SYMMETRIES

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

SPRINGER SCIENCE+BUSINESS MEDIA, LLC

 Library of Congress Cataloging-in-Publication Data
On file

Proceedings of a Symposium on Symmetries in Science IX, held August 6-10, 1996,

in Bregenz, Austria
ISBN 978-1-4613-7715-3 ISBN 978-1-4615-5921-4 (eBook)
DOI 10.1007/978-1-4615-5921-4

1997 by Springer Science+Business Media New York

Origina11y published by Plenum Press New York in 1997
Softcover reprint of the hardcover 1st edition 1997

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.

The Symposium was sponsored by

Southern Illinois University at Carbondale,
the Land Vorarlberg, Austria, and
the Landeshauptstadt Bregenz, Vorarlberg, Austria.

I wish to thank Chancellor Emeritus John C. Guyon of Southern Illinois Univer-

sity at Carbondale, Dr. Hubert Regner of the Amt der Vorarlberger Landesregierung,
and Biirgermeister and Landtagspriisident Dipl.-Vw. Siegfried Gasser for financial as-
sistence for the meeting and Frater Albin Printschler of the Cloister Mehrerau for his
continued support of the Symposium series.

Bruno Gruber

v
CONTENTS

Classical vs. Quantum Groups as Symmetries of Quantized Systems 1

M. Arik and C. Vnel

Algebraic Model of an Oblate Top 9

R. Bijker and A. Leviatan

The Mass-Squared Operator and the Einstein-Hilbert Action for Rescaled

Lorentz Metrics ..................... 25
E. Binz and P. Oellers

Multichannel Dynamic Symmetry 37

J. Cseh

Kazhdan-Lusztig Polynomials, Subsingular Vectors and Conditionally

Invariant (q-Deformed) Equations ........... . 47
V. K. Dobrev

On a Path to Nonlinear QuantuIll Mechanics 81

H.-D. Doebner and J.-D. Hennig

Quantum Mechanical Problems with q-Deformations and over the p-Adic

Number Fields . . . . . . . . . . . . . . . . . . . 93
1. H. Duru

A Symmetry Adapted Algebraic Approach to Molecular Spectroscopy 99

A. Frank, R. Lemus, R. Bijker, F. Perez-Bernal, and J. M. Arias

Dyson Boson-Fermion Realization of Lie (Super)Algebras 117

D. V. Fursa, Yu. F. Smirnov, and V. N. Tolstoy

Formal Languages for Quasicrystals 139

J. Garda Escudero

vii
On Quadratic arid Nonquadratic Forms: Application to R 2m ----t R 2m-n
Nonbijective Transformations . . . . . . . . . . . . . . 153
M. Kibler

Quantization of Systems with Constraints . . . . . . . . . . . . . . . 167

J. R. Klauder

Automorphisms and Discrete Fiber Bundles 179

P. Kramer, U. Beck, and S. I. Ben-Abraham

Algebraic Approach to Baryon Structure 193

A. Leviatan and R. Bijker

Discrete Reflection Groups and Induced Representations of Poincare Group

on the Lattice . . . . . . . . . . . . . . . . . . . . . . . . 211
M. Lorente

Quantum Mechanics and Classical Probability Theory 225

V. I. Man'ko

Supermultiplet Formulation of the Dirac Oscillator of Arbitrary Spiin 243

M. Moshinsky and Yu. F. Smirnov

Construction of Schrodinger- and q-Schrodinger Invariant Differential

Operators . . . . . . . . . . . . . . . . . . . . . . . 251
C. Mrugalla

Automorphisms of Free Groups and Universal Coxeter Groups 263

B. M iihlherr

Generalized Quantum Mechanics and Nonlinear Gauge Transformations 269

P. N attermann

A Solution of an Extended Hartree-Bogoliubov Equation on the Coset Space

SO(2N+2)/U(N+1) for Unified Description of Bose and Fermi Type
Collective Excitations . . . . . . . . . . . . . . . . . . . . . 281
S. Nishiyama

Quantum Space-Time Groups and Beyond 297

M. A. del Olmo

Quantum Mechanics on the Torus, Klein Bottle and Projective Sphere 313
C. Schulte

Application of the Supersymmetric Model to Exotic Oxygen Nuclei 325

S. Szpikowski, L. Pr6chniak, and W. Berej

viii
q-Quantum Mechanics on Sl 331
R. Twarock

List of Participants 345

Contributors 349

Index . . . 351

ix
CLASSICAL VS. QUANTUM GROUPS AS
SYMMETRIES OF QUANTIZED SYSTEMS

Metin Arik and Gokhan Unel

Bogazi<;i Gniversity
Department of Physics
and
Centre for Turkish-Balkan Physics
Research and Applications
istanbul, Turkey

INTRODUCTION

Quantization of nonlinear systems has been one of the fundamental problems of

quantum mechanics. The idea that classical integrable nonlinear systems should be
solvable after quantization, has lead to the introduction of quantum groups as "gen-
eralized symmetries" of a physical system. Quantum groups [1] can he obtained by
applying the idea of q-deformation either to the Lie algebra or to a matrix represen-
tation of the corresponding classical group. However, as we will show a q-deformation
of the algebra of quantum observables does not necessarily imply a q-deformation of
the symmetry group. To clarify this point, we discuss the quantum group invariant
and the classical group invariant q-oscillators separately in the second section. In the
third section, we consider a specific quadratically constrained model [2J and show that
the classical group invariant q-oscillator can be derived using Dirac bracket (diracket)
quantization [3J with some modifications.

QUANTUM GROUP AND CLASSICAL GROUP INVARIANT q-OSCIL-

LATORS

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

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997
one dimensional simple q-oscillator [4] defined by

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)

It follows that the number of n-particle states is given by

(
n + d -1 ) = (n + l)(n + 2) (n + d - 1)
(8)
d-l 12(d-l)

which counts the number ofindependent components of an nth-rank symmetricaltensor

in d dimensions.

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):

a, b) aia; - q a}ai = Dij (9)

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

a-+ Ma ME U(d) (10)

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.

A SPECIFIC QUADRATIC CONSTRAINED MODEL

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

as primary constraints, and

<P3 = X"P,. ~ 0 (14)
as the secondary constraint. The total Hamiltonian of (12) is made by the primary
constraints only, as it should for a Lorentz-invariant system. The Poisson brackets of
the canonical quantities satisfy

For the classification of these constraints, we evaluate their Poisson brackets and obtain

{<PI, <P2} -4<P3 , (16)

{<PI, <P3} -2(<PI +m 2 ) ,
{<P2, <P3} 2( <P2 + r2) .
The above equations show that we have one first class constraint and two second class
constraints. According to the constraint analysis theory, we should have as many gauge
conditions as first class constraints. Thus we will formulate first and second class
constraints plus the gauge condition needed in the definition of dirackets. We denote
the first class constraint as

(17)
the second class constraints as

-r 2 tan{a)<pI + m 2 cot(a)<p2 ~ 0, (18)

+
_r2 tan( a)<pI - m 2 cot( a)<p2 2mr<p3 ~ 0

Here a is a parameter which describes the 50(2) symmetry associated with the choice
of second class constraints. We choose

X2 == mcos(a)X5 - rsin(a)Ps (19)

as the gauge condition associated with the first class constraint. This gauge condition,
since it only contains P5 and X s , is invariant under 50(3,1) and it is chosen so that
its Poisson bracket with the first class constraint is nonzero, i.e.

(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

If we examine the properties of these quantities under the interchange of i and j, we

see that the left hand side of the first two equations are antisymmetric and the right
hand side of the third equation is symmetric. Before passing from dirackets to quantum
operators, we note that the quantities on the right hand sides of the equations (24) are
classical, i.e. commuting ones. They should be normalized in the conventional way
while passing to the quanturn case. Thus we symmetrize all the products on the right
hand side. After this procedure, the right hand side of the first two equations become
anti symmetric whereas the right hand side of the third equation remains symmetric
under the interchange of i and j. The passage from dirackets to commutators is ac-
complished by replacin?'; the diracket of two momentum or two position variables by
-i times a commutator whereas the diracket of a momentum and a position variable is
replaced by i times a commutator

[Xi,Xj] i2~2([Pi,Xjl+ - [Pj,Xd+) , (25)

[Pi,Pj ] i2~Z([Pi,Xj]+ - [Pj,X;]+) ,

. .
[Pi,Xj ] -i Oij - 2:. 2 ([Xi, X j ]+) - 2~2 ([Pi, Pj )+)

where [A, B]+ == AB + BA. These equations can be simplified by scaling Xi and Pi as

= (J Xi, Pi
Xi Pi = 7i (26)

and choosing (3 such that

1 (32
82m2 = ;:z == k . (27)

By defining the creation and annihilation operators iiI and ai as

Xi +iPi
ai == ~-'--.-::. (28)
V2

5
we obtain that (25) is equivalent to

[d';,d'j] = [iiLiij] =0 , (29)

(1 - k) Ilia} - (1 + k) a}ai = Dij (30)

Since both Ilia; and a;ai are positive semi-definite, the i =j components of (30) are
inconsistent unless k < 1 . This allows us to rescale
_
ai
ai
= "'1- k
(31)

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

H rv XoPs - PoXs . (34)

Using only the constraint equations, we obtain the Hamiltonian of the system as:

(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

where N is defined through the inner product of oscillator states

(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

[aLa}] = 0 and thus [aj,aj] = 0 .

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

We would like to thank George Jaroszkiewicz, Rufat M. Mir-Kasimov, Vladimir

1. Manko, and Engin Arik for discussions, and Bruno Gruber for his hospitality and
encouragement.

7
REFERENCES

[1] 1. D. Faddeev, N. Y. Reshetikin, L. A. Tahtajan, Algebr. Anal. 1 (1989), 178.

[2J G. Jaroszkiewicz, J. Phys. A28 (1995), L343.
[3J P. A. M. Dirac, Phys. Rev. 114 (1959), 924; P. A. M. Dirac, Lectures on Quantum
Mechanics. New York, Yeshiva University 1964; L. D.' Faddeev, Theoret. Math.
Phys. 1 (1970), 1; P. Senjanovic, Annals of Phys. 100 (1976), 227.
[4J M. Arik, D. D. Coon, J. Math. Phys. 17 (1976), 524.
[5J P. Kulish, E. Damaskinsky, J. Phys. A23 (1990), L415.
[6J W. Pusz, S. L. Woronowicz, Rep. Math. Phys. 27 (1989), 231.
[7J M. Arik, Symmetries in Science VI, B. Gruber (ed.), Plenum Press, New York, 1992,
47.
[8J M. Baker, D. D. Coon, S. Yu, Phys. Rev. D5 (1972), 1492.
[9J M. Arik, D. n. Coon, Y. M. Lam, J. Math. Phys. 16 (1975), 1776.
[10J A. M. Polyakov, Phys. Lett. BI03 (1981), 207.
[l1J M. Arik, M. Mungan, G. Unel, Phys. Lett. B321 (1994), 385.
[12J M. Arik, M. Mungan, Phys. Lett. B282 (1992), 101; R. M. MirKasimov, J. Phys.
A24 (1991), 283.

8
ALGEBRAIC MODEL OF AN OBLATE TOP

R. Bijker 1 and A. Leviatan 2

1 Instituto de Ciencias Nucleares

U.N.A.M., A.P. 70-543,
04510 Mexico, D.F., Mexico
2 Racah Institute of Physics
The Hebrew University
Jerusalem 91904, Israel

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.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 9
ALGEBRAIC TREATMENT OF A THREE-BODY SYSTEM

The geometry of a three-body is characterized by two relative coordinates which we

choose as the relative Jacobi coordinates

(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

btp,m' btA,m' S t (m=-l,O,l). (2)

The scalar boson does not represent an independent degree of freedom, but is added
under the restriction that the total number of bosons N = np+n,\ +ns is conserved. This
procedure leads to a compact spectrum generating algebra of U(7) whose 49 generators
are

(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]

FI L (b!,mb,\,m + bl,m bp,m)

m

Fz -i L (b~,mb,\,m - bl,m bp,m)

m
F3 L (b~,mbp,m - bl,m b'\,m)
m
(5)

10
The operator it counts the total number of dipole bosons

it = itp+it>. = L(b!,mbp,m+b1.mb>.,m) (6)

m

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

if = Ho + ts sts - tp (b! . bp + bl b>.)

+ Uo (ststss) - UI st(b!. bp + bl b>.)s
+vo [(b!. b! + bl bl)ss + stst(bp' bp+ b>. . b",)] + Cl (b1 x b1)(1) . (b;>, x bp)(1)
+ L CI [(bb x b! - bl x b1)(I) . (bp x bp - b;>, x b>.)(I) + 4 (b! x bl)(I) . (b;>, x bp)(l)]
1=0,2

+L WI (b~ x b! + bl x b1)(I) . (b p x bp + b", x b>.)(I) . (7)

1=0,2

Here bp,m = (_l)l-mbp,_m , b>.,m = (-l)l-mb;>,,_m and s = s. The dots indicate

scalar products and the crosses tensor products. The Hamiltonian of eq. (7) con-
tains several interesting limiting situations. The harmonic oscillator model arises for
va = 0, i. e. no coupling between different harmonic oscillator shells. This scenario cor-
responds to the U(7) :) U(6) group reduction. On the other hand, for the choice Va =I- 0
the eigenfunctions are collective in nature, since they are spread over many oscillator
shells.
The eigenvalues and corresponding eigenvectors can be obtained exactly by diago-
nalization in an appropriate basis. The wave functions obtained in this way have by
construction good angular momentum L, parity P, and permutation symmetry t.

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

I1/JA, ) 1 o o 0 11/J A,)

11/JA,) o -1 0 0 11/JAJ
P(12) (8)
I1/JEp) o 0 -1 0 I1/JEp)
I1/JEJ o o o 1 I1/JEJ

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)

The equilibrium shape of an equilateral triangle is a result of the underlying D3 sym-

metry.
In order to analyze the vibrational and rotational excitations it is convenient to
split the Hamiltonian into an intrinsic (vibrational) and a collective (rotational) part
[14]. The intrinsic part of eq. (7) by definition annihilates the condensate of eqs. (14)
and (15) and is given by [7]

Hint = 6 (R2 S t s t - b~ . b~ - bl . bl) (R2 55 - bp. bp- bA. bA)

+6 [(b~.b~-blbt) (bp.bp-bAb) +4(b~.bl) (bAb p)]. (16)

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

[-Rst + (b~,y + bl,x)/V2] /Vl + R2 ,

(bl,x - b~,y)/V2 ,
(bL + b~,x)/V2 . (18)

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)

for which we have

(-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

Figure 1. Schematic representation of rotational spectrum of a (VI, V~=O) vibrational band.

The states are labeled by Lf. Here K denotes the absolute value of the projection of the
angular momentum L on the symmetry axis.

The angle n specifies the orientation of the intrinsic state

IN, V!, V2, 1; R, n) R(n) IN, VI, V2, I; R) ,
R(n) e -i.pL z e -ioL. e -iq,L z
(28)

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

_2~ -+.~ -~tA2

._2A2Al
_2~

._lAlA2
_ot
J{=l J{=2

Figure 2. As Figure 1, but for a (VI, v~=I) vibrational band.

These wave functions are characterized by m F = J{ + 21 = v (mod 3). For J{ =I 0

and l =I 0 there are two extra linear combinations with mF = J{ - 2l = v (mod 3)

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 ")
' .......... _-_ ..... "..,... -................ _--_ ....
\
/

Figure 3. Geometry of collective model of baryons.

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

Helicity amplitudes for electromagnetic and strong couplings can be expressed in

terms of some elementary spatial matrix elements or form factors. For nonstrange
resonances these are the matrix elements of the operators [7, 20, 21]

(33)

18
 N(1710)
3

oTH
N(1440)
2

1 1
_N_(9_39--...:

Figure 4. Schematic representation of the vibrational spectrum of nucleon resonances. The

resonances are labeled by the usual spectroscopic notation [19] and their oblate top classifi-
cation (VI, V&); K, Lf .

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 ,

(bl x s- st x b).,)~) . (34)

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)

In the large N limit (and R2 > 0) it reduces to

. NRV2
hm Xv
N~oo
= -R2'
1+
(36)

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

l1{lo) = IN,vl = O,V2 = 0,/ = O;R;K = O,L = O,M = 0) (37)

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

~ J do' 'D~l(n) eikf3sinBcos

.K V"'(L=-+---=-:K=)!-;-:(L=-----=-:K=)! 1 L-K .
= OM,OZ v'2L+1 (L+K)!!(L-K)!! 2[1+(-1) lJdk,8). (39)

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'

Final state F2(k) G1;p,( k )jm3kOj3

[56,0+](0,00);0 jo( kj3) 0
[20,1 +ho,oO);O 0 ~j1(kj3)
[70, 1- ho,oO);! -i v'3 j1 (kj3) i:16 [2jo( kj3) - j2( kj3)]
[56,2+](0,00);0 if j2( kj3) 0
[70, 2-] (0,00);1 0 z. ~ . (kj3)
v'2 J2
[70,2+] (0,00);2 -~ j2(kj3) =f Jw [3h(kj3) - 2h(kj3)]

and

2L + 1 J(L + K)!(L - K)! K

2(1 + OK,O) (L + K)!!(L - K)!! JL(L + 1)
X [1 + (_l)L-Kj ~ [(L + 1)jL-1(kj3) - LjL+1(kj3)]

2L + 1 J(L + K)!(L - K)! 1

+OM,1 m3koj3 i K
+ OK,D) (L + K -l)!!(L - K - I)!! JL(L + 1)
2(1
x [1- (_l)L-K] jL(k(3) ,
G 2 ;p,m(k) = O. (44)

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.

SUMMARY AND CONCLUSIONS

In this contribution, we presented an algebraic treatment of the three-body problem.

We used the method of bosonic quantization, which for the two relative coordinates of
the three-body system gives rise to a U(7) spectrum generating algebra. The model
space is spanned by the symmetric irreps [N] of U(7).

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.

Final state F2(k) G 1 ;p,(k )/m3ko(3

[56,0+ b,OO);O -Xl k(3 jl (k(3)

,fi; k(3 [2jo( k(3) -
[20,1+](1,0 0 );0
(70,1-](1,00 );1 -i Xl ~k(3(jo(
k(3) - 2j2( k(3)) Xl j2( k(3)]
i X1,fi; (2jo( k(3) +2j2( k(3) - 3k(3 j1 (k(3))

-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

It is a pleasure to thank A. Frank for interesting discussions and a careful read-

ing of the manuscript. This work is supported in part by CON ACyT, Mexico under
project 400340-5-3401E and DGAPA-UNAM under project IN105194 (R.B.), and by
grant No. 94-00059 from the United States-Israel Binational Science Foundation (BSF),
Jerusalem, Israel (A.L.).

