalling that A~/A~' = 8~.
From the total difference of a function of several variables f (x, y, z, t),
218
it follows that the coefficients of the I-forms, namely
(25)
(26)
where Al, A2 are two integral Lorentz transformations described in section 2, and al, a2
are two sets of four integer numbers that define the parameters of the translation groups
T4 in Minkowski space.
In order to construct the unitary representation of the Poincare group, we use the
standard method of induced representation [10]. All the properties of these repre-
sentations can be translated into the language of integral representations. It is well
known that the irreducible representations of the inhomogeneous space-time groups
can be realized on different kind of states such as Mackey, Wigner or covariant
states.
Consider now the representations acting on the Wigner states, corresponding to
massive particles. Following standard procedure we need:
i) the unitary representation of the translation group
Pi' being the discrete momentum and a = (ao, al, a2, a3) the parameters of the group.
ii) The unitary representation of the Wigner little group (in our case SU(2)) cor-
responding to integral rotations in 3-dimensional space. They are only 24 different
elements of this type.
iii) The representative momentum is (m, 0, 0, 0) == PI" and the orbit generated by
this vector is given by the boost
PI" = A~Pv. (28)
Pv is defined on the Minkowski lattice:
(29)
the points (t, x) lying in the integral hyperboloid generated by the reflection S4 of f 3 ,1
(see section 2).
219
With these ingredients we can write down the transformation properties of the
Wigner functions in momentum space under the transformations of the restricted Poin-
care group, namely [ref. 10, formula 16.2]
(30)
The integral representation of the SU(2) group can be given by the use of Cayley
parametrization
with det = n6 + ni + n~ + n~ = 1 or 2.
For representation of higher dimension we have to substitute the generators 5 by
the corresponding representations of 5.
For the covariant states the unitary representations of the Poincare group restricted
to integraltransformations is given by the following transformations [see ref. 10, formula
(1 +fE:
17.1]
3 1 P )a~
U(a,A)1j;(P)=II I' D(A)1j;(A-lp) , (31)
1'=0
1- -ZE:P
2 I'
where D (A) is an integral representation of the restricted Lorentz group. In our case
it is a representation of the subgroup SL (2, C) n GL (2, Z) which is homomorphic to
the integral Lorentz group as described in section 2.
In order to construct covariant functions that transform under irreducible represen-
tation of the Poincare ,e;roup we need a subsidiary condition on the functions:
(32)
where Q is a projection operator that restricts the components of the vector 1j; (p) to
one representation of unique spin.
For example, for the Dirac representation of SL (2, C), which contains D1 / 2 (SU (2))
twice, we take
Q= ~ (1 13), (33)
Taking p= (mo,O,O,O), this equation leads to the wave equation in the rest system
(34)
220
and ex = (~ ~), ~ = (~ ~I ) , we obtain
In order to construct the wave functions in configuration space we solve the Klein-
Gordon and Dirac wave equation on the lattice. This procedure can be easily generalized
to higher spin representations.
We define the scalar function on the (3 + 1) dimensional cubic lattice
'f/ + =
- rr-fl.1"
1'=0
Then the Klein-Gordon wave equations defined on the grid points of the lattice can
be read off
(38)
It can be verified by direct substitution that the plane wave solution satisfy the differ-
ence equation
(40)
221
From section 5 the Klein-Gordon equation is invariant under finite Lorentz transforma-
tions.
The discrete version of the Dirac wave equation can be written as
(41)
where 1M J.l = 0,1,2,3, are the usual Dirac matrices. Applying the operator
from the left on both sides of (41), we recover the Klein-Gordon equation (38). Let now
construct solutions to (41) of the form
where the f (1, n) are given in (39). The four-component spinors w (k, E), with spatial
momentum k == (k1) k2' k3 ), must satisfy
(42)
(43)
The transformation properties of the wave functions in the configuration space are
given, as in the continuous case, as follows:
(44)
(45)
ljJ(P) = "L.J
00 (1 + iiep)i
~.
1- -uP
ljJ(ej) (46)
j=-oo 2
L
00
222
Using summation by parts and the boundary conditions we derive that (36) is the
Fourier transform of (44), that (37) is the Fourier transform of (45), and that (42) is
the Fourier transform of (41). The consistency of the new discrete Fourier transform in
(47) is proved with the help of the 8 sequence:
SL (j) = L
L
k=-L
(1 + rk)
i
1 - '2Ek
j
N
r 2N
N:..:! 1 " SL (J') = 1
~
j=-N
for all L.
ACKNOWLEDGMENT
The author wants to express his gratitude to professor Peter Kramer for many
illuminating critics and remarks.
This work has been partially supported by spanish grant of D.G.I.C.Y.T. (Project
PB 94 - 1318).
REFERENCES
[1 J M. Lorente and P. Kramer, in:' Symmetries in Science VIII (B. Gruber ed.), Plenum,
New York, 1995, 315
[2] M. Lorente, Quantum Mechanics on Discrete Space-time, II Int. Symposium on
Fundamental Problems of Quantum Physics (Oviedo 1996).
[3] M. Lorente, J. Group Theor. in Phys. 1 (1993), 105.
[4] 1. Montvay, G. Munster, Quantum Fields on a Lattice, Cambridge University Press,
1994.
[5] J. G. Ratcliffe, Foundations of Hyperbolic Manifolds, Springer, 1994, 294
[6J J. E. Humphreys, Reflection groups and Coxeter groups, Cambridge University
Press, 1990, 142-144.
[7J V. Kac, Infinite dimensional Lie Algebras, Cambridge University Press, 69-71.
[8] A. Schild, Can. J. Math. 1 (1948),29 (appendix).
[9] See M, Lorente, Discrete Differential Geometry and Lattice Field Theory, in: VII
International Conference on Symmetry Methods in Physics (Dubna 1995)
[10] U. H. Niederer and L. O'Raifeartaigh, Fortschr. Phys. 22 (1974), 11.
223
QUANTUM MECHANICS AND
CLASSICAL PROBABILITY THEORY
V. I. Man'ko
INTRODUCTION
It was shown [7], that for the generic linear combination of quadratures which is a
measurable observable (1i = 1 )
(1)
where q and p are the position and momentum, respectively, the marginal distribution
W (X, ft, v, 8) (normalized with respect to the X variable), depending upon three
extra real parameters ft; v; 8, is related to the state of the quantum system expressed
in terms of its Wigner function W(q, p) as follows
W (X, ft, v, 8) = J .
exp [-zk (X - ftq - vp ~ 8)] W (q, p)
dk dq dp
(2 Jr)2 . (2)
As it follows from this formula, the marginal distribution depends on the difference of
the variables X - 8. The physical meaning of the parameters ft; v; 8 is that they
describe ensemble of shifted, rotated, and scaled reference frames in which the position
X is measured. Formula (2) can be inverted and the Wigner function of the state can
be expressed in terms of the marginal distribution [7]
(3)
where WF(X, a, b, s) is the Fourier component of the marginal distribution (2) taken
with respect to the parameters ft; v; 8, namely,
WF(X, a, b, s) = J .
w(X, ft, v, 8) exp[-z(fta+vb+8s)]
dft d/J d8
(21ly . (4)
Since the Wigner function determines completely the quantum state of a system and,
on the other hand, this function itself is completely determined by the marginal distri-
bution, one can understand the notion of the quantum state in terms of the classical
marginal distribution for squeezed and rotated quadrature.
So, we say that the quantum state is given if the position probability distribution
W (X, ft, v, 8) in ensemble of rotated, squeezed, and shifted reference frames in classical
226
corresponds to optical tomography scheme [10, 11], i.e., ft = cos <p and the rotation
angle <p labels the reference frame in classical phase space.
So, we formulate also the notion of quantum states as follows:
We say that the quantum state is given if the position probability distribution 'W(X, <p)
in ensemble of rotated reference frames in classical phase space is given.
Since the quantum state is defined by position distribution, one could associate an
entrophy with the state using standard relation known in classical probability theory,
i.e., the entrophy S (fl, v, 8) is given by the formula
If we use the distribution 'W (X, <p) , the entrophy S (<p) depends only on the rotation
angle.
One could also use the introduced formulation of notion of the quantum states to
describe situations in which the states are either close or essentially different. 'vVe say,
that two states are close if their distributions are close, i.e., all the highest momenta
of the distributions differ very slightly. We also say, that two states are essentially
different if their distributions differ essentially, i.e., there are highest momenta for the
two distributions with the large corresponding difference.
(6)
the marginal distribution satisfies quantum time-evolution equation, being the integral
equation determined by the imaginary part of the potential energy considered as the
function of a complex coordinate. The evolution equation is
(10)
227
Eq. (10) follows from the evolution equation (7) for the marginal distribution of the
quantum system [5], if the marginal distribution does not depend on time. Thus, the
normalized marginal distributions of stationary states of quantum systems satisfy the
system of two equations (9) and (10).
We consider an example of the quantum harmonic oscillator since it is one of the
most important quantum systems. For this case, using the Hamiltonian
(11)
. {) {) 0
11! - I)' {)v 11! + v {)fJ, 11! = . (12)
Thus, we obtain the solutions for the alternatives to Schrodinger equations for the
harmonic oscillator [Eqs. (9) and (10)] given by
[ (X _ 8)2]
(OS)
11!o
,
(X, fJ" v, 8) = J rr (fJ,21 + v2) exp - 2
fJ, + V
2 (14)
These solutions correspond to the quantum energy levels En = n + 1/2. Since the
problem of the harmonic oscillator is commonly used in quantum physics, below we
derive expressions (13) and (14) in detail by two methods.
QUANTUM TRANSITIONS
For the transition probability, one can obtain, in view of (17), the expression in terms
of the marginal distributions
P 12 = J
Wi(X, fJ" v, 8) W2(Y, -fJ" -v, 8) exp [2 (X
. +Y - 28)]
dfJ,dvdXdY
2rr
. (18)
228
As follows from (18), any pure normalized quantum state is described by the marginal
distribution wp (X, {-t, v, 8) , which satisfies the additional condition
The complex wave functions, which belong to different energy levels of a quantum sys-
tem, are orthogonal. This orthogonality condition is expressed in terms of the classical
marginal distribution as the relation
where the labels m, n correspond to the energy levels Em, En. The pure states 1 n)
satisfy the completeness relation
L 1n)(n 1= 1. (21 )
n
This relation may be rewritten as the condition for the marginal distributions of the
pure states with the energies En
(22)
n
In this section, we derive the expression for the marginal distribution W (X, {-t, v)
for the energy level of the harmonic oscillator En = n + 1/2. First, we use the known
expression for the oscillator wave functions. The wave function of the excited state of
the harmonic oscillator
(23)
yelds the following expression for the density matrix of this state
(24)
229
To evaluate this integral, we obtain the marginal distribution w",(X, p" 11) for the
coherent state of the oscillator with the wave function
(26)
We have
(27)
Since the coherent state is generating function for the excited states
(28)
we have
n*m
v'
00 00
Q9 J [-l- :2 -
exp ik(X - p,q - vp) - iPU]
Then
(31)
The integral of (27) has the Gaussian form and for the marginal distribution of the
coherent state we obtain the expression
(32)
(33)
we obtain
Wnm(X, 11) [7r (p,2 + V2)tl/2 2-(n/2+m/2) (n! m!t1/2 (v + ip,)n (1/ - ip, )n
p"
(p,2 + v2 )n/2+m/2
Q9 exp (- p,2:2V2) Hn (k) (vp,:+ Hm 11 2). (34)
230
From this expression, in view of (31), we obtain the marginal distribution for the n-th
excited state of the harmonic oscillator
which coincides with expressions (13) and (14). This eigendistribution function satisfies
the eigenvalue equation
EWE(X,~,v), (36)
which has solutions (35) only for the discrete values of the energy En = n + 1/2.
For completeness, we derive (35) directly solving eq. (36). This equation may be
rewritten for the Fourier component of the marginal distribution
wE(k, ~, v) = _1
271"
. J
WE(X, ~, v) exp (-ikX) dX (37)
in the form
Since the marginal distribution of the stationary state of the harmonic oscillator must
satisfy the stationarity condition [9]
(39)
eq. (38) is equivalent to the equation for axially symmetric wave functions of the two-
mode harmonic oscillator with the mass m = k 2 , the frequency w = 1/2, and the
angular momentum M = O. The wave function corresponding to zero angular momen-
tum is expressed in terms of Laguerre polynomials
(40)
The main quantum number n of the one-dimensional harmonic oscillator under dis-
cussion is equal to the integer radial quantum number nr of the artificial two-mode
oscillator
nr = 0, 1, 2, .... (41)
The energy level of the artificial symmetric two-mode oscillator labeled by the radial
quantum number n" and the angular momentum M as
(42)
231
J
To find the marginal distribution, we have to calculate
1:
In view of the integral
211l' exp ( - ~2 + ikX) Ln (~2) dk = 1l'-1/2 Tn (n!)-l exp (_X2) H~(X), (45)
We take solutions (40) without using the normalization condition in terms of the vari-
ables /L and v of the artificial two-mode oscillator, but using the normalization
condition of the marginal distribution in terms of the variable X and the correspond-
ing property of its Fourier component.
ALGEBRAIC DESCRIPTION
One can construct analogs of creation and annihilation operators acting on the
marginal distribution. To do this, we use the operator Pnm = In)(ml which has matrix
elements (in the coordinate representation)
Pnm(X, X') Wn(X) Wm(X')
1l'-1/2 2-(n/2+ m / 2) (n! m!t1/2
(48)
and
b= J2 a) ;
X + ax'
1 (, (49)
act on the matrix elements (47) as
apnm = VnPn-lm;
(50)
at Pnm = Vn+1 Pn+1m; bt Pnm = y'm + 1 Pnm+1 ,
one can rewrite this action in terms of the arguments of the marginal distribution as
1 {a /L + i v (a)
a =
J2 ax - 2- - ax aILa, a)}
+ av ;
-1 (
Z
1 { a /L - i v (a)
- J2 ax - 2 - + ax aILa - z,a
ov ) }-1 (
(51 )
a*',
232
To obtain the explicit form of the operators under discussion, we used the following
correspondence of the action of the operators on the Wigner function W (q, p) and
the marginal distribution
qW(q,p) ( a ) -1 a
-----+ - ax a(1W(X, (1,v);
0 0
oq W(q, p) -----+ (1 ax w(X, (1, v);
(52)
( a ) -1 a
pW(q,p) -----+ - ax avW(X, (1, v);
a a
ap W(q, p) -----+ v ax W(X, (1, v).
Here the correspondence of the action on the density matrix p (X, X') and on the
Wigner function was also used:
and act on the function wnm(X' f.l, v) given by eq. (34) as follows
A ab
1 {( a )
"2 ax
2~-4~
f.l2 + v 2 (a) -2 [( a ) 2 (a ) 2]
+ ax af.l + au
a a }
- f.l a/l - u au - 1 ;
(56)
At at bt
-1 { ( -a ) 2+ 2 (a)
2 ~+ - -2 [( -a ) 2+ (a
- ) 2] a
+f.l-+ u -a+ 1} .
2 ax 4 ax af.l au af.l au
233
These operators act on the marginal distribution (35) of quantum states as
Thus, the operators realize dynamical symmetry of the quant.um harmonic oscillator
in terms of classical formulation acting on the positive marginal distribution of the
quantum states.
CLASSICAL PROPAGATOR
In [6], the classical transition-probability density from an initial position X' mea-
sured at time t = 0 in the reference frame in the classical phase space labeled by the
parameters fl'; v'to the position X measured at time t in the reference frame in the
classical phase space labeled by the parameters fl; v was introduced. This classical
transition-probability density is the propagator for the evolution equation (29) for the
marginal distribution and the propagator is the kernel of the integral relation
W (X, fl, v, t) = JII (X, fl, v, X', p', v', t) W (X', l, v', 0) dX ' dl dv ' . (58)
Here we put 15 = 0 since the marginal distribution depends only on X - 15. The
classical propagator may be related to a quantum propagator (Green function) for the
density matrix p (X, X', t) in the coordinate representation. For a pure state wit.h
the wave function W (X, t) , we have
The Green function of the Schrodinger equation G (X, X', t) connects the wave func-
tions at the initial time moment t = 0 and at time t
We have for the density matrix (59), in view of relation (60), the expression
p (X, X', t) = J J( (X, X', Y, V', t) P (Y, V', t = 0) elY elY' , (61)
where the propagator J( (X, X', Y, V', t) is expressed in terms of the Green function
(for unitary evolution)
1
J((X,XI,Z,Z',t) =
(2 7r)2
(>9 J X-
1 exp {i (Y - fl _.
-
Vi
+-
2
XI) - i -
Z-- Z'
v'
- y' +i Z2 - ZI2
2 v'
l }
(>9 II (Y, fl, X - X', V', l, Vi, t) dfl dfl' elY dY' dv ' . (63)
234
Thus, given the classical propagator for the classical marginal distribution the propa-
gator for the density matrix is also given. Formula (63) may be converted.
Deriving (63) we used the relations
W (X, p, v) = 1
27rv J p(Z, ZI} exp -i -[
Z_Z' Z+ZI)]
v - (X - P - 2- dZ dZ ' .(66)
The last formulae give some connections of the marginal distributionw (X, p, v) to
the Wigner function and to the density matrix in the coordinate representation.
where the functional S[q] being the classical action gives the possibility to relate the
introduced classical propagator II (X, p, v, X', p', v', t) to the path integral. Expres-
sion for the density matrix propagator (62) may be rewritten through the path integral
JJ
as
J{ (X, X', Y, Y', t) = d[q] d[qj exp (i S[q]- i S[q1) (68)
Relation (63) gives the expression of the path integral in terms of the classical propa-
gator. Thus, the symplectic tomography map provides the connection of the classical
probability densities to the quantum transition amplitudes expressed in terms of func-
tional path integrals.
If one writes the classical propagator as a function of the initial time moment i l
and the final time moment t2 (i.e., i l #- 0), relation (58) may be rewritten as
w (X, p, v, t 2 ) = JII (X, p, v, X', p', v', i 2 , t l ) w (X', p', v', t l ) dX' dp' dv' . (69)
From the physical meaning of the classical propagator, the nonlinear integral relation
follows
II (X, p, v, X', p', v', t 2 , tt) = J II (X, p, V, X", P", v", t 2 , t/)
I8i II (X", P", v", X', J-t' , v', t', tt) dX" dp" dv" . (70)
This relation means, that if the system is initially located in the point X' at time
il in the reference frame in the phase space labeled by the parameters p'; v', the
possibility for the system to arrive at the point X in the reference frame in the phase
space labeled by the parameters p; v at time t2 equals to probabilities to arrive
at an intermidiate point X" in a reference frame in the phase space labeled by the
parameters P"; v" at time t' integrated over all the intermidiate positions and all the
intermidiate reference frames. Also the propagator satisfies the differential equation [6]
all _ p ~ II _ i
at2 av
[v (__1_~ ~ i !!..~)
a/aX ap 2 aX
_V ( __1_~ + i !!..~)]
a/aX OJ-t 2 aX
II
235
The classical propagator satisfies the initial condition
II (X, Il, v, X', Il', v', t, t) = 8 (X - X') 8 (Il - p.') 8 (v - v') . (72)
The relation which is used to express the classical propagator in terms of the func-
tional integral may be also rewritten
fn N-l
k=l
{II (Xk+l' Ilk+b Vk+1, Xk, Ilk, Vkj tk+l, tk) dXk dll k dvd , (73)
There exist time-dependent quantum integrals of the motion [17,18] I (t) satisfying
the equation
dIet) = aI(t) [H I()] =
dt at + Z , to, (74)
or
d
dtTr[p(t)I(t)] =0, (75)
where p (t) is the density operator in the Schrodinger representation. For unitary
evolution of the system. the kernel leX, X', t) of the operator, which is the integral
of the motion, is expressed in terms of the quantum propagator (62) as follows
T(X, X', t) = f I< (X, X', Y, yI, t) Io(Y, yI) dY dY ' , (76)
where the function Io(Y, yI) is the initial value of the kernel
Formula (76) is analogous to formula (61) giving the density matrix in the coordinate
representation. This analogy exists because the density operator itself is the integral of
the motion.
Since the quantum propagator is expressed in terms of the classical propagator, the
kernel of the quantum integral of the motion may be expressed in terms of the classical
propagator.
For time-dependent Hamiltonians, the marginal distribution depends on time as,
for example, for an ion in a Paul trap considered in [14]. For such cases, one can
find the operators (integrodifferentialones) corresponding to quantum integrals of the
motion, the kernels of these operators being satisfied (76). These operators (integrals
of the motion) determine the time-dependent eigendistribution functions due to the
eigenvalue equation
and the eigenvalue n does not depend on time. The time-dependent eigendistribution
functions also. satisfy eqs. (20) and (22), in which one has to put 8 = o. The indices
236
I and r mean that there exists a pair of commuting operators Jz(t) and Jr(t) , the
marginal distribution wn(X, fl, v, t) being the eigenfunction for the both operators.
QUADRATIC SYSTEM
(79)
where one has the vector-operator Q = (p, q). The symmetric 2x2-matrix B and real
2-vector C depend on time. The system has the linear integrals of the motion [17, 18J :
II (X, fl, V, X', fl', v', t) = 8(X - X' +N A-1/:;')8(N' - N A-I), (83)
For the quadratic systems without linear terms (C = 0) , thp classical propagator is
II (X, fl, v, X', fl', v', t) = 8(X - X') 8 (N' -N A-I). (84)
Thus, if one knows the linear integrals of the motion, i.e., the matrix A (t) and the
vector /:;. (t), one knows the classical propagator.
For free motion with the Hamiltonian
(85)
(87)
Thus, we have
N A-l(t) = (v + flt, fl). (88)
Consequently, the classical propagator of free motion has the form
II(f)(X, fl, v, X', fl', v', t) = 8 (X - X') 8 (v' - v - fll) 8 (fl-Il')' (89)
237
For the harmonic oscillator with the Hamiltonian (ll), linear invariants are known
[17, 18] and the matrix A (t) is
rr(OS)(X, It, v, X', f.l', v') = 8(X -X')8 (v'-'-lICOS t+p sin t) 8 (p' -v sin t- p cos t).
(92)
For free motion, we will show how the classical propagator (89) determines the
quantum propagator. In fact, the Green function of free motion (see, for example, [17,
18] ) is
G(X, Y, t) = v'1 , exp [~(X _ y)z] . (93)
21f'zt 2t
It means, that the quantum propagator for the density matrix of a free particle is
f{ (X, X', Y, Y', t) = 2~t exp [2it (X - y)2 - 2\ (X' - Y')2] . (94)
In view of eq. (89) for the classical propagator of the free particle and of eq. (63)
relating the classical propagator to the quantum propagator, we rederive the known
expression (94).
In notations used in (63), the classical propagator (89) looks as follows
n (Y, f.l, X - X', f.l', v', t) = Ii (Y Y') Ii (f.l - f.l') Ii [v - (X - X') - f.lt]. (95)
f{
( , ') 1
X,X, Z, Z, t = (21f')2
J df.ldY
X -X'+f.lt
_1
21l'
J 8(1 _
dY --7 Z - ZI
X-X'+f.lt
)
we do integration over the variable f.l and after some algebra obtain exactly (94). The
relation
f.lot = Z - Z' - X + X'
238
was used to reduce the result of integration to the form of the quantum propagator (94).
We show also, that the classical propagator for free motion (89) satisfies nonlinear
integral relation (70). In fact, if tt < t' < t z , we have
rr(f)(X, p" v, X", p,", v", t z, t') = {j (X - X") {j (p, - P,") {j (v" - v - P, (tz - t') (97)
and
rr(f)(X", tt", v", X', 11-', v', t', t 1) = 8 (X" - X') 8 (tt" - tt') {j (v' - v"(t' - tl))' (98)
Then, in view of (98), we check that
239
Thus, we avoid the paradox of the quantum world which needs for its explanation
measurements by a classical apparatus accepted in frame of standard treatment of
quantum mechanics. But the difficulties of the quantum approach are present, since we
need to understand better the procedure of measurement in a rotated reference frame in
the phase space of the classical system. The problem of wave function collapse [19, 201
reduces to the problem of a reduction of the probability distribution which occurs as
soon as we "pick" a classical value of the classical random observable in the classical
framework of [5, 6].
The approach developed in [5, 6] gives the possibility to transform such unpleasant
problem of standard quantum mechanics as necessity of a classical device and reduction
of wave packets into standard problem of classical measurements of classical random
variables. In fact, it means that the problem of classical measurements is as difficult
as the problem of quantum measurements. An important analogy with methodology
of special relativity arises: It turns out that it is necessary to introduce the consider-
ation of events in the set of moving reference frames in space-time in order to explain
relativistic effects, as well as, it is necessary to introduce the consideration of events in
the set of rotated and scaled reference frames in phase space in order to explain the
nonrelativistic quantum mechanics in terms of only classical concepts of classical fluc-
tuation theory. But these reference frames are the reference frames in phase space (not
in space-time). Possibly, a combination of these two approaches may be generalized to
give a classical description of relativistic quantum mechanics.
CONCLUSION
One can conclude, that the stationary states of quantum systems (for example, of
the harmonic oscillator) may be obtained using classical-like alternative equations to the
Schrodinger equations. The new type of eigenvalue problem for real positive marginal
distribution is formulated. The analogs of orthogonality and completeness relations for
the wave functions are formulated in terms of conditions for the marginal distributions
as well as the transition probabilities among the energy levels. The criterium for deter-
mining the pure states of the quantum system is given in terms of the classical marginal
distribution.
Thus, using the marginal distribution one can formulate the standard quantum
mechanics without the complex wave function and density matrix. But the position
distributions in ensemble of classical reference frames in phase space play important
role.
There are problems to be solved in order to make complete classical formulation of
quantum mechanics. We have discussed only one-dimensional systems with continuous
observables like position and momentum. But there exists purely quantum observable
like spin. Thus, a formulation of the spin density matrix in terms of positive marginal
distribution must be done to take into account all types of quantum states. To refor-
mulate the quantum field theory in terms of the classical distributions, one needs to
extend the approach developed to the case ofinfinite number of degrees of freedom.
It should be pointed out, that in standard formulation of quantum mechanics there
exist different representations such as coordinate representation, momentum representa-
tion, etc. The counterpart of this variety in classical formulation is related to different
tomography schemes like optical tomography [10, 11], symplectic tomography [6, 7],
and photon number tomography [21-23]. The photon number tomography uses the
marginal distribution of des crete variable which corresponds to number representation.
240
As well as different representations in standard formulation of quantum mechanics are
related by some transformations in the Hilbert space of states, the marginal distribu-
tions of different tomography schemes may be transformed each into the other. This
transformation consists of two steps. First, one makes the map of the marginal dis-
tribution (in one of the tomography schemes) onto the Wigner function and then one
makes another map (of the different tomography scheme) of this Wigner function onto
the corresponding marginal distribution.
ACKNOWLEDGMENTS
This work has been partially supported by the Russian Foundation for Basic Re-
search under Grant No. 96-02-17222.
The author would like to acknowledge Prof. B. Gruber and Prof. M. Ramek for the
kind hospitality at the Bregenz Conference "Symmetry in Science."
REFERENCES
241
[14] O. V. Man'ko, "Symplectic tomography of nonclassical states of a trapped ion,"
contribution to the Adriatico Research Conference "Interferometry 2" (ICTP, Tri-
este, March 1-11, 1996); Preprint IC/96/39, ICTP, Trieste (1996); Los Alamos
Report No. quant-ph/9604018; J. Russ. Laser Res., 17,439 (Plenum Press, New
York, 1996); "Tomography of a trapped ion," contribution to the XXI Intern.
Colloquium on Group Theoretical Methods in Physics (Goslar, July 15-20, 1996).
[15] R. 1. de Matos Filho and W. Vogel, Phys. Rev. Lett., 76 (1996), 608.
[16] M. M. Nieto, Phys. Lett. A, 219 (1996), 180.
[17] V. V. Dodonov and V. 1. Man'ko, Invariants and Evolution of Nonstationary Quan-
tum Systems, Proceedings of the Lebedev Physical Institute, 183, Nova Science,
New York, 1989.
[18] I. A. Malkin and V. I. Man'ko, Dynamical Symmetries and Coherent States of
Quantum Systems, Nauka Publishers, Moscow, 1979, (in Russian).
[19] J. A. Wheeler and W. H. Zurek (eds.), Quantum Theory and Measurement, Prince-
ton University Press, Princeton, 1983.
[20] J. S. Bell, Speakable and Unspeakable in Quantum Mechanics, Cambridge Univer-
sity Press, Cambridge, 1987.
[21] S. Wallentowitz and W. Vogel, Phys. Rev. A, 53 (1996), 4528.
[22] K. Banaszek and K. Wodkiewicz, Phys. Rev. Lett., 76 (1996), 4344.
[23] S. Mancini, P. Tombesi, and V. I. Man'ko, "Density matrix from photon number
tomography," Europhys. Lett. (to appear, 1996).
242
SUPERMULTIPLET FORMULATION OF THE
DIRAC OSCILLATOR OF ARBITRARY SPIN
INTRODUCTION
In a recent publication [1] Moshinsky and Del Sol Mesa discussed the equation for
a Dirac oscillator with arbitrary spin, and developed a procedure for determining its
spectra through a recurrence procedure, restricting the analysis to symmetric solutions
in the spin variables.
Later Moshinsky and Smirnov [2] showed that one can develop a supermultiplet
picture for relativistic problems by the introduction, besides the ordinary spin, of a
new concept that they called sign spin. They applied this procedure to a relativistic
particle of arbitrary spin in a constant magnetic field and got the. spectra associated
with several supermultiplets i.e. different representations of the U(4) symmetry group
of the problem.
In this note we will apply the supermultiplet theory mentioned [2] to a Dirac oscil-
lator of arbitrary spin and start by introducing the idea in the simplest case i.e. when
the spin is s = 1/2.
Several years ago Moshinsky and Szczepaniak [3] introduced the concept of Dirac
oscillator through the equation
with r, p the coordinates and momenta, w the frequency of the oscillator and
j (3)
j = [ o1 0]
1 ,tl = 2"1 [01 1O] , t2 = 12"i
[0 -i] ,t3 = 2"1 [1
0 0 (4)
(5)
The matrices Si, i = 1,2,3 are those of the ordinary spin, while ti, i = 1,2,3 which
we distinguish by a square instead of round brackets, have the same definition of Si but
playa very different role and we shall call them sign spin as it was shown [2] that they
are associated with the sign of the energy.
The set of matrices in eq. (3) and (4) are identical to those appearing in supermul-
tiplet theory as introduced by Wigner [4], but in which t; i = 1,2,3 were interpreted as
the components of the isotopic spin.
We would like now to write the Dirac oscillator eq. (1) in a supermultiplet form
using the relation (5). There is though also the presence of a term a(3 in (1) which
requires a careful analysis if later we wish to extend our procedures to a Dirac oscillator
of arbitrary spin. For this purpose we note from (4) that in the present problem the
sign spin can only have the value 1/2 and so its eigenstates could be denoted by the ket
(6)
where
(7)
(11)
We shall now proceed to write the eq. (1) in the supermultiplet form using (5) for
a, (3 though later, when we consider the Dirac oscillator of arbitrary spin, it will also
be convenient to use the form (9) for (3.
244
To achieve our purpose besides the relation (5) for a, (J we need also to consider the
product a, (J which appears as the coefficient of r in (1). From (5) we have
(12)
which follows from the fact that from (4) we have that tIts = -i/2h.
Thus we can write (1) as
(13)
To calculate now the energy spectrum E within the supermultiplet picture (13) it
is convenient to express it in terms of creation and annihilation operators defined by
(14)
From (14) we can obtain r, p in terms of 11, eand substituting in (13) we arrive at
the equation
As t;,i = 1,2,3 are the 2x2 matrices of (4), the 7/.J can be interpreted as a two
component vector and so we have
tl . =
+ Zt2 [00 1]
0 ' (16)
(17)
(18)
(19)
where
(20)
Eq. (18) is identical to (4.3) of ref. 1 and thus by the same reasoning that takes us
from eq. (4.3) to (4.13), in ref. 1 we conclude that the energy in terms of the number
of quanta N and the total angular momentumj, is given by the relations
Having then solved the spin 1/2problem in the supermultiplet formulation we pass
to consider the problem for arbitrary spin.
245
SUPERMULTIPLET FORMULATION OF THE DIRAC OSCILLATOR
OF ARBITRARY SPIN
with I being a 4x4 unit matrix and the ,B, given in (2), is in the position u, while a
similar relation holds for au. On the other hand [1]
To write it in supermultiplet form we use the relations (5) but now for each value
of u, which take the form
(25)
where Siu, tiu are direct products of the form (23) but where now I stands for (2x2)
unit matrices and in the u position we have Si or ti given by (3) or (4).
As in ref. 2 it is convenient to introduce now the operators
n n
where i,j = 1,2,3 and which, together with the unit operator i0i are the 16 generators
[2] of a U(4) group.
From eq. (22), taken in terms of the components of the vectors that appear there,
we see that we need to evaluate
n n
in terms of the generators (26) of U(4) group, if we wish to rewrite eq. (22) III a
supermultiplet language as was done in (12) when n = l.
From (26) we obtain immediately that
n n
but that the last term in (27) requires first the determination of B of (24) in terms of
the generators of U(4).
For the purpose mentioned we note that from (11) we can write ,Bu as
, 1
,Bu = I 0 ( -1) 'i- 13u (29)
and thus from (24) we get
246
so that finally
n
L>l:iuB=4Ril(-1)~-T3 . (31)
u=l
The fact that the operator T3 appears now as an exponent will not cause us any
problem as it will be applied later only to eigenstates of T3
We are now in the position of writing eq. (22) as
(32)
where Xi,Pi,i = 1,2,3 are the components of r,p, and introducing creation and anni-
hilation operators of (14) we can express it as
To find the eigenstates 'lj; and eigenvalues E of eq. (33) we first note that it is
convenient to consider a chain of groups relevant to the problem [2], which turns out
to be
U(4) :> U(2) Q9 U(2), (34)
where the generators of U( 4) are those in (26) while those of U(2), U(2) are respectively
Si,Ti,i = 1,2,3.
The states of the chain (34) will be characterized by a partition {h} = {hlh2h3h4}
where the h's are non-negative integers satisfying the relations
(35)
and also by the eigenvalues of the Casimir operators of the ordinary and sign spin
s(s + 1), t(t + 1). Furthermore the U(2) groups have an 0(2) subgroup whose irreps
could be characterized by the index (J in the case of ordinary spin and r for the sign
spin. Thus the basis of irreducible representations (BIR) of the chain of groups (34)
can be denoted by the ket [2]
l{h}-ys(Jtr) , (36)
where I is an index that distinguishes the representations (s,t) in {h} when they are
repeated.
We note now that eq. (33) contains only the generators (26) of the U(4) group as
well as the creation and annihilation operators, 7]i, ~i. For the latter we could consider
harmonic oscillator states of the form
(37)
247
where RNl(r) is the radial wave function of the.oscillator and YiI"(O,cP) the spherical
harmonic.
A complete set of states, on which we could represent the operator H of eq. (33) as
a numerical matrix equation, will then be of the form
The partition {h} is an integral of motion of the problem [2], and thus is fixed and
the same for the bra'8 and ket's of the form (38), with which we wish to obtain the
matrix elements associated with eq. (33). The effect of the operator T3 (38) is just to
give the eigenvalue T, while
(_I)~-T3 -+ (_1)~-1", (39)
and thus the only type of terms that remains to be calculated are
(40)
As the matrix elements of the second term in (40) can be obtained from those of
the first one by hermitian conjugation, we only need to consider operators of the form
+
L( -IF Rq>Jl-q == R,\ . 11, (41)
q=-
where, as in [2] , we have spherical rather than Cartesian components that take the
values q, A = +,0, - and we use the fact that
(42)
As the term in (41) clearly commutes with the total angular momentum operator
Ji = Li + Si , (43)
where Li = (r x p)i and Hi is given by (26), then its clear that J2, h are also integrals
of motion and we would have the same value j, m in the ket (38) and its corresponding
bra. We obtain then, from well known Racah analysis and the Wigner-Eckart theorem
for the sign spin part [5] , that
(N'l'\\11\\Nl)
_6 {[(N+[+3)(l+I)]1/2 6 [(N-l+2)[]1/2 6 }
- N'N+1 (21 + 3) {'{+1 + (21- 1) {'{-1
(45)
248
while that of R was discussed in ref. 2.
For the full analysis of the matrix elements with respect to the states (38), of the
H on the left hand side of (33), we need to consider also the eiqenvalue of the operator
(46)
with H given in (33), it is convenient to enumerate in the rows and columns first the
e
states where ~ - T is even, in which case, from (46), only the operator Rl . survives,
and then those in which ~ - T is odd, where Rl . 1] is the one left. In this way we can
see, as shown in ref. 1, that only the states (38) with Nand N - 1 are required and
thus we have a finite matrix as
is also finite.
We are now in a position to calculate our matrices for arbitrary n and partition
{h} and get the energy E as function of N, j. As this was already done in ref. 1 for
n = 2, {h} = {2}, n = 3, {h} = {3} using another method, we shall not repeat it here
and just finish by noting that Dirac oscillator of arbitrary spin can be fully discussed
within the supermultiplet formulation.
Clearly the whole procedure also holds if we put w = 0 in (32) and introduce the
interaction through an arbitrary potential V(r). Thus we can, for example, analyze the
problem of hydrogenic relativistic atoms of arbitrary spin, as we plan to do in future
publications.
REFERENCES
[1] M. Moshinsky and A. Del Sol Mesa, J. Phys. A: Math. Gen. 29 (1996), 4217.
[2] M. Moshinsky and Yu. F. Smirnov, J. Phys. A: Math. Gen. 29 (1996), 6027.
[3] M. Moshinsky and A. Szczepaniak, J. Phys. A: Math. Gen. 22 (1989), L817.
[4] E. P. Wigner, Phys. Rev. 51 (1937), 106.
[5]M. Rose, Elementary Theory of Angular Momentum, John Wiley & Sons, New York,
1951, 115-120.
[6] M. Moshinsky and Yu. F. Smirnov, The Harmonic Oscillator in Modern Physics,
Hardwood Academic Publishers, The Netherlands, 1996 35.
249
CONSTRUCTION OF SCHRODINGER- AND
q-SCHRODINGER INVARIANT DIFFERENTIAL
OPERATORS
Christian Mrugalla
INTRODUCTION
Historical Background
In 1972 U. Niederer [3] determined the maximal kinematical invariance group of the
free Schrodinger equation in 3+ 1 space-time dimensions:
S1jJ == .a +
( z-
at
-~
2m
'Ii) 1jJ o (1)
[S,X] = A(t,X)S
with some a priori arbitrary function A and are calculated as (i = 1,2,3):
Pt = at, Pi = ai,
Gi = itai + r: Xi , j = -i xx V
55 = i (2tat + Xi a;+ D (2)
+ tXiai + ~t)
m
K=
~2
-i (t 2 at 21i x .
He proposed to call the corresponding Lie algebra the Schrodinger algebra which we
denote by s(n) (n = 3). By exponentiating the generators (2) one obtains (a unitary
representation of) the Schrodinger group 6(n) which endows for n = 3 the following
(local) IRj XiR;transformations:
252
Definition and Basic Properties
Some of its subalgebras that are also interesting from a physical point of view are
As already mentioned, s( n) is not semisimple; it allows for any n exactly one central
extension realized by the additional relation
(5)
Pt Ot,
Pi Oi,
D 2tot +XiOi -d,
(6)
J ik XkOi - XiOk
Gi tOi + jJXi
K t20t + tXiOi + ~XiXi - td
253
Pt time-translations (t, x) --+ (t + a, x)
Pi space-translations (t, x) --+ (t, x + b)
D dilatations (t, x) --+ (d 2 t, dx)
space-rotations (t,x) --+ (t,Rx) (R E SO(n))
Galilean boosts (t, x) --+ (t, x + vt)
K special conformal transformations (t x)
,
--+ (_t_ ~)
l+d'l+d
Symmetry Properties
The Schrodinger algebra is not only a symmetry of the free Schrodinger equation
but also generates symmetry transformations for several physically important equa-
tions. Rideau and Winternitz [6] have given a classification of all second order evolution
equations that are invariant under '5(1). Some examples are:
for /1 = i m/n and d = -n/2, s(n) is a symmetry algebra for the (n + 1)-
dimensional free Schrodinger equation;
it has been shown [7] that for a subclass of the nonlinear Doebner jGoldin
equations [8] the Schrodinger algebra (in a different realization) also gives sym-
metries;
'5( n) is also the invariance algebra of the Schrodinger equation with harmonic
oscillator and linear potentials [9].
Niederer [10] and Barut {Il] demonstrated that '5(3) can be obtained also as a "non-
relativistic limit" of the conformal algebra (by a construction different from the Wign-
er/Inonii contraction method).
Construction of Verma-modules
254
D acts as a grading operator, i.e., for any generator A of 5(1) we have
[D, A] = deg(A) A.
The sign Ef) is meant to indicate a direct sum of vector spaces (not of Lie algebras).
(8) resembles the Borel-decomposition of a semisimple Lie algebra.
The basic ingredient for the construction of a Verma module of s~l) with fixed d is
the vacuum vector Va, which is defined as an eigenvector of DE 5(1) (with eigenvalue
d) annihilated by s(1r, i.e.,
D Vo = -d Vo, Pt Va = Px Va = O. (9)
(10)
The Verma modules are thus labeled by the same number d as the representations (6).
Note also that a possible realization of Va is the constant function 1. This fact will be
used later.
By virtue of (4) one finds as module-relations:
G Vk,l = Vk+l,l,
]{ Vk,l = Vk,l+1 ,
For an eigenvector V of D, called homogeneous vector, we define its degree as the sum
of the corresponding eigenvalue and d (i.e.,degv(vk,z) = k + 21). It follows from (11)
and (7) that for any generator A E 5(1) the relation
holds.