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

E. Binz 1 and P. Oellers2

1 Fakultiit
fiir Mathematik und Informatik
Universitiit Mannheim
D-68131 Mannheim, Germany
2Kopernikusstr. 52
D-68165 Mannheim, Germany

In memoriam Rudi Domiaty

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
)

as quantum operator of the mass-squared function defined on T M. Here Og and Rg are

the d'Alembertian and the scalar curvature of g, respectively. We set ft = l.
The expectation value ~g (Og + ~ . Rg, 1/;) of the mass-squared operator Og + ~ . Rg
at a smooth real state 1/; on a bounded region D of M is

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/.-

Symmetries in Science IX, edited by Gruber and Ramek

Plenwn Press, New York, 1997 25
the basic observation to our note (cf. Oellers 1985). This shows that up to the factor
6 the Hilbert action for Einsteins vacuum equation is nothing else but the expectation
value of the mass-squared operator (For simplicity we set 167l'G = 1).
Thus Einsteins vacuum equations for 9 determines a geometry, i. e. a metric g under
which this expectation value is observed optimally.
This remark extends to the space time with matter: Adding a selfadjoint operator
Ho to Og !. !.
+ Rg and a matter density o:(g) to R g, the above variational problem
yields, as usual, an Einstein equation (in this case) for the metric 9

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

Ricg = 6 . (T(g) - ~ . trgT(g) . g).

If therefore T(g) is trace free then Rg = 0, yielding the Klein-Gordon equation

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

"F;g(H,'ljJ):= (H'ljJ,'ljJ) (2)

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),

as to be found in Sniatycki (1980). For the sake of simplicity we set = 1. Inn

these notes, this operator is hence forth called the mass-squared operator. Og is the
d'Alembertian. The map Rg : M --+ JR. is the scalar curvature of g. For both of these
notions we refer to the Appendix.
The expectation value "F;g( Og + ~ Rg, 'ljJ) of the mass-squared operator (given by (2))
at a nowhere vanishing 'ljJ E COO(D, JR.n) turns out to have a special form: Setting

(3)

this expectation value is by (30) in the Appendix

(4)

(d. Oellers, 1995). Here fllJ = 'ljJ4 . fl9 is used.

Equation (3) expresses that 9 is rescaled by 'ljJ. Hence 'ljJ plays the role of a state
and yields a rescaling of the metric, on the other hand. Thus 'ljJ is also referred to as a
rescaling function, in the sequel.
If Ho on COO(D, JR.) is a linear operator selfadjoint with respect to the scalar product
(d. (1)) we let
1
H := Og + [; . Rg + H o. (5)
In what follows we will study the expectation value "F;g(H, 'ljJ) of H at 'ljJ.

AN INTERPRETATION OF THE EINSTEIN-HILBERT ACTION

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)

This allows us to state the following interpretation of the Einstein-Hilbert action

of the vacuum (cf. Wald (1984), with 167rG = 1) defined on the conformal class of g:
Proposition 2
For any Lorentz metric 9 the Einstein-Hilbert action JD Rg . /1g for the vacuum is SIX
times the expectation value of the operator 0g' + ~ . Rg" at l/J formed with 1'espect to the
Lorentz metric g' := l/J-2 . g.
The following is now immediate from (6) and the fact that Rg = 0, provided 9
satisfies Einsteins field equations for the vacuum:
Corollary 3
If the Lorentz metric 9 satisfies Einsteins field equations for the vacuum then
Sg,(Og' + ~. Rg"l/J) = 0 for any rescaling function l/J E Coo(M,][~n) and hence

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

~ .L( + Rg 6 . a(g)) . /1g,

where a(g) is a smooth g-dependent density on D.

Defining Ho(g) on COO(D, R), for a given nowhere vanishing l/J E Coo(D, R), by the
equation
Vip E COO(D,R),
we observe that

(8)

for H := Og + ~ . Rg + Ho(g). Therefore, we have the following general interpretation

of the Einstein-Hilbert action on the conformal class of g:
Theorem 4
Any Einstein-HilbeTt action associated with a given Lorentz meb-ic 9 is six times the
expectation value Ofog,+~.Rg,+Ho, formed with Tespect to g' = l/J-2' g at an approp1'iate
nowheTe vanishing l/J E Coo(D, R).

VARIATION OF Eg(H,'t/J) WITH RESPECT TO g

For a Lorentz metric 9 let H := Og + ~. Rg + Ho where Ho acts on Coo(D, R) and is

selfadjoint with respect to (6). Moreover we let JD l/J. Hol/J' /1g = JDa(g) /1g. Motivated
by Hilbert's derivation of Einsteins field equations (cf. Wald (1984)) we will vary

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

By using (30) in the Appendix we deduce immediately the following:

Theorem 5
Let oo(g) be a smooth g-differentiable density of $g(Ho,7j;) with 7j; a rescaling func-
tion. The expectation value l}g(H, 7j;) is stationary at 9 := 7j;2 . 9 iff
.
R~cg - 2'1 . Ri) . 9 = 6 . T(g). (10)

The g-trace of (10) is

Rg = -6 trgT(g) (11)
(here trg denotes the g-trace) and (10) reformulates as

Ricg = 6 . (T(g) - ~ . trqT(g) . g). (12)

By (30) equation (11) turns into

(13)

iftrgT(g) = 0 then Rg = 0 by (11) and hence (13) yields the Klein-Gordon equation
1
(Og + '6 . Rg)7j; = O.

Finally, the expectation value of the mass-squared operator is

( 14)

in case lfJg (H, 7j;) is stationary at g.

Example: Let F be a two-form on D satisfying Maxwell's equation. We assume
T(g) to be given by

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

a := 2"1 . 'Ij;- 3 . Ho'lj; (15)

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

hold true. Therefore

Ricg = 3 . 'Ij;-3 . Ho'lj; . g. (19)
In case of 9 = 9
Rg = 12 Hal. (20)
Comparing (18) with (13) yields

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

provided -fg (H, 'Ij;) is stationary at g.

Remark: The variation of ~g(H, 'Ij;) = JD'Ij;-3.H'Ij;'/-lg = JD 'Ij;H'Ij;/-lg with respect to
'Ij; (variations are supposed to have compact support in the interior of D (d. Appendix))
yields (Og + ~Rg + Ho)'Ij; = O.
The coefficient 2 of Ho in (18) is due to the variation of the volume-form.
The following is an immediate consequence of the corollary 5 (d. Oellers, 1995).
Corollary 6
If Ho = rr;2 . id with m 2 E IR a constant then (13) reads as

.
RZCg - 2"1 . Rg . 9_ = -2"3 . m 2 . 'Ij; - 2. -9 (22)

30
and hence
(23)

holds true. Thus

(24)

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

and require from cp E Coo(D, 1R) that 'lj; is a rescaling function.

Equation (18) reads then as
1 1
(Og + "6 . Rg - 2 . Ho)cp + "6 . Rg - 2 . HoI = O. (26)

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

Ricg = 3 (1 + cpr 3 Ho(l + cp). 9

holds true. If Jj)g(H, cp) is stationary at 9 := (1 + cp)2 . 9 then

1
(Og + "6 . Rg - 2 . Ho)cp = O. (27)

To extend (13) to cp E COO(D,IR) we use (21) and (26) to obtain

trgT(g) = trgT(g) + trgTl(g) (28)

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

By (28) we therefore conclude

Hocp = -~. (tr 9T 1 (g) + cp(3 + 3cp+ cp2). trgT(g))

which corresponds to (13). We may thus state the following for a more special situation:

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

Let M be a smooth, connected and oriented manifold of dimension four and g. be a

Lorentz metric on it.
In this section we will define some basic concepts and will derive some of the
fundamental formulae used in these notes.
The Hodge star operator:
At first we define a metric on A1(M), the collection of smooth one-forms on M as
follows: Given any two a, f3 E A1(M), there are smooth unique vector fields XCV, X(J
such that
a(X) = g(X" , X) and f3(X) = g(X8, X)
for all X in the collection fT M of all smooth vector fields of M. We set

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

0'1\ *f3 = g(a, (3). fL9

with fLg being the metric volume-form of 9 on M. In particular we have

*l=fLg and *fLg=-l

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)

150' = (_1)4(k+1)+1+ Ind g * d * a

where Indg is the indix of g, (the number of - signs), which is 1 in our case, and dis
the exterior derivative. Hence 150' = *d * a.

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

with Lxthe Lie-derivative, then

divgX = -og(X, .. .),

zX/tg = *g(X," .J
and
Lx = do ix +ix 0 d,

here iXPgC,,,J = /t(X,"',"'," .). All these equations are easily verified.
The last formula yields immediately

divgX . p(Xo, Xl, X 2 , X 3 ) = d(ixp(Xo, Xl, X 2 , X 3 )

I:;=1( _1)i. 'Vxi(ixp)(Xo,'" ,Xi,'" X 3 )
p(Xo,"', V XiX"" ,Xo)
trV X . I1(Xo, Xl, X 2 , X 3 )

showing
diVg X = trV X.
For a orthonormal basis Eo,E1,E2,E3 with g(Eo, Eo) = -1 we deduce from above
3

tT' V X = -g(V EoX, Eo) + L g(V EiX, Ed

i=l

To compute Dgf with f E COO(M,Rn ) we first observe that

gmdgf being the gradient of f formed with respect to g. We hence deduce

Dgf = *divg9radgf . /tg = divg gradg! . */tg

and therefore
Dgf = -divggradgf = -trV gradgf.
Remark: In physics literatnr often the d' Alembertian of a function! is defined by
divggmdgf
Dirichlet integral:
Given two smooth one-forms 00 and (3, then

r doo 1\ *(3 lDr001\*8(3+ laDr i*ool\(3

lD
=

where D C M is a compact smooth four dimensional submanifolds and i : D --+ M

denotes the inclusion. If 1;, 1;' E CoorD, R) then the above formula yields

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.

The covariant derivative Vi) of !J reads as

V~Y = V x Y + df(X)Y + df(Y)X - g(X, Y) . gradg f.

The curvature tensor Rg of!J is expressed in terms of Rg for any X, Y, Z E rT M by

Rg(X, Y)Z = R(X, Y)Z + Hg(f)(X, Z) . Y - Hg(f)(Y Z) . X

g(Y, Z) . V xgradgf + g(X, Z) . Vygradgf
+ df(X) . df(Z) . Y - df(Y) . df(Z) . X
+ 2 df(Z) (df(Y). X - df(X) Y)

(g(y, Z) X - g(X, Z) . Y) 'lgradgfl2

(df(Y) . g(X, Z) - df(X) . g(Y, Z)) . gradgf.

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)

e 2J Ri) = Rg + 6 ogf - 6 . Idfl2.

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:

Ricg = Ricg + 8- 2 . d8 d8 - 28- 1 . Hg(B)

+ (B-1. OgB - B-2 'ldBI 2 ) 9 (29)

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

R. Abraham, J. E. Marsden, and T. Ratin, Manifolds, Tensor Analysis and Applica-

tions, Addison-Wesley Publishing Company, Reading MA, London, 1983
A. L. Besse, Einstein Manifolds, Ergebnisse der Mathematik und ihrer Grenzgebiete,
Band 10, Springer Verlag, Berlin, Heidelberg, New York, 1987
P. Oellers, Das Einstein-Hilbert Funtional und der Bezug zur geometrischen Quan-
tisierung, Dissertation Universitiit Mannheim, 1995
J. Sniatycki, Geometric Quantization and Quantum Mechanics, Applied Mathematical
Sciences 30, Springer Verlag, New York, 1980
R. M. Wald, General Relativity, The University of Chicago Press, Chicago, London,
1984

35
 MULTICHANNEL DYNAMIC SYMMETRY

J. Cseh

Institute of Nuclear Research of the Hungarian Academy of Sciences

Debrecen POB 51, Hungary-4001

INTRODUCTION

A quantum mechanical system is said to have a continuous symmetry desrcibed by

a Lie group G, if its Hamiltonian commutes with all the generators of the group, i.e.
it can depend on the generators only through the Casimir invariants of the group. If
both the potential and the total energy is invariant, the symmetry is called geometric,
contrary to the dynamic symmetry which leaves invariant only the total energy [1].
Well-known examples are the 0(4) dynamic symmetry of the Coulomb problem, and
the U(3) dynamic symmetry of the harmonic oscillator problem. In both cases 0(3)
is a geometric symmetry. These kind of exact dynamic symmetries hold only for very
special forces, therefore in this strict form they are not very helpful in building up
models of few and many-body systems.
It is a generalized version of the dynamic symmetry that proved to be-very successful
in various areas of physics: when the Hamiltonian can be expressed in terms of Casimir
operators of a chain of nested subgroups (see e.g. [2]):

G => Gf => G" => . .. . (1)

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.

Symmetries in Science IX, edited by Gruber and Ramek

Plenwn Press, New York, 1997 37
 In nuclear physics extensive studies of this kind have been carried out in relation
with the quadrupole collective motion of nuclei, including the question of connecting
the collective model to the. shell model. Relatively less attention is paid so far to the
description of clusterization in this way, and to the investigation of the common part
of the three basic structure models, i.e. the cluster model, the shell model and the
collective model.
Here I would like to report on some recent attempts that show that the concept
of dynamic symmetry is very useful also in cluster studies, in each respect mentioned
above: i) It is applicable for the description of experimental data. ii) For the case of
dynamic symmetry the interactions of different approaches can be related to each other.
iii) It connects the cluster model to the shell and collective models (in the region of light
nuclei); in particular the common intersection of these three descriptions is a dynamic
symmetry [3].
Furthermore, there exits the possibility for a new type of composite symmetry, called
multichannel (or multi configurational) dynamic symmetry, which connects with each
other different cluster configurations of a nucleus, and as a consequence, the different
reaction channels associated to these configurations. This dynamic symmetry puts a
very strong constraint on the relation between the interactions of different cluster con-
figurations; occasionally the interactions in one channel can completely determine that
of the other one. Therefore, the distribution of states with different cluster structure is
governed by the same Hamiltonian. This enables us to analyse in a unified framework
experimental data that are usually treated separately.

RELATION BETWEEN THE STRUCTURE MODELS OF LIGHT NU-

CLEI

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)

and taking into account the simple

(4)

equation, it can be written as:

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.

THE SEMIMICROSCOPIC ALGEBRAIC CLUSTER MODEL AND ITS

U(3) DYNAMIC SYMMETRY

In the Semimicroscopic Algebraic Cluster Model [6, 7] the internal structure of

the clusters is described by the Elliott model of UgT (4) 181 Uc (3) group structure [4],
where U ST (4) is Wigner's spin-isospin group [8]. The relative motion is treated in
terms of the UR (4) vibron model [9]. (Due to the anti symmetric requirement of the
total wavefunction, that will be discussed below, the nuclear physical application of
the Vibron Model differs considerably from the applications in molecular and hadron
spectroscopy [10].) The group structure of the model for a system of J{ clusters (having
J{ - 1 independent space vector of relative motion) is:

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:

UR(4) :> UR(3) :> OR(3) ,

I N n " , L ). (7)

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)

The corresponding eigenvalue is:

(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:

VCr) = -63.998 + 1.9078(1' - 1'0)2 - 0.0118(1' - 1'0)\ (10)

where the 1'0 = 5.. 01fm shift relative to the origin is due to the Pauli-blocking effect
(and it compares quite well with the 4.93fm value of two touching spheres, calculated
with a radius parameter of 1.2fm).

Application of the dynamic symmetry

A consistent description of the energy spectra of the A = 16 - 20 nuclei is given in

ref. [14] in terms of core-plus-alpha-particle configurations: 12C + 0', 14C + 0', 15 N + 0'
and 16 0 + 0', with a standard form of the Hamiltonian:

H = E + ,/Rn + 8RC1~R(3) + 8cc1~c(3) + 8C;'~(3) + f)/{2 + /3L2 + eL S.

Jr (11)
Here the subscript C refers to cluster, S is the (core) spin, and /{ distinguishes the
different bands having the same SU(3) quantum numbers. The parameters have a
smooth A-dependence, based on which an interpolation could be made for those of the
13C + 0' system.

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.

MULTICHANNEL DYNAMIC SYMMETRY

Harmonic oscillator limit

When harmonic oscillator interactions are applied, not only the cluster Hamiltonian
can be obtained from that of the shell model

Hi/oCM = HHOSM, (12)

but, as a consequence, Hamiltonians of different clusterizations can be rewritten into

each other:
H Cj
HOCM = HC-
HacM (13)
The corresponding energy spectra are identical again, and a basis state of any cluster
model can be obtained as a linear combination of those shell model basis states that
belong to the same energy-eigenvalue. In case there is only a single shell model sate of
this kind and there are different cluster configurations, their wave function may differ
only in a normalisation factor.
We can say that for the harmonic oscillator cluster model a multichannel dynamic
symmetry holds. By this we mean the following:
i) The energy spectrum of different cluster configurations are identical. (If we do not
take into account all the possible internal excitations of the clusters, we may exclude
parts of the spectra. Nevertheless, the remaining part must be identical. In such a case
the multichannel dynamic symmetry is called partial. )
ii) The wavefunctions of different cluster configurations overlap largely, in some cases
they differ only in a normalisation factor.
Further important characteristics of this symmetry are (as found before):
iii) The same relation holds between any cluster configuration and the shell model
configuration (thus we can consider the latter one as a special zero cluster (i = 0)
configuration) .
iv) The shell model connection associates to the cluster configuration a well-defined
quadrupole deformation [4].
However, the harmonic oscillator interactions are too primitive, they are very far
from being realistic. One can ask the question, whether the nice features mentioned
above hold for more realistic interactions as well, and if they do, for what kind of
interactions?

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)

The energy eigenvalues corresponding to the dynamical symmetry (14) up to linear

terms in n". can be written as:

E = t + '"'(n". + ;3L(L + 1) + Bn"L(L + 1) + F(A,Il,L), (15)

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

'"'(1 = '"'(2 = '"'( , t1 = t2 + ,",(n"o , 81 = 82 = 8 ,

;31 = ;32 + 8n rro , F1(A, Il, L) = F2 (>', Il, L) (16)
equations have to hold. They guarantee the identical energies for the common states of
the two fragmentations, and give a unique relation between the cluster-cluster interac-
tions of two different configurations.
The multichannel dynamic symmetry builds up a connection between different clus-
ter configurations (as well as between the cluster states, shell model sates, and collective
model states). Again the remarkable new feature is that it does so for largely anhar-
monic interactions, that seem to be reasonable good for the description of realistic
spectra.

Applications of the multichannel dynamic symmetry

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].

ON THE STRUCTURE OF COMPOSITE DYNAMIC SYMMETRIES

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

[1] L. I. Schiff, Quantum Mechanics, McGraw-Hill, New York, 1968.