D VB = (p - d) VB, Pt VB = Px Vs o. (12)
255
is an invariant submodule (here U(:5(I)+) denotes the universal enveloping algebra of
:5(1)+). Note that pEN because of (11).
To determine such vectors Vs explicitly one considers the most general homogeneous
vector of degree p
[~l
vP L
8~O
as V p -2s,s
v;: = aD f
1=0
(-2/l)1 0) Vp -21,1 (14)
We focus our attention on the case jJ 1= 0, which is more interesting from a physical
point of view (see later).
For /l 1= 0 and d = P;3 (p even) one finds that the modules
(15)
where
I'?' == U(:5(I)+) v;:
are irreducible (for other d the Verma modules V d themselves are irreducible). ~ is
generated by
k,l E fiI, l<'f.-I.
- 2
The module relations for =2 are the same as (11) except for:
t (!)
"_1
[{~ Vk,1 = - (_2/l)i-~ Vk+p-2i,i
,=0
Invariant Operators
The singular vectors are determined (in the cases where they exist) by the operator:
(16)
For semisimple (resp. reductive) Lie algebras 11 Dobrev [2] has shown that inserting
a realization of the Lie algebra with at most first order differential operators into a
defining equation for singular vectors like (16) S(p) becomes a 11 invariant operator, in
the sense that the solution variety of the equation S(p)1jJ = 0 is invariant under the
256
corresponding algebra. We will demonstrate that this is also the case for :5( n) with
n = 1 (this statement is also true for any n).
Inserting the realization (6) into (16) we obtain:
(17)
and
which is a 1-1 map. In special cases these basis vectors can be expressed using degenerate
hypergeometric polynomials:
fo,1
(18)
the factor-relation (15) is fulfilled, so that the basis fkl gives indeed a realization of the
module ~.
Moreover we have:
Summarizing it can be stated that the realizations of the 5(1) invariant modules
~ consist of solutions of the equations:
01.
'f/ E 1'=
L 2 , p even (19)
257
q-DEFORMATIONS
Motivated by results (cf., e.g., [15]) from the investigation of quantum groups a
theory of generalized symmetries we are looking for a q-deformation of :5'(1) which can
be represented by differential operators and q-difference operators. For the latter we
use the following notation in COO(R y ):
v f( ) = f(qy) - f(q-ly)
(20)
y Y - Y ( q - q -1) ,
for q --+ 1 one has: V y, V~ --+ ay. With the "number" operator Ny for the coordinate
y, i.e. Nyyk == kyk, we have for suitable (e.g. analytical) functions:
or V'y qCNt+eNx .
As such a general replacement by itself would not give well defined results, we need
additional requirements-which could be called q-assumptions-in order to give some
meaning to this procedure. We choose the following rules:
1. In the limit q --+ 1 we should have the classical algebra with representation (6).
3. The resulting q-deformed algebra should be closed under the formation of com-
mutators and q-commutators.
258
Under these conditions we obtain for .fq(1) the following non-trivial relations (we denote
the deformed generators by calligraphic symbols in order to distinguish them form the
un deformed ones):
We have the following "simple" realization (for a more general one d. [14]):
Pt V t qNt+Nx, Px = V~ q~Nx,
D 2Nt + Nx d,
t V~ q-~Nx + q-~ flx q-Nx , (22)
JC qd t 2 V t q-Nt + qd tx Vx q-2Nt -Nx
Note that .fq(1) is different from the q-deformation of:5(l) given by Floreanini and Vinet
[16] as a symmetry algebra of a q-deformed heat equation defined in an ad hoc fashion.
The main difference is that this algebra closes only on the space of solutions of the
latter equation.
q- Verma-modules
As the q-algebra .fq(l) (21) also admits the gmding of:5(l) given in (7), we can define
Verma modules ~d in the same way as in the undeformed case. The module structure
is (with v k,l =- gk JCl V 0 E Vqd )
As in the undeformed case we have a singular vector (for the usage of singular vectors
p-3
in the context of (standard) quantum groups d. [18]) Vs E Vqd iff d = -2- with p
even. It is given by (7r == ~)
(24)
As in the undeformed case we have for J1 =f. no singular vector with odd degree p and
E=.l
again we can obtain irreducible lowest weight modules Lq 2 by factorizing the invariant
submodules given-as in formula (15)--by the relation (24).
259
Invariant q-difference operators
(25)
CONCLUSION
The method of constructing invariant partial differential operators via singular vec-
tors of Verma modules originally introduced for semisimple Lie algebras by Dobrev [2]
has proven to be fruitful also for the non-semisimple Schrodinger algebra, as not the
complete structure of semisimple Lie algebras is needed for it, but only the triangular
decomposition, which is given for 5(1) in (8). It leads us to a family of operators in-
variant under realizations of this 5(1) with d = ~, p even. The family contains the
free Schrodinger equation as one member (p = 2).
This method could also be applied to a q-deformation .5;(1) determined with the
help of certain q-assumptions from 5( 1) .. The result is a family of q-difference equa-
tions containing a q-analogue of the free Schrodinger equation different from other
proposals in the literature. This again demonstrates the non-uniqueness of methods to
construct q-deformed partial differential equations via q-difference operators. To check
the relevance of such q-deformations one has to derive consequences from the various
q-deformed Schrodinger equations, which could be tested experimentally through preci-
sion experiments in order to decide which q-deformation-if any of them-is appropriate
for the description of real physical systems.
260
ACKNOWLEDGMENTS
The results of this article have been obtained in joint work with H.-D. Doeb-
ner (Clausthal) and V. K. Dobrev (Sofia). The work was supported by Deutsche
Forschungsgemeinschaft under contract Do 155/17-1.
REFERENCES
[1] B. Kostant, in: Lect. Notes in Mathematics 466, Springer, Berlin, 1975, 101.
[2] V. K. Dobrev, Rep. Math. Phys. 25 (1988), 159.
[3) U. Niederer, Helv. Phys. Acta 45 (1972), 802.
[4] H.-D. Doebner and H.-J. Mann, Journ. Math. Phys. 36 (1995), 3210.
[5] W. I. Fushchich, W. M. Shtelen, and N. I. Serov, Symmetry Analysis and Exact
Solutions of Equations of Nonlinear Mathematical Physics, Kluver, Dordrecht, 1993.
(6] G. Rideau and P. Winternitz, Journ. Math. Phys. 34 (1993), 558.
[7] P. Nattermann, Rep. Math. Phys. 36 (1995), 387.
[8] H.-D. Doebner and G. A. Goldin, Phys. Lett. A 162 (1992), 397.
[9] U. Niederer, Helv. Phys. Acta 47 (1974), 167.
[10] U. Niederer, Helv. Phys. Acta 47 (1974), 119.
[11] A. O. Barut and B. W. Xu, Phys. Lett. 82 A (1981), 218.
[12] J. Dixmier, Enveloping Algebras, North Holland, New York, 1977.
[13] V. K. Dobrev, H.-D. Doebner, and C. Mrugalla, Lowest weight representations
of the Schrodinger algebra and generalized heat/Schrodinger equations, preprint
ASI/TPA 15/95, TU Clausthal, July 1995, to appear in Rep. Math. Phys.
(14) V. K. Dobrev, H.-D. Doebner, and C. Mrugalla, Journ. Phys. A 29 (1996),5909.
[15] V. Chari and A. Pressley, A Guide to Quantum Groups, Cambridge University
Press, Cambridge, 1994.
[16) R. Floreanini and L. Vinet, Lett. Math. Phys. 32 (1994), 37.
[17) J. A. Minahan, Mod. Phys. Lett. A5 (1990), 2625; T. T. Truong, Jorn. Phys. A27
(1994), 3829; A. Hebecker and W. Weich, Lett. Math. Phys. 26 (1992), 245.
[18) V. K. Dobrev, Journ. Phys. A25 (1992), 149.
261
AUTOMORPHISMS OF FREE GROUPS AND
UNIVERSAL COXETER GROUPS 1
Bernhard Miihlherr
Fachbereich Mathematik
Universitiit Dortmund
Vogelpothsweg 87
D~44221 Dortmund
Germany
INTRODUCTION
(ii) we is isomorphic to Fn - I ! and the set {Xi I 1 :::; i :::; n - I} generates we! where
Xi = SiSi+l
(iii) If n =I- 2 and if a E Aut(W) is such that a Iwe= idwe! then a = idw .
Theorem B: Aut(W) is generated by the set {OO1f I 7r E 5 n } U {O"ij 11 :::; i =I- j :::; n}.
Acknowledgements
My research on automorphisms of Coxeter groups was motivated by disussions with P.
Kramer. The idea to consider Aut(Wn+!) as a subgroup of Aut(Fn) is due to him.
PRELIMINARIES
Graphs
Given a set X, then P2(X) denotes the set of all subsets of X having cardinality 2.
A graph is a pair (V, E) consisting of a set V and a set E <;;;: P2(V). The elements
of V and E are called vertices and edges respectively.
Let r = (V, E) be a graph. If v E V, then Ev denotes the set of edges which contain
v; r has order n if lEvi = n for all v E V. Let v,w be two vertices of r. They are
called adjacent if {v, w} E E. A path from v to w is a sequence v = va, VI, ... , Vk = w,
where Vi-I is adjacent to Vi for all 1 :::; i :::; k; the number k is the length of the path.
The distance between v and w (denoted by 8 (V, w)) is the length of a shortest path
joining them; if there is no path joining v and w, we put 8(v,w) = DO. The distance
between two edges e, e' E E is defined as the minimal distance between two vertices
v, Vi satisfiying vEe, Vi Ee'; we will denote it by 8(e, e' ).
A circuit is a path va, ... ,Vk satisfying Vi-1 =I- Vi+! for all 1 :::; i :::; k and va = Vk.
The relation ReV x V defined by R = {(v, w) I 8( v, w) =I- DO} is an equivalence
relation whose equivalence classes are called the connected components of r. A graph
is said to be connected if it has only one connected component.
A tree is a connected graph having no circuits. Let r be a tree and let e = {v, w} be
an edge of r. The graph obtained by removing the edge e has two connected components
which we will denote by H( e, x) and H( e, y) respectively. Given any set F of edges in
a tree, then we say that f = {x, y} E F is on the boundary of F if F \ if} is contained
264
in H(f, x) or H(f, y). The set of edges being on the boundary of F will be denoted by
of and we put FO = f \ of.
Proposition 1: Let r be a tree and let G ::; Aut(r) be such that each element of
G \ {IG} fixes neither a vertex nor an edge. Then G is a free group.
1. Given an edge e E E, then there exists a unique involution O"(e) E W fixing e and
switching the two vertices contained in e.
2. Given an involution SEW, then there exists a unique edge E(S) which is fixed by
s.
Lemma 2: Let f = {x,y} be an edge ofr and let t = O"(f). Suppose that e,e' are
edges ofr such that e EH(f, x) and e' E H(f,y). Then 8(e,t(e' )) < 8(e,e /).
Proof: Let vEe, v' E e' be such that 8( v, v') = 8( e, e' ) and let v = vo, VI, Vk =
v' be the unique shortest path from v to v'. As v E H(f,x) and v' E H(f,y) it
follows that Vi = x and Vi+! = Y for some 0 ::; i ::; k - 1. Now v = Vo, VI, . ,Vi =
t( Vi+!) , t( Vi+2), . ,t( Vk) = t( V') is a path from v to t( v') of lenth k - I and as v E
e, t( v') E t( e') the claim follows. 0
265
PROOF OF THEOREM A
Lemma 5: we is the free group on n - 1 generators, and the set {Xi I 1 :::; i :::; n - I}
generates we where Xi = SiSi+! for 1 :::; i :::; n - 1.
Proof: Let 1:::; i < j:::; n, then SiSj = XiXi+! ... Xj-1 and SjSi = (SiSj)-1. Now, if
wE we, then W = SijSi2 ... Sik for some even k and hence wE (Xi 11 :::; i :::; n -1).
As \O( t) = -1 for each involution t E W (by Assertion 4. in Section "Preliminaries" it
follows that there is no involution in we. Therefore the sabilizer of an edge in we is
trivial. As W acts regularly on the vertices of r we can apply Proposition 1 to see that
we is a free group and by the previous lemma and the considerations above it follows
that we is the free group on n - 1 generators. 0
Proof: Let", be an automorphism centralizing we. For each 1 ::; i ::; n put Wi .-
S;1",(Si). As ",(SiSj) = SiSj it follows that SjWiSjWj = 1 and hence SjWiSj = wj1.
Applying this identity twice we obtain SiSjWiSjSi = Wi. Hence Wi commutes with SiSj
for each j E I. Given i E I then it is easily verified that we = (siSjlj E 1). Thus Wi is
in the center of we. But we is a free group of rank n - 1 and has therefore a trivial
center if n #- 2. Hence Wi = 1 for all i E I and we are done. 0
Conclusion: Parts (i),(ii) and (iii) follow from Lemmas 5, 6 and 7 respectively.
PROOF OF THEOREM B
Bases
A subset F of E will be called a base if it has cardinality n and if cr( F) := {crU) If E F}
generates W. If F = Ev for some v E V, then F will be called a root base.
266
LeIllIlla 8: Let F be a base, then the following are equivalent:
(aJ f)F=F.
(b J F is a 'rOot base.
Let F be a base. By the previous lemma there exists for each i E I a unique element
fi such that T(Ji) = i. We define the distance matrix of F by D(F) = (8(J;, /i))i,jEI'
Let i E I and let J S;; 1\ {i} then F( i, J) denotes the set {f~ IkE I} where f~ = fk
if k rf. J and f~ = (a(f;))(Jk) if k E J.
LeIllIlla 10: Let F be a base and let i E I,JS;; 1\ {i}. Then F(i,J) is a base.
AutoIllorphisIllS of W
By Assertion 4. of Section "Preliminaries" two involutions t, t' E Ware conjugate if and
only if T(c(t)) = T(c(t')). Given an automorphism of W then it induces a permutation
on the conjugacy classes of involutions in W. Thus we have a cononical homomorphism
from Aut(W) into Sn, whose kernel will be called the group of special automorphisms
and which will be denoted by Spe(W). The group of inner automorphisms of W will
be denoted by Inn(W). Given i E I and K ~ 1\ {i} then we put aiK := IIkEKaiK.
The following observations are immediate:
3. Let F be a base and suppose that F = {c( a( a(J))) I f E F} for some a E Spe(W).
Then a is the identity.
267
4. If F is a root base, then there exists an inner automorphism a of W such that
(a(S)) = F.
5. The automorphism UiK where I< = 1\ {i} is the conjugation with Si; in particular
Inn(W) :::; (Uij 11 :::; i #- j :::; n).
Lemma 12: Let a E Spe(W) and let F = t(a(S)), then the following are equivalent:
(cJ a E Inn(W).
Lemma 13: Let a E Spe(W), let F = t(a(S), let i E I and lei I< <;;: 1\ {i}. Then
F(i,I<) = t(aouiK(S)),
Proof: Let F' = F(i, I<), then fi = Ii jf j E 1\ I< and fj= (U(fi)(fj) if j E K.
For j E I let ii' = t(a 0 UiK(Sj). If j ~ I< then 0'0 UiK(Sj) = a(uiK(Sj) = a(sj) and
hence ii' = t(a(sj)) = Ii = ij. If j E I<, then (a 0 UiK)(Sj) = a(uik(Sj) = a(siSjSi) =
a(si)a(Sj)a(si) = lJ"(fi)IJ"(Ii)IJ"(fi) and hence fj' = t(IJ"(j;)cr(li)cr(fi)) = (cr(fi(Ii) = ii
(the second equality follows from Assertion 3. in Section "Preliminaries" Thus F" = F'
and we are done. 0
REFERENCES
268
GENERALIZED QUANTUM MECHANICS AND
NONLINEAR GAUGE TRANSFORMATIONS
Peter Nattermann1
INTRODUCTION
(1)
and the one derived by Doebner and Goldin [2-5) from the representation theory of the
kinematical algebm for a single particle on JR3,
1 aspn@pt.tu-clausthal.de
2 D' being; a constant with the dimensions of D, to make Cj dimensionless.
(4)
But evidently under a nonlinear evolution of the pure states 1/;j and 1/;; the two expansion
evolve differently,
(5)
270
topology of ::r will provide a notion as to which states are physically "similar", and
thus it will have to be determined by the physical interpretation of the pure states. We
should stress here, and this will be the case in the examples below, that the points of
::r are not necessarily in one-to-one correspondence to the pure states.
The second ingredient is the set of time evolutions. In any experiment the ex-
perimentalist influences the system by changing the external conditions. For a non-
relativistic (semi-classical) theory we might think of external classical fields being ap-
plied, changing the evolution of the system. We describe this dependence by elements
Ext of the set C of all external conditions that are possibly applicable. Thus in general
the time evolutions are given by homeomorphisms of ::r,
depending on the initial time t, the final time t', and the external conditions Ext E C.
Now the experimentalist will be free to change the external conditions at different
times, and the combination of these evolutions will have to be considered as a possible
evolution of the system as well. Thus we call the smallest group containing all such
homeomorphisms, closed in the topology of pointwise convergence 3 , the motion group
M of the system.
Finally, we need a notion of observables. As shown in the introduction the set
of observables will have to be invariant under time evolutions, i.e. under the motion
group M, and we may generated it closing a set of primitive observables!P under
the action of the motion group. As mentioned by various authors (see e.g. Feynman
and Hibbs [20] or Mielnik [14]) the positional measurement are distinguished in a non-
relativistic quantum theory, and all other observables can be obtained from positional
measurements. Adopting these observations we consider as primitive observables the
set !P of positional observables
PB : ::r ---; [0,1]' (7)
labeled by Borel-sets B E B(M) and interpreted as the probability of finding the system
in the region Be M of physical space M, i.e. for every,p E ::r, P.(,p) : B(M) ---; [0,1] is
a probability measure, i.e.
(8)
Using these notions we define a quantum system to be a triple (::r, M,:P) of
a manifold of pure states ::r, a motion group M, and a set of positional (primitive)
observables !P.
For any such quantum system (::r, M, !P) we are able to define observables and mixed
states as derived concepts.
Continuing the above argument any physical observable can be obtained as a com-
bination of time evolutions and a subsequent positional measurement, i.e. the set of
observables is the completion 4 in the topology of pointwise convergence (p.c.) of
the set of all combinations of time evolutions (under various external conditions) and
positional (primitive) measurements 5 ,
271
The set of mixed states of the quantum system (:T, M, :P) depends on 0 in an obvious
manner: Consider first statistical mixtures of pure states, i.e. probability measures 7r
on the pure state manifold:T. For any observable f E 0 the measure 7r defines an
expectation value
(10)
The first requirement on 7r has to be that these expectation values are finite for all
observables, i.e. we consider the set of finite probability measures
II(::r):= {7r prob. measure on :T17r(J) < 00 Vf E O} . (ll)
Furthermore, we have to identify those probability measures that are indistinguishable
by any kind of measurement, i.e. lead to the same expectation values for all observables.
This establishes an equivalence relation among the finite probability measures,
(12)
and the set of mi;Eed states or statistical figure of the quantum system (:T, M,:P) is
the set of equivalence classes
S := II(::r)/ '" . (13)
The set S inherits a natural convex structure from II(::r), and the pure states are in
one~to~one correspondence to the extremal points [; of S.
Summing up, the concept provides a consistent description of observables and mixed
states for a quantum system with any sort (linear or nonlinear) of time evolutions.
Before we turn to two examples, we note, that in classical statistical mechanics
the equivalence relation would be 'trivial; the statistical figure there - generated by
probability measures on phase space - is a (generalized) simplex, i.e. every state has
a unique decomposition in terms of pure states (single point measure). Hence, a non-
trivial equivalence relation reveals one characteristics of quantum mechanics.
(16)
where the scalar potential6 V represents a (sufficiently large) class of external conditions
C. If we allow for arbitrary smooth potentials V the motion group is the unitary group
on :J-C [14, 21]
M = U(:J-C). (17)
6 One may add a magnetic vector potential as well, but it turns out that the scalar potentials are
already large enough to give the full unitary groupas a motion group of the system.
272
As a consequence the set of observables is given by the set of orthogonal projectors on
9-C,
o ~ Proj (9-C) , (18)
i.e. for all obscrvables 0 E 0 there is a projector E E Proj (9-C) such that
(19)
Thus generalized quantum mechanics naturally yields the lattice of closed subspace of
9i, the quantum logic of propositions [22]. Then due to Gleason's theorem, the mixed
states are density matrices and the pure states one-dimensional projectors (rays)
~ P(9-C) , (20)
.
zftot1jJt = (ft2 )
- - , 6 . + V 1/;t + A~ . -1/;t
1; . (21 )
2m Pt
Both, the scalar potential V and the vector field X represent the external conditions C
of the system. Thus the motion group still contains as a sub-group the unitary group
on 9-C (for X:= 0), but is effectively larger than in the purely linear case (17),
M::) U(9-C). (22)
Consequently, the set of observables is also larger than (18), and contains non-quadratic
forms as well,
o ::) Proj (9-C) . (23)
Haag and Bannier showed that the set of observables for this quantum system is large
enough to "separate" states, i.e. the equivalence relation of probability measures is
trivial (on rays), and we get - as opposed to (20) - a generalized simplex
s= II(), = P(9-C). (24)
Thus the quantum system (9-(, M, 3J) has the classical feature of a uniq ue decomposition
of mixtures into pure components!
GAUGE EQUIVALENCE
273
L 3>(1) and 3>(2) are positional observables over the same space M;
Conditions 1 and 2 guarantee that the physical interpretation of the observables is the
same, whereas condition 3 establishes an isomorphism N of the two quantum systems.
We call N a generalized gauge transformation as it is a generalization of U B in
orthodox quantum mechanics.
As we will see in this section the nonlinear gauge transformations introduced pre-
viously [15-17] are a special case of generalized gauge transformations. We can thus
adopt the arguments of [17] to obtain nonlinear quantum systems gauge equivalent
to the linear quantum system, a gauge generalization of which yields a larger class of
substantially nonlinear quantum systems.
We shall start with linear quantum mechanics, i.e. a quantum system (1-C, M, 3
with 1-C, 3>, and M given as in the example 1 (eqs. (14)-(17)). As explained in this
example, the set of pure states of the quantum system (1-C, M, 3 turns out to be the
projective Hilbert-space P(1-C), i.e. wave functions 1jJ and c1jJ, c E t, in 1-C are identified.
The generalized gauge transformation
N:1-C->T, (26)
that we are looking for, where T is the manifold of pure states of the nonlinear theory,
will thus have to respect this structure in order to be consistent, so we assume
IclHh (D"2 ,
A
c'(c) = (29)
where 6, I E ~ and A E Z. To be more specific, we will have to fix the manifold of pure
states T of the nonlinear theory we are aiming for. We take the simplest choice and
assume that the nonlinear theory is defined on the Hilbert space,
Furthermore, we will restrict our attention to strictly local gauge transformations, i.e.
274
for a continuous function n: <C x ~3 ----f <C. Identifying n(l, x) = K:( x) exp (iO( x)) it
follows immediately from (27) and (29) that N depends on the parameters 15", A and
the two functions K: : ~3 ----f ~, 0 : ~3 ----f ~,
(31)
11.= 1, (32)
and K: has to be positive and bounded from both sides. N is even norm-preserving iff
K == 1, and the strictly local norm-preserving generalized gauge transformations
(33)
on L2(~3, d3x) coincide with the nonlinear gauge transformations introduced 7 in [15,
16, 17]. In general, the parameter, and the function () can be time dependent and the
transformations (33) are indeed continuous [24], i.e. and thus N b ,II.;9) is a homeomor-
phism of:T = L2(~3, d3x).
The set of these generalized gauge transformations forms a group 9 under composi-
tions; it is actually a semi-direct product of the groups of pUl'e (() == 0) nonlinear gauge
transformations N and the local unitary transformations U/ OC1
9 = N 098 U/ oc , (34)
(35)
In view of this group structure and the fact that A = -1 corresponds to complex
conjugation, it suffices for our purpose to use purely nonlinear gauge transformations
(36)
Starting with linear Schrodinger equations (16) we obtain for the transformed wave
function 1/;; = N,[1/;) evolution equations of Doebner-Goldin and Bialynicki-Birula-
Mycielski type:
7 Actually, arbitrary, non-vanishing A have been considered there. Then N will only be defined
on some subset of L2(1IJ(3, d3x).
275
Gauge generalization [17) now involves breaking the constraints among the coeffi-
cients of the various terms in (37) a generalization of the parameters of the equation
(37): the coefficient of iR2[7jJ) + RI[7jJ)- R4[7jJ) depends linearly on it, whereas that of
Rd7jJ)- 2Rs[7jJ] is quadratic in it, and that of In 17jJ1 2 is proportional to the derivative of
it. Thus as in a first step we could choose these three coefficients independently. This
leads to a family of equations
that is invariant under gauge transformations NT In (39) we have used the expansion
(40)
NONLINEAR OBSERVABLES
276
The maps are idempotent, Eo E = E and to every E there is a unique -oE E Ont such
that -oE 0 E = Eo -oE = O. Thus E is a generalized projection onto a nonlinear
submanifold of L2 (IR3 , d3x). Furthermore, Ont inherits the logical structure of Proj (X),
it is a nonlinear realization of a quantum logic, and the motion group is contained in
its automorphism group.
The expectation values of E w.r.t. 1/J and E w.r.t. N[1/J] are the same,
(45)
that determines the distribution of the values of the observable A on the real line lR..
With the above definition of generalized projections the nonlinear analogue to these
projection valued measures is a generalized projection-valued measure [18],
(46)
(i)
/I B f-+ /I (B)'= IIEA(B)[1/J]112 (47)
rV;' rV;' 11~)1I2
(ii)
(48)
(iii)
(49)
277
FINAL REMARKS
The purpose of this contribution has been to merge Mielnik's concept of general-
ized quantum mechanics with the concept of nonlinear gauge transformations. Hav-
ing introduced generalized gauge transformations for the quantum systems of gener-
alized quantum mechanics, we were able to identify nonlinear gauge transformation
recently introduced in the context of Doebner-Goldin equations as strictly local, norm-
preserving generalized gauge transformations on L2(JR3 , d3x). On the one hand these
nonlinear gauge transformations were subsequently employed to obtain a unified class
of Bialynicki-Birula-Mycielski and Doebner-Goldin evolution equations.
On the other hand generalized norm-preserving gauge transformations were used
to construct a nonlinear (linearizable) realization of a quantum logic on L2(JR3 , d3x),
and led to the definition of generalized projection-valued measures and nonlinear ob-
servables. These may serve as candidates for observables in truly nonlinear quantum
theories. Following the arguments of section the set of all these nonlinear observables
will have to be invariant under all time evolutions admissible. However, for the formal
evolution equations obtained in section via gauge generalization these time evolutions
are still an open problem. So far, the Cauchy problem for the Doebner-Goldin sub-
family of (41) for certain time-independent coefficients can be solved only on a certain
dense subset of non-vanishing wave functions in L2(JR 3 , d3x) [26]. But then there cannot
be any ideal position measurements in this theory as such a measurement would result
in a Cauchy initial wave function outside the domain of definition for a further time
evolution.
Another motivation for the present article was the prediction of supraluminal com-
munications in nonlinear quantum theory. The review of Mielnik's generalized quantum
mechanics in section showed that the formal obstacle that led to this conclusion is due
to an improper definition of observables and mixtures. Furthermore, the equivalent
nonlinear description of linear quantum mechanics of section shows that there 'are
(seemingly) nonlinear time evolutions with no supraluminal communications at all!
For truly nonlinear time evolutions, however, the question whether supraluminal
communications are possible remains unsolved. The crucial point in order to answer
this question will be the description of two (or many) particle systems and, in particular,
of correlations in such a nonlinear theory8.
The formalism of gauge generalizations employed in section can - at least formally
- be extended to more general cases. For instance, we could follow [14] and consider
as a manifold of pure states of the nonlinear system the set of k-integrable complex
functions (k > 0),
(50)
where 11,6( X')Ik /n(k) [1,6] is taken as a position-probability density. The resulting strictly
local gauge transformations and the formal nonlinear Schrodinger equations will be
treated elsewhere [27, 28].
8 A hint in this direction has been made in [23] where the authors refrain from calling their theory
classical because of problems of interpretation in a many particle theory.
278
ACKNOWLEDGMENTS
REFERENCES
[1)1. Bialynicki-Birula and J. Mycielski, Ann. Phys. (NY) 100 (1976), 62.
[2] H.-D. Doebner and G. A. Goldin, Phys. Lett. A 162 (1992), 397.
[3] H.-D. Doebner and G. A. Goldin, in: Annales de Fisica, Monografias, Vol. II, Pro-
ceedings of the 17th International Conference on Group Theoretical Methods in
Physics, Salamanca, 1992, CIEMAT, Madrid, 1993, 442-445.
[4] H.-D. Doebner and G.A. Goldin, in: Proceedings of the First German-Polish Sym-
posium on Particles and Fields, Rydzyna Castle, 1992, World Scientific, Singapore,
1993, 115 ff.
[5] H.-D. Doebner and G. A. Goldin, J. Phys. A: Math. Gen. 27 (1994), 1771.
[6] G. A. Goldin, R. Menikoff, and D. H. Sharp, in: Measure Theory and Its Applications:
Proceedings ofthe 1980 Conference, G. A. Goldin and R. F. Wheeler (eds.), DeKalb,
Illinois: Dept. of Mathematical Sciences, Northern Illinois University, 1981, 207-218.
[7] B. Angermann, H.-D. Doebner, and J. Tolar, in: Non-Linear Partial Differential Op-
erators and Quantization Procedures, volume 1037 of Lecture Notes in Mathematics,
S. Anderson and H.-D. Doebner (eds.), Springer, Berlin, 1983,171-208.
[8) H.-D. Doebner and J.-D. Hennig, these proceedings.
[9) S. Weinberg, Phys. Rev. Lett. 62 (1989), 485.
[10) S. Weinberg, Ann. Phys. (NY) 194 (1989), 336.
[11] N. Gisin, Phys. Lett. A 143 (1990), 1.
[12) J. Polchinski, Phys. Rev. Lett. 66 (1991), 397.
[13) S. Weinberg, Dreams of a final theory, Pantheon Books, New York, 1992.
[14) B. Mielnik, Comm. Math. Phys. 37 (1974), 221.
[15] H.-D. Doebner, G. A. Goldin, and P. Nattermann, in: Quantization, Coherent
States, and Complex Structures, J.-P. Antoine, S. Ali, W. Lisiecki, I. Mladenov,
and A. Odzijowicz (eds.), Plenum Publishing Corporation, New York, 1996, 27-31.
[16] H.-D. Doebner and G. A. Goldin, Phys. Rev. A 54 (1996), 3764.
[17] H.-D. Doebner, G. A. Goldin, and P. Nattermann, "Gauge Transformations in
Nonlinear Quantum Mechanics and the Unification of Nonlinear Schrodinger Equa-
tions". Clausthal-preprint ASI-TPA/21/96, to be submitted, 1996.
[18] W. Liicke, in: Nonlinear, Deformed and Irreversible Quantum Systems, H.-D. Doeb-
ner, V. Dobrev, and P. Nattermann (eds.), World Scientific, Singapore, 1995, 140-
154.
279
[19] B. Mielnik, Rep. Math. Phys. 12 (1977), 33l.
[20] R. Feynman and A. Hibbs, Quantum Mechanics and Path Integrals, McGraw-Hill
Book Company, 1965, 9l.
[21] J. Waniewski, Rep. Math. Phys. 11 (1977), 33l.
[22] W. Lucke, Acta Phys. Pol. 27 (1996), 2357.
[23] R. Haag and U. Bannier, Comm. Math. Phys. 60 (1978), l.
[24] W. Lucke, private communication.
[25] P. N attermann, in: Proceedings of the XXVI Symposium on Mathematical Physics,
Torun, December 7-10, 1993, Nicolas Copernicus University Press, Torun, 1994,
47-54.
[26J H. Teismann, in: Group21 - Physical Applications and Mathematical Aspects of
Geometry, Groups and Algebras, H.-D. Doebner, P. Nattermann, W. Scherer, and
C. Schulte (eds.), World Scientific, Singapore, 1997, to appear.
[27J P. N attermann, in: Group21 - Physical Applications and Mathematical Aspects
of Geometry, Groups and Algebras, H.-D. Doebner, P. Nattermann, W. Scherer,
and C. Schulte (eds.), World Scientific, Singapore, 1997, to appear.
[28] H.-D. Doebner, G. A. Goldin, and P. Nattermann, "Projective Structure and Non-
linear Gauge Transformations" , work in progress.
280
A SOLUTION OF AN EXTENDED HARTREE-
BOGOLIUBOV EQUATION ON THE COSET SPACE
SO(2N+2)/U(N+l) FOR UNIFIED DESCRIPTION OF
BOSE AND FERMI TYPE COLLECTIVE EXCITATIONS
Seiya Nishiyama
INTRODUCTION
The time dependent Hartree-Bogoliubov (TDHB) theory is the first standard ap-
proximation in the many-body theoretical description of a superconducting fermion
system [1]. The HB wave function for the fermion system represents Bose conden-
sate states of fermion pairs. It is a good approximation for the ground state of the
fermion system with a short-range pairing interaction that produces a spontaneous
Bose condensation of the fermion pairs. The fermion number-nonconservation of the
HB wave function is a consequence of the spontaneous Bose condensation of fermion
pairs, which causes coherences in the phases of superconducting (Bose condensed)
fermions. The SO(2N) Lie algebra of the fermion pair operators contains the U(N)
Lie algebra as a subalgebra. Here SO(2N) and U(N) denote the special orthogonal
group of 2N dimensions and the unitary group of N dimensions, respectively (N is
the number of single-particle. states of the fermions). The canonical transformation
of the fermion operators generated by the Lie operators in the SO(2N) Lie algebra
induces the well-known generalized Bogoliubov transformation for the fermions. The
TDHB equation is derived from the classical Euler-Lagrange equation of motion for
the SO(2N)/U(N) coset variables [2]. The TDHB theory is, however, applicable on-
ly to even fermion systems. For odd fermion systems we have no TD self-consistent
field (SCF) theory with the same level of the mean field approximation as the TDHB
theory.
282
BRIEF REVIEW OF THE TIME DEPENDENT EXTENDED HARTREE-
BOGOLIUBOV EQUATION ON THE COSET SPACE SO(2N+2)/ U(N+l)
We consider a fermion system with N single-particle states. Let c'" and c~, a =
1, ... , N, be the annihilation and creation operators of the fermion, respectively. The
SO(2N + 1) canonical transformation of U(G) is generated by the fermion SO(2N + 1)
Lie operators consisting of the set {c, c~, E~ = C~Cil -1/2 b",il, E"'il = C"'Cil, E',il = c~cb;
a, fJ = 1, ... , N}. The U(G) induces the inhomogeneous linear transformation having
the q-number gauge and is specified by an SO(2N + 1) matrix G as
G =- [~
y/V2 -y*/V2
~: -:;!f],
z
GtG = GGt = 1 , (2)
The SO(2N +1) group is embedded into the SO(2N +2) group. The embedding leads
us to an unified formulation of the SO(2N + 1) regular representation in which paired
and unpaired modes are treated in an equal manner [5]. We define the (N + 1) X (N + 1)
matrices A and B as
A = [a-x*Y/2(I+Z) -X*/2]
y/2 (I+z)/2'
B [ b+xY/2(I+Z) X/2]
-y/2 (I-z)/2'
y (9)
283
Imposing the orthonormalization of the matrix G, the matrices A and B are shown to
satisfy the orthonormalization condition for the (N + 1)-dimensional HB amplitudes
and to form an SO(2N + 2) matrix g. The representations for the matrix 9 and for
the SO(2N + 2)jU(N + 1) coset variables are given as
from which the SO(2N + 1) variables qa.(3 and ra. are shown to be just the independent
variables of the SO(2N + 2)jU(N + 1) coset space. The paired mode and the unpaired
mode variables qa(3 and ren respectively, in the SO(2N + 1) algebra are unified as the
paired variables in the SO(2N + 2) algebra. We denote the (N + I)-dimension of the
matrix Q by the index 0 and use the indices p, q, ... running over 0 and the single-
particle states a, f3, .... Then, Q01(3 = qa(3 and Q",0 = r",
The expectation values of the fermion SO(2N + 1) Lie operators by the I G> are
calculated easily as
(12)
An expectation value of a two-body operator is given as
(14)
where the single particle Hamiltonian h,,(3 includes a chemical potential and
(17)
284
where the operators E$) E"/3, and E"/3 are images of the fermion SO(2N) Lie operators.
The SO(2N + 1) TDHB equation can be derived from the Euler-Lagrange equation
of motion for the SO(2N + 2)jU(N + 1) coset variables. However, we must start from
the following equations of motion
+ Ha;3ho](Rexa.Ry6 - V<:,K6/3) ,
h"/3R"/3
k"/3<c/3>G + 1,,/3< ch >G . (19)
The classical equation of motion for the SO(2N + 2)jU(N + 1) coset variables is
calculated to
z
-1i[FQ + QF* + D + QD*Q]pq
where the SCF parameters F = (F"/3) and D = (D"/3) are defined by the functional
derivatives
8<H>G
F,,/3 h"/3 + [af3ho].Ry6 , Ft F,
8Rcx/3
8<H>G 1
Da/3 = 2" [a'Ylf3o] (-K6,) , DT = -D, (21 )
8K:/3
which involve effects of unpaired mode amplitudes as well as those of paired mode.
From eqs. (20) and (10), we get the equations of motion with respect to the variables
q and r, respectively, as
1: 1:
Using the definitions (6) and (7) for the variables q and r, eq. (22) is rewritten as
285
and their complex conjugates. From the above eq. (23) and the compatibility with
the orthonormalization condition, we obtain the equations of motion for the proper HB
amplitudes a and b and the unpaired mode amplitudes x and x* in the following forms:
ina =
ini (24)
Thus we could reach to an extended TDHB theory in which the paired and the unpaired
modes are trated in an equal manner [3]. This extended TDHB theory applicable to
both even and odd fermion systems is a TDSCF theory having the same level of a mean
field approximation as the usual TDHB one for even fermion systems.
By putting a andb as eiEt/lia and eiEt/lib, we obtain a static extended HB equation
[
F*-l!Z(X*MT+M*XT)D*-l!Z(X*Mt-M*X t )] [a] =E[a] , (25)
_D+_l
l+z
(xMT-MxT)-F+-
1 (xMt+Mxt)
l+z
b b
F
[ -D*
D
-F*
v2M] [ x /
v2M* -x*/v2
v2] = 0. (26)
v2Mt v2MT 0 Z
Note that, though we use the same symbols, the new a, b, x and z in eqs. (25) and (26)
are time independent amplitudes. The constraint term in (17) as a whole should vanish
in the physical fermion space, but their classical part in (18) might not. Through
eq. (19), we must determine the parameters k and I in the above M.
286
into
The time reversal symmetry of F is broken in general, but is recovered if the SO(2N + 1)
density matrix has the time reversal symmetry R~f3 = sO'sf3Rc,jj. The SCF parameters
FO'f3 and DO'(3 are decomposed as
where fO'f3 and dO'(3 are the usual SO(2N) SCF parameters defined as
(31)
The expectation values of the annihilation-creation operators themselves, by the I G>,
can be expressed in terms of the PO'(3 and KO'(3 and the XO' and x~ as
(33)
Due to the property of time reversal symmetry, the time-reversed Hartree state i is
degenerate in energy with the state i:
ei . (34)
287
The state tiw",i corresponds to the time reversal of the state W",i. The phase ti, the
counterpart of s'" in the Hartree frame, satisfies the relations ti = -ti and tfti = 1.
The Hartree amplitude W",i satisfies the relation W~i = sCitiwCii'
Using the Hartree basis, we put the HB amplitudes for the paired mode into product
forms
We also introduce the following quantities associated with the unpaired mode ampli-
tudes
(37)
(38)
We further assume that the pairing potential D = (Dc;f3) is constant. This makes the
situation very simple as the BCS theory does [10]. Then, the pairing potential has the
following form:
(39)
where 9 is the strength parameter for the pairing force. By substituting eq. (35) into
eq. (27) and using the Hartree equation (33), the extended HB eigenvalue equation (27)
under the assumption (39) is converted to the following form:
eiUi - -1- (*
XiMi + Mi*)
Xi Ui + 6 *Vi + -1- (*-
x;lvli - Mi*_Xi ) Vi EiUi,
l+z l+z
1 - -) 1(-
-eiVi + --(X1Mi
l+z
+ M(Xi Vi + 6ui - - - X1Mi - M(Xi)Ui
l+z
EiVi, (40)
where we have used the orthonormalization condition appearing in eq. (33) and the new
quantities defined in eq. (37). Similarly, multiplying eq. (28) by w~, and summing over
the index a, we obtain the equations
0,
0, (41)
e1 + 6*M
- 2z -',--;~'-;-:-.,-::-....:.'
Xi (42)
ef + 161 2
288
In the second step we diagonalize the pairing interaction contribution only to the
SCF pairing potential D = (Dcx(3). To diagonalize the extended HB eigenvalue equa-
tion (40), it is very useful to notice some important relations satisfied with Xi, Xi, Mi
and Mi. Using the above formal solutions, we obtain the relations
2ei1M;l2 - t:,.MTMi - t:,.*MTMi
-2z e; + 1t:,.12 '
2eilMil2 + t:,.MTMi + t:,.*MTMi
- 2z
e; + 1t:,.12
t:,.*(IMiI2 + IMiI 2)
2z er + 1t:,.12
2 t:,.(IM;l2+ IM;l2)
(43)
- Z e; + 1t:,.12 .