[2] F. Iachello and A. Arima, The interacting boson model, Cambridge University Press,
Cambridge, 1987.
[3] J. Csen, Proc. 4th Int. Wigner Symposium, Guadalajara, Mexico, 1995.
[4] J. P. Elliott, Proc. Roy. Soc. A 245 (1958), 128 and 562.
[5] K. Wildermuth and Th. Kanellopoulos, Nncl. Phys. 7 (1958), 150.
[6] J. Cseh, Phys. Lett. B 281 (1992), 173.
[7] J. Cseh and G. Levai, Ann. Phys. (N.Y.), 230 (1994),165.
[8] E. P. Wigner, Phys. Rev. 51 (1937), 106.
[9] F. Iachello and R. D. Levine, J. Chern. Phys. 77 (1982), 3046.
[10] F. Iachello, J. Cseh and G. Levai, APH N.S. Heavy Ion Phys. 1 (1995), 91.
[11] H. Horiuchi and K. Ikeda, Cluster Models and Other Topics, World Scientific,
Singapore, 1986.
[12] K. Kat6, K. Fukatsu and H. Tanaka, Prog. Theor. Phys. 80 (1988), 663.
[13) K. Varga and J. Cseh, Phys. Rev C 48 (1993), 602.
[14) G. Levai and J. Cseh, Phys. Lett. B381 (1996), 1.
[15] P. O. Hess, G. Levai and J. Cseh, Phys. Rev. C54 (1996), 2345.
[16] G. Levai, J. Cseh and W. Scheid, Phys. Rev. C46 (1992), 548.
[17] Zs. Fulop, G. Levai, E. Somorjai, A. Z. Kiss, J. Cseh, P.Tikkanen and J. Keinonen,
Nncl. Phys. A604 (1996), 286.
[18) J. Cseh, G. Levai and W. Scheid. Phys. Rev. C 48 (1993), 1724.
[19] J. Cseh, Fourth Int. Conf. on Radioactive Nuclear Beams, Omiya (1996).
[20] J. P. Draayer, Nucl. Phys. A 520 (1990), 259c.
[21] J. Cseh, R. K. Gupta, and W. Scheid, Phys. Lett. B 299 (1993), 205.
[22] J. Cseh, Phys. Rev. C 50 (1994), 2240.
[23] J. Cseh and W. Scheid, J. Phys. G18 (1992),1419.
[24) J. Cseh, J. Phys. G19 (1993), L97.
[25] J. Cseh, G. Levai and K. Kat6, Proc. 9th Int. Symp. on Capture Gamma Ray,
Budapest, 1996.
[26) A. Algora and J. Cseh, J. Phys. G22 (1996), L39.
[27) J. H. Hamilton, A. V. Ramayya, J. Kormicki, W-C. Ma, Q. Lu, D. Shi, J. K. Deng,
S. J. Zhu, A. Sandulescu, W. Greiner, G. M. Ter Akopian, Yn. Ts. Oganessian,
G. S. Popeko, A. V. Daniel, J. Kliman, V. Polhorsky, M. Morhac, J. D. Cole, R.
Aryaeinejad, I. Y. Lee, N. R. Johnson and F. K. McGowan, J. Phys. G20 (1994),
L85.
[28) A. B. Balantekin, I. Bars and F. Iachello, Nncl. Phys. A370 (1981), 284.

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

Bulgarian Academy of Sciences

Institute of Nuclear Research and Nuclear Energy
BG-1784 Sofia
Bulgaria

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.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 47
 In the present paper we start the systematic study of subsingular vectors of
Verma modules of q-deformations of simple Lie algebras (Drinfel'd-Jimbo quantum
groups), though it is clear that these considerations can be easily extended to affine
Kac-Moody algebras, W-algebras and their corresponding quantum algebras. In the
following section we give an exact and constructive definition of the notion of sub-
singular vector. Naturally, this is just an explication of the folklore definition that
a subsingular vector is vector which is not a singular vector of the Verma module,
but becomes a singular vector in a factor-module after some (or all) invariant sub-
spaces built on singular vectors are factored out. This is illustrated with two explicit
examples of subsingular vectors occurring for Uq (sl(4)) (one of them summarized in
Proposition 1). Since we would like to make the relation of subsingular vectors and
Kazhdan-Lusztig polynomials more precise we recall in detail these polynomials in
the third section in a simplified exposition different from the original one [24] and suf-
ficient for our representation-theoretic purposes. Our other aim is to provide explicit
formulae not only for subsingular vectors but also for related character formulae. For
this purpose we give explicit character formulae in the fourth section including all oc-
currences of non-trivial Kazhdan-Lusztig polynomials for Uq (sl(4)) (Proposition 2).
We also show that a non-trivial Kazhdan-Lusztig polynomial does not necessarily
guarantee the existence of a subsingular vector (section 5). More precisely, the vec-
tor with the prescribed weight may turn out to be a linear combination of singular
vectors. Based on the examples considered we make a conjecture when a subsingular
vector necessarily exists. We hope to show this conjecture in a subsequent publica-
tion. We give also a systematic discussion of the relation between subsingular vectors
of Verma modules over semisimple Lie algebras 9 and differential equations which
are conditionally g-invariant (section 6). We treat in detail the conformal algebra
su(2,2) and its complexification sl( 4). The conditionally invariant equations are the
d'Alembert equation and a new equation arising from a subsingular vector proposed
by Bernstein-Gel'fand-Gel'fand. In the final section, this treatment is extended to the
Drinfeld-Jimbo q-deformation Uq (9) of 9 and we give also the q-difference analogues
of the equations mentioned above.

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)

I)-1)m(n) (xt)mxt(X!r-m=O, j#k, n=l-ajk, (lb)

m=O m qi

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:

mi = miCA) ~ (p - A)(Hi) = 1 - A(Hi) = 1 - (A, an, i = 1, ... , r (2)

(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)

where pOl,m a is a homogeneous polynomial in its variables of degrees mni, where

ni E 1+ come from Ct = 2:: niCti, Ct; - the system of simple roots. The polynomial

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.

Definition: Let VA be a reducible Verma module. A vector Vsu E VA is called a

subsingular vector if Vsu ~ fA and the following holds:

x Vsu E VX E g- (7)

Remark 2: The image of a subsingular vector in the factor-module FA is a singular

vector of FA. For shortness we shall say the subsingular vector 'becomes' a singular
vector in the corresponding factor-module. <>
We need to be more explicit even in the general case. First of all it is clear that
it is enough for a vector to be subsingular if (7) holds for the negative simple root
vectors X i- . We can rewrite (7) in the following way:

X i- Vsu = L QiOl vOl,m a (8)

OIELli

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).

The Bernstein-Gel'fand-Gel'fand Example

The first example of a subsingular vector was given for the sl(4), i.e., for q = 1
[2]. It occurs for A(H1 ) = A(H3) = 1, A(Hz ) = 0, i.e., X = (ml,mZ,m3) = (0,1,0).
Thus there are four positive ma E N from (11): m2 = mlZ = m23 = mI3 = l.
Correspondingly, there are four singular vectors:

(13a)

V~2 = xtxivo, mlZ =1

V~3 = X: xi vo, m22 =1 (13b)
, -X+X+X+
v13 - } 3 zvo, m13 = 1

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:

Vbgg = (xtx;tX: -X:x;txt) Vo (15a)

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:

Vbgg = (XliX: + x;t;xt) vo, q =1 (15c)

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

Consider first an arbitrary Verma module VA and the following vector:

Vf = P Va , (18)

where P is a homogeneous polynomial in Uq (9+):

P -- X+X+
13 2 - q-1/2X+X+
12 23 (19)

Below we shall need the following technical result:

X:;Vf = q-1 ([A(H2) + l]qXi xt - [A(H2)]qXt Xi) Xiva +

+ q-1 xi ([A(H2) - l]qXt xi - [A(H2)]qXi xi) Va = (20a)
= q-1 ([A(H2) + l]qxi xi - [A(H2)]qxi xi) xtva +
+ q-1 xt (-[A(H2) - l]qXt xi + [A(H2)]qxi xi) Va (20b)

For future reference we note several equivalent forms of the polynomial P:

P = xl'3xi - q-1/2 xi; xi; = (21a)

-X+X+
- 13 2 - q
1/2X+X+-
23 12- (21b)
=q-1 (xt Xi xi xi + xi xi xi xt - [2]qXi xt xi Xi) = (21c)

=q-1 (xi xi xt xi + xi xt xi xi - [2]qxi xt xi xi) (21d)

and two forms valid if a == A(H2) -=I- 1:

P = [q-1/~
a-I q
(Xi xi - [2]qxi xi) ([a - l]qXt xi- [aJqXi xi) +

+ [a -1I ]q Xi ([a - l]qXt xi - [aJqXi xi) xi = (22a)

-1/2
[q ] (xt Xi - [2]gxi xi) ([a -1]qxi xi - [a]qXi xi) +
a - 1 q

+ [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:

vW = xi ([a - l]q[a - 2]q(Xt?(Xi)2 - [2]q[a]q[a - 2]qXt(X;)2 xt +

(25)
+ [a]q[a -1]q(X;?(Xi)2)Xi vo, m13 = 2, ml = a, m3 =1
This vector gives no new condition since factoring out the sub module Ima we factor
out also the sub module I m,a (it is a sub module of Ima)' Denote by h the submodule
generated by these singular vectors, II = Ima U I m12 , and by FI = V A / h the cor-
responding factor-module. Let 11) denote the lowest weight vector of Fl. Then the
expressions in (24) become null-conditions, namely:

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

However, if a -I- 1 then Vj itself belongs to the maximal invariant submodule jA

generated by the singular vectors, and thus is not a subsingular vector by definition.
Explicitly we note that if a -I- 1 then using (22a) we have:

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

Thus, indeed, Vj belongs to jA = Im '2 U Im a, for A from (23).

We now write down systematically all situations.

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)

If a E -N then in addition to (24) there is one more singular vector [21,28]

corresponding to m2 = 1 - a E N + 1:

(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)

(xi)I-a II" ) = 0 (31c)

If a = 0 then there is a singular vector corresponding to m2 = 1 and given by

(30) with a = O. Here also (24b) is descendent and the maximal invariant submodule
is generated by the singular vectors (24a) and (30), II. = It = lCia U l Ci 2 We denote
by L~' = VA / It the irreducible factor-module; 11"1 ) the lowest weight vector of L~/.
Then the null-conditions are:

xii 1"1 ) =0 (32a)

xii 1"1 ) = 0 (32b)

If a E N + 1 then there exists another singular vector [21,28]:

(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)

Finally, if a = 1 then the non-descendent singular vectors are V3 = xi Vo, cf.

(24a), and VI = xtvo, cf. (33) with a = 1, while (24b) is descendent of (33), and
there appears also a singular vector v~3' cf. (13b), corresponding to mn = 1 which is
descendent to (24a). Here we have also the subsingular vector Vj, cf. (18) and (27);
from the latter the essential one simplifying here to:

(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)

The vector V f becomes a singular vector in F I , which we denote as:

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:

(xt xt - [2]qXt xt) xi XiiI) = 0 (38c')

where we have used (38b) and (21c).

Analogously consider a Verma module VA with lowest weight A satisfying the

conditions:
A(Hd = 0 <==? m1 = 1 (39a)
A(H2 + H 3 ) = 1 <==? m23 = 1 (39b)
X3(a)=x(A)=(1,I-a,a), aEC (39c)
cf. (lla, d). From these conditions follow that there are two singular vectors which
are explicitly given by [28]:

VI = xt Vo, m1 =1 (40a)
V23 = ([a - l]qXi xt - [a]qxt xi) Vo, m23 = 1 (40b)

Denote by 13 the submodule generated by these singular vectors, 13 = 1 m , U Im 23 ,

and by F3 = VA /13 the corresponding factor-module. Let 13) denote the lowest
weight vector of F3 . Then the expressions in (40) become null-conditions, namely:

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

Thus, indeed, v f belongs to jA =:= I m 23 U 1 m !, for A from (39).

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 :

h n I3 = I m '2 U I m ,., a = A(H2) = 1, A(Hd = A(H3) = (44)

X = Xl (1) = X3(1) = (1,0,1)
showing which we have used:

To summarize, we have shown:

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

The Kazhdan-Lusztig polynomials (KLP) were introduced [24] in the construc-

tion of an alternative basis in Heeke algebras. Here we shall int.roduce them using only
a recursive procedure, (also contained in ref. 24), independent of the Heeke algebra
context.
Let W be a Coxeter group and S be the set of simple reflections; let c is the unit
element of W.
Next we need the Bruhai order on W, (d., e.g., [1]): Let w,w ' E W. One
writes w' f-- w if ( w) = fi( WI) + 1 and w = sw ' , with s a simple reflection. One
writes w' ::; w if either w' = w or exist elements Wk E W, k = 1, ... , n, so that
w' = W1 f-- Wz f-- , f-- Wn = w. Thus, the Bruhat order w'::; W is a partial ordering
of W. We shall also write w' < w if w' ::; wand w' =F w.
Let y, w E Wand y ::; w in the Bruhat order on W. The KLP Py,w( u) in an
indeterminate u may be introduced in a recursive procedure with the following initial
conditions:
Pe,e = 1 (46a)
Py,w = 0, y i w (46b)

In the recursive procedure [24J the following additional relation in W depending

on the KLP is used:

y -< w, ify<w, Cy=-cw, degPy,w=!U~(w)-fi(y)-l) (47)

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):

Py,w = u 1 - c Psy,v +Uc Py,v + 2:=

z
yS;w
y<z-o{v
s-z<z

if sy <y (48)
c= {~ if sy >y
ux == uf(x) , x EW

where /-l(z, v) is the coefficient of the highest power of u in Pz,v.

Note that from (46a) and (48) follows that Py,w' are polynomials in u. Indeed
each term on the r.h.s. adds non-negative integer powers of u (for the summation
term note that 1!(w) - 1!(z) E 2N).
Consider (48) first for y = w. On the r.h.s. in the second term Py,v = 0 since
y = w = sv > v. For the same reason the summation term is empty. Then c = 1 since
sy = sw = v < w = y and the r.h.s. is Psy,v = Pv,v, while the 1.h.s. is Py,w = Psv,sv.
Thus, (48) becomes: Psv,sv = Pv,v, Continuing in the same way one obtains, using
(46a):
Pw,w = 1 (49)
Next we consider (48) for 1!( w) = 1!(y)+ 1. Let first w = sy, as reduced expressions
with s E S. Then in (48) sy > y = v and because of this one has: c = 0, in the first
term in the r.h.s. Psy,v is zero, and the summation is empty. Thus, we are left with
Py,w = Py,v = Py,y = 1. Let now w = ys', as reduced expressions with s' E S. Write
y = sy', then w = sy' s', v = y's'. Then in (48) c = 1 since sy < y and the summation
is empty since sy' is not comparable with y's'. For the latter reason the second term
Psy' ,y's' is zero. Thus (48) is reduced to Py,ys' = Py' ,y's" Continuing in the same way
we can show that Py,ys' = Pe,s', which is equal to 1 by the already shown first case
w = sy, (replacing y = e, s I-> s'). Thus, we have shown [24]:

Py,w = 1, if1!(w)=1!(y)+l (50)

From (47) and (50) follows:

y ~ w, if1!(w) =1!(y) + 1 (51)

Analogously using (48) may be shown:

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):

Py,w = Psy,w, if y < w, and 8 E S is such that sw < w, sy > y (55)

From this follows (cf. Lemma 2.6.vi of ref. 24) if W is finite with longest element Wo:

Py,wo = 1, VyEW (56)

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.

CHARACTERS OF LWM AND NON-TRIVIAL KL POLYNOMIALS

Preliminaries on Characters of Lowest Weight Modules
Let again 9 be any simple Lie algebra. Let r, (resp. r +), be the set of all
integral, (resp. integral dominant), elements of?-{*, i.e., A E?-{* such that (A,an E
Z, (resp. Z+), for all simple roots ai. We recall that for each invariant subspace
V C Uq (9+)vo ~ V"I\. we have the following decomposition:

V;, = {u E V I Hu = (A + 11)(H)u, V H E ?-{} (59)

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:

ch V = L (dim VJl) e(A + {t) = e(A) L (dim VI') e({t) (60)

Thus e(A) is an overall factor characterizing the Verma module VA of which V is

a submodule. We recall that for V = VA we have dim VI' = P({t), where P({t) is
a generalized partition function, P({t) = # of ways {t can be presented as a sum of
positive roots (Jj, each root taken with its multiplicity dim9j3j (= 1 here), P(O) == l.
Analogously one uses [1] to obtain:

(61)

Further we recall the standard reflections in 1{*:

(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

where the action "." is defined by w . A = W(A - p) + p. If q is not a root of unity

this formula holds for the finite-dimensional irreducible LWM over Uq (9) (this can be
deduced from the results of [35, 36]). For future reference we note:

(64)

which follows easily using the definition (4).

A more general character formula involves the Kazhdan-Lusztig polynomials
Py,w(u), y, w E W [24], which we have recalled in the previous section. Namely,
the character formula of the irreducible LWM when A E f is [24,36]:

ch LA = L (_l)fCw"w) Pw,w,,(l) ch v w .(w':;:'. A ) , A Ef (65)

wEW
W~WA

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').

The Case U q (sl(4))

We restrict now to 9 = 51(4), which is exactly the simplest case in which non-
trivial KLP appear. These are given by (d. [24] and the previous section):

Pt ,S,SlS"S2(U) = l+u, t=e,52 (67a)

Pt,s, S3S2S1 S3 (u) = 1 + u, t = e, 51,53,5153 (67b)

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:

Lemma 1: Fix a semi-dominant signature XO = (ml,m2,m3), mk E Z+, with

m13 = rrt} + m2 + m3 > O. Let A', A" be the lowest weights with signatv.res (resp.):

x' (m~,m;,m~) = (-m1,m13,-m3), m;=l-A'(Hi ) (69a)

X" (m~,m~,m~) = (m12,-m2,m23), m;'=l-A"(H;) (6gb)

with m}2 = m} + mz , m23 = m2 + m3 .

i) The Verma modules V Ao , V Ao , VA', VA" enter the following embedding diagrams
depending on the values of Tnk:

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:

for mI, m3 2:: 0, mz > 0

3Z03 1 3 3 for m.1,m3 > 0, mz = 0
WA' = 32 3 3 for m1 > 0, m2 = m3 = 0
(71a)
32 3 1 for m3 > 0, m1 = mz = 0
W" for m1, m3 > 0, m2 2:: 0

31 3 3 3 2 3 3 for m1 = 0, m2, m3 > 0
31 3 3 3 2 3 1 for m3 = 0, m1, m2 > 0
WAil (71b)
31 3 3 3 2 for m1 = m3 = 0, m2 > 0
31 3 2 3 3 for m1 = m2 = 0, m3 > 0
33 3 2 3 1 for m2 = m3 = 0, m1 > 0

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:

Ao = wb . A' = 32 (31 . (33 (32 A') = (72a)

= W~ A" = 31 (33 (32 (31 (33 A" (72b)

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:

A' = 31 (83 Ao)

. = 33 . (31 . Ao) = Ao +mlG'l +m3G'3 (73a)
A" = 82 . Ao = Ao +in2G'2 (73b)

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:

1\0) = ml + m3 > 0, m2 = (76a)

X(S2 S 1 S 3 X' ,

1\0) = X' , ml > 0, m2 = m3 = (76b)

X(SZ S 3

= X' ,
1\0) m3 > 0, ml = mz = (76c)

X(SZSl

= X",
1\0) ml = 0, m2, m3 > (77a)

X(SlS3 S ZS 3'

X(SlS3 S ZS 1 1\0) = X" , m3 = 0, ml, m2 > (77b)

X(SlS3 S Z' 1\0) = X", ml = m3 = 0, mz > (77c)
X(SlSZ S 3 1\0) = X , = mz = 0, >
II
ml (77d)
>
m3
X(S3 S ZS 1 1\0) = X" , mz = r113 = 0, ml (77e)

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:

ch VA = e(A)j(l - tl)(l - iz)(l - h)(l - t12)(1 - b)(l - t13) (78)

Now we can prove:

Proposition 2: The character formulae of the irreducible LWM LA', LA" with
signatures in (69) are explicitly given as follows:

+ t';'32t';"3 + t';'13t';"2 + t'{'2t;';12 + t'{;,23t';'2 + t'{' 23 t';"3 _

(79)
_ tm2tm13tm12 _ tmntm13tm2 _ 2 tmntm,3tm,2 + 2 t m23 t m12 )
lZ 2 3 1 2 23 1 2 3 lZ Z3

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)

For m = 1 formula (85e) appeared in [32].

Remark 3: The above Proposition gives (for Ug (51(4))) an exhaustive parametriza-

tion of all occurrences of KLP in character formulae. It shows also, together with
Lemma 1, the relation between the two principal situations, since, a priori there is no
reason to use the same numbers in the signatures of A', All. The Verma modules V A' ,

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

are no KL multiplicities, (d. (85)), each character of a Verma module is contributing

once. We also have:

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:

= 51 . (523' A") = All + m13Lt1 + m3Lt23

A~ = 52 . A' = A' + m13Lt2 =

= 51'(53 '(52 .Ao)) (86)

A~ = 523 . A' = A' + m12Lt23
= 53 . (512 . All) = All + m1Lt12 + m13Lt3
A~ = 813 - A' = AI + m2Lt13 =

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).

Related Character Formulae

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):

Al = 81838283 . Ao = 5253 . Ao (88a)

A3 = 81838281 . Ao = 8281 . Ao (88b)

Next we need the following:

Lemma 2: For fixed mk E Z+ there exist the following embedding diagrams:

n n (89a)

VA; = VAs C VAl = VA", m2 E Z+, ml EN, m3 =0 (89b)

VA; = VAl C VAs = VA" , m2 E Z+, m3 EN, ml =0 (89c)

VA; = VAl = VAs = VA" m2 E Z+, ml = m3 = 0 (89d)

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)

Proof: The proof is analogous to that of Proposition 2 and Lemma 1.

(Formulae (90a) and (90b), resp., for m1 = m2 = 1 and m2 = m3 = 1, resp., have
appeared in ref. 32.) Now we have the desired formulae:

Corollary 2: The character formula of LA" when m1m3 = 0 is given by:

(9Ia)

valid when m1,m2 EN, m3 = 0

(9Ib)

valid when m2, m3 E N, m1 = 0

(9Ic)
valid when m2 E N, m1 = m3 = O. <>

KL POLYNOMIALS AND SUBSINGULAR VECTORS: A CONJEC-

TURE

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:

A~2 - A' = a13 (93a)

Ao - A' = a13 + a2 (93b)

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)

We have given explicitly the subsingularvectors corresponding to (93a) and (94a).

Case (94a) is the vector Vbgg, d. ref. 2 and (15). Case (94a) corresponds to the
subsingular vector v J , d. (18).
We start checking (93) and (94). Consider first (93b). It is easy to see that there
is no subsingular vector. Indeed, it is easy to check that any non-zero vector VI rt. CVo
of this weight which satisfies the properties: Xkv l E jA = 12 , k = 1,2,3, (for 12 cf.
(14, will necessarily be itself in h. An example of such a vector is a descendant of

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:

V~gg = xi (X;;Xi + Xi; X: ) Vo =

(96)
= (X;;Xi + Xi;xt) xi Vo E 12

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

These descendants come from singular vectors of V

All
'1'3 which are: xtvoA",,'-3,
A" A'!
xivo ,,,-3, xt xivo ,,,-3. The descendants are given explicitly by:

xt Vf Xl+ PVo = P Xl+ va E I~"

xi Vf Xi P Vo = P Xi Vo E jAil (98)
V~/3 = xi xt Vf = xi xt P va = P xi xt Vo E jAil

and we have also demonstrated that they belong also to jAil = 1 m , U 1 m3 .

Thus, from the six non-trivial KLP Py,w in (67) only two generate subsingular vec-
tors, namely, those for which (y,w) = (82,82818382) and (y,w) = (8183,8183828183).
It is important to notice that these are exactly the only two cases of pairs of elements
of W (for 8l(4 which fulfil the special relation y -< w, d. section 3. Thus we are
able to conclude by formulating the following:

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 . (;

CONDITIONALLY INVARIANT DIFFERENTIAL EQUATIONS

Preliminaries on Conditionally Invariant Differential Equations
In this section we write down explicitly the conditionally invariant equations
related to the subsingular vectors considered in the second section. For simplicity we
treat the case q = 1 first and the q-deformed analogs in the next section.
We use some results from ref. 21 which we give in a condensed form here. We
work with induced representations, called elementary representations (ERs). The
functions of the ERs can be taken to be complex-valued Coo functions on the group
G. The representation action is given by the left regular action, which infinitesimally
IS:
d
(XL'p)(g) == dt 'P(exp(-tX)g) It=o (99)

where X E 9, 9 E G, 9 is the Lie algebra of G. These functions possess the proper-

ties of right covariance [21]. For our purposes it is enough to consider holomorphic
elementary representations for which right covariance means:

X'P=A(X).'P, XEH (100a)

X'P = 0, X E 9- (100b)

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)

Thus, instead of (102) we have (d. (92)):

DSli C- A ---7 CA'

, A' = w . (w AI . A) (106a)
Dsu 0
A
XL tp = XL
A A'
0 Dsu rp , VX E y , (106b)

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)

while z, z encode the inducing Lorentz representation as explained below. In partic-

ular, one may use the following covariant representation for R2 [37] employing the
Pauli matrices 0'1':

(Z 1) 0' I' 81' (~)

( ~ ~), 0'1 = (~ ~) 0'2 = ( 0i -i)
0

Note also that under the natural conjugation:

w(X) = X, w(v) = v, w(z) (ll2)

Y is also a flag manifold of the conformal group SU(2,2).

Remark 6: As we noted if one wants to treat the case of a real non-compact
algebra go one has to use also the results for its complexification g. The application
of these results to go has some subtleties [21]. However, in the case at hand when
go = su(2,2) and g = sl( 4) the passage to su(2,2) is straightforward [37]. Also
considering representations of the corresponding groups (which are used here only to
provide the representation spaces) involves some subtleties [21], which, however, are
not felt in the case under consideration [37].
The reduced function spaces of the ERs in which our equations are defined are
complex-valued Coo functions on the flag manifold. The holomorphic ERs of sl( 4)
are labelled by the signature X = (ml' m2, m3).
In the su(2, 2) case most applications in physics are in the case when ml, m3 E N
and one uses reduced functions which are polynomials in the variables z, z of degrees
ml - 1, m3 - 1, resp. These then carry finite-dimensional irreducible representations
of the Lorentz algebra of dimension ml m3. Let us stress that this is an indexless
notation on which all Lorentz components of the fields are gathered together by the
polynomial dependence in z, z. To restore the components one has to take the entries
of the independent terms in z, z, cf. ref. 37. Note that in the physics literature often
is used, instead of (ml' m2, m3), the labelling [d,h,jz], where d = 2 - (m13 + m2)/2
is the conformal weight, h = (ml -1)/2, jz = (m3 -1)/2, so that finite-dimensional
Lorentz irreps one has: jk E 7l+/2.

The. Equations Arising from the BGG Example

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:

(RIRzR3 - R3RzRI) 1j; ==

= (OvO. - OijOz + (zo. - zOz) 0+) 1j; = 1j;' (113b)

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)

The Equations Arising from the Other Archetypal sl( 4) Example

We pass now to equations arising from the other archetypal 8l( 4) example. We
consider the case when the lowest weight satisfies conditions (23).
We shall substitute the operators Rk into the null-conditions (29), (31), (32),
(34), (36), and (38). In all cases we have the equation arising from the singular vector
V3 = xivo, (null-conditions (29a), (31a), (32a), (34a), (36a), and (38a:

(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:

(ozt 1j; = 0, (116)

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 :

(a -1)RIR2 - aR2RI) 1j;

(117)
= (Ca-l)(ov+zO+)-R20z) 1j; = 0

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)

Finally, we obtain the conditionally invariant equations corresponding to the

subsingular vector vf' Let us denote by P the polynomial P with X:; replaced by
Rk. Now we shall obtain this operator in explicit form:

(R3 R2 - 2R2R3) RI R2 0 = (119a)

(iO+ + OJ; - R20z) Oz R2 0 = (119b)
((za+ + av) az R2 - R2 Oz (zo+ + av)) 0 (119c)
(ov Ov - 0_ 0+) 0 = 0 0 (119d)

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)

where 0 is as in (120c) and again satisfies (120a, b).

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:

(a -1- zOz) (0_0+ - ovov) 0

= (a-l-zoz) 00 = 0 (121a)

73
This still follows from (120d). Analogously, taking 0+ derivative from (l1Sa) and ov
derivative from (l1Sb) and subtracting the two we get:

Oz(O-O+ - ovov) <P

= oz 0 <p = 0 (121b)

This also follows from (120d). Now, clearly from (121a, b) follows:

(a-1)D<p=0 (121c)

which implies the d' Alembert equation if a -I- l.

Using the conjugate situation with signature X3(a) we recover the d'Alembert
equation on functions which do not depend on z and satisfy

(a - 1)R3R2 - aR3Rl) <p =

(122)
= (a-1)COv+zO+)-R20z) <p = 0

instead of (117). Furthermore the analogs of (l1Sa, b), (114), (116), resp., are:

(Ca - 1 - ZOE)OV - O_OE) <p =0 (123a)

( (a - 1 - ZOE)8+ - Ov Oz) <p =0 (123b)

oz <p =0 (123c)
(ozt <p = 0, aEN (123d)

Thus if a E N, then the functions of the irreducible representations are polynomials

in Z of degree a - l.
If a E Z_ our functions satisfy (115) as those with signature Xl (a).
Finally the d'Alembert equation (120d) follows from equations (123a, b) (a -I- 1).
We do not need to consider a = 1 since the two signatures coincide.

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)

which is not zero unless a = 1.

The cases a > 1 are interesting in other contexts, especially, if we consider
together the representations with the conjugated signatures XI(a) and X3(a) with
the same a E N + 1. In particular, in the case a = 2 the two conjugated fields are
two-component spinors and (117) and (122) are the two conjugated Weyl equations.
The cases a = 3 are maybe most interesting. The Lorentz dimension is 6 (= 2a)
and the resulting field is the Maxwell field. As was shown in detail [38], equations
(117) and (122) are just a rewriting of the free Maxwell equations:

aP,Fp,v=O, aP,*Fp,v=O (125)

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]). <)

CONDITIONALLY INVARIANT q-DIFFERENCE EQUATIONS

The q-Deformed Setting
We give now the treatment of the conditionally invariant equations in the q-
deformed case. We need first to introduce our reduced representation spaces G A with
signatures X = XCA) = (ml, m2, m3), cf. [23,38]. The elements of G A, which we shall
call (abusing the notion) functions, are formal power series in the noncommuting
variables z, v, x _, x+, v, z, which generate the q-deformation Yq of the flag manifold
Y (the commutation relations of these variables using the same notation are given in
ref. 38). More explicitly, these reduced functions are given by:

cp(Y)
i,j,k,R.,m,nE'iL+ (128)
CPijklmn = zi v j x~ x~ vm zn

where Y denotes the set of the six variables.

Next we introduce the following operators acting on our functions:

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)

where K = z, , v, v, z, and the explicit action on CPijkfmn is defined by:

Mz CPijklmn = CPi+l,jklmn , (130a)

Iv CPijkfmn = CPi,j+l,klmn , (130b)
M-CPijklmn = CPij,k+1,lmn , (130c)
M+ CPijklmn = CPijk,R+l,mn , (130d)
If) CPijklmn = CPijkl,m+l,n , (130e)
Mz CPijklmn = CPijkm,n+1 , (130f)
Tz <Pijklmn == qZ r.pijklmn , (131a)
Tv CPijkfmn = qj CPijklm n , (131b)
T_ CPijkfrnn = l CPijkrnn , (131c)
T+ CPijkfmn = l CPijkimn , (131d)
Tv r.pijkmn == qm ~ijkmn , (131e)
=q n '
r.pijkmn (131)
Now we define the q-difference operators by:
, ,- 1 , -1
V" r.p(Y) = ~ MK (T" _T;;l) cp(Y), K z,,v,V,z. (132)

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)

To obtain the (conditionally) invariant q-difference equations amounts now sim-

ply to substitute X: with R% in the expressions of the (sub )singular vectors for
general q.

The q-Difference Equations Arising from the BGG Example

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_ +

+ Mz Mz V+ (TvL)-1 Tv + q-1 Mz Vv - (134a)

The subsingular vector Vbgg gives rise to the following conditionally invariant equation:

(R~ R~ Rj - R~ R~ Rj) <j; =

4-- 222 AA
= { q Vv V z T z Tv T_ - Vv Vz Tz Tz +

+ l V+ T+ T; (Tz Tz + (q-1 Tz T;l - q Tz T z- 1)/A) +

+ 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 -

- >-.. Mv Mi V_ V+ Tv) V z Vz Tz} (TvLT+)-l 0 = <j;' (134b)

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 -

-). Mv M .. V_ V+ Tv) Vz) Tz (TvT_T+)-1 0 o (135b)

(R~ R~ - [2]q R~ Rj) R~ R~ 0 = 0 (135c)

As in the q = 1 case we use (135a) to split (135b) in two equations and to simplify
(135c). Finally, we have:
(136a)

( qa+2 [a - l]q 'Dv Tz Tv2 T_ - A

- (q Mz VV T; T_ +V_) Vz) Tz (TvT+)-1 0 0 (136b)

(qa+l [a-l]q V+ Tz (TvL)-1 T v -

- (Mz V+ (TvT_)-1 Tv + q-l Vv -

-). Mv V_ V+ Tv) V z ) Tz (TvT_T+)-1 0 0 (136b')

3 -1 -2) -1
'Dv 'Dv Tv Tv - q 'D_ 'D+ T_ Tz Tv -
A A A A
{ (

- qV_ V+ (TvT z- 1 - T v- 1T z )} (TvT+)-1 Tz L T; 0 o (136c)

In addition, if a E N we have also:

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
{ (

- q). V_ V+ Mv V v } (TvT+)-1 T_ T; 0 = 0' , a =1 (136e)

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

[1] J. Dixmier, Enveloping Algebras, North Holland, New York, 1977.

[2] I. N. Bernstein, I. M. Gel'fand and S. r. Gel'fand, Funkts. Anal. Prilozh. 5 (1)
(1971), 1; English translation: Funct. Anal. Appl. 5 (1) (1971), 1.
[3] M. Bauer, P. Di Francesco, C. Itzykson and J. B. Zuber, Phys. Lett. 260B
(1991),323; Nucl. Phys. B362 (1991), 515.
[4] P. Di Francesco and P. Mathieu, Phys. Lett. 278B (1992), 79.
[5] A. Kent, Phys. Lett. 273B (1991), 56; Phys. Lett. 278B (1992), 443.
[6] A. Ch. Ganchev and V. B. Petkova, Phys. Lett. 293B (1992), 56; Phys. Lett.
318B (1993), 77.
[7] B. Gato-Rivera and A. M. Semikhatov, Nucl. Phys. B408 (1993), 133.
[8] L. Benoit and Y. Saint-Aubin, Lett. Math. Phys. 23 (1991),117; Int. J. Mod.
Phys. A 7 (1992), 3023; Int. J .. Mod. Phys. A9 (1994), 547.
[9] M. Dorrzapf, Int. J. Mod. Phys. A10 (1995), 2143; DAMTP-94-53, May 1995.
hep-th/9601056.
[10] A. M. Semikhatov and I. Yu .. Tipunin, hep-th/9512079.
[11] M. Bauer and N. Sochen, Phys. Lett. 275B (1992), 82; Comm. Math. Phys.
152 (1993), 127.
[12] A. M. Semikhatov, Mod. Phys. Lett. A9 (1994), 1867.
[13] V. Schechtman, H. Terao and A. Varchenko, J. Pure Appl. Alg. 100 (1995), 93.
[14] V. K. Dobrev, Lett. Math. Phys. 22 (1991), 251.
[15] F. Malikov, RIMS-835, Oct 1991, RIMS91 project "infinite analysis", Jun 1 -
Aug 31, 1991, contributed paper #50.
[16] P. Furlan, A. Ch. Ganchev and V. B. Petkova, Phys. Lett. 318B (1993), 85;
Nucl. Phys. B431 (1994), 622.
[17] Z. Bajnok, Phys. Lett. 329B (1994), 225.
[18] N. Chair, V. K. Dobrev and H. Kanno, Phys. Lett. 283B (1992), 194.
[19] G. Felder, Nucl. Phys. B317 (1989), 215 & erratum -ibid. B324 (1989), 548.
[20] D. Bernard and G. Felder, Comm. Math. Phys. 127 (1990), 145.

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

H.-D. Doebner and J.-D. Hennig

Arnold Sommerfeld Institut fur Mathematische Physik

Technische Universitiit Clausthal
D-38678 Clausthal-Zellerfeld, Germany

THE LINEAR CREED

Quantum mechanics is an intrinsically linear theory. Us mathematical framework

is based on three related building blocks. We consider a system localized on a smooth
(finite dimensional) space-manifold M moving (non-relativistically) with time t; the
physical space-time is M x R}. The following building blocks are modeled on a separable
Hilbert space H chosen as L2(M,df-l) with elements depending on t:

(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

EXPW t (A) = Tr (WtA) . (1)

Projection operators P,pt on one dimensional subspaces {c0tl0t E H, c E C}

are special density matrices, denoted as pure states. Any density matrix can be
decomposed into pure states 0j,t with j element from some countable index set I,

Wt = L AjP,pj,t' L Aj = 1, Aj > O. (2)

jEI jEI

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.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press. New York, 1997 81
 (3) We want that the time dependence of density matrices is (uniquely) given through
the time dependence of pure states, i. e. we introduce the principle that only pure
states rule the system; the Aj are time-independent.
The evolution of 1/;t is given through a smooth operator Ut2 ,t, or propagator

(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

1/;t = U(t)1/;o, (5)

with a self-adjoint H as an observable. The corresponding evolution of the density

matrix is the von Neumann equation

(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.