Substituting eq. (43) into eq. (40), we have a secular equation
-+ 2z 2eiIMiF-t:,.MtMi-t:,.*MtMi E t:,.*(1+..1LIMiI2+IMiI2)
e, Hz e; + It:,. 12 , Hz e; + 1t:,.1 2
=0, (44)
t:,.(l 2z IMiI2+IMiI2) _ -+..1L 2ei IM;!2 +t:,.MTMi+t:,.*AlI[Mi Ei
+ Hz e; + 1t:,.1 2 e, Hz e; + 1t:,.1 2
from which we obtain an eigenvalue
In the above quasi-particle energy, we must adopt the solution with upper sign, because
it reduces to the usual HB quasi-particle energy when z = 1 [ll]. The quasi-particle
energy involves unpaired mode effects in contrast to the usual HB theory, which is
unable to describe the unpaired modes.
With the help of the extended HB eigenvalue equation (40), the orthonormalization
condition (33) and the set of relations (43), the eigenvectors are obtained as
289
(47)
where we have used the normalization condition I:i Ix;j2 = I:i IXil2 = 1 - Z2.
Up to the present stage, the SCF Hartree potential F,,{h the chemical potential
>. included in the single-particle Hamiltonian h",{3 and the pairing potential Do remain
undetermined. With the use of eqs. (29)-(32), the product form of the HB ampli-
tude (35), its time reversal ansatz (36) and the time reversal property for the unpaired
mode amplitude (38), the SCF Hartree potential given through eq. (29) is determined
as
and the pairing potential (39) and the chemical potential are determined by the condi-
tions
We have yet an unsolved problem, namely to determine the parameters k",{3 and
l"'{3 in M". They cannot be determined in the classical limit only. The determina-
tion of them requires to take quantum mechanical fluctuations into account but has
a difficult prescription. Instead, we will take here another approach. Using the rela-
tions
(51)
we obtain the identities
290
We have used the property U(9)IO> = U(G)IO> and eq. (7) repeatedly. On the
U(9)1 0 >, the fermion operators Ca and ct are shown to satisfy exactly the anti-
commutation relations
(C~CiJ+CiJc~)U(9)10> = DaiJU(G)IO> ,
(c~c1+c1c~)U(9)10> = (caciJ+ciJca)U(9)10> = 0, (53)
which, however, make no roles to determine the parameters kaiJ and laiJ in Ma. At a
region near to z = 1, using an approximate function
U(9)IO> ~ iP~o(G)(I+r",c~)exp(I/2qaiJC~cb)IO>
= ~(Jl;Z+I/Jl;Z)U(G)IO>. (55)
(c~CiJ+CiJc~)U(9)IO> ~ (c~CiJ+CiJC~)U(G)IO>
(c~cb+cbc~)U(9)IO> ~ (c~cb+cbc~)U(G)IO> =
where we have used eqs. (52) and (55). We should demand that the expectation
values of the anti-commutators of the Ca and cl by the approximate U(G)IO> of
U(9) 10> satifsy the results deducible from the corresponding anti-commutation rela-
tions
291
from which we can obtain the following relations:
Z+ 1I (1 ; Z) _2 _ 4 1 ; ZI (1 ; Z+ 1)] . oa(3
2 [3 1 ;
The expectation values < c" >a and < c~ >a given by eq. (32) are expressed in terms of
the P,,(3 and and the x., and x:. The p.,(3 and
/\'.,(3 become time independent if we
/\',,(3
put a and b as e a and e
iEt / ll b . The x." x: and are time independent originally.
iEt / ll Z
Then, the expectation values < >a and < c~ >a become also time independent. We
Ca
call the set of time independent relations the quasi anti-commutation relation approx-
imation for the differential fermion operators Ca and c~. We should stress that this
quasi anti-commutation relation approximation can be regarded as a good approxima-
tion at the region very near to the point Z = 1 (the proper HB wave fundion). From
the set of eqs. (60) and the definition of M., in eq. (19), the conditions to determine
the parameters in the Lagrange multiplier terms are calculated to be
- c,,>ax~) x~k;jJ- c,,>a*x.,) x~I'YjJ = Ix,,12MI (not summed for a). (61)
They may play crucial roles for a unified self-consistent description of Bose and Fermi
type collective excitations at the region very near to the point Z = 1, since they are
the necessary and sufficient conditions to determine the unknown parameters in the
Langrange multiplier terms. Until now we have not had a unified and self-consistent
method to describe the Bose and Fermi type collective excitations, except the work
of Fukutome and the present author [5]. It is a very interesting problem to consider
another approximate SO(2N + 1) wave function U( G)I 0 > of the U(9)1 0> at a region
very near to Z = 0, because at the limiting case of Z = 0 the SO(2N + 1) wave function
corresponds to the situation that we have the largest contributions from the unpaired
mode amplitudes owing to the relation x:x.,
+ z2 = L Such a physical situation is very
contrary to the one at the limiting case Z = 1 (the proper HB wave function), in which
we have no effects due to the unpaired modes.
CONCLUDING REMARKS
The extended HB theory for a fermion system with N single-particle states has been
derived from the extended TDHB theory. The extended HB and TDHB theories have
292
been constructed from a group theoretical deduction starting with the fact that the
fermion annihilation-creation and pair operators form a Lie algebra of the SO(2N + 1)
group. The induced representations of the SO(2N + 1) group was found by a group ex-
tension of the SO(2N) Bogoliubov transformation for the fermions to the new canonical
tranformation group. Embedding the SO(2N + 1) group into the SO(2N + 2) group
and using the variables of the coset space SO(2N + 2)jU(N + 1), we have obtained the
extended TDHB equation from the classical Euler-Lagrange equation of motion for the
SO(2N + 2)jU(N + 1) coset variables. We have expressed its final form through the
variables as the representatives of the paired modes and the unpaired modes. From the
extended TDHB we have derived the static extended HB theory in which the paired
modes and the unpaired modes are treated in an equal manner. The extended HB the-
ory applicable to both even and odd fermion systems is also the SCF theory with the
same level of the mean field approximation as the usual HB theory for the even fermion
systems. However, it includes the unknown parameters which originate from the La-
grange multiplier terms involved in the image of the Hamiltonian in order to select out
the physical spinor subspace. They cannot be determined in the classical limit only, and
a complete determination of them requires to take the quantum mechanical fluctuations
into account. We have, instead, determined the parameters with the aid of the quasi
anti-commutation relation approximation for the differential fermion operators. Both
the extended HB eigenvalue equation and the quasi anti-commutation relation approx-
imation provide a group theoretically transparent kinematical frame, which is able to
describe both the paired and unpaired modes in many fermion dynamics. They may
work well in both even and odd soft nuclei with strong collective correlations in which
the effect of the unpaired modes appears apparently. They play important and cru-
cial roles for a unified self-consistent description of the Bose and Fermi type collective
excitations in such nuclei.
It is a very exciting problem to suppose an approximate SO(2N + 1) wave function
at the region very near to the point z = O. We have never experienced the physical
situation arising from such an SO(2N + 1) wave function, because in this limiting case
most physical contributions originate from the unpaired mode effects. This situation is
in striking contrast to the circumstance, in which we have no effects due to the unpaired
modes at the limiting case of z = 1 (the prape". HB wave function). A study of such a
problem can be expected to open quite a new field for the exploration of the fermion
dynamics including the stronger unpaired mode effects. In a forthcoming paper, we will
investigate the above mentioned problem and propose a detailed solution procedure to
the extended HB eigenvalue equation unter the two kinds of the quasi anti-commutation
relation approximation at the regions very near the points z = 1 and z = 0, respectively.
We will treat, as an illustrative example, a problem of a schematic model of nuclei with
pairing plus quadrupole interactions.
Finally, we should point out that it is also an interesting and important task to
study an interrelation between the present theory and the work by Fukutome and the
present author [5]. It will become possible in the near future.
ACKNOWLEDGEMENTS
An early stage of this work was performed during the author's stay at the Physik-
department der Technischen Universitat Miinchen (TUM). This work was also support~
ed in part by the Portugese program PRAXIS XXIjBCCj4270j94. The author would
like to express his sincere thanks to Profs. P. Ring and J. da Providencia for kind and
293
warm hospitality extended to him at the Physikdepartment der TUM and the Centro
de Fisica Te6rica, Universidade de Coimbra. He also thanks Prof. H. Fukutome of the
Kyoto University for his comments.
APPENDIX
In this appendix, we give the SO(2N + 2) Lie operators and their representations
in terms of the SO(2N + 2)/U(N + 1) coset variables Qpq. All the fermion SO(2N + 2)
Lie operators {E~, Epq , Epq} are spanned by the fermion SO(2N) Lie operators and the
operators with index 0
Q' a Q a '0 a
;r aQ~r - qr aQpr - Z pq aT '
a - aQ;q
Qpc Qsq aQrs a - Z'Q pq aTa '
(63)
The phase variable T becomes identical with the phase variable T (= i /2 . In[det( a*) /
det(a)]) of the U(N). The images of the fermion SO(2N + 1) Lie operators can be
represented with the aid of the images of the fermion SO(2N + 2) Lie operators as
E (3o: = e '"(3 , E 0:(3 = eo:(3, E",(3 = ~",(3,
~
(65)
where we have used the relation (EO:O)2 = O. Eq. (65) leads to the property U(G)I 0> =
U(9)1 0> where G E SO(2N + 1) and 9 E SO(2N + 2). From eqs. (63) and (64), we
can get the representations of the SO(2N + 1) Lie operators in terms of the variables
qo:(3 and T",:
294
(66)
(67)
(68)
With the aid of the property U( G) 10> = U(g) 10> (G E SO(2N + 1) and 9 E
SO(2N + 2)) and eq. (65), the classical Lagrangian of the system embedded into the
SO(2N + 2) group is given as
L(g(t)) < 0 Iut(g(t)){ in! - H}U(9(t))1 0 >
in tIt t n
"2Tr{(1 + Q Qt (Q Q - Q Q)} -"2+ - <H>G(t) . (69)
Using the expressions for the matrices Rand K in eq. (12), the relations (70) and
their derivative formulae (71) given below
Rt R, KT = -K,
Q(l- R) (l-R*)Q = K, Q*K K*Q -R, (70)
(71)
the Euler-Lagrange equation of motion for the SO(2N + 2)/U(N + 1) coset variables
is calculated to be
d ( 8L )
dt 8Q* - 8Q*
8L
=
--2\n [Q(l- Q*Q)-l + (1- QQ*)-lQ
+ Q(l- Q*Qt 1 Q*Q(1 _ QQ*)-l
+ (1 - QQ*t 1 QQ*(1 - QQ*)-lQ] + 8~~:G = 0, (72)
and its complex conjugate. From eq. (72), we can obtain eq. (18) without the additional
term.
Finally, we show the following formulae to be very useful to evaluate the expectation
values of anti-commutators of the differential operators Ca and c~ by an approximate
U(G)I 0> ofthe SO(2N + 1) wave function U(G)I 0> (= U(Q)I 0:
C Jl+Z __l{-+Z
- - { rT t
(l+qq)-l+rtq(l+qtq)-l} = _ T {-+Z
x 01
2(1 + z)
__
01 2 2 2 2'
l{-+Z{ Tq (1 + qqttl} {-+Z
0/
t~ - (1 + qq t
-- rt t 1 - r t = xt01
cOIV~-2- 2(1 + z)
--
2 2 a 2'
295
c(3{rT(1+ qtqtl + rtq(1 + qtqtl} a =
{(I + qqttlr - (1 + qqttlqr*} (3 . {rT(1 + qtq)-l + rtq(1 + qtqtl} a '
2
c(3{rt(1 + qqt)-l - rT qt(1 + qqt)-l} = (2 - --)60/(3 +
a 1 +z
{(I + qtq)-lr* + (1 + qtqtlqtr t {rT(1 + qtqtl + rtq(1 + qtq)-l}(3 ,
2
c~{rT(1 + qtqtl + rtq(1 + qtqtl}(3 = -(2 - --)6a (3-
l+z
{(1 + qtq)-lr* + (1 + qtqtlqtr} a ' {rT(1 + qtqtl + rtq(1 + qtq)-l}(3 ,
and
REFERENCES
[1] N. N. Bogoliubov, Sov. Phys. Uspekhi 67 (1959), 236; J. P. Blaizot and G. Ripka,
Quantum Theory of Finite Systems, MIT Press, Cambridge, MA, 1986.
[2] S. Nishiyama, Prog. Theor. Phys. 66 (1981), 348.
[3] H. Fukutome, M. Yamamura and S. Nishiyama, Prog. Theor. Phys. 57 (1977),
1554.
[4] J. Schwinger in: Quantum Theory of Angular Momentum, L. Biedenharn and H. van
Dam (eds.), Academic Press, New York, 1965, 229; M. Yamamuraand S. Nishiyama,
Prog. Theor. Phys. 56 (1976), 124.
[5] H. Fukutome, Prog. Theor. Phys. 58 (1977), 1692; J. Dobaczewski, Nucl. Phys.
A3S0 (1982), 1; H. Fukutome and S. Nishiyama, Prog. Theor. Phys. 72 (1984),
239.
[6] M. Baranger, Phys. Rev. 122 (1961), 992.
[7] S. Nishiyama, talk at the 21st Internat. Colloquium on Group Theoretical Methods
in Physics, Goslar, Germany, July 15-20, 1996.
[8] H. Fukutome, Int. J. Quant. Chern. 29 (1981), 955; H. Fukutome, Prog. Theor.
Phys. 65 (1981), 809; J. Dobaczewski, Nucl. Phys. A369 (1981), 213, 237.
[9] S. Nishiyama and H. Fukutome, J. Phys. G: Nucl. Part. Phys. 18 (1992), 317.
[10] J. Bardeen, L. N. Cooper and J. R. Schrieffer, Phys. Rev. 108 (1957), 1l75.
[1l] P. Ring and P. Schuck, The nuclear many-body problem, Springer, Berlin, 1980.
296
QUANTUM SPACE-TIME GROUPS
AND BEYOND
INTRODUCTION
1 olmo@cpd.uva.es
298
THE PSEUDO-ORTHOGONAL CAYLEY-KLEIN SCHEME
The N-dimensional CK algebras are real Lie algebras of dimension N(N + 1)/2,
characterized by N fundamental real parameters (WI, ... ,WN) [13). The most natural
way to obtain this family of algebras is by graded contractions (a generalization of the
idea of IW contraction keeping invariant a given grading of the original Lie algebra
introduced in [14)) of the Lie algebra so(N + 1) [15). So, considering a 7J,~N grading of
so(N + 1), the CK algebras appear as a particular set of the 7J,~N graded contractions of
so(N + 1). We will denote them by SOW1,W2, ... ,WN(N + 1), and their associated CK groups
by SOW1,W2, ... ,WN(N + 1). The real parameters Wi can be rescaled to +1 or -1 if they are
nonzero. So, we have 3N different N-uplas (WI,'" ,WN) and the same number of CK
algebras, but not all of them are non-isomorphic. However, from a geometrical point
of view, the CK groups SOW1,W2, ... ,WN(N + 1) are the group of motions of 3N different
geometries. We do not pay more attention to the geometrical aspects that can be found
in Ref. [13).
Let us consider the following well adapted basis {Jij ; i < j, i, j = 0, 1, ... ,N} to
the grading (7J,~N) of SOW1,W2,,,,,WN(N + 1). The nonvanishing Lie brackets of this family
of algebras are
(1)
(2)
i :::; t, j :::; m.
Among these groups we can mentioned for their interest, the Euclidean group in N
dimensions, which appears when (W2,'" ,WN) = (1, ... ,1), the Poincare group in (N-
1,1) dimensions,which is is obtained several times, for instance, for (-1,1, ... ,1), and
the (N - 1,1) Galilei group that corresponds to (0,1, ... ,1). In general, a determined
group can appear several times in this CK description, but in each case the geometry
determined by it is different.
From relations (2) it is clear that the groups ISO(w2,,,,,wN)(N) == SO(0,W2, ... ,WN)(N +1)
have a semi direct product structure
(3)
299
The second order central element for the algebra iso W2 ,... ,WN(N) is
N-l
C(W2, ... ,WN) = P;, + LWiNPl- (4)
i=l
Remark that the Casimir has only generators of the Abelian translation algebra.
As an example, we consider the nine two-dimensional CK geometries. There are on-
ly five non-isomorphic motion groups for them: 80(3) for (Wl,W2) = (1,1); 80(2,1) for
(1, -1), (-1,1), and (-1, -1); the Euclidean group in two dimensions for (1,0), (0,1);
the (1 + 1) Poincare group for (0,-1), (-1,0); and the (1 + 1) Galileigroup for
(0,0). 'rhe Lie commutators of the generators of the CK Lie algebras sO(wl.,W:l)(3)
are (1)
[J12 ,JOl] = J02 , [J12 , J02 ] = -kVOl, [JOl ' J02 ] = kd12' (5)
and the Casimir is
(6)
First at all, we remember the idea of deformed Hopf structure that we will use
along this work. By a deformed algebra or quantum algebra Uz(g) we understand the
Hopf algebra of formal series on the deformation parameter z with coefficients on the
universal enveloping algebra U(g) a Lie algebra 9 [4].
There are several forms to obtain a deformatio11 of the family of CK algebras, such
that for all the CK algebras the Lie algebra level or the coalgebra level or both are
deformed [10]-[12]. An interesting way [16] is based is to make use of the concepts of
Lie bialgebra and contractions of Lie bialgebras [17]. In low dimensions, N = 2,3, the
semisimple algebras so(p, q), p+q = N +1, have been also deformed within this scheme,
however for N 2:: 4 only for those algebras having WI = it has been possible to obtain
an explicit deformation. Note that inside the Drinfel'd-Jimbo scheme of deformations
of semisimple Lie algebras, a deformation for the Cartan families B/ and D/ is given
explicitly in the Chevalley basis, but to obtain explicitly the algebraic structure on the
basis {Jij} has not been completely achieved [18]. However, despite this setback, from
a physical point of view inside the cases with WI = we find kinematical groups of
great interest like Poincare and Galilei.
In Ref. [12] the general expression of a standard (i. e., associated to a quasi-triangular
r-matrix) deformed Hopf structure equipping all the affine N-dimensional (N 2:: 2) CK
algebras isO(W2, ... ,WN)(N) has been introduced. It corresponds to a quantization of the
coboundary Lie bialgebra (SO(0,W2, ... ,WN)(N + 1),15), generated by the (non-degenerate)
classical r-matrix
N-l
15 =ZL J sN /I. P., (7)
s=1
300
Coproduct:
Counit:
i,j = 1, ... ,N; (9)
Antipode:
')'(X) = _e(N-1)~PN X e-(N-1)~PN
Deformed commutators:
i,j,k = 1, .. . ,N - 1. (12)
On the other hand the set of the N basic IW contractions relating the algebras of
the CK family can be implemented to the q-deformed case. The crucial point is that the
deformation parameter z must be transformed: the action of the new q-contractions on
the generators is as in the non-deformed case, and the deformation parameter transforms
as z --+ z / c since the generator PN under the action of these contractions goes to cPN.
The exponential e ZPN appears in the Hopf structure, only if z and PN transform under
contractions in a opposite way this exponential does not diverge or disappear when
c --+ O. An analysis in terms of Lie bialgebras over the role played by the deformation
parameter z under contractions has been carried out in [17].
We would like to do some remarks. Thus, for instance, quantum contractions are
not a direct consequence of the IW contractions, since in a q-algebra there are two
algebraic levels where the transformation acts: the sector relative to the Lie algebra,
i.e, the commutation rules (deformed and nondeformed), and the coalgebra sector, i. e.,
the Hopf algebra structure. The IW contractions deals at the first level, but their
q-generalizations concern both levels and are not, in principle, longer unique. The
deformation parameter z is also involved, and when is contracted acquires, usually,
dimensions.
301
It is worthy to mention that Uq isO(W2, ... ,WN)(N) has a structure of bicrossproduct [19]
302
[JI, Pz] == ~ sinh(zP3), [Jz, Pl ] == -~ sinh(zp3 ),
[I<3' H] = ~ sinh(zP3), [Jl,Jz] = J3 cosh(zP3) + ~HW4' (18)
7
[Jl , I<3] == -I<z cosh(zP3) + Pl w4, [J 2 , I<3] = I<l cosh(zP3) + ~ Pz w4,
where W 4 = -~H J3 - PlI<z + PZI<l is the (non-deformed) fourth component of the
Pauli-Lubanski vector. Note that PI, Pz , H, I<l, I<z, J 3 and P3 span the maximal Hopf
subalgebra.
The second order q-Casimir is
(19)
In this case we can write in an explicit way the Casimir associated with the Pauli-
Lubanski vector
1 q2
Czq = -2"Wl + W 2q2 + W3q2 + [cos h( zP3) + -Z24 (P21 + Pz21HZ)]
- 2" W24' (20)
C c
W lq = -Pdl - P2 J 2 - ~sinh(zP3)J3'
Wi = -~H Jl - P2I<3 + ~sinh(zP3)I<2' (21)
Wj = -~H J 2 + Pl I<3 - ; sinh(zP3)I<1'
With respect to the Galilean case (Uz isO(O,H,+l)(4 note that the non-relativistic
limit of the q- Poincare is obtained from the above deformation making W2 = 0 (or taking
the limit W2 = -1/ c2 -+ 0). Coproduct and counit remain invariant since their expres-
sions do not depend on W2' The only Lie brackets different from those of Poincare are
i,j=I,2,3, (22)
and all the deformed ones (18) since the Galilean version of W4 differs from the corre-
sponding to Poincare. Now the deformed central elements are
where
Other interesting quantum algebras are the 4D q-deformed Euclidean algebra and
the (3 + 1) K- Poincare [8J. The first one is related with the set of values of the
(W2,W3,W4) = (1,1,1). The K-Poincare appears for (1,1, -1) or if one considers di-
mensions W4 = _c 2 , and assigning the time translation generator to the original P4 ,
the remaining generators are identified according with this assignation. Note that the
non-relativistic limit c -+ 00 is not well defined for the K-Poincare and the K-Galilei
algebra does not appear in a natural way [20J.
303
q-DEFORMED EQUATIONS FROM q-CASIMIRS
(26)
obtained by means of the q-Casimir acting on the wave functions W( x, t). Taking into
account the differential realization (25) and the symmetric derivative
l I z
2 (W(x + z, t)
+ W(x - z,t) - 2W(x, t)) - 20t W(x, t) = czW(x, t), (28)
z c
where Cz is a real number. This equation (28) can be considered as an imbeding on the
304
continuous of the finite difference-differential equation (Db-E)
(29)
on a lattice of length spacing z, where j is a discrete variable indexing the points of the
lattice. This equation is a discretization in the spatial axis of the (1 + 1) Klein-Gordon
equation. For the massless case (c z = 0) this equation describes a one-dimensional infi-
nite harmonic chain. The (deformed) q-symmetry of this equation allows to explain the
fusion of phonons by means of the deformed coproduct. A more detailed explanation
of this system was made in 1992 in this series of Conferences by Celeghini [23]. In a
similar way the same author and collaborators [24] proved that the quantum extended
Galilei(1 + 1) is the symmetry algebra of the Heisenberg ferromagnet (the 1D Heisen-
berg spin chain of th~ XXZ type), incidentally the magnon wave function verifies a
q-deformed Schrodinger equation.
In the N-dimensional case, i. e., for the affine CK algebras Uz (is0 W2 , ,wN(N, there
exists the following differential realization
where x == (Xl, X2, ... , XN) = (x, XN)' Thus, the q-Casimir equation
N-l
42 (sinh 2 (zaN)) "II1(x, XN) +L WiNa;"II1(x, XN) = czili(x, XN). (32)
Z i=1
L wiNalilij(x) = czilij(x).
N~1
305
differential realization (30) becomes now in coordinates (t, xl, X 2 , X 3 ):
which gives rise via the q-Casimir a differential-difference version of the usual (3 + 1)
Klein-Gordon equation with the spatial coordinate x 3 discretized
where x = (t, xI, x 2 ). By considering the limit W2 = -1/c2 - t 0 we obtain the non-
relativistic limit of this equation, i. e., a differential-difference version of the usual (3+ 1)
Schrodinger equation
Note that the non-relativistic limit of the differential realization (35) allows to obtain
a differential realization for the (3 + 1) q-Galilei algebra.
Another case to mention is the corresponding to the (3 + 1) ,,;- Poincare [8], where
the ,,;-deformed Klein-Gordon equation takes the form
(38)
Also a ,,;-Dirac equation is defined by taking a square root of the ,,;-Klein-Gordon oper-
ator
,,; 4 at ) <p =
( "!
If)
i + 2-;;-,,! 8m 2k - m 0, (39)
where the matrices,,!!', f.l = 1, ... ,4, verify b!',"!V} = 2g!'v withgw = diag (1,1,1, -1).
Here the discretization is in the time instead of a spatial coordinate, and the deformation
parameter,,; has dimensions of inverse of time.
On the other hand, the field of D.6.E's is very active since is a way to approach
in many cases to physical situations by numerical methods. Among the applications
we can quote: dissipative systems, nuclear physics, onedimensional chains of spins
(magnons), onedimensional harmonic chain (phonons), etc. For the two last examples,
as we mention before, their relevant structures are explained by q-deformed algebras.
An important subject of investigation in this field is the study of the symmetries
of D.6.E's in order to find their solutions. There are different methods to attack the
problem, from those that try to mimic the standard Lie infinitesimal procedures used
for the differential equations, up to new ways to proceed in the spirit of the D.6.E's. We
can mention, for instance, the works by Negro et a1. presenting linear equations [25],
Winternitz and coworkers [26] (and Ref. therein) studying non-linear situations.
Another point to mention is that the ,,;-Poincare algebra (group) is a subject of
great activity in order to find physical applications. In the last years many papers
have appeared trying to relate it with quantum field theory, [27], nuclear physics [28],
astrophysics [2], non-commutative differential calculus [29], particles [30], etc.
Finally, to note that there is another deformation of the Poincare algebra [31, 32].
306
This is a non-standard coboundary deformation (i. e., its corresponding r-matrix is
triangular) called "null-plane" quantum Poincare algebra because is a deformation of
the Poincare algebra in such a basis that provides a null-plane description [33]. This
Poincare invariance splits into a kinematical and a dynamical part, which originates a
Schrodinger equation that gives the time evolution of the wave functions living on the
null-plane. In the deformed case a q-Schrodinger equation appears, whose associated
q-Hamiltonian is a deformation of the null-plane Hamiltonian of quantum chromody-
namics [34]. The null-plane deformation includes a potential part to add to the quantum
chromodynamics interaction.
~)
T=(aed' (40)
From the original action of SL(2) on 1R 2 (i. e., S L(2) x 1R 2 __ 1R 2) by matrix multipli-
cation we can try to induce a coaction of SLq(2) on the algebra of polynomial functions
on 1R 2 , M (M -- SLq(2) x M)
(42)
"",P
where (AB)ij = L..k=l Aik B kj . After a straightforward computation one obtains
that x and fj ought to verify
xfj=qfjx, (43)
if x' y fj', defined by (42), accomplish a relationship analogue to (43). So, we get a
deformed algebra Mq generated by x and fj, which do not commute (43). There is a co-
action Mq -- Mq SLq(2) compatible with the multiplication on M q. The q-deformed
algebra Mq is called the quantum space associated to the quantum group SLq(2).
As is well known the homogeneous spaces play an important role in mathematical
physics. An implementation of this concept to quantum group theory seems natural.
However, the usual geometrical meaning disappears and it is necessary an algebraic
approach. So, one needs to find an appropriate sub algebra of the quantized functions
on the group realizing the functions on the homogeneous space itself.
We present now the study of the quantum equation associated in a natural way
with a quantum homogeneous space of the 2D Euclidean quantum group Ez(2) [6, 37]
(a similar study for another q-deformation of E(2) was presented in Ref. [38]). This
307
quantum group is generated by e-" at, a2, with commutators, coproduct and antipode
b.( e- i ) = e- i @ e- i ,
we can obtain a quantum homogeneous space as the invariant subalgebra and left coideal
The local coordinates x = al - ia2 and x = al + ia2 generate this quantum plane. The
coaction of E z (2) on this q-space is
Ox = e- i @ x + x @ 1, Ox = ei @ X + X @ 1 . (50)
On the other hand, the following "infinitesimal" action of Uz e(2) on the q-plane
(52)
308
where H+ = l(e ZP2 - 1) - !ie ZP2 PI and H- = l(1 - e- ZP2 ) + lie- zP2 PI are the
2z 2 2z 2
deformations of the holomorphic and antiholomorphic operators P2 /2 =F iPI/2. The
z-deformed free Schrodinger equation is:
(53)
The way of solving this equation is diagonalizing the operator on the right hand side
of (53) in analogy to quantum mechanics. Considering "plane wave" states, i. e., eigen-
functions of H+ ()..(H+)V\h_ = h+'I/;h+h-) and H- ()"(H-)'I/;h+h- = h-'I/;hh-)' we
get that the functions
(54)
(55)
Note that the these eigenfunctions can be expressed as hypergeometric series. In the
classical limit z -+ 0 we recover the usual plane waves ei(h+x-h-x).
On the other hand, on the angular momentum states, in this case eigenfunctions
of the operator :J = e- ZP2 / 2 (J - i(z/4)PI)' ()..(:J)'I/;r = 'I/;r) , we obtain that the
hypergeometric functions
'1/;-. = t; (-kz/lY
00 _
l! (l + r)! (X)I (1 - X)I(X)r (56)
diagonalize the Casimir, i. e., )..(H+ H-) 'I/;-r = k'l/;-r. A similar expression is obtained
for theeigenfuctions '1/;.. The classical limit z -+ 0 gives rise to the standard Bessel
functions.
CONCLUSIONS
The panorama presented in this paper, obviously, does not cover all the aspects
of the quantum kinematical groups and physics, and therefore quantum groups and
physics. With respect to the physical applications of the deformed kinematical algebras
this is an open research field, in the sense that at the present time the role of these
q-symmetries in physics has not been clearly established: a experimental support is
lacking. The possible relevance of the quantum algebra in lattice physics was showed in
Ref. [221-[241 but only in (1 +1) dimensions. For higher dimensions, as we have seen, only
a spatial coordinate or in other cases the time coordinate appear discretized, and this is
the reason that only onedimensionallattices has been obtained in connection with these
quantum algebras. Another point to clarify is the real meaning of these onedimensional
lattices, since we do not obtain in fact a discretization of the continuous. So, starting
from any point Xo E 1R we can go only to the points Xo + nz, nEZ, by means of the
action of the discrete derivatives as translations of length nz such as we have showed
in Sect. 4. Choosing another origin x~ we have a "different" lattice {x~ + nz, n E Z}.
On the other hand, the use oLquantum homogeneous spaces as .has been presented
in Sect. 5 is very recently; there the main difficulty is to deal with non-commutative
309
objects that force to work in an algebraic context. Moreover, it is necessary to obtain
a suitable q-version of the Haar measure for non-compact groups (for compact groups
there is a q-theory of the measure), in order to compute expectation values. This fact
is crucial to establish a Quantum Mechanics with quantum homogeneous spaces as
configuration spaces .
All that shows that there is a wide field of research in the interface between quantum
groups and space-time symmetries. We hope that the problems enunciated above will
be solved in the next years, physicists and mathematicians are putting forth many
efforts in this task.
ACKNOWLEDGMENTS
REFERENCES
310
[15] F. J. Herranz, M. de Montigny, M. A. del Olmo and M. Santander, J. Phys. A:
Math. Gen. 27 (1994), 2515.
[16] A. Ballesteros, Contraction of Lie Bialgebras and Quantum Deformations of Kine-
matical Symmetries, PhD Thesis, University of Valladolid, 1995 (in spanish).
[17] A. Ballesteros, N. A. Gromov, F. J. Herranz, M. A.del Olmo and M. Santander,
J. Math. Phys 36 (1995), 5916; A. Ballesteros and M. A. del Olmo, in: Quantum
Groups and Quantum Spaces, Banach Center Public. 40, Warsow, 1997.
[18] A. Ballesteros, N. Gromov, F. J. Herranz, M. A. del Olmo and M. Santander,
in: Quantum groups, formalism and applications, J. Lukierski, Z. Popowicz, J.
Sobczyk (eds.), Polish Scient. Publ., Warsow, 1995,325.
[19] J. A. de Azcarraga, M. A. del Olmo, J. C. Perez Bueno and M. Santander, Grad-
ed contractions and bicrossproduct of deformed inhomogeneous algebras (preprint
1996).
[20] S. Giller, P. Kosinski, M. Majewski, P. Maslanska and J. Kunz, Phys. Lett. B286
(1992), 57.
[21] J. F. Carifiena, M. A. del Olmo and M. Santander, J. Math. Phys. 26 (1985),
2096.
[22] F. Bonechi, E. Celeghini, R. Giachetti, E. Sorace and M. Tarlini, Rev. Lett. 68
(1992), 3178.
[23] E. Celeghini, in: Symmetries in Science VI, B. Gruber (ed.), Plenum, New York,
1993, 149.
[24] F. Bonechi, E. Celeghini, R. Giachetti, E. Sorace and M. Tarlini, J. Phys. A 25
(1992), L939; Rev. Lett. B46 (1992), 5727.
[25] J. Negro and L. M. Nieto, Lett. Math. Phys. 36 (1996), 351; J. Phys. A: Math.
Gen. 2a (1996), 1107.
[26] D. Levi and P. Winternitz, J. Math. Phys. 37 (1996), 5551.
[27] A. A. Kehagias, P. A. A. Meessen and G. Zoupanos, Phys. Lett. B346 (1995),
262.
[28] A. Delfino, J. Dey and M. Malheiro, Phys. Lett. B348 (1995), 417.
[29] C. Gonera and P. Kosinski, J. Math. Phys. 37 (1996), 5820.
[30] J. Dey, M. Dey, P.Leal-Ferreira and L. Tomio, Phys. Lett. B365 (1996), 157.
[31] A. Ballesteros, F. J. Herranz, M. A. del Olmo and M. Santander, J. Phys. A: Math.
Gen. 28 (1995), 941.
[32] A. Ballesteros, F. J. Herranz, M. A. del Olmo and M. Santander, Phys. Lett. B
351 (1995), 137.
[33] H. Leutwyler and J. Stern, Ann. Phys. (N.Y.) 112 (1978), 94.
[34] 1. B. Kogut and D. E. Soper, Phys. Rev. Dl (1970), 2901; H. H. Liu and D. E.
Soper, Phys. Rev. D48 (1993), 1841.
[35] Y. 1. Manin, Quantum groups and non-commutative geometry, Publ. CRM, Uni-
versite de Montreal, Montreal, 1988.
[36] A. Connes, Non-commutative geometry, Academic Press, London, 1994.
311
[37) A. Ballesteros, F. J. Herranz, M. A. del Olmo and M. Santander, J. Math. Phys.
36 (1995), 631.
[38) F. Bonechi, E. Celeghini, R. Giachetti, E. Sorace and M. Tarlini, Commun. Math.
Phys. 169 (1995), 463.
[39) M. S. Dijkhuizen and T. H. Korwinder, Geo. Dedicata., 52 (1994), 291.
312
QUANTUM MECHANICS ON THE TORUS,
KLEIN BOTTLE AND PROJECTIVE SPHERE
Christoph Schultel
INTRODUCTION
The Borel quantization shows that there is a topological dependence of the "free"
dynamics on the configuration space M, on which a quantum mechanical system is
localized. Unitarily inequivalent quantization mappings are dassified by elements (0, D)
in 7r;(M) X R. In the frame work of Borel quantization the quantization parameter D
gives rise to a non-linear Schrodinger equation [1, 2], which reduces to a linear one for
D = O. Our procedure is motivated by the isomorphism of elements in 7r~(M) in the
set of equivalence classes of complex line bundles with fiat connection [3]. Using these
fiat connections we will construct a Laplacian on the complex line bundle.
On the 2-dimensional torus T2, only the trivial complex line bundle allows fiat con-
nections and - using the restriction of the evolution equation given by the generalized
Ehrenfest relation [4] used in Borel quantization - the construction of the Hamiltonian
and its dependence on the topological parameter 0 E 7r;(M) has been investigated in
[5]. Here, we will consider manifolds which involve a non-trivial complex line bundle as
well, i. e. the Klein bottle J{2 and the real projective sphere RP2.
Quantization methods on the topologically non-trivial manifolds S2 and T2 have
also been discussed in [6, 7]. In contrast to our approach, they are based on geometric
quantization [8], the manifolds describe the phase space instead of the configuration
space and the evolution equation, e. g. the Laplacian, is not derived. In [7] the case
of S2 is treated and it turns out, that there is no complete quantization, whereas it is
shown in [6], that there exists a complete quantization on the torus.
In a first step we describe and analyze the topology of Tn, J{2 and Rp 2 and furnish
these manifolds with a Riemannian structure by projection from R n, for Rp2 also
by projection from S2. Then equivalence classes of complex line bundles with fiat
connection are introduced and locally parallel sections are explicitly derived. Using
1 christoph.schulte@tu-clausthal.de
these parallel sections, we can locally identify arbitrary sections on the line bundle with
a function on the manifold. In the function space on the manifold, the metric induces
a Laplacian via the Laplace Beltrami operator. Using the parallel sections once more,
the resulting function is lifted to the complex line bundle and yields a definition for a
Laplacian on the Hilbert space of square integrable sections on a complex line bundle.
This operator depends on the chosen flat connection on the line bundle and on the
metric on the manifold.
The treated manifolds will be constructed by equivalence relations on Rn and we
will choose the metric induced by Rn. The eigenvalues and eigensections of the corre-
sponding operators are derived and it turns out that they depend, as expected, on the
topological quantum number 0' E 7f;(M).
In the final section a comparison of the results.for the eigenvalues is presented.
Concerning physical applications, Tn is often considered in the context of Hamil-
tonian systems and RP2 appears in the effective configuration space of two identical
particles in R3. An application of the Klein bottle is not at hand.
In this section, we will construct a Laplacian on the complex hermitean line bundle
over the n-dimensional torus. It will turn out, that - due to the non-trivial topology
of the torus - there exist different Laplacians on the complex line bundle that may
be characterized by elements in 7f;(Tn), the group of characters of 7f1 (Tn) given as the
unitary representations of 7r1(Tn).
Topology of Tn
The n-dimensional torus Tn, an orient able manifold, can be modeled with perimeters
ai > 0 as the n-dimensional space Rn together with the identifications (or equivalence
relation)
(1)
i. e. Rn /"-" Its fundamental domain is given in Figure 1.
Let Ij denote the path connecting the origin in Rn with aj ej ERn, ej the unit
vector in direction j. The equivalence relation (1) identifies /j with a path from akek
to akek + ajej (k E {1, ... , n}). Hence the first fundamental group of Tn is given by2
7f1 (T n) = \/1, = zn ,
/ ... ,In Ilillizlj,-1 Ij2-1) "-' (2)
2 (a, bla b a-I b- I >is the group generated by the letters a, a-I, band b- I subject to the condition
a b a- 1 b- 1 = e.
314
and Rn is a universal covering of Tn. The group of characters, which characterizes
the equivalence classes of isomorphic complex line bundles with unitarily equivalent fiat
connection, is
(3)
(Xl, ... ,Xn,Z)'" (Xl, ... ,xn,z)+(mlal, ... ,mnan,z) zEC (4)
The action of a connection \7 on the line bundle can be related with the parallel
transport Ilx',x of an element in a fiber Fx over x to fiber Fx' over x' along a path con-
nection x with x'. If the connection is flat, the parallel transport is locally independent
on the path connecting x and x' and there exists a smooth function h : Rn --+ R onthe
covering space such that
(5)
(X,c) 1---7 (X',C' ) = (x',exp(ih(x') - ih(x))c).
6. 9 f = L..
~ 1
~-8
~ 9 kl8
8 ( Vdetg -8 f ) . (8)
k,l=l ydetg X k Xl
In a second step, we need a connection on 17. Here we choose - following our arguments
in the introduction - the previously discussed flat one \7. Consider now a locally finite
315
partition of unity, i. e. open sets Ui, i E I, that cover Rn /"', such that each x E Rn /'"
is only in finitely many Ui , and a set offunctions ki with suppki CUi, 0 ki(x)s: 1 s:
and 2:iEI ki ( x) = 1 for all x = (Xl,' .. , xn). Then the Laplace operator on (T/, 'V) acting
on arbitrary section a is given as
(9)
where a p is given in (7). One can show, that (9) is independent on the chosen partition
of unity, but depends on a p
The motivation for the lift defined in (9) is as follows: given a section ao with
ao(x) =I- 0, any other section can locally be written as a(x) = J(x)ao(x), where the
function J is uniquely determined by ao and a. Now calculatel:,g (~) := l:,gJ. To
have as a result a section a', multiply with ao to obtain
tJ.' a ~ a 2a _ 2i ah aa _
( Z.aax2h + (~)2)
= 2
9 ~ax2 ax.ax ax a (11)
j=l J J J J J
Two flat connections on line bundles over a manifold are called equivalent, if there
exists a line bundle isomorphism such that parallel sections are transformed into parallel
sections.
Bundle isomorphic flat connections give rise to unitary equivalent quantization map-
pings [9J. It was shown in [9], that two flat connections are unitarily equivalent, if the
difference of their differentials are logarithmically exact [10J, i. e. there exists in our
setting a mapping k : Tn -+ 51 such that
(13)
This is the case if and only if the functions h1 and h2 satisfy the conditions
316
i. e. the parameters aj,l of hI and aj,2 of h2 in (6) differ by multiples of 21T. Hence
the functions h with (6) split into a family of equivalence classes [hal with a E [0, 21T)n
(compare with (3. A member of each class is e. g. he< (Xl,'" ,X n ) = 2:7=1 ajXj with
aj fixed in [0, 21T), a = (a1," . ,an)'
We summarize the result: Pure states of a free particle moving on Tn evolve on
L;( 1], df-L) with
(15)
The evolution equation of unitarily inequivalent quantization is labeled by ha, a E
[0, 21T)n.
Certainly, for a fixed metric g, the eigenvalues of the lifted Laplacian depend only
on the chosen equivalence class and are independent on the representant h a .