BEYOND THE LINEAR CREED

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

- from Borel quantization of a system localized and moving on a manifold M

[7] through a generalized Ehrenfest relation [8], or
- equivalently for M = Rn, from a representation of an inhomogeneous non-
relativistic current algebra on Rn [9], through probability conservation [10].

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.

2. A generalization of the linear framework is necessary which allows for nonlinear

evolutions via a generalized notion of states and observables. An attempt in
this direction is due to B. Mielnik [14] and R. Haag and U. Bannier [15], and
is presented, adapted to the DG family, in this volume by P. Nattermann [16]
including a discussion of observables following W. Lucke [17].
It seems that there is a sound background for nonlinear quantum mechanics. There
is a respectable DG family of nonlinear evolution equations, a formalism that allows
an evolution with DG family is at least in principle known and should be worked out.
However, the geometry which implies the nonlinearity through Borel quantization is
not completely understood and - essentially - there is up to now no experimental
evidence for a nonlinear behaviour.

83
A PATH TO NONLINEAR EVOLUTIONS: BOREL QUANTIZATION

Quantum Borel Kinematics

We start with a system localized and moving on a manifold M as mentioned above

and with a certain set of classical observables O(M) relevant for this system. A quan-
tization (in the linear framework) is defined as a map Q of O(M) into the set SA(H)
of self-adjoint operators on H - - quantization map -

Q: O(M) --7 SA(H) , (9)

such that it respects at least some part of the mathematical structure of M and of O( M).
Additional mathematical assumptions are in general necessary to construct Q which
have no classical counterpart, the so called quantum assumptions. The applicability of
the different quantizations depends on the structure of the system and on its physics.
Borel quantization [7-9] is a general quantization method. It is based on first princi-
ples in a two step procedure. In a first step the kinematical structure of the system (for
fixed t) is quantized. In a second step a dynamics is introduced in analogy to classical
Newton mechanics and refering to the quantization of.the kinematics.
To explain Borel quantization remember that according to Wigner's definition of
elementary systems the physics for the free particle in (R3, 90) with metric 90 together
with the time evolutions is based on a (linear) irreducible representation of its symme-
try group acting on R~ x Rj. The symmetry group is here the central extension Ge (3)
of the Galilei group with Lie-algebra ge(3). The numbers that classify its inequivalent
representations are denoted as. quantum numbers in the sense of Wigner. The kinemati-
cal algebra of the free particle S C ge(3) is spanned by the Killing vectorfields of 90 and
interpreted as positions Qi, rotational and translational momenta Li and Pi, i = 1,2,3,

S = {Qi} EEls {Pi, Ld = a 3 EEls e(3) , (10)

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

S(M) = C=(M) EEls Vectc(M) (11)

with Lie brackets (if [Xl,X2J E Vectc(M)

[UI, Xd, (h, X 2)1:= ([XJ2 - [X 2 !!, [Xl, X 2]) , (12)

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

Q: S(M) --7 SA(L2(M,p,)

(13)
U,X) f--4 (QU),P(X)).
We assume for Q:

84
 (I) The "partial" Lie-algebra structure of S(M) should survive, i. e., Q is a partial
Lie-algebra homomorphism.

(II) Q(f) is a multiplicative operator acting on 1{ as

(Q(f)ljJ) (m) = f(m)ljJ(m) , mEM. (14)

(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.

P(X)ljJ =0 if suppljJ n suppX = 0. (15)

The motivation is that quantum mechanical operators fullfil this condition. Lo-
cality for the Q(f) is valid by construction.

A more detailed physical justification of these assumptions can be found as a conse-

quence of the mathematical modeling of non-relativistic quantum systems in [7]. (III)
and (IV) are characteristic quantum assumptions.
From (I), (III) and (IV) it can be shown that

TJ has to be a hermitian line bundle with flat hermitian connection form w. It

is known that fiat hermitian line bundles (7/,W) are classified up to isometric
isomorphisms through the dual of the first fundamental group of M,

(16)

P(X) is a first order differential operator with respect to V acting on C=-sections

ljJ
(17)

c E R is an arbitrary real number with dimension [kg;2]. 1i is due to the quantum

mechanical scale. The divergence term with respect to the measure /1 which is
proportional to ~ appears naturally because of the self-adjointness of P(X). The
divergence term proportional to c reflects thf' fact, that we passed from Killing
vectorfields to all vectorfields. A more geometrical view on c can be found in [18].
For trivializable bundles (TJ, w) the covariant derivative 'i7 X decomposes

= x + TiW(X) ,
~
'i7 X (18)

with W as a closed one-form. For non-trivializable (TJ, w) - standard case - equation

(18) is valid locally.

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:

Theorem 1 (Angermann, Doebner, Tolar) Irreducible selfadjoint representations

oj S(M) with (I)-(IV) on L 2(7],w) are classified through

(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)

We express (20) as an equation on the phase T* M as follows: Let Y E Vect(TM) be

the "second order" vectorfield on T M corresponding to the given differential equation
on Mandl the canonical lift of J on M to T M, then equation (20) implies

(21 )

U sing the isomorphism

g':TM3Vf-7PVET*M,pV(w):=g(v,w), V,WETmM, (22)

equation (21) becomes

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)

Observe that J in (23) is independent of the specific "second order" evolution Y on

T*M.
To be able to relate in the third step the representation Q(5(M)) with (23) we
apply the natural isomorphism (see, e. g., [19])
~
C=(M) Efls Vect(M) ---> Fc(T* M) Efls Fz(T* M)
f Efl X f-+ J + B(X),
which gives via Q a Lie algebra representation

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

.From (23)-(25) we get

d
dt Tr (WtQ(f)) = Tr (WtP(gradgf)) for all f E C=(M) (26)

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.

Pure states. For W t = P,pt the generalized Ehrenfest relation reads

for all f E C=(M).

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)

with (gU = (g')-l) and

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)

This give a constraint for any evolution of sections . The ansatz

(30)

with a linear operator H and a nonlinear operator F[t, ;j,tlleads to [21]

(31 )

H contains the expected Laplacian on M lifted to sections on 1). A fixed nonlinear

imaginary part proportional to the quantum number c appears and an arbitrary real
nonlinear part. For worked out examples on two dimensional M with c = 0 and R = 0
we refer to the contribution of C. Schulte [22] in these proceedings.

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)

or mass or particle conservation as a result. An alternative derivation of (31) based on

(32) can be found in [10].
For a system of N distinguishable or identical particles a corresponding nonlinear
evolution equation holds, which can be shown, e. g., if we choose as configuration space
(M x ... x M - D) or (M x ... x M - D)/ SN (D is the diagonal and SN the symmetric
group). The resulting equation seperates in one particle equations because the nonlinear
term induces no interaction between different particles.
For M = (R3 ,go) the nonlinear evolution equations (31) are known as Doebner-
Goldin equations or DG family, with R[?ji,."b] choosen in analogy to 1m F[."bt, ?jit] with five
free parameters CI, .. , Cs. There are many results on the DG equations, especially on
their symmetries, on the Cauchy problem, on stationary and non-stationary solutions,
its interpretation in connection with a generalization of the linear framework and on
various other generalizations. We refer to [24] and to the contribution of P. Nattermann
[16] in these proceedings.
The DG class can be derived also through a physically and mathematically com-
pletely different path. The following method was used in [13]:
It has been remarked by various authors that all actual quantum mechanical mea-
surements consists or are obtained from positional measurements performed at vari-
ous times [25]. Adopting this point of view, quantum theories - - whether linear or
non-linear - for which corresponding time dependent wave functions ?ji give the same
probability density l?ji(x, t)12 in space at all times are "in principle" equivalent. This
means that a (not necessarily linear) operator N acting in some dense set in L2(R;, d3 x)
yields physical equivalent theories, if

N[?ji](X',t)N[?ji](X', t) = ?ji(X', t)?ji(X', t). (33)

Now if all actual measurements (outcomes of experiments) are obtained from positional
measurements performed at different times, it can be argued that a system with pure
states -1jJ obeying the Schrodinger equation, and one with pure states N[1jJ) obeying
the transformed equation, have the same physical content. But we make two essential
observations:
(a) Equation (33) does not require N to be a linear transformation - nonlinear N are
also possible.

(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

ifdJtl,b(x,t) = (-;~b. + V(x)) I,b(x,t).

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

Wt = -k[H, W t ] + CL[aj, [OJ, Wt]] + FtlWtJ, (34)

j=l

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

W t = -k[H,Wt ] + C L[aj, [OJ, Wtll (35)

j=1

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

'vVe tried to answer partially the question:

Is there a phys%cally motivated proposal for a nonlinear Schrodinger equation and is
there a possibility to generalize the intrinsically linear framework of quantum mechanics
such that nonlienar evolutions are allowed?
The reviewed results and arguments suggest that such evolution equations for pure
states based on a quantization procedure, the Borel quantization, exists with a nonlinear
term proportional to a quantum number in the sense of Wigner and enforced through the
quantization. This evolution equation can also be obtained from the observation that all
actual physical measurements are measurements of position and time; this observation
yields nonlinear gauge transformations, which applied to the linear Schrodinger equation
leads to nonlinear evolutions. Both derivations depend on the notion of pure states.
An attempt for a generalized framework of quantum mechanics of Mielnik, Haag and
Bannier is based on pure states and is discussed in the contribution of P. Nattermann
in these proceedings [16].
For mixed states the result of Borel quantization is a generalization of the von
Neumann equation, with the property that pure states evolve into mixtures.
The quantum system described by the DG family and by the generalized von Neu-
mann equation is dissipative. This can be traced back to the Fokkcr- Planck like equation
(29) where a quantum diffusion term appears. The physcial background for this is not
yet completely understood. It seems that a generalized interaction between the system
and its configuration space exists.

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

[1] E. Wigner, Ann. Math. 40 (1939), 149.

[2] S. Weinberg, Ann. Phys. (NY) 194 (1989), 336.
[3] 1. Bialynicki-Birula and J. Mycielski, Ann. Phys. (NY) 100 (1976), 62.
[4] A. Shimony, Phys. Rev. A 20 (1979), 394; C. G. Shull, D. K. Atwood, J. Arthur,
and M. A. Horne, Phys. Rev. Lett. 44 (1980), 765; R. Gaehler, A. G. Klein, and
A. Zeilinger, Phys. Rev. A 23 (1981), 161l.
[5] N. Gisin, Phys. Lett. A 143 (1990), l.
[6] J. Polchinski, Phys. Rev. Lett. 66 (1991), 397.
[7] B. Angermann, H.-D. Doebner, and J. Tolar, in: Non-Linear Partial Differential
Operators and Quantization Procedures, S. Anderson and H.-D. Doebner (eds.),
Lecture Notes in Mathematics 1037, Springer, Berlin, 1983, 171-208; U. A.Mueller
and H.-D. Doebner, J. Phys. A: Math. Gen. 26 (1993), 719.
[8] H.-D. Doebner and J.-D. Hennig, in: Symmetries in Science VIII, B. Gruber (ed.),
Plenum, New York, 1995, 85-90.

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

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 93
the problems defined over unusual configuration spaces such as the non-commutative
geometries or the p-adic number field. In practice it is not even necessary to formulate
the path integration explicitly. The final form of the formulae, which are derived for
the usual quantum mechanical problems, may directly be adopted. At this point we
specially like to emphasize that in formulating the problems over the non-commutative
geometries, calculations of the Green functions over the q-group spaces look promis-
ing [SJ.
It is important to stress that studying the non-relativistic potentials should not
be considered as the simple exercises. Many field theoretical problems, including the
pair creations in the given cosmologies or in the external electromagnetic fields, Casimir
interactions, in the process of calculations may formally become equivalent to some non-
relativistic potential problem. For example the pair production in the Robertson- Walker
space-time expanding with the factor a(t) is formally reduced to the calculation of the
Green function of a particle moving in the one-dimensional potential V(t) = a- 2 (t)
with the time t playing the role of coordinate [9J.
In this note we study a Kepler-like toy potential for the purpose of giving a simple
illustration.
In the following section we present the derivation of the Green function for the q-
deformed version of the Kepler-like potential. Section III is devoted to the study of the
same potential over the p-adicnumber field.

q-GREEN FUNCTION FOR A KEPLER-LIKE POTENTIAL

To illustrate the ideas summarized in the introduction we study a simple and a

rather special one-dimensional potential (n = c = 1)

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

K(x, Xl; T) = to ds /00 dE e-iET e- 4ic," ( uu')1/2

io -00 21r
(Koss( u, U'; s) + Koss( -u, U'; s) + Koss( U, -u'; s) + Koss( -u, -U'; s)) (4)

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

x 2)Wn(u) + Wn( -u(Wn(u') + Wn(u' (5)

n=O

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

vcosh(kj3(flw)1/2U) hn (sinh(kj3(flW)1/2U) I q) (10)

with k = -log q, j3 = (2(1 - q-1/2, we arrive at a q-deformation of the potential (1).

After q-integration over dq~, the q-Green function (8) becomes

Kq(x, x'; T) = ~
00 1
00 dE
-
e- iET
n=O -00 271" wq-4' w([i(n + 1/2)]- [4ia/w])
'Q .

(XX')1/44>~(x )4>~(X') (11)

The usual integration over dE displays the correct energy poles at [E] = [en + 1/2)-2].
Note that the "physical" energy of q-Kepler-like potential is Eq = -2a 2 [n + 1/2J-2
which is the e-value of the time evolution operator zD';, z = e- 2i ,,2 t with respect to the
wave functions z(n+1/ 2)-2 Xl/44>~(x).

p-ADIC KEPLER-LIKE POTENTIAL

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

v> 0 for p = l(mod4)

// = odd for p = 3(mod4) (16)
v> 3 for p= 2

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

](~ss(u, u'; s) =L X(WO'ps)I]!"p/Jp (u) I]! "p/Jp (u') (17)

"p/Jp
and insert it into the p-adic version of (5), we obtain

j{P(x,x';T) = 2?r JJ ds dEX(ET)x(-4as)

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)

which is representing the p-delta function, (18) becomes

I{P(x, x'; T) = 2?r L

CXp/Jp
J dEX(ET)8(wO'p - 4a)!"p(3p(x )!cxp/Jp(x') . (21)

Changing the integration variable E --+ W = 2";- 2E / f-t we get

(22)

or

(23)

which displays the e-value of the p-adic Kepler-like potential

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

A. Frankl ,2 , R. Lemus l , R. BijkerI, F. Perez-Bernal3 , and J. M. Arias 3

1 Institutode Ciencias Nucleares, U.N.A.M.

A.P. 70-543, 04510 Mexico D.F., Mexico
2Instituto de Fisica, Laboratorio de Cuernavaca
A.P. 139-B, Cuernavaca, Morelos, Mexico
3Departamento de Fisica Atomica, Molecular y Nuclear
Facultad de Fisica, Universidad de Sevilla
Apdo. 1065, 41080 Sevilla, Espana

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

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 99
[15] to use a U(k + 1) model for the k = 3n - 3 rotational and vibrational degrees of
freedom of a n-atomic molecule. This model has the advantage that it incorporates all
rotations and vibrations and takes into account the relevant point group symmetry, but
for larger molecules the number of possible interactions and the size of the Hamiltonian
matrices increase very rapidly, making it impractical to apply.
Although the algebraic formulations have proved useful, several problems remained,
most important of which is the absence of a clear connection to traditional methods.
On the other hand, a related problem is the lack of a systematic procedure to construct
all physically meaningful interactions in the algebraic space. In this paper we show that
both these issues can be resolved by considering a symmetry-adapted version of the U(2)
algebraic model for the analysis of molecular vibrational spectra .. In this approach
it is possible to construct algebraic operators with well defined physical meaning, in
particular interactions fundamental for the description of the degenerate modes present
in systems exhibiting high degree of symmetry. The procedure to construct them takes
full advantage of the discrete symmetry of the molecule and gives rise to all possible
terms in a systematic fashion, providing a clear-cut connf'ction between the aigebraic
scheme and the traditional analyses based on internal coordinates, which correspond to
the harmonic limit of the model [16].
As a test of the symmetry-adapted approach we discuss an application to three
:D3h-triatomic molecular systems, namely Ht, Be3 and Nat, and to two tetrahedral
molecules; the Be4 cluster and the methane molecule. Since small molecules can in
general be well described by means of ab initio calculations [17, 18], we emphasize the
basic purpose of this work. We establish an exact correspondence between configuration
space and algebraic interactions by studying the harmonic limit of the U(2) algebra.
This general procedure not only allows to derive algebraic interactions from interactions
in configuration space, but can also be applied to cases for which no configuration space
interactions are available. The 'D3h-triatomic molecules constitute the simplest systems
where degenerate modes appear and where the new interactions in the model become
significant. In the caseof Be4 we present a comparison with ab initio calculations, while
for CH 4 we present a detailed comparison with experiment. The application of these
techniques to other molecules, as well as a more complete presentation can be found in
references [16,19,20,21].

THE U(2) VIBRON MODEL

The model is based on the isomorphism of the U(2) Lie algebra and the one dimen-
sional Morse oscillator

1- = _ 1i,2 ~ + D(e- 2:c/d _ 2e~:c/d) , (1)

2p, dx 2

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

which corresponds to a two-dimensional harmonic oscillator (in units where 1i, = p, =

e = 1) associated to a U(2) symmetry algebra [23]. By carrying out a change of variable

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)

from which we immediately obtain

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

[Ni' lit,;] = 0, (11)

with It = , O. As mentioned before, for the symmetric irreducible representation [Ni , 0]

of Ui (2) one can show that the Casimir operator is given by 1';2 = Ni(N; + 2)/4 [23],
from which follows the identification j; = N;/2. The 80;(2) label is denoted by mi.

THE Be4 CLUSTER

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)

The Hamiltonian operator can be constructed by repeated couplings of these tensors to

a total symmetry AI, since it must commute with all operations in To. [14].
All calculations are carried out in a symmetry-adapted basis, which is projected
from the local basis

@ U6 (2) @ 806 (2) J 80(2)

L L L (14)
[N6] V )
in which each anharmonic oscillator is well defined. By symmetry considerations, Ni =
N for the six oscillators, Vi = N i /2 - mi denotes the number of quanta in bond i and

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

I \?~) = L>~~,i Ii) . (15)

i=l

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)

The relevant commutator can then be expressed as

1 A
1- 2Vi
-2.],0 . (23)
Ni " Ni '

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

iI1 = 9 22 fA2 iA2 +933 ~

L iF! iF! "I "I . (26)
"I

With this method we obtain an algebraic realization of arbitrary configuration space

interactions. As a simple example, a one-dimensional harmonic oscillator Hamiltonian
iIi = (b;b i + bi bt)/2, transforms into
(27)

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]

iI = WlHA! + W2 HE + W3 HF2 + X 12 (HA! itE) + X 13 (itA! itF2) + X33 (HF2) 2

+933 L i~'1~! + t33 6 33 + t23 6 23 , (28)
"I

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.