For the physical interpretation and applications the self-adjoint operator 6 hgu , its
(discrete) spectrum and the degeneracy of the eigenvalues have to be known. Here, we
discuss in Lfoc(Rn, dJ1)
(16)
for a given E E R, and restrict the solutions to Rn/~ by (1). For the solutions, we use
as an ansatz
IY (Xl, ... , xn) = exp (i t fi(Xj) + ih", (Xl, ... , xn) ) , (17)
j=O
From the restriction of Rn to Rn I~, the translational symmetry of Tn (see (4) we have
Hence, the eigenvalue problem has, for m = (m1, . .. , m n) E zn. the solutions
with eigensections
(21)
317
'/'2
0
'/'1
2
11,
___ - l
~,/,2
'/'1 '/'1
Figure 2. Identification of the Klein bottle and its covering by the 2-dimensional torus.
One can show, that the linear hull of the eigensections is a dense subset of L2(Tn, dll)'
Hence this is a complete set of solutions.
The eigenvalues Em depend only on the perimeters ai of the torus and the param-
eters a1, ... , an which classify unitarily inequivalent quantization mappings; they are
quantization parameters. They are independent on the choice of hex within an equiv-
alence class. However, the fiat connection occurs explicitly in the expression for the
eigensections.
In connection with the degeneracy of spec (D,~a) with the metric gij = IJij , we note
that if one of the quantization parameters ai vanishes or equals 7r, this leads to a degen-
eracy of the eigenvalues and doubles the dimension of the eigenspaces. The dimension
of the eigenspaces for arbitrary choices of the perimeters and quantization parameters
is not at hand. However, for a generic choice of the perimeters, such degeneracies will
not occur.
Topology of K2
l('!'l) E [0,27r)
(25)
[('!'2) E {1} ~ Z2
318
and the group of characters of K2 is
(26)
71";(K2) classifies line bundles with flat connections. Elements in Z2 give rise to two
non-isomorphic line bundles TJ given by the identifications
On each of the line bundle exist [0; 271") unitarily inequivalent flat connections parame-
terized by a E [0,271") and associated with functions hOI : R -+ R
(28)
A useful element of each class is hex = aXI, a E [0,271") fixed.
The relation between the equivalence classes of flat connections and different choices
of function hOI is similar as in (14). They yield on each of the complex line bundle
different Laplacians in the same manner as before. As hOI now depends only on Xl, the
formal expression for the lifted Laplacian is given by
(29)
Its form is independent on the chosen line bundle, but the domains differ for the two
line bundles.
(30)
amI ,m2 and aml ,-m2 are solutions of the eigenvalue problem on the tours with the same
eigenvalue. 3
a
Under the symmetry of the Klein bottle the functions ml ,m2 transform according
to
aml ,m2(al + xl,aZ - Xz) a
(_l)ml m ,,-m2(XllX2)
(31)
am1 ,m2(XI, a2 + X2) (_l)ml aml ,m2(Xl, xz).
3 Notice, that the first term in the exponent now reads 2;:":1:1, as the covering torus has a
perimeter of double the size of the Klein bottle.
319
The solutions (j on the line bundle 'Tj have to satisfy (27). We construct the solutions
by linear combinations in the degeneracy spaces of the eigenvalues on T2 and get on 'Tj
(32)
(33)
In both cases, the eigensections constitute a dense subset of sections in L2( 'Tj, dfJ,). The
degeneracy can be discussed similarly as in the case of the torus.
Topology of RP2
The projective sphere Rp2 is a non-orient able manifold that can be constructed as
a subset of the 2-dimensional sphere 52 by identifying diametral points. It appears in
a physical context as the configuration space of two identical particles in R3 which is
given by
A natural choice of the metric on Rp2 in this context is the restriction of the metric on
R3 to Rp2 as a subset of 52. We will briefly discuss this geometry at the end of this
section.
An alternative construction of RP2 as a subset of R2 is given by the equivalence
relations
(x!, X2) (0,1 + XI, 0,2 - X2)
(35)
(XI, X2) (0,1 - XI, 0,2 + X2).
U sing this construction, we are able to compare the results obtained for the mani-
folds T2, (2 and Rp2, which will be done in the section "Conclusion".
As is indicated by the dashed box in Figure 3, the Klein bottle is a two-fold covering
of Rp2 and the torus is, via the covering of the Klein bottle, a four-fold covering of
RP2.
The first fundamental group of Rp2 and its characters are
(36)
320
Figure 3. Identification of Rp2 and its covering by the Klein bottle.
i. e. there are two equivalence classes of line bundles with fiat connections. As in the
case of the Klein bottle, the elements in Z2 give rise to two non-isomorphic line bundles
and each of them allows exactly one fiat connection. The function hex in the case of f{2
and Tn is zero. The following equivalence relations characterize the line bundles TJ:
(37)
As there is only one fiat connection on each line bundle, the formal expression for
the Laplacian 6 g is the usual one in R2 with a dense domain of functions invariant
under (38). Since the torus (with perimeters 2al and 2a2) covers the projective sphere,
the solutions of the eigenvalue problem on Rp2 are elements in the four-dimensional
spaces spanned by the degenerate eigensections ere1 m, ,<2 m 2 on TJ(T 2 ) (Ej E {l}, mj E Z,
j = 1, 2) where
~ (
a m },m2 xI,X2 )= expz. (7rml 7rm2) .
--Xl + --X2 (39)
al a2
The eigensections a of 6 g on L 2 (TJ,dp), i. e. with the symmetry (38), are
( 40)
RP2 as a Subset of 8 2
321
both constructions, the Laplacian is also different. The two complex line bundles T/
on Rp2 are given by the identifications
on the line bundle over 52. The natural choice of the metric on 52 is the restriction
of the metric on R3 to 52 and the Laplace Beltrami operator on 52 is described by its
angular components in spherical coordinates, i. e.
(43)
The eigenfunctions are given by the spherical harmonics Yzm with eigenvalues
(44)
The eigensections on T/+ (resp. T/-) are related to the symmetric (resp. antisymmet-
ric) functions on 52 and the spherical harmonics are symmetric for l even and anti-
symmetric for 1 odd.
A comparison with (41) shows the influence of the metric.
CONCLUSION
For a generic choice of the perimeters and quantization parameters, the eigenvalues are
not degenerate.
For the quantization parameters equal to 0, we choose the perimeters such that
b1_- a1
-
2
(45)
For these non-generic parameter values the spectra of the Laplacians on the three
discussed manifolds are the same; hence a measurement of the eigenvalues cannot give
any information on the underlying configuration space.
322
To determine the configuration space from a Laplacian it is necessary to know its
eigensections and the degeneracy: on the projective sphere there is no degeneracy, on
the Klein bottle a two-fold and on the torus a four-old degeneracy.
It is possible by an appropriate choice of the coefficients that the eigenvalues of the
torus and the Klein bottle coincide.
Whereas the eigenvalues are independent on the quantization parameter , which
classifies the non-isomorphic line bundles, the eigensections depend on it. In partic-
ular, if the eigensections are symmetric, one has the +-quantization, if they are anti-
symmetric, one has the --quantization.
ACKNOWLEDGEMENTS
I am very grateful to H.-D. Doebner and J. Hennig for valuable discussions. I would
like to thank the Studienstiftung des deutschen Volkes for financial support.
REFERENCES
[1] H.-D. Doebner and G. A. Goldin, Phys. Lett. A162 (1992), 397.
[2] H.-D. Doebner and G. A. Goldin, J. Phys. A: Math. Gen. 27 (1994), 177l.
[3] B. Kostant, in: Lecture Notes in Mathematics 170, Springer, Berlin 1970, 87.
[4] H.-D. Doebncr and J. Hennig, in: Symmetries in Science VIII, B. Gruber (ed.),
Plenum Press, New York, 1995, 85.
[5] C. Schulte, The Proceedings of the VIIth International Conference on Symmetry
Methods in Physics 2, A. N. Sissakian and G. Pogosyan (eds.), Publishing Depart-
ment JINR, Dubna, 1996, 487.
[6] M. J. Gotay, in: Quantization, Coherent States and Complex Structures, J.-P. An-
toine, S. Twareque Ali, W. Lisiecki, 1. M. Mladenov, and A. Odzijewicz, (eds.),
Plenum Press, New York, 1995, 55.
[7] M. J. Gotay, H. Grundland, and C. A. Hurst, Trans. Amer. Math. Soc. 348 (1996),
1579.
[8] N. Woodhouse, Geometric Quantization, Clarendon Press, Oxford, 1980.
[9] B. Angermann, H.-D. Doebner, and J. Tolar, in: Lecture Notes in Mathematics
1037, Springer, Berlin, 1983, 17l.
[10] R. S. Palais, Proc. Amer. Math. Soc. 12 (1961), 50.
323
APPLICATION OF THE SUPERSYMMETRIC MODEL
TO EXOTIC OXYGEN NUCLEI 1
Institute of Physics
M. Curie Sklodowska University
20-031 Lublin, Poland
INTRODUCTION
1 The work is partly supported by Polish Committee for Scientific Research (KBN) under Contract
No.2 P03B 19909.
THE SUPERMULTIPLET N =5
The supermultiplet which has been considered in [3] has the following members:
E=E~ + Ar(r+3)+BL(L+1)+(-1)Tfaolj(Ij+1)+
+ ,8J(J + 1) + ,Tj(Tj + 1). (2)
The last two terms of (1), being constant for a given nucleus, have been put in Eb. The
adjusted phenomenological parameters are given in Table 1.
In the excited energy calculation relative to the ground energy there is only the
combination llao - , which enters the calculation.
The Eb involves the core contribution C6 0) as well as the Casimir invariant con-
tributions for such groups whose only one irreducible representation enters the calcula-
tions. For example, for the nucleus i~Mg13 there are one fermion and four bosons and
these numbers are kept constant for the nucleus. Hence, the single particle energies of
fermions and bosons do not enter the excited energy calculation. For the same reason
the contribution of the total isospin term (E) does not appear in the relative energy
calculation (2). However, in the calculation of the total binding energies of the different
supermultiplet members they have to be taken explicitly into account. Hence, in the
binding energy calculation
326
where Eo is the binding energy of the core 16 0 (with some small contribution of the
other constant terms) and a and b have a meaning of the single particle energies of
fermions and bosons, respectively. The structure of (3) has to be taken in (2).
Because the values of Go and 'Y taken separately do not spoil the excited energy
calculation unless we change the relation of Table 1, we assume, for simplicity and to
avoid one free parameter:
Let us notice, that for mirror nuclei the EB - Eo is practioally the same. We
precisely fix the values:
327
Following our programme we adopt the fixed N = 5 supermultiplet parameters to
the exotic oxygen nucleus 26 0 which belongs to the supermultiplet. Now we have for
the ground state of 26 0:
T = L = J = Tj = 0 T=5
and
E6 = Eo + 5b + 30c = -184.31 (MeV)
Let us compare the calculated value of binding energy of 26 0 with - from one side
- the experimental value of 24 0 and - from the other side - with one of the last
theoretical prediction for 26 0 [6J obtained in the relativistic mean field approach.
The numbers show that in the relativistic mean field approach the nucleus 26 0 seems
to be stable against the neutron pair emission while in the supersymmetric model it is
not bound.
OTHER SUPERMULTIPLETS
Let us make another exercise taking the other supermultiplets with different oxygen
isotopes. For simplicity we take only even-even nuclei for which the binding energy
simply reads
E = Eo + bNb + cT(T + 1). (ll)
After detailed consideration we have found that for the supermultiplets N = 2,3,4 and
6 the parameter c remains almost the same as for N = 5 i.e. c :;::j 2.41 MeV while the
parameter b changes for different supermultiplets which is a natural behaviour in the
symmetery consideration. However, the change of b is very regular, namely
which is rather compact with the supersymmetry assumption. It means that the super-
symmetric model with only one free parameter, flb, is able to reproduce the binding
energies of the even-even nuclei belonging to the supermultiplets N = 2,3,4 and 6.
In Table 3 we show the values of band c and in Table 4 we compare the experimental
and calculated binding energies for nuclei under consideration.
The values of the parameter b show quite reasonable single-boson energies and the
positive value of the parameter c reflects the destructive action of the isotopic term in
(11).
N 2 3 4 5 6
b -20.34 -20.94 -21:54 -22.14 -22.74
f 2.41
328
Table 4. Comparison of the experimental and calculated binding energies for the
even-even nuclei of the supermultiplets N =2,3,4 and 6
N == 2 N=3
~o012 ~gNelo ~2014 i6 Ne 12 gMg 10
(EB - Ec)exp -169.66 -186.56 -180.16 -203.68 -203.68
(EB - EC)th -172.13 -186.59 -179.81 -203.91 -203.91
N == 4 N == 6
~6Ne14 r!Si 1O ~iMg12 ~4016 i~Ne18 i~Mg16 r~Si14 ~8020
(EB - Ec)th -217.61 -217.61 -232.07 -183.87 -234.15 -267.83 -282.35 -181.13
2.8
2.6
c
2.4
2.2
2 3 4 5 6
N
Figure 1. The small changes of the parameter f versus the supermultiplet number N.
0
24 26 0 280
Eexp -168.48 ? ?
Etheor (this work) -168 -166 -163
Etheor [5] - -168.43 -
329
Our supersymmetric calculations are expressed only in integer numbers because of
necessary approximations of the model. However, it is clearly seen, that our predictions
put both of "exotic" isotopes A = 26 and A = 28 as unbound.
In the evaluation [5], based on systematic trends, the isotope 26 0 is unbound but only
by 50 keY. The extensive shell model calculations with so called WBMB interaction [8,
9] and the latest results of droplet model (FRDM) [7] predict 26 0 to be stable. However,
the more recent shell model hamiltonian (SPODATA) [10] gives this isotope as unbound
but again only by 20 keV (after [11]). It means that from theoretical calculations one
could not get clear conclusion about the isotope 26 0.
Just after preparation of the present contribution to the Bregenz Conference, it has
been brought to our attention the last experimental results concerning the isotope 26 0
[11]. The authors state that the heaviest bound oxygen isotope is 24 0 and the strong
statement has been given, as they wrote: with an increase in the confidence level of
more than 1 order of magnitude. It is in accord with our supersymmetric calculations.
The line of this approach is going to be extended toward calculations of binding
energies of the full sd-shell nuclei. It demands both: considering other supermultiplets
with inclussion of even-odd and odd-odd nuclei, and also improving numerical results
by detailed discussion of other, although very small, terms which are still present in
the Eo but which are not coming from the core but from the valence nucleons through
some Casimir eigenvalues.
Such a work is now under progress.
ACKNOWLEDGMENT
We have really appreciated the Professor Klaus Dietrich suggestion to perform the
present work. The final version of this contribution has been prepared during the stay of
one of the authors (S.S.) in the Max Planck Institut fiir Kernphysik in Heidelberg. The
author (S.S.) is very grateful to Professor H. Weidenmiiller for his invitation and also
to both: Technische Universitiit in Miinchen and Max Planck Institut fiir Kernphysik
in Heidelberg for the support.
REFERENCES
[1] S. Szpikowski, P. Klosowski and L. Prochniak, Nucl. Phys. A487 (1188), 301;
L. Prochniak and S. Szpikowski, Acta Phys. Pol. B24 (1993), 557.
[2] S. Szpikowski, P. Klosowski and L. Prochniak, Z. Phys. A335 (1990), 289.
[3] S. Szpikowski, Y. S. Ling and 1. Prochniak, Acta Phys. Pol. B26 (1995), 1403.
[4] P. W. M. Glaudemans, G. Wiechers and P. J. Brussard, Nucl. Phys. 5,6 (1964), 529.
[5] G. Audi and A. H. Wapstra, Nucl. Phys. 565 (1993), l.
[6] Zhongzou Ren et al., Phys. Rev. C52 (1995), R20.
[7] P. Moller, J. R. Nix, W. D. Myers and W. J. Swiatecki, At. Data Nucl. Data Tables
59 (1995), 185.
[8] B. H. Wildenthal, Prog. Part. Nucl. Phys. 11 (1984), 5.
[9] E. K. Warburton, J. A. Becker and B. A. Brown, Phys. Rev. C 41 (1990), 1147.
[10] B. A. Brown, W. A. Richter, R. E. Julies and B. H. Wildenthal, Ann . Phys. (N.Y.)
182 (1988), 191.
[11] M. Fauerbach et al., Phys. Rev. C 53 (1996), 647.
330
q-QUANTUM MECHANICS ON SI
Reidun Twarock
INTRODUCTION
The purpose of this presentation is to construct a discrete quantum mechanics in
which differential quotients appear instead of differentials, because these can be mea-
sured directly in experiments. This means to derive a theory which is based exclusively
on the following objects (Xi = Xi(X), i = 1, 2, X E R);
D.j ;= I(Xl) - I(X2) (1)
Xl - X2
where real Coo-functions 1 should be defined on a discrete set. We want this set to be a
discretization of a manifold M with non-trivial topology (e. g., with non-vanishing first
fundamental group), because in this case, the structure of the position and momentum
operators given by Borel quantization and the corresponding dynamics derived from it
is richer, i. e., it contains not only an additional constant but also a topological potential
which is related to the topology of M.
Here, we consider the compact Sl as the simplest example of such a configuration
manifold M and the N -point discretization SJ.r given as embedding in the complex plane
C by z = exp(i, > E [0,211') and, respectively, the N-th roots of unity z/ ;= exp (2;/1),
1 = 0, ... , N - 1. The natural choice for differential quotients (1) in this setting are
q-derivatives
(2)
where q is an N-th root of unity, because then qz/ and q-1z{are again in SJ.r. (For
notational convenience we will drop the subscript I in the following.) One can express
Dq also in terms of the homogeneity operator N z ;= zOz as
(3)
if we use the following q-number notation
(4)
(5)
Since differentials on 8 1 , which correspond to the usual parameterization of 8 1 by the
angle </; E [0,211"), can be expressed in terms of the coordinate z = exp( i</;) as
(6)
differential quotients [Nz ], i.e., discretizations of N z via q-deformation, are appropriate
objects for the formulation of a discrete quantum mechanics.
Based on similar considerations, q-quantum mechanics on the circle has been derived
by [1], but without the use of a particular quantization method which takes the topology
of the underlying configuration space into account.
The difficulty in the construction of a q-quantum mechanics stems from the fact that
there is no unique prescription on how to replace differentials by q-derivatives. Different
methods lead to completely different evolution equations and thus to different physics.
In this ansatz, we take the point of view that a mere substitution of differential" by
q-derivatives is not enough, but that the introduction of q-derivatives should be linked
to a corresponding choice or deformation in the sense of Gerstenhaber [2], denoted as
a (special) q-deformation of a group or an algebra structure, e. g., as presented here to
a deformation of the kinematical algebra of a system localized on 8 1
(7)
which is given in terms of COO-functions and of vector fields on 8 1 . In addition, some
physically or mathematically motivated assumptions which we will call q-assumpfions
will be necessary. We note that there is an approach for a derivation of a q-Schrodinger
equation based on a q-deformation of the Schrodinger algebra [3]; this result is valid only
for the topologically trivial configuration space Rn and not connected to a quantization
procedure, especially, it gives no information on generalized position and momentum
observables, i. e., on the quantum kinematics.
In this section we show how the quantum kinematics can be derived from Borel
quantization on 8 1 , i. e., the representation theory of (7). We first sketch the un-
deformed case and then tailor the corresponding q-deformed version along these lines
in the second subsection the dynamics is discussed in a second step in the following
section.
Borel quantization for the kinematics has been introduced for general configuration
spaces in [4] and applied to 8 1 in [5]. The position and momentum operators are given
as a representation of 5($1) in the Hilbert space 1-( = L2(8t,d</;) (</; E [0,211")) by
332
o yield unitarily representations. 0 comes in because 51 is topologically non-trivial
(7r1 (51) = Z). The reality of D guarantees that P is selfadjoint on a suitable dense set.
If the functions f( rp) and the vector fields X = X( rp) 14" where f( rp), X( rp) E c oo (5 1 , R),
are expressed by a Fourier expansion
00
(9)
=
n:::::-oo
n=-oo
(10)
n=-oo
with
zn (z~ + ~ + 0)
dz 2
(ll)
[Tm,Tnl o
[Ln,Tml (12)
[Lm' Lnl
i. e., the operators L~ are realizations of the Witt algebra. Otherwise, we find
[Tm,Tnl o (13)
[L.~, Tm] (14)
--I. 0 n02 - mOl
n r m, 3 =
(15)
n-m
n=m
The infinite dimensional Lie algebra defined in (15) may be called a generalized Witt
algebra.
333
counter part. q-assumption (q2) guarantees that no new parameters occur which are
not algebraically motivated, i. e., we only allow new parameters which an:) necessary in
order to close the algebra to a Lie algebra.
If the notion of the derivative is changed, it is reasonable to change also the notion
of the integral which realized the inner product
The assumption (q2) is fulfilled, because the extra parameter j is necessary to close
the q-deformed Witt generators };:,B) to a Lie algebra, as can be seen frqm the commu-
tation relations below:
(18)
with
jl Bl + hB2
jl + j2
(19)
jl Bl - j2 B2
)1 - J2
(20)
However, it does not fulfill (q1) and (q2) and is thus not appropriate for our purpose.
Since the operators Tn do not contain derivatives, they remain undeformed. The
first relation in (12) is thus not changed and the second relation becomes (using (17))
(22)
[ (j,B)
m,n
T] = Tn ((j,9+
m
n) _ (j,B))
m' (23)
334
The q-deformation of {Tn}EE{L n } is thus an infinite dimensional vector space spanned
by );{,8) and Tn and is closed as a quadratic algebra.
The construction of the above operators is the first step in the derivation of a q-
quantum Borel kinematics and leads to operators of the form
00
n=-oo
00
(24)
n=-oo
(25)
p(X)i
{Tn}
= i (Nz) f
n=-oo
X"DTn. (26)
pg(X)i
{Tn}
= i [NzJ f
n=-oo
Xn DT" =i f
n=-oo
Xn[nJDTn . (27)
L
n=-oo
iSn (= QU))
00
(28)
PJ(X)
n=-oo
DYNAMICS
In the undeformed case, the quantum Borel kinematics (10) can be used to derive
a dynamics via a generalized version of the Ehrenfest relation [7]. On Sl this relation
implies, that a time evolution for pure states 1{; E 'H. = L2(SI, dtjJ) is restricted by the
following equation between matrix elements (Schrodinger representation)
(29)
335
It leads to a Fokker-Planck-type of equation [8] of the form (the prime denoting deriva-
tion with respect to the variable ):
p. -_ - (Jo'8)' + DP,,_
- - ('8), Jo - D'
J , J8 -_'8 P, (30)
where
jg = ~(1f;'1f;-1f;1f;') + 8p (31)
is the quantum mechanical current density. With the ansatz (w.l.o.g.)
which contains a nonlinear term proportional to iD and an arbitrary nonlinear real part
which cannot be derived from (29) without further assumptions.
2: fnTn (= O(f))
n=O
N-l
(34)
Ptd(X) = 2: Xn (L~,8) + i[nJDTn) .
n=O
In particular, the range of the parameter j has been restricted to integers in order to
guarantee that the action of the q-derivatives contained in the L~,8) on functions defined
on the lattice points is well defined, i. e., gives an expression which is given in terms of
lattice points. This condition is q-assumption
(q4) j E Z.
The Hilbert space on Slv is a finite dimensional Hilbert space of sequences W
(1f;(O), .. ;,1f;(N -1)), denoted by H N . Functions are of the form (di,:crete Fourier
transform)
N-l
1 N-l _
(1f;,)d:= NL1f;(l)(l) (36)
1=0
336
with
N-I
m) d = N1
( z/k , z/ '"'" -k m_
~ zi z/ -
C
Ukm . (37)
1=0
In this setting, the q-analog of (29) reads
where D t denotes a time derivative which can be either the usual time derivative at or
a q-deformed time derivative giving at for q -71. Furthermore, Pg,d(Xgrad.f) is given
by
N-I
Pg,d(Xgrad) = L i[nJfn (~,B) + i[nJDTn) . (39)
n=O
The fact that the latter contains [nJ instead of n is again a consequence of q-assumption
(q3).
From (38) one can derive an equation for Dtp [10J which can be viewed as a q-analog
to (30). It is given by
(41)
(40) implies restrictions on the parameters j and ti, i = 1, ... ,4.
Since (j ~] is composed of shifts q~Nz, it is only well-defined in our discrete setting
if we restrict q-assumption (q4) (j E Z) further to
(q4') j E 2Z.
Furthermore, the constants ti, i = 1, ... 4, are a freedom in the theory. They appear
because of (4.12) in [10J. The choice
in (40) is physically motivated by the fact that it implies atp to be real. With this,
also the quantum mechanical current, derived via the relation
(42)
337
is real, and one obtains for the q-analogue j: of the quantum mechanical current density
(31) the expression
) .(q,O)
o
_
-
(43)
which reproduces the current (31) on "5';'''= 51 in the limit q ---; 1 (i. e.,N ---; 00).
In analogy to the construction in the undeformed case, an evolution equation for
W from (40), can be derived. For this one makes the following ansatz (compare with
(32)), where Ht
is linear in Wand G{[0,w] nonlinear in 1/J, 0,
5 and R are shift operators like in (41), which typically occur in a q-deformed theory.
(44) re<!uces to (32) in the limit q ---; 1, if Hg and G~[0,w] are such that they give H
and G[W, W] in this limit.
One can use (44) to express OtP in terms of Hg = HI + iH2 - called linear part -
G~[0, W] = G l + iG 2 (we skip the index j in the following) - called nonlinear part -
5 = 51 T i52 and R = Rl + iR2. For W= WI T iW2 one obtains
(Ot0)w + (Otw)0
i {(Hg0)(Sw) + (G~0)(RW) - (HN)(50) - (G~w)(R0)}
2 {(HIWd(52Wl - 51 W2) + (H2W2)(51 W2 - 5 2Wl)
(45)
+ (HZWl)(51W1 + 521/J2) + (HIW2)(51Wl + 5 2 W2)}
+ 2 {(Gl 1/Jl)(RzWl - R1~)Z) + (Gz1/J2)(Rl1/J2 - R 2 Wl)
+ (Gz~)I)(Rl1/J1 + RZW2) + (GIWz)(RIWI + R2W2)}.
(47)
(48)
have been introduced in order to facilitate notation in the following. Three different
combinations for HI and H2 are possible, which we discuss separately.
Case a: HI = V, H2 =0
338
Here, (46) reduces to
Hence,
W
(52)
2
o. (53)
Case b: HI = 0, H2 = V
With
one has
which implies
o (58)
W
(59)
2
Case c: HI = V, H2 = V
Here, the defining relation is
which leads to
(61)
339
With
which leads for all j in the limit q ~ 1 to the Hamiltonian - ~ d~2 obtained in the
undefortned case. 2
The problem of deducing the nonlinear part G1 + iG2 of (44) can be reduced to an
identity, in which all expressions are given in leading orders of h, where h is introduced
via q := exph such that the limit q ~ 1 corresponds to h ~ O. If the shift R is denoted
by R = Rl + iR z = (aqaNz + bq{fNz ) + i (cq'YNz + dqliNz) with a, b, c, d, a, (3",6 E R,
then this identity is given by
(64)
with
340
From (64) and (65) itis clear that only the combination
Al
(68)
2BI
Az
(69)
2B2
produces well-defined expressions in the limit h ---'> 0 or q ---'> 1, respectively. The
corresponding limits are given by
(75)
(76)
(77)
341
with the limits
Dp//
(78)
2p(c + d)
c2j (1/;// If;' - {;//1/;')
(79)
8( c"( + dEi) (1/;1/;' - 1/;1/;') .
Together with R2 -+ i( c + d) for q -+ lone finds for the nonlinearity
As before, the imaginary part coincides with the usual one, whereas the real part looks
- up to the fore factor - like the one obtained in case a.
(81 )
Case e: R I , R2 I' 0
In this case both BI and B2 are of order 0(1) and one has
This allows several possibilities to derive G I and G 2 from (64). To treat the most
general case, we introduce the following partitions of unity
h+12 (84)
PI + P2 (85)
and split (64) such that
(86)
(87)
holds. One deduces that in the limit q -+ 1 (notice that thecontribution of Al vanishes)
PIDp"
(88)
2(c + d)p
P2Dp"
(89)
2(a+b)p
342
With R = Rl + iR2 = (a + b) + i( c + d) for q ~ lone obtains the following expression
for the nonlinearity in the limit q ~ 1:
(90)
Also in this case, the usual expression for the imaginary part is obtained. The real part
is equal to the imaginary part up to a factor which depends on the coefficients a, b, c
and d as well as the partition chosen above. For the special cases Rl = R 2, PI = 1
and P2 = 0 and Rl = -R2' PI = 0 and P2 = 1 real and imaginary part coincide. It
is interesting here, that the real part lies in the Doebner-Goldin family. This result is
always obtained when the shift R in (44) is chosen maximal in the sense that Rl =I 0
and R2 =I o.
The q-Schrodinger Equation on 81
Summarizing the result, the following statements can be made:
In all cases, the limit of the imaginary part of the nonlinearity coincides with the one
derived without q-deformation. In addition, three types of real part occur in dependence
on the shift R:
1.
'IjJ"'0'_'0I'IjJ'
A -'---'=----'=---'- (91 )
'IjJ'IjJ'-'ljJ'IjJ' ,
. c2j(a + b) = 0 and A = -c2j(c+d)
wIth A = 8( aCt + b(3) 8(c'Y+ dS)
A = 1/2 holds.
2.
Dp"
B-
2p , (92)
. 1 h 1 . f h . h B (a + b) (c + d)
1. e., equa to t e lmlt 0 t e lmagmary part, WIt = (c + d) PI - (a + b) P2
3. For R = 1 (trivial shift operator) the real part remains undetermined by (40) like
in the un deformed case.
+FNL ,
where FNL is given by (66). For a discussion of the q-Schrodinger equation, also with
respect to a possible q-deformation of the time-derivative, see [10J.
343
CONCLUSION
ACKNOWLEDGEMENTS
The presented results have been obtained in joint work with H.-D. Doebner (Claus-
thal) and V. K. Dobrev (Sofia). I am grateful to the Studienstiftung des deutschen
Volkes for financial support.
REFERENCES
344
LIST OF PARTICIPANTS
345
Juan Garcia Escudero Romuald Lenczewski
Departamento de Fisica Institute of Mathematics
Universidad de Oviedo Technical University of Wrodaw
E-33007 Oviedo Wybrzeze Wyspianskiego 27
Spain PL-50-370 Wrodaw
and Poland
Departamento de Fisica
Universidad Europea-CEES
Campus Universitario Amiram Leviatan
E-28670 V. de Od6n, Madrid Racah Institute of Physics
Spain The Hebrew University
Jerusalem 91904
Gerald A. Goldin Israel
Departments of Mathematics and
Physics
Rutgers University Miguel Lorente
Busch Campus Departamento de Fisica
Piscataway, N J 08855-11 79 Facultad de Ciencias
USA Universidad de Oviedo
Calvo Sotelo 18
E-33007 Oviedo
Bruno Gruber Spain
Southern Illinois University at Car-
bondale in Niigata
Nakajo
Vladimir I. Man'ko
Niigata 959-26
Lebedev Physical Institute
Japan
Leninsky Prospekt 53
117924 Moscow
Maurice Kibler Russia
Institut de Physique Nucleaire de and
Lyon Osservatorio Astronomico di Capo-
IN2P3-CNRS et Universite Claude dimome
Bernard Via Moiariello 16
43 Boulevard du 11 Novembre 1918 1-80131 Napoli
F-69622 Villeurbanne Cedex Italy
France
Peter Kramer
Marcos Moshinsky
Institut fiir Theoretische Physik
Instituto de Fisica
Universitiit Tiibingen
U.N.A.M., A.P. 20-364
Auf der Morgenstelle 11
01000 Mexico, D.F.
D-72076 Tiibingen
Mexico
Germany
346
Christian Mrugalla Christoph Schulte
Arnold Sommerfeld Institut Institut fur Theoretische Physik
Technische Universitiit Clausthal Technische Universitiit Clausthal
Leibnizstrafie 10 Leibnizstrafie 10
D-38678 Clausthal-Zellerfeld D-38678 Clausthal-Zellerfeld
Germany Germany
Yu. F. Smirnov
Bernhard Muhlherr
Institute for Nuclear Physics
Fachbereich Mathematik
Moscow State University
Universitiit Dortmund
Moscow 119899
Vogelpothsweg 87
Russia
D-44221 Dortmund
Germany
Stanislaw Szpikowsky
Institute of Physics
Peter Nattermann M. Curie-Sklodowska University
Institut fur Theoretische Physik PL-20-031 Lublin
Technische Universitiit Clausthal Poland
Leibnizstrafie 10
D-38678 Clausthal-Zellerfeld
Germany Reidun Twarock
Arnold Sommerfeld Institut
Technische Universitiit Clausthal
Seiya Nishiyama Leibnizstrafie 10
Centro de Fisica Teorica D-38678 Clausthal-Zellerfeld
Universidade de Coimbra Germany
P-3000-Coimbra
Portugal Gockhan Unel
and Department of Physics
Department of Physics Bogazi<;i University
Kochi University and
Kochi 780 Centre for Turkish-Balkan Physics
Japan Research and Applications
istanbul
Turkey
Mariano A. del Olmo
Departamento de Fisica Teorica
Universidad de Valladolid Sujeewa Wickramasekara
E-47011 Valladolid Department of Physics
Spain The University of Texas at Austin
Austin, TX 78712
USA
Michael Ramek
Institut fur Physikalische und Theo- Nihan Yenigul
retische Chemie Fizik Bolumu
Technische Universitiit Graz Fen Fakultesi
Brockmanngasse 27 Ege Universitesi
A-80lO Graz TR-35100 Bornova-Izmir
Austria Turkey
347
Serif Yenigiil
Matematik Bolumu
Fen Fakiiltesi
Ege Universitesi
TR-35100 Bornova-Izmir
Turkey
348
CONTRIBUTORS
J. M. Arias 99
M. Arik 1
U. Beck 179
S. 1. Ben-Abraham 179
W. Berej 325
R. Bijker 9, 99, 193
E. Binz 25
J. Cseh 37
V. K. Dobrev 47
H.-D. Doebner 81
1. H. Duru 93
A. Frank 99
D. V. Fursa 117
J. Garcia Escudero 139
J.-D. Hennig 81
M. Kibler 153
J. R. Klauder 167
P. Kramer 179
R. Lemus 99
A. Leviatan 9, 193
M. Lorente 211
V. 1. Man'ka 225
M. Moshinsky 243
C. Mrugalla 251
B. Miihlherr 263
P. Nattermann 269
S. Nishiyama 281
P. Oellers 25
M. A. del Olma 297
F. Perez-Bernal 99
L. Prochniak 325
C. Schulte 313
Yu. F. Smirnov 117, 243
S. Szpikowski 325
V. N. Tolstoy 117
R. Twarock 331
G. Dnel 1
349
INDEX
351
Dynamics Hurwitz transformations, 156-164
Riemannian, 86
quantum, 86 Icosahedral symmetry, 144
Dyson realizations, 117 Inhomogeneous Lie algebra, 298
Integral linear group, 185
Ehrenfest relation, 87-88. 313
Einstein equation, 25 Jacobi vector, 194
Enveloping algebra, 298, 308 Kac-Moody algebra, 48, 118
Equation of motion, 281-295 Kazhdan-Lustig polynomials, 47-79
Euclidian Lie algebra, 253 Kepler problem, 164
Euclidian Weyl- Lie algebra, 253 Kepler-like potential, 94-96
Euclidian algebra, 298, 302-303 Killing vectorfields, 84
Euclidian group, 300 Kinematical algebra, 269, 298, 302
Exotic nuclei, 325-330 Kinematical groups, 298
Klein bottle, 31~:-323
Factor group, 182 Klein-Gordon equation, 25
Factor-module, 48 Kustaanheimo-Stiefel transformation, 164
Faddeev-Popov determinant, 169
Fermi type excitation, 281-293 Langrange multipliers, 3, 167-175,284-292
Fermion Levi-Civita transformations, 157
Lie operators. 285-294 Lie algebra, 84, 251-260, 281-293, 3:14
system, 281-293 constraint, 160
Fiber hllndles, 153, 179-190 inhomogeneous, 298
Fibonacci sequences, 140 q-deformation, 1
Fokker-Planck equation, 336-343 semisimple, 47-69, 298-300
Form factor, 18-19 simple, 48 69
of nucleons, 195-1\)\) symplectic, 160
Free group, 182-185 Lie bialgebra, 300-301
Lie group, 181
Giirsey Radicati form, 195 Lie operators, fermion, 285-294
Galilean Lie algebra, 253 Lie superalgebra 117-,118, 126
Galilei algebra, 298, 302-303 Line bundle, 31::-323
Galilei group, 84, 300 Lorentz group, 82, 219
Gauge transformation, 270-275 Lorentz manifold, 27
Generators Lorentz metric, 25-28
Cartan-Weyl,51 Lorentz representation, 71-76
Chevalley, 51 Lorentz transformation, 215-222
Gleason's theorem, 273 Lowest weight module 49-69
Golden section, 140
Grassmann algebra, 120-122 Magnon,306
Grassmann envelope, 121-122 Maxwell equation, 75, 218
Grassmann variable, 169 Maxwell field, 75
Greens function, 94-96, 234-238 Methane, 107-113
Minkowski lattice, 219
Minkowski space-time, 71,218
Hi, 106
Molecular dynamical group, 107
Hadron, 203, 205
Molecular spectroscopy, 99-114
Hamiltonian functional, 83
Morse oscillator, 100-104
Hamiltonian operator, time-dependent, 170
Morse potential, 93
Harmonic oscillator, 38,93,100-104,165,
Motion group, ~m-272
226-240, 254
quark model, 206 Nat, 106
Hartree-Bogoliubov equation, 281-293 Non-commutative space groups, 184-190
Hecke algebra. 57 Nonbijective transformations, 153-165
Heisenberg ferromagnet, 305 Nuclear states. 41
Heisenberg-Weyl superalgebra, 126 Nucleon
Helicity amplitudes, 200-203 resonances, 19
Hodge star operator, 32 form factor, 195-199
Hopf fibration, 153-159
Hopf structure, 300-302 0(3), 41
Hurwitz theorem, 162 Oblate symmetric top, 13
352
Oblate top, 9 Riemannian dynamics, 86
Observables, 86, 226, 273 Robinson decomposition, 141-142
nonlinear, 276-277 Rovibrational states of molecules, 9
primitive, 271
Octonions, 154 Schrodinger algebra, 251-253, 332
Operators, generalized, 86 Schriidinger equation, 93, 164, 165, 272,
Oscillator, 1-7 304-307
Coon-Yu-Baker,6 free, 252-260, 309
Dirac, 243-249 nonlinear, 82, 313
Morse, 100-104 Schriidinger group, 251-253
Posz-Woronowicz,7 Schriidinger operator, 251-260
harmonic, 38, 93, 100-104, 165, 226-240 Schriidinger representation, 335
Oxygen, 325-330 Schur lemma, 304
Ozone, 112 Semisimple Lie algebra, 117,298-300
Singular vector, 48, 255-256
Pattern SL(4),53-79
Penrose, 141-149 SLq(2), 307
Pauli-Lubanski vector, 303 SO(1,l), 6
Pauli-blocking effect, 41 SO(2), 6, 101
Penrose pattern, 141-149 SO(2,2),4
Phonon, 306 SO(3), 6
Pisot-Vijayaraghavan number, 148 SO(3,1),4
Pochhammer polynomial, 257 SO(3,2), 4, 6, 9
Poincare algebra, 298, 302-307 SO(4),4
Poincare group, 219-220, 300 SO(4,1), 4,6
Posz-Woronowicz q-oscillator, 7 SU(2), 10
Potential SU(2,2),71-77
Kepler-like, 94-96 SU(3), 39, 40
Morse, 93 Space-time gronps, 297-310
Propagator, 171,234-239 Spectrum generating algebra, 9
Spin-flavor symmetry, 194-207
Q-Schrodinger equation, 332 344 Spinor, 75
Q-conformal algebra, 57 Structures
Q-deformations, 93-97 aperiodic, 140
Q-oscillator, 1-7 quasiperiodic, 149
algebra, 6 Subsingular vector, 48-69
Q-quantum mechanics, 331-344 Superalgebra, Heisenberg-Weyl, 126
Quantization Supermultiplet, 326-329
Borel, 83-91 Supersymmetry, 325-330
bosonic, 21 dynamic, 44
constrained systems. 167-178 Symmetry
Quantum Borel kinematics, 84 adaption, 99-108
Quantum algebra, 48 dynamic, 37-44
Quantum dynamics, 86, 227-228 icosahedral, 144
Quantum groups, 1-7,297-308 spin-flavor, 194-207
Quantum kinematics, 332 Symplectic Lie algebra, 160
Quantum mechanics
generalized, 270 Tiling theory, 139-149
nonlinear, 81 Time-dependent Hamiltonian, 170
Quantum transitions, 228-229 Transformation
Quark models, 193-209 Bogoliubov, 282
Quasicrystals, 139-149, 179 gauge, 270-275
Quasiperiodic structures, 149 Hurwitz, 156-164
Quaternions, 154 Kustaanheimo-Stiefel, 164
Levi-Civita, 157
Realization Lorentz, 21.5-222
boson, 126 nonbijective, 153-165
Boson-fermion, 1l7, 129
Dyson, 117 U(2), 100-107, 247
Reflection group, 212- 214 U(3), 39,41, 133
Representation, Bargmann, 126 U(4), 9, 243, 247
353
U(6), 9, 11 Vibron model, 9, 39, 40, 100
U(7), 9, 11, 209 Von Neumann equation, 82, 90
Universal enveloping algebra, 119
Usui operator, 126-132 W-algebra,48
Weyl equations, 75
Vector coherent state theory, 118 Wigner function, 226-233
Verma module, 47-68,117,119,254-260 Wigner's spin-isospin group, 39
Vibrational spectra, 99-114 WigIier-Eckhart theorem, 248
354