V (vl,v;',vD r Ab initio Fit V (Vb v;', vD r Ab initio Fit

N-+oo N=44 N-+oo N =44

1 (1,0,0) Al 638.6 637.0 3 (1,0,2) Al 2106.8 2105.6

(0,1\ 0) E 453.6 455.0 (1,0,22 ) E 2000.1 1999.8
(0,0,11) F2 681.9 678.2 F2 2056.8 2052.8
2 (2,0,0) Al 1271.0 1269.2 (0,3\ 0) E 1341.3 1343.7
(1,1\ 0) E 1087.1 1087.0 (0,3 3 ,0) Al 1355.5 1352.5
(1,0,11) F2 1312.6 1308.3 A2 1355.5 1354.4
(0,2,0) Al 898.3 901.4 (0,2,2,11) F2 1565.5 1565.7
(0,2 2 ,0) E 905.4 906.1 F2 1584.4 1583.1
(0,1\11) Fl 1126.7 1125.1 (0,22,11) Fl 1578.5 1578.0
Fz 1135.5 1134.1 (0,1 1 ,2,2) E 1821.4 1821.6
(0,0,2) Al 1484.0 1483.0 E 1929.5 1929.0
(0,0,22) E 1377.3 1373.9 (0,11,22) A2 1813.3 1813.1
F2 1434.1 1429.6 Al 1830.8 1831.7
3 (3,0,0) Al 1897.0 1896.7 F2 1874.4 1873.2
(2,11,0) E 1714.3 1714.3 Fl 1883.2 1883.0
(2,O,t!) F2 1937.0 1933.7 (0,0,31,3) F2 2136.5 2134.2
(1,2,0) Al 1526.6 1529.2 F2 2327.3 2326.9
(1,2 2 ,0) E 1533.7 1532.8 (0,0,33 ) Fl 2199.8 2197.1
(1,11,11) Fl 1752.2 1749.7 Al 2256.5 2254.4
Fz 1761.0 1759.8

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

1 (0,11) E -1.55 0.51 0.93

(1,0) Al 0.42 0.02 1.95

2 (0,2) Al 7.48 -0.74 0.37

(0,22) E -5.69 0.17 0.84
(I,ll) E -0.61 0.82 1.68
(2,0) Al -0.11 -0.04 1.26

3 (0,3 1 ) E -4.46 -2.05 -1.19

(0,3 3 ) Al 3.18 -1.23 -0.34
(0,3 3 ) A2 2.44 0.61 -0.33
(1,2) Al 0.66 1.90 -U.Ol
(1,22) E -5.00 -1.36 0.34
(2,rt) E 4.07 0.79 -0.19
(3,0) Al -1.23 -1.66 -2.06

r.m.s. 5.84 1.35 1.33

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

For V 3h molecules we follow a similar procedure, namely, we construct the 'D3h

symmetry-adapted operators and states analogous to eq. (13,15) and carry out the
building up procedure to construct the Hamiltonian and states with a higher number of
quanta with the appropriate projection operators and Clebsch-Gordan coefficients [19].
In Table 2 we present the fits to the spectra of Be3, Nat and Ht up to three quanta.
While remarkably accurate descriptions of the first two molecules can be achieved using
a four-parameter Hamiltonian, we were forced to include four additional higher order
terms in the Ht Hamiltonian in order to properly describe this molecule. This is in
accordance with the work of Carter and Meyer [18], who were forced to include twice
as many terms in the potential energy surface for Ht than for the Nat molecule. The
Ht ion is a very "soft" molecule which, due to the light mass of its atomic constituents
carries out large amplitude oscillations from its equilibrium positions [18].

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)

As before, the algebraic Hamiltonian can be constructed by repeated couplings of these

tensors to a total symmetry AI.
The methane molecule has nine vibrational degrees of freedom. Four of them cor-
respond to the fundamental stretching modes (AI ED F 2 ) and the other five to the fun-
damental bending modes (E ED F2 ) [27]. The projected tensors of Eqs. (30) and (31)
correspond to ten degrees of freedom, four of which (AI ED F 2 ) are related to stretching
modes and six (AI ED E ED F2 ) to the bendings. Consequently we can identify the tensor
T:~,b as the operator associated to a spurious mode. This identification makes it pos-
sible to eliminate the spurious states exactly. This is achieved by (i) ignoring the TA~,b
IL,

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 zeroth order vibrational Hamiltonian is riow written as

fIo = WI HAl" + W2 HEb +W3 Hp2" + W4 itP2 ,b + W34 itsb

+0:2VEb + 0:3 Vp + 0:4 VP2,b + 0:34 Vsb .
2" (34)

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

corresponds to a constant Ns + 2. A similar relation holds for the bending interactions,

but in this case the interaction VA, b has already been excluded in order to remove
the spurious Al bending mode. The subsqipts of the paraineters correspond to the
(VI, V;" V!4)
v~", labeling of a set of basis states for the vibrational levels of CH4 . Here
VI, V2, V3 and V4 denote the number of quanta in the AI,., Eb, F 2 ,s and F 2 ,b modes,
respectively. The labels Ii are related to the vibrational angular momentum associated
with degenerate vibrations. The allowed values are Ii = Vi, Vi - 2, ... ,1 or 0 for Vi odd
or even [27].
In the harmonic limit the interactions of Eqs. (32) and (33) again attain a partic-
ularly simple form, which can be directly related to configuration space interactions.
This limit is obtained, as before, by rescaling J+,i and L,i by ,;N; and taking Ni --+ 00,
so that

.
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

b~xt == L o~~ bt , (38)

i;;::l

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]

if = WI HAl" + 6.12 HE. + 6.13 HF", + 6.14 HF2,b + 03 VF 2"

,)2
+Xll (?iA
" ,, + X 22
(,)2
?iEb + X33 (?iF2"
,)2 + X (,)2
?iF2,b 44

+X12 (HAl,' HEb) + Xu (HAl,' HF"b)

+X23 (HE. HF2,,) + X24 (HEb HF2,b) + X34 (HF2" HFZ,b)
+g22 'A) 2
(1 Z + g33 '~ ls,~ ls,~( + g44 '~
" 'F 'F " 'F 'F " 'F 'F
lb,~ lb,~ + g34 '~ ls,~ lb,~
~ ~ ~

+t33 OS8 + t44 Obb + t34 Osb + t 23 O2 + t24 02b . (:39)

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 ,

+i V2 ~x [t~2,X x t:2'X]~1 (40)

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 .

Parameter Fit 1 Fit 2 Fit 3 Fit 4 Fit 5 Fit 6

Ns 43 43 43 00 00 00

Nb 150 150 150 00 00 00

WI 2977.60 2966.17 2970.03 2967.40 2966.81 2969.75

W2 1554.83 1549.96 1550.88 1558.38 1558.51 1548.30
W3 3076.45 3076.41 3079.33 3081.34 3082.00 3060.48
W4 1332.22 1329.52 1337.69 1337.51 1337.54 1338.64
C3 582.87 480.33
Xll 3.69 10.23 6.06 -21.30 -21.19 -21.59
Xn 1.30 1.32 1.37 -1.17 -1.17 -0.28
X33 5.43 6.97 5.34 -10.79 -11.12 -7.89
X 44 -3.47 -3.64 -4.41 -6.26 -6.27 -7.26
Xl2 -3.60 -0.94 -1.47 -3.39 -3.28 -2.80
X I3

X I4 -2.86 -0.49 -2.30 -3.10 -3.00 -4.48

X 23 -11.14 -8.75 -10.68 -7.97 -8.10 -3.80
X 24 1.00 0.91 2.03 -5.37 -5.37 -4.74
X 34 -5.60 -3.97 -6.50 -3.46 -3.50 -1.21
g22 -0.46 -0.46 -0.41 0.37 0.37 -0.62
g33 0.19 -1.23 0.25 -4.35 -4.22 -5.49
g44 4.07 4.11 3.79 4.98 4.98 5.30
g34 -0.65 -0.72 -0.56 -0.74 -0.87 -0.67
i33 0.10 0.16 -l.25 -1.27
i44 1.00 1.00 0.56 0.56
h4 0.21 0.24 0.24 0.25
in -0.39 -0.39 -0.39 -0.39
t24 0.13 0.13 0.91 0.91
834 -0.11

r.m.s. 1.16 4.49 7.81 20.42 20.90 21.00

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.

SUMMARY AND CONCLUSIONS

In summary, in this paper we have studied the vibrational excitations of sever-

al molecules in a symmetry-adapted algebraic model. In particular, for the methane
molecule we find an overall fit to the 44 observed levels with a r.m.s. deviation of 1.16
cm- l , which can be considered of spectroscopic quality. We pointed out that for this
calculation the (\;3 term in eq. (39) in combination with the anharmonic effects in the
other interaction terms plays a crucial role in obtaining a fit of this quality. Purely an-
harmonic terms of this sort arise naturally in the symmetry-adapted algebraic model,
but vanish in the harmonic limit. Physically, these contributions arise from the anhar-
monic character of the interatomic interactions, and seem to play an important role
when dealing with molecular anharmonicities, especially at higher number of quanta.
This conclusion is supported by our other applications of the symmetry-adapted model
to the Be4 cluster [16] and the Ht, Be3 and Nat molecules [19], as well as our study of
two isotopes of the ozone molecule [20].
These studies suggest that the symmetry-adapted algebraic model provides a nu-
merically efficient tool to study molecular vibrations with high precision. The main
difference with other methods is the use of symmetry-adapted tensors in the construc-
tion of the Hamiltonian. In this approach, the interactions can be constructed in a
systematic way, each term has a direct physical interpretation, and spurious modes can

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

Al (1000) 2916.32 2916.48 -0.16 (01U) 5844.98

E (0100) 1533.46 1533.33 0.13 (1200) 5974.81
F2 (0001) 1309.86 1310.76 -0.90 (IOU) 7147.49
(0010) 3018.09 3019.49 -1.40 (0021 ) 7303.38
(2100) 7315.60
Al (0002) 2587.77 2587.04 0.73 (0120) 7479.48
(0200) 3063.66 3063.65 0.01 (0120) 7557.17
(0011) 4323.81 4322.72 1.09 (1020) 8833.05
(2000) 5790.13 5790 0.13 FI (0003) 3920.46 3920.50 -0.04
(0020) 5966.57 5968.1 -1.53 (0102) 4128.38 4128.57 -0.19
E (0002) 2624.14 2624.62 -0.48 (0201) 4364.39 4363.31 1.08
(0200) 3065.22 3065.14 0.08 (0012) 5620.08
(0011) 4323.09 4322.15 0.94 (0012) 5630.76
(UOO) 4446.41 4446.41 0.00 (U01) 5755.58
(0020) 6045.03 6043.8 1.23 (0111) 5829.79
Fl (0101 ) 2845.35 2846.08 -0.73 (0111) 5848.94
(OOU) 4323.15 4322.58 0.57 (0210) 6061.57
(0110) 4537.57 4537.57 0.00 (1011) 7147.53
F2 (0002) 2612.93 2614.26 1.33 (0021) 7303.29
(0101) 2830.61 2830.32 0.29 (0021) 7343.21
(1001) 4223.46 4223.46 0.00 (1110) 7361.79
(0011 ) 4321.02 4319.21 1.81 (0120) 7518.70
(0110) 4543.76 4543.76 0.00 (0030) 8947.65 8947.95 -0.30
(1010) 5845.53 F2 (0003) 3871.29 3870.49 0.80
(0020) 6003.65 6004.65 -1.00 (0003) 3931.36 3930.92 0.44
(0102) 4143.09 4142.86 0.23
Al (0003) 3909.20 3909.18 0.02 (0201) 4349.01 4348.77 0.24
(0102) 4131.92 4132.99 -1.07 (0201) 4378.38 4379.10 -0.72
(0300) 4595.26 4595.55 -0.29 (1002) 5523.80
(1002) 5498.66 (0012) 5594.92 5597.14 -2.22
(0012) 5617.16 (0012) 5620.68
(0111 ) 5836.11 (0012) 5632.36
(1200) 5973.26 (1101) 5740.86
(1011) 7147.56 (0111) 5830.28
(0021) 7300.85 (0111) 5848.46
(0120) 7562.91 (0210) 6054.58
(3000) 8583.81 (0210) 6067.03
(1020) 8727.97 (2001) 7094.16
(0030) 8975.64 8975.34 0.30 (1011) 7145.84
A2 (0102) 4161.52 4161.87 -0.35 (0021) 7266.11
(0300) 4595.28 4595.32 -0.04 (0021) 7303.38
(0111) 5844.61 (0021) 7344.87
(0120) 7550.53 (1110) 7365.83
E (0102) 4105.22 4105.15 0.07 (0120) 7514.67
(0102) 4152.15 4151.22 0.93 (2010) 8594.90
(0300) 4592.13 4592.03 0.10 (1020) 8786.05
(1002) 5535.04 (0030) 8907.91 8906.78 1.13
(0012) 5620.36 (0030) 9045.36 9045.92 0.56
(0111) 5836.45

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

D. V. Fursa\ Yu. F. Smirnov2* and V. N. Tolstoy3

1 School of Physical Sciences

The Flinders University of South Australia
G.P.O. Box 2100, Adelaide 5001, Australia
2Instituto de Ciencias N ucleares
U.N.A.M., A.P. 70-543
04510 Mexico D.F., Mexico
3Institute of Nuclear Physics
Moscow State University
Moscow 119899, Russia

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-

tragredient Lie super algebras. )

Among various papers concerned this question two papers [5] and [6] should be
mentioned, in which two universal approaches to the construction of Dyson realizations
for semisimple Lie algebras were suggested. In ref. [5J the generalized coherent state
(GCS) [7] technique was used and the algorithm of construction of a single definite
Dyson type boson realization for a Lie algebra was formulated. The method of ref. [6J
was based on the properties of the elementary representation (or Verma module) of a
Lie algebra and permits to obtain a number of equivalent Dyson type boson realizations
for the same Lie algebra.
In present paper the generalization of these two approaches on the arbitrary contra-
gredient Lie superalgebras is given and the interrelations between them are discussed.
The new element of our aproach is the application of the factorized form of the GCS. It

On leave of absence from Institute for Nuclear Physics at Moscow State University, Moscow
119899, Russia

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press. New York. 1997 117
should be noted that factorized form of the Lie algebra representation operator, in fact,
was used in ref. [8] in connection with the solution of the time-dependent Schrodinger
equation.
The GCS and BFR of Lie superalgebras can be used as the powerful tools for the
development of the Lie superalgebras representation theory. The impressive advances
in the solution of the bosonization problems were obtained using the vector coherent
state (VCS) theory and mixed boson realizations [9].
The problems discussed in this paper are of physical interest. The general prob-
lem of realization of the Lie superalgebras in terms of polynomial combination of some
number of the boson and fermion creation and annihilation operators arises in con-
nection with the microscopical foundation of the interacting boson-fermion model and
various supersymmetry schemes in nuclei [10, 11]. Mixed boson realizations proved to
be very useful in the microscopical nuclear collective model Sp(2d, R) [12]. In higher
energy physics boson representations of the affine Kac-Moody algebras are a subject of
intensive study [13].
The paper is organized as follows. In sect. 2 some necessary elements of the fi-
nite dimension contragredient Lie super algebra theory are presented. A symmetric
and ordered bases in the elementary representation of the finite dimension contragre-
dient Lie super algebra are introduced. In sect. 3 the functional realization of the Lie
superalgebras are constructed using the GCS. In sect. 4 it is shown that the matrix
representation [14, 15, 16] of the Lie superalgebras can be easily found if the functional
realization is known. (The explicit formulae of the matrix representations are given in
appendices A and B). In sect. 5 the Usui operator technique is discussed and an ele-
mentary BFR (EBFR) and Dyson BFR (DBFR) of the Lie superalgebra generators are
described. The generalization of the adjoint operation [17] permits to analyse the con-
nection between EBFR and DBFR. The main points of the VCS theory and mixed BFR
are formulated in sect. 6. In sect. 7 we discuss the construction of the boson realizations
of the su(3) algebra and the boson-fermion realization of the osp(2/1) superalgebra as
illuminating examples.

PRELIMINARY

Let g = go + gl be a finite-dimensional contragredient Lie (super)algebra [17]; IT :=

aI, a2, ... , aT be a system of simple roots with respect to IT; fl. = fl.+ u (-fl.+) be the
total system of roots; 0(e) be a parity function on g (0(go) = O,0(gl) = 1) and let
hi, e" (a E fl., i = 1,2, ... , 1') be the "normalized" Cartan-Weyl basis of g. This basis
satisfies the following relations [18]:

(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

[a, b] == (-1 )0(a)0(b)+1 [b, a] , (5)

(-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)

[hi, e~l = n/3(hi)e~ , (9)

[e,6,e~,6J = ne":.~l(/3;hi - ~(n -.1)/3;/3i) (10)

for even elements e,6.
In this paper we shall use the two following standard involutive antiautomorphisms
in U(g), wich are defined by the formulae

(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

pj(h;)u = [h;,uJ + j;u , (15)

(16)

pA e",)u = L pA[e", e,6])u~ + pj( e_(6)pj( eo:)u~ (17)

,6E~+

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 )

is a (super )symmetrizator summing over all permutations P of the generators; 0( P) is

the parity of the permutation. The powers n", of even generators E" can be arbitrary
non-negative integers. For odd generators EO! the power n" can only be 0 or 1, because
the element E~ is either zero or proportional to the even generators in this case.

GENERALIZED COHERENT STATES AND FUNCTIONAL REALIZA-

TION OF THE (SUPER)ALGEBRA GENERATORS

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)

we can rewrite (22) in the form

(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)

Let g(9) = go(9o) E9 gl (9d be the Grassmann envelope of Lie superalgebra 9 =

go + gl The algebra g(9) consists of formal linear combinations L: j 1]jej, where {ej}
is a basis of g, 1]j E 9. The elements ej and 1]j with the same indices j are assumed to
be of the same parity. It is convenient to use the following permutation relation for the
generators e of superalgebra 9 and the element 1] of the Grassmann algebra 9

(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

9; = 90, 9; = 91 and 9~ = 90, 9f = 9l,

(V1])* = 1]*v*, (c1])* = C1]* and (V1])O = (_l)B('1)B(v)1]Vo,

(C1])O = c1], (1]*)* = 1] and (1]0)0 = (-1/(1)1] (29)

for homogeneous V,1] and complex c.
The adjoint operation "+" and grade adjoint operation "+"
in the superalgebra 9
in combination with operations "*,, and "0,, in the Grassmann algebra 9, induce the
adjoint operation "+,, in the algebra g(9) [20]

(30)
and

(1]l e 1 + 1]2 e 2 )+ = e 1t1]1o

+t
e 2 TJ2' (31)
It can be readily verified that

(x+)+ = x, [x,y]+ = [y+, x+], x,y E g(9). (32)

The supergroup G is the exponential mapping of g(9)

G = {exp(x), x E g(9)}. (33)

Therefore the GCS for supergroup G could be introduced in analogy to formula (24).
However, it is more convenient to introduce the two types of GCS (normal ordered and
symmetric ones) in the space of the elementary representation of the superalgebra g.
They are determined by the relations

(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

AIC} = IC} f (A); (37)

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

70 (e_ ap ) =8"p + L Bal B;;/xp((a), a p E !ltv, (43)

1=1

122
 N

fo (e ap ) = 2:= 8al B~i((a), a p E tl(ginv), (44)

1=1

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

Here {kn are non-negative integers satisfying the equation

I-I

al = a p + 2:= k~as , (49)

5=1

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]

fs (hi) = ji - 2:= (ai,,8) 8(3 t(3 . (50)

(3EL1;nv

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

fs (e_,,) = Ldk L 8(3k+l e"l .. e"'k N -C>1,-(31 , .. N_ OIk ,_(3k' (52)

k=O O'l"'CXk

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

is (eO') = -(a,j)~a - L 8,,-"1 ~'YN_'Y'O'

'YEtltnv

+ (L d~ L 8Pk+l ~'" ... ~akNpl,al ... Npk,ak)

k=O 0:1"'CXk

x ( L L ~P~O'+Ii_pNP-Ii-O',aN-p,P-1i - L (a, .5)~Ii~O')' (54)

PEi'l.tnv (Ii-P}Ei'l.(ginv) liE i'l. tnv

where (31 = 5, (3k = .5 + ~7~} al , and

, (_l)k k-l, (_l)k-l ,

dk = (k + 2)! - t;d1 (k -I + l)!,d k = 1/2,1/12,0, ... (55)

The functional realization of raising generators corresponding to composite roots

can be found from (46) and (54) using the commutation relations
+-
IC)[f (e,,), f (ep)] = IC) f (eO'+p)NO'"i3 . (56)

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).

GENERALIZED COHERENT STATES AND MATRIX REPRESENTA-

TION OF THE (SUPER)ALGEBRA GENERATORS

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.

Aln) = L An"nln') L An+k,nln + k).

= (59)
n' k

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

where the evident relation

(62)
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

Ie) = LL In)(nl (64)

n

Now we substitute this expression in the definition (26) of the functional realization

L L Aln)(nl = L L In)(nl 7(A). (65)

n n n n

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

(nl f (A) = L(n + kIA~+k,n . (66)

k
Using eqs. (59)-(66), we can conclude that

Aln) = L A~_k,11ln - k), (67)

k

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)

(nl aap= (n - 1ex pln exp ( _l)B(a p)I:f:,' B(",;) , (69)

it is easy to find the explicit expression for coefficients A~_k n and, consequently, the
matrix form of the generators Ea or Hi in ordered and sym'metric bases. The corre-
sponding formulae for coefficients A~_k11,A~_k11 in the expansions (59) are given in
appendices A and B, respectively. ' ,
It is possible to formulate a formal receipt to find the matrix representation of some
operator, when its functional representation is known: Each term aa (rya)k containing
the derivative should be substituted by the factor (n" + 1 - k), and in the vector In)
the substitution n -> 11 + 1", - k" should be done. Each term (7)",)k without derivative
leads to the substitution n -> n - kG' in the vector In).

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

[COt, ct] = D",(3 , [C"" C(3] = [ct, ct] = 0 ,

[A, B] = AB - (_1)O(A)8(B) BA, (72)
where a parity function fJ is defined in following way: fJ( C"!) = 1 if C"! is a fermion
operator, and fJ( C') = 0 if C' is a boson operator. The operators COt, C: form the
Heisenberg-Weyl (HW) superalgebra [3].
For the HW superalgebra as well as for the arbitrary Lie superalgebras two types of
adjoint operations can be introduced. The first type corresponds to ordinary Hermitian
conjugation

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+

where 10) is a boson-fermion vacuum (C'IO) = 0).

It is easy to find the functional representation (Grassmann-Bargmann representa-
tion) of the operators COt and C;;

(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

form a dual basis with respect to the basis (77)

(nlm) = cnm . (79)

The linear mapping U defined by the formula

In) = Uln) , (80)

is realized by the so-called Usui operator.

It is easy to demonstrate that the Usui operators Uo and Us are of the form

Uo = LLln;o)(nl = (Ol{ II exp(E_"C,,)}li) (81)

n n O'Etltnv

and
Us = L Lin; s)(nl = (01 exp( L E_"Ca) Ii). (82)
n n CtEA.tnv

It follows directly from eq. (80) that the following formula

(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

[Ca , A] =0. (85)

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)

Formula (90) leads to a result of similar structure

(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

U: = L L In)(n; 01 = UI II exp(C: Ea)IO), (93)

n

(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)

where 1L (n+) is a nilpotent subalgebra of lowering (raising) generators that do not

belong to the subalgebra K. The elements of n_(n+) can be choosen as components of
irreducible tensors with respect to the subalgebra K. Some IR Dj of algebra 9 can be
decomposed into the direct sum of the IRs DA of the K algebra

(97)

Let us take some basis vector IjAA) belonging to some definite IR DA of the K sub-
algebra and construct the expressions

IGA,o) = II exp(Ko(o)ljAA), (98)

E"En+

IGA,s) = exp (L K,,(a)ljAA). (99)

EaEn+

A normal ordering is used in (98).

The ves are defined as a set of vectors (98) and (99) with all possible A values.
The Usui operators and bases in the universal enveloped algebra can be constructed in
analogy with the previous section. The modification consists in the substitution of a
highest weight vector Ij) in (19), (20), (34), and (35) by the vectors IjAA).
In principle, the choice of the IR DA is arbitrary, but it is convenient to use in some
sense "highest" ("lowest") representation defined by the relations:

(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

The general scheme of construction of DBFR is illustrated in this section by the

examples of the U(3) algebra and the osp(2/1) superalgebra.
The generators of the U(3) algebra Aik (i, k = 1,2,3) satisfy the commutation
relations
(103)
The generators A 13 , A 23 , Al2 are called the raising generators and the generators A31,
A 32 , A21 the lowering generators. As for generators Au, An, A33 they form the Cartan
subalgebra of U(3).
Two variants of normal order for raising generators are possible, namely: A 13 , A 23 ,
Al2 and A l2 , A 13 , A 23 As a result two ordered forms of the Usui operator U+ can be
written
ut = UI exp(A12Bin exp(A13Bit) exp(AZ3Bit) 10), (104)
Ui = UI exp(A23B it) eXp(A13Bit) exp(A 12 BinIO), (105)
while the symmetrical form of U+ is an unique one and can be written as

(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)

UI exp(A12Bit) A 12 CXp(A13Bit) exp(A23Bit)lO)

- UIexp( A 12 Bit) exp( A13Bit)A13Bit exp( A23Bit) 10)
(Bi2 - Bit Bi3)Ut .
Thus we may conclude that

(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)

PsD(A 12 ) -_ BS12 - 2"1 BS+Bs

23 13' (127)

p~(A23) = B;3 + ~B:t B:3 , (12S)

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

+ (Btt + ~Btt B~t)Bt2 , (129)

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)

(11. - )2. )Bs+

12 -
BS+Bs+Bs
12 12 12 -"21 (B13+ + 2"1 B 12S+B23+)B12s+B13s
8 8

S
-
(8 8+
. 13 -"21 BS+Bs+)B
12 23 23' (131)

p~(All) = j1 - Nt2 - Nt3 , (132)

P~(A22) = h + Nt2 - N;3 , (133)
p~(A33) = j3 + N;3 + Nt3' (134)
The BCH formula gives a possibility to find the relations between three different
forms of DBR for the U(3) algebra.
To find the relation pP(A;j) with p?(A;j), one should combine the three exponents
inU: into a single exponent using the BCH formula:

exp(A12B~n exp(A13Bttl exp(A23Bitl =

- (A 12 B1+
exp 12 + A 13 B1+
13 + 23 23 + 2" 12
1+A13
.. A B 1 + 1 B1+B23 ).

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):

exp( A nBtn exp( A13Bit) exp( A 23 Bit) =

exp(A23Bit) exp(AdB~t + Bit Bit)) exp(A12Bii)
As a result we have

Bit = B;t , Bit = B~t , B;t = Bit + BIt Bit ,

Bt2 = B;2 + B~tB;3 , Bi3 = B~3 + B;t B;3 , Bi3 = B;3 . (136)

Similarly the relation between pf (Aij) and p~ (Aij) can be obtained

= B 122+ , B 2s3+ = B223+ 13 - ~

Bft , B 13
8+ -
- B2+ 2 B2+ 2+
12 B 23,

B212 = B 12 -"21 B 23S+B 13'

8 S 8 23
2
=.as23 - 2"1 B 12S+B 13, S B213 = BS13' (137)

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:

[Jo, hJ = h , [h, LJ = 2Jo , [V, hJ = 0, [J,!" VJ = ~7,!, ,

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)

where b is a boson operator and a is a fermion operator, 10) is the boson-fermion

vacuum, and /j) is the highest weight of the representation of the superalgebra osp(2/1).
Similarly to the previous consideration the EBFR can be written in the form:

(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)

pE(V+) = ~b+ba - a+b - ~ja . (157)

4 2
It is well known that the osp(2/1) superalgebra has only the grade adjoint operation

(158)

If the grade adjoint operation for the fermion operator is taken in the form

at = (_l)ll+1 a+ , (a+)t = (-l)lla, p = 0 or 1, (159)

the DBFR can be obtained immediately using formula (92):

pD(J+) = b, (160)

pD(Jo) = j - b+b - ~a+a, (161)

2
pD(J_) = b+(2j - b+b - a+a) , (162)
pD(V+) = (_1)It+f+1(a + ~a+b) , (163)

pD(V- ) = (-I)'*(~b+ba
4
+ b+a - ~ja+)
2' (164)

ACKNOWLEDGMENTS

One of us (D.V.F.) acknowledges financical support from The Flinders University

of South Australia. This work was also supported by the Russian Foundation for
Fundamental Research, Grant No. 96-0101421.

APPENDIX A: ELEMENTARY REPRESENTATION IN ORDERED

BASIS

From formula (43) it follows that

p-1

P(KCip)jn,o) = (nap + 1) In + l,'p, 0) + L B;:ICip(k~, n)ln - kl, 0), (165)

l=l

while for Cl'p E 6. (gin.) we have from (44)

N

p(eCip) = LB~i(k~,n)ln - kl,o). (166)

[=1

According to eq. (45) we obtain

N
p(Hdln,0) = (ji -L(Cl'I,Cl'i)n"I)ln,o). (167)
[=1

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)

APPENDIX B: ELEMENTARY REPRESENTATION IN SYMMETRIC

BASIS

From eq. (50) we have

p(Hi)ln,s) = (ji - L (ai,,8)nl1)ln,s}. (170)

I1Ef:l;nv

According to eq. (51) we find for E", E ginv

p(Ea)ln, s) = L N- I1 ,a(nl1+a + l)ln + 1,,-11 - 111, s}. (171)

I1Ef:l;nv

For lowering generators E_",(a E ~t.v) we obtain from eq. (52) the result

p(E_a)ln, s) = Ld k L N- cxl ,-111 ... N- CX k,-l1k

k=O OftCXk

k k

X (n(3k+l + 1- LO(,8k+l - ai))ln - L 1"" + l(3k+l}, (172)

i;::;;:l i=l

For simple roots E",(a E ~t..) we have according to eq. (54)

p(E,,)ln,s) = (a,j)ln -1",} - ~ N_'Y,ex(n a + 1) + In + la -l",-'Y's}

'YELl.tnv

+ 2: N I1 - 5- a,ex N -I1,I1-5 2: d~ 2: N I1I ,al ... N I1 k. ak (nl1k+1 +1 - 2: 8(,8k+! - a;)

I1Ef:l;nv k=O exl .. ak i=1

- 8(,8k+! - ,8) - 6 (,8k+! - (a + 6 - ,8)) ) In - L la, + 111k+1 - 1a+5-I1}

i:::::l

- L (a, 0) L d~ L N I1I ,QI Nl1k ,ak (nl1k+1 +1

oEf:l;nv k=O al ..ak

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

p(A13 )/n,s) = (n13 + l)/n + h3,S), (175)

1
p(A12)/n, s) = (n12 + l)/n + 113 , s) - 2(n13 + l)/n - 123 + 113 , s), (176)
1
P(A23 )/n,s) = (n23 + l)/n + 123 ,s) - 2(n13 +l)/n - h2 + 113 ,s), (177)

P(A32 )/n, s) = (j2 - h + 1 - 2n23)/n - 123 , s) - ~(n13 - n12)/n - In, s)

1
- (n12 + l)/n + 112 - h3, s) + 4(n13 + l)/n - h2 + h3 - 223, s), (178)

P(A3d/ n , s) = (jl - j3 + 1 - n12 - n13 - n23)ln - 113 , s) + ~(j1 + h - 2j2

1
+ n23 - n12 - 2)ln - 123 - 112 , s) - 4(n13 + l)/n - 212 + h3 - 223 , s), (179)

P(A21)ln,s) = (j1 - j2 + 1- nd/n - hz,s) - ~(n13 - ~n23)/n -l}2,s)

1
- (n23 + l)ln + b3 -113 ,s) - 4(n13 + l)/n - 212 + 113 - In,s) (180)
p(All)ln,s) = (jl - n12 - n13)ln,s), (181)
P(A22) In, s) = (j2 + n12 - n23) In, s), (182)
P(A33)ln, s) = (h + n23 + ndln, s). (183)

REFERENCES

[1] F. J. Dyson, Phys. Rev. 102 (1956),1217.

[2] D. Janssen, F. Donau, S. Frauendorf and R. V. Jolos, Nucl. Phys. A 172 (1971),
145.
[3] F. A. Berezin, Introduction to the algebra and analysis with anticommuting vari-
ables, Moscow State University Press, Moscow, 1983 (in Russian).
[4] M. Scheunert, Lecture Notes in Mathematics, Springer, Berlin, Heidelberg, New
York, 1979.
[5] J. Dobaczewski, Nucl. Phys. A 369 (1981), 213,237; Nucl. Phys. A 380 (1982), 1.
[6] Yu. F. Smirnov, V. N. Tolstoy, A. A. Saharuk, Me Havlicek and C. Burdic in: Group
Theory Methods in Physics, 2 (Nauka, Moscow, 1983) 517; Yu. F. Smirnov and

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

Juan Garcia Escudero

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

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 I 139
dimensions, then the p-facets transforming under the subgroup, are projected on tiles
of the same form and different orientations.
Given a set of tiles it is sometimes possible to construct aperiodic tilings, by using
matching rules for the p-facets. In two dimensions these matching rules are usually
indicated by placing different arrows on the edges of the basic tiles that constrain the
way two tiles appear together edge to edge. However it is difficult for instance to get
a growth of a Penrose tiling without defects, adding one tile at a time in accordance
only with the edge matching rules. In order to get indefinite growth it is necessary to
use the so called vertex matching rules which are a catalogue of allowed configurations
around one vertex [7].
It is possible to obtain also aperiodic tilings by a procedure known as substitution
which is a combination of decomposition and inflation. This method can be applied
when every basic tile can be subdivided into smaller copies of themselves and other
basic tiles.
In this work I consider the problem of the description of aperiodic tilings that can
be generated by substitution, by means of a type of formal grammars known as DOL
systems. These are the simplest type of L Systems introduced by Lindenmayer in 1968
to model biological developments in which parts of the developing organism change
simultaneously.
A DOL system (see [8]) is a triple G = {L;, q'>, w } where L; is an alphabet, q'> is
an endomorphism defined on the set L;* of all the words over the alphabet L;, and
w referred to as the axiom is an element of L;*. The word sequence generated by G
consists of the words q'>0 (w) = w, q'>( w), q'>2 (w), q'>3 (w), ... and the language of G is defined
by L(G) = {q'>i(w)/i ~ o}. In the sequel the endomorphism q'> will be defined by listing
the productions for each letter. In the abreviation DOt, 0 means that the rewriting
is context independent (originally, communication between the individual cells is zero-
sided in the development) and D stands for deterministic: there is just one production
for each letter, i.e" the totality of all productions defines an endomorphism on L;*. For a
DOL system with alphabet L; = {at, a2, . .. ,ad, the ij entry of the substitution matrix
Mequals the number of occurrences of aj in q'>[a;].

GRAMMARS FOR APERIODIC STRUCTURES IN DIMENSION ONE

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

U J----+ UVW, V J----+ Uv, W J----+ U (2)

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:

ai J----+ (iJ>p[ai]) = (biHai+6ai-3) bi J----+ (iJ>p[biD = (biHai+6)

ai J----+ (iJ> p[ail) = (biaiHai+3) hi J----+ (iJ>p[hiD = (aib;+6) (3)
(J----+ ( ) J----+)

There is a Coxeter group H(n) = < Rr, R2/(RIR2t = Ri = R~ = e > in relation

with every planar pattern with n-fold symmetry considered in this work. The group
contains two reflections Rl and R2 and a n-fold rotation R I R 2. For Penrose patterns if
the action of the generators of H(S) on the symbols representing tiles is defined in the
following way:

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).

PLANAR PATTERNS WITH ODD SYMMETRIES HIGHER THAN

FIVE

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

WO f---+ </>[WO] = U1 , VO f---+ </>[V0] = VI UO, UO f---+ </>[u0] == WI U1VO ,

(4)
WI f---+ </>[Wl] = Ua , VI f---+ </>[Vl] = U1VO, U1 f---+ </>[U1] = V1UoW a .
The basic triangles T(UO, WO,UO), T(VO, WO,U 1), T(Vt, Wt,UO), T(Wl, wI, Va)
and T(Vl, VI, U1) are represented by letters of type a, b, b, c and d respectively. Tiles
t
of type t and with t = a, b, c, d have the same form, but a different "colour" which
takes into account that the arrowing is different~ The tiles band b have also the same
form but different colours. The triangles band b can be obtained by mirror reflection
on an arbitrary side of the triangles band b respectively, having also different colours.
For the tile T(A, B, C) the vertex An C is placed on the origin and the edge C
on Uo where Uk with 0 :s: k :s: 6 are unit vectors, so that the angle between Uk and
the positive x-axis is equal to br/7. If r represents a 14-fold rotation around the
origin, then a rotation ri followed by a traslation v is denoted by [ri, v]. If we choose,
for simplicity, (length of U)=(3, (length of V)=(3 - (3-1 and (length of W)=l, the
geometrical description of the tiling growth can be expressed in the following way:

cpn[a] = [1, O]<l>n(b] U [r 4 , -(3n+1U6]cpn-l[a]

cpn[b] = [r 6 , U6((3n-2 - (3n)]cpn-l[C] U [r 8 , uo((3n + (3n-l)]cpn-l[b]

U[r 6, U6((3n-l - (3n+l)]cpn-l[d] U [r 9 , _(3n+1uo]cpn-l[d]

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:

Now we assume the equality to be true for n - 1:

p(ijin(U0)) = p(ijin-l((WlU l Va))

= P( ijin-l( (WI) )P( ijin-l ((U l ) )P( ijin-l (V0))
= Mir(P(ijin-l(WO))) Mir(P(ijin-l(UO))) Mir(P(ijin-l(V l )))
(9)
= Mir(P( ijin-l(VlUOWO)) = Mir(P( ijin(U l )))

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)

ICOSAHEDRAL DANZER TILINGS

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]

with the shorthand notation: a = 232312, f3 = 32312132132312, I = 31231, 5 = 3121,

E = 2321, ( = 132323 and Ri1Ri2 ... Rin = R i1 i2 ...in or ri1i2 ...in' depending whether n is
odd or even respectively.
Within this squeme concatenation of words is converted into gluing of tetrahedra.
In analogy with the haptagonal tilings (see eq. (5)) an alternative geometric description
can be given where the tilings are obtained through union of sets of points. The
tetrahedra have a fixed position in 3D space and the bracket structure is not necessary.
If there is no translation involved, the vertex lying on the origin is t 1 n t 2 n t 3 for the tiles
of type A and Band t 1 n t 2 n t 4 for C and K. If we take u = a, b, c, k for A, B, C, K
respectively the unit vectors Uij lie on the edge t i n ti. Valid expressions are obtained
by making the transformations T -+ T and U -+ ii :

iIln[K] = [r", Tn- 1 pk 14 ]iIl n - 1 [K] U [r;3, 0]iIl n- 1[B]

iIln[c] = [Rl,O]iIl n- 1[K] U [R3, Tncn]iIln-1[K] U [Rr,T n+1(TC14]iIl n- 1[C]

U[r5, Tn+1 (TC14]iIl n- 1[C] U [1, Tn+1(TC14] iIl n- 1[A]

iIl"[B] = [r-Y1' Tn+1(Tbn]iIln-1[K] U [R61,Tn+l(Tbn]iIln-1[K] U [r"Tn+1pb 32 ]<l>n-1[B]

U[Rd' Tn+1 pb 32 ]<l>n-1[B] U [R-y, T,,-l pb 32 ]iIl"-1[K] U frs, Tn- 1pb 32 ]iIl n- 1[K]
U[I,O]iIln-I[C]

iIln[A] = [r(, T"+1 pa23]iIln[B] U [R(3' Tn-lpadilln-1[C] U [RI" , T"pa23]iIln-1[K]

U[R3" , Tnpadilln-1[K] U hI, O]<I:>,,-l[B]
(15)
where the unit vectors k14' C12, C14 and b 13 , b 32 , a23 have positive sense outwards the
origin of their respective tiles.

TILINGS AS MODELS FOR QUASICRYSTAL STRUCTURES: THE

FOURIER TRANSFORM

In 1D the simplest model for a quasicrystal is obtained by placing atoms on the

extremes of the segments.
Consider the alphabet ~ = {a, b}, the set of production rules a f - + q,[a] = aab, b f - +
q,[b] = a and the axiom w = a. The characteristic equation of the substitution matrix
is x 2 = 2x + 1. If w" = .pn[a] and ~n is the number of letters of Wn then W n +1 = W~Wn-1
and ~n+I = 2~n + ~n-l'

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):

I; = {L, S},L f---+ LS, S f---+ L ,W = L

in+! = in + in-l exp( iqin),tn+! = in + in- 1
I; = {a, b, c },a f---+ aac, b f---+ ac, c f---+ b ,w = a

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):

An( q) = Bn( q) + exp( -iq,en+!uo)An_l (r 4 q)

Bn(q) = Cn-1(q) + exp(iqu6(,en + ,en-l))Bn_l (r 8 q)+
exp( iquo(,en-2 - ,en) )Dn- l (q) + exp( iquo(,en-l - ,en+! ) )Vn- 1 (r 3 q)

Bn(q) = exp(iquo(,en-l - ,en+I))Bn_l (r 3 q)+

exp( iqu6(,en+! - ,en-I) )(Dn- 1 (r 9 q) + Cn(r 8 q))
Cn(q) = exp( iquo(,en-l - ,en+!) ).An- l (r 6q) + Bn- l (q)
Vn(q) = Bn( q) + Bn- l (rq) + exp( iqus(,en+l - ,en-I) )(An- 1 (r 9 q) + Dn- l (rIOq))
(19)

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)

Cn(q) = K n_1(R11q) + exp(-iqrnc12)Kn_1(R31q) +

exp( -iqrn+1ac14) (Cn_1(R:;-lq) + Cn_1(ri 1q) + A n- 1(q))
Bn(q) = Cn- 1(q) + exp( -iqr n+1 abd(Kn_1(r:;}q) + Kn- 1(Ri11q)) +
exp( -iqrn+1pb 32 )(Bn_1(r;lq) + Bn- 1(R;/q) + Kn- 1(R:;-lq) + Kn_1(ri 1q))
An(q) = exp( -iqrn+1pa23) (Bn (r.;-l q) + Cn- 1(R.;-iq) + Kn- 1(R 1;q) + Kn- 1(R 3;q)) +
Bn- 1(r311q)
(20)
Eq. (19) and (20) are valid for generic mass distributions. In order to iterate the
recursion relations we need a set of initial conditions which depend on the density of
matter on the tiles. Every term of the form 8(x - v) in the density function leX),
contributes with exp( -iqv) in the Fourier transform. We can put for instance an atom
of unit mass at each facet of the tetrahedra in the tiling: if the facets are scaled down
a factor r, the atom is placed in the middle of an edge. In Fourier space,

Ka(q) = Hexp( -iq(~k12 + T~lkd) + exp( -iq(~k12 + T~2k23))

+ exp( -iqGk12 + T;I k13 + 9k34)) + exp( -iq( r- 1pk14 + T~2pk34)))
Ca( q) = Hexp( -iq( C12 + T~2p C13)) + exp( -iq( r- 1PC24 + ~C23))
+ exp( -iq(PC24 + r- 1PC23 + 9C13) + exp( -iq( r- 1 PC24 + T~I" C34)))
Ba( q) = ~(exp( -iq( ab13 + 9b14)) + exp( -iq(pb23 + T~I"b24))
+ exp( -iq(Pb23 + T~ Pb 34 )) + exp( -iq( r Pb23 + r- 1Pb 34 + T~2PbI4)))
A o (q") = 21( exp (-zq
. (T + -Z-a14
T-'P)) + exp (-zq
-1 pa13 . ( -1 a1Z + -z-aZ4
T
I ,,))
r-

+ 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

For the heptagonal patterns the characteristic polynomial is A3 - 2A2 - A + 1 with

roots Al = (3, A2 = 1 - (3 + (3-\ A3 = (3((3 - 2).
If we define the numbers

(22)

with So = SI = 0, S2 = 1 then by induction on n:

(23)

On the other hand, the general solution to the difference equation T( n + 3) =

2T(n + 2) + T(n + 1) - T(n) is T(n) = aA~ + bX~ + cA~ with a, b, c arbitrary constants
to be determined by the initial conditions. Having in mind that A2, A3 are less than one
in modulus:
S - = .
m
1I Sn (3 ( 24 )
n-l
A wave vector q corresponds to. a Bragg peak if the growth for the amplitudes is
proportional to the number of tiles. This occurs generically if and only if all the phases
entering in eq. (19) vanish modulo 27r in the n ---+ 00 limit [27]. This condition is
satisfied if the wave vector is of the form

(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

Quasicrystals in 3D are basicaly of three types: i) icosahedral, ii) periodic in 1D

and quasiperiodic in a perpendicular plane with rotational symmetries of order five or
higher than six, and iii) periodic in 2D and quasiperiodic in a perpendicular direction.
Icosahedral symmetry is found in the diffraction diagrams of AI-Cu-Cr, AI-Pd-Mn and
many other alloys. Systems with one-dimensional quasicrystalline (Fibonacci) phases
are AI-Pa and AI-Ni-Si. A two-colour BKSZ pattern has been used by Burkov in his
model of the AI-Cu-Co decagonal phase [29J. On the other hand two-colour Penrose
patterns are closely related to the AI-Mn decagonal phases. Systems with octagonal and
dodecagonal phases are Mn-Si and AI-Cu-Fe, respectively ([7J and references therein).
Penrose tilings have been proposed by A. Connes in a different context [30], as
examples of noncommutative or quantum spaces. Within the non-commutative crys-
tallography developed by P. Kramer [15], the generators of free groups are associated
with directed segments or paths.
Two different algebraic descriptions have been given in this work. The interpretation
in terms of DOL systems is based on concatenation of words which is converted into
gluing of tiles. This is a non-commutative and non-associative squeme, independent of
a coordinate system. In the second interpretation (obtained along the lines of [27]), the
tilings are obtained through union of sets of points with a fixed position in the space
and this is a commutative approach.
In this work, I have also presented a method to obtain planar patterns with matching
rules and odd symmetries higher than five. The algebraic conjugates of the scaling
factor Al are Ap (see section 4) such that 2p - 1 and d are co-primes. The scaling
factors are Pisot numbers at least for d = 7,9,15, and therefore the corresponding
patterns can be used as structure models of quasi crystals in case these symmetries
are found in diffraction experiments. The patterns with nine-fold symmetry can be
described in terms of DOL systems. This approach can not be extended to eleven-fold
symmetry, unless other shapes than triangles are considered. However for d = 11,13
the scaling factor is not a Pisot number and the patterns obtained with this method do
not correspond to structure of quasi crystals.
Many questions remain unanswered about these objects that, bored by living in
periodic spaces, have decided to inhabit our space which is more interesting. If they
correspond to real quasi crystals, the interpretation in terms of DOL systems suggest
that their growth description could show similarities with the growth of some living
beings.

APPENDIX: CONTEXT FREE LANGUAGES AND THEIR A,UTOMATA

An alternative way to describe the tilings is through context free languages [14, 15J.
A context free grammar is a quadruple

(26)

where ET is the alphabet of terminal symbols and EN the alphabet of nonterminal

symbols, SEEN is the start symbol or sentence symbol and P is a finite set of
production or substitution rules. We take the alphabets ET = < ai, hi, ai, bi , (, ) > and
EN = < Ai, Bi, iL, Bi > with i E ZlO. Every element belonging to EN can be used as

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)

A physical situation of the automaton is a triple{ x, [y] , z} where x E L;T, [y] is an

internal state and z is (j or a string of symbols belonging toL;T U L;N. A situation of
the automaton is either the triple of a physical situation {x, [y] ,z} or a triple obtained
by replacing the x or z (or both) by the neutral word e. If the automaton is in the
physical situation {x, [y] , z} then it is at the same time in the four situations: {x, [y] , z},
{e,[y],z}, {x, [y],e} and {e,[y],e}.
An instruction is a quintuple of the form: {[u], v} 1----+ {a, [b] , c}, where {a, [b] , c}
is a situation, [u] an internal state and v E ET U EN U {a}. If the automaton is in
the internal state [u] and at the same time it is scanning the symbol v written on
the memory tape then it can choose the instruction {[u] , v} 1----+ {a, [b] ,c}. In that
case
a) Internal state: the computing unit switches to the state [b].
b) Input tape: According as a E ET or a == e then the corresponding head writes
the symbol on the left of the last written symbol on the input tape or else does not
move the tape and writes nothing.
c) Memory tape: the head scanning the memory tape acts upon it in the following
way:

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 = e then it does not move the tape and writes nothing;

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:

1: {[O]'o-} f--+ {e, [1], At}; 2i : {[I],A} f--+ {e, [Ai],o-};

3i : {[l],B;} f--+ {e, [Bi],o-}; 4i: {[I],;q f--+ {e, [Ai],o-};
5i: {[I],B;} f--+ {e, [Bi],o-}; 6i : {[Ai],e} f--+ {e, [1], (Bi+4Ai+6Ai-3)};
7i : {[Ai],e} f--+ {e, [1]'ai}; 8i: {[Bi], e} f--+ {e, [1], (Bi+4Ai+6)};
9i : {[Bi],e} f--+ {e, [l]'b,}; 10i : {[AiJ,e} f--+ {e, [IJ,(BiAiHAi+3)};
11i : {[Ai],e} f--+ {e, [l]'a;}; 12i: {[Bil,e} f--+ {e, [1], (ABi+6)};
13 i : { [Bd , e} f--+ {e, [1], bi}; 14i : { [1], ai} f--+ {ai, [1], 0- };
15i : { [1], b,} f--+ {bi, [1] , 0-}; 16 i : { [1], ai} f--+ {ai, [1],0- };
17i : { [1] ,bn {b;, [1], 0-}; 18 : {[I],O f--+ {(, [1],0-};
n,
f--+

19 : {[I],)} f--+ [1]'0-}; 20 : { [1] , 0- }

f--+ {[So], 0- }.
(29)
Not all the words belonging to the language generated by the grammar characterize
pieces of the pattern; only those appear with the same derivation length for the elements
of ~ n ~T.

REFERENCES

[1] M. Senechal, Quasicrystals and Geometry, Cambridge University Press (1995).

[2] R. Penrose, Bulletin of the Institute for Mathematics and Applications, 10 (1974),
266.
[3] D. Shechtman, L Blech, D. Gratias and J. Cahn, Phys. Rev. Lett. 53 (1984), 1951.
[4] N. G. de Bruijin, Math. Proc. A84 (1981), 38.
[5] P. Kramer and R. Neri, Acta Cryst. A40 (1984), 580.
[6] A. Katz and M. Duneau, Phys. Rev. Lett. 54 (1985), 2688.
[7] C. Janot, Quasicrystals: a primer, Oxford University Press, 1993.
[8] G. Rozenberg and A. Salomaa, The Mathematical Theory of L Systems, Academic
Press, New York, 1980.
[9] A. Beutelspacher and B. Petri, Der Goldene Schnitt, Wissenschaftsverlag Zurich,
1989.
[10] J. Garda Escudero, J. Phys. Soc. Jpn. 64 (1995), 2317.
[11] M. Baake, P. Kramer, M. Schlottmann and D. Zeidler, Int. J. Mod. Phys. B4
(1990), 2217.
[12] B. Grunbaum and G. C. Shephard, Patterns and Tilings, Freeman, New York,
1989.
[13] J. Garda Escudero, J. Phys. Soc. Jpn. 65 (1996), 20l.
[14] J. Garda Escudero and P. Kramer, J. Phys. A: Math. Gen. 26 (1993), LI029.
[15] P. Kramer and J. Garda Escudero, in: Beyond Quasicrystals, Springer, Berlin,
1995, 55,

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

The application of (Hopf) fiber bundles is well developed in theoretical and

mq,thematical physics [1,2]. Along this vein, the Hopf fibrations on spheres lead
to nonbijective canonical quadratic transformations useful in classical and quantum
mechanics [3-13]. The Hopf fibrations on spheres, as well as their extensions on
hyperboloids [10], yield the concepts of "constraint Lie algebra" and "Lie algebra un-
der constraints" , which are of importance for connecting invariance or noninvariance
algebras of dynamical systems [11].
On the other hand, there is a close connection between number theory and quadrat-
ic mappings, especially those mappings sending the (2m-1 )-sphcre on the (2m-1-n)-
sphere [14,15]. In this direction, one can easily imagine that the replacement of
mappings on spheres by mappings on hyperboloids might produce interesting re-
sults.
It is the aim of this work to describe how fibrations on spheres and hyperboloids
naturally arise in the framework of Cayley-Dickson algebras. This will provide us
with quadratic [10] and nonquadratic [12] forms which generate nonbijective trans-
formations from R 2m onto R 2m-n. We shall give two prototypical applications of
these quadratic and nonquadratic transformations, viz. one to number theory (in the
spirit of ref. 15) and one to dynamical systems. Much of this work takes its origin
in references 5-13. The application to number theory constitutes the premises of a
more complete work.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 153
CAYLEY-DICKSON ALGEBRAS

The Cayley-Dickson (C-D) algebras generalize the algebras of complex numbers,

quaternions and octonions. A 2m-dimensional C-D algebra may be obtained from a
m-dimensional (m = 1,2,4,) C-D algebra by a 'doubling' process (cf. C = R+iR).
In the present paper, we limit ourselves to C-D algebras of dimensions 2m = 2, 4 and
8. They are denoted as A(c) with c == CI, C == CI,C2 and C == Cl,C2,C3 for 2m = 2,4
and 8, respectively, where Ci = 1 for i = 1,2,3. Note that A(-I), A(-I,-I) and
A( -1, -1, -1) are nothing but the algebra C of usual complex numbers, the algebra
H of ordinary (or Hamilton) quaternions and the algebra 0 of ordinary (or Cayley)
octonions, respectively. The algebra A(I) is the algebra n of hyperbolic complex
numbers. The three algebras A(CI' C2) with (CI' C2) i- (-1, -1) are isomorphic to
the algebra N I of hyperbolic quaternions and the seven algebras A( CI, C2, C3) with
(CI' C2, C3) i- (-1, -1, -1) are isomorphic to the algebra 0' of hyperbolic octonions.
Let
2m-1

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

eiej = -gij + 2:= a7 j ek i and j = 1,2, .. , 2m - 1 (2)

k=1

for the generators of A( c) is given. The constants at

(totally anti symmetric in ij k)
appear in ref. 10 for 2m = 2, 4 and 8. The constants gij are defined by the matrix

(3)

for 2m = 8. The matrix 9 for 2m = 4 (or 2m = 2) follows by restricting the matrix

in eq. (3) to its first four (or two) lines and columns.
The product W = uv can be represented in a matrix form as

w = H(u;c)v (4)

where w, u and v are column vectors defined by

tw = (Wo, WI,, W2m-l) tu = (uo, Ul,, Uzm-J) tv = (Vo, VI,, V2m-J) (5)

and H( u; c) is a 2m X 2m matrix generalizing the Hurwitz matrix [10] that occurs in

the search of bilinerar forms w'" = rf",( ufj, vfj) with (\' = 1,2, ,n and (3 = 1,2, .. , n
such that

WI2 + W2z + ... + Wn2 = (Ul

2 + Uz2 + ... + Un
2) (VI
2 + V22 + ... + Vn
2) (6)

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

Uo ClUl C2 U 2 -ClC2 U 3 C3 U 4 -ClC3 U 5 -C2 C3 U S ClC2 C3 U 7

Ul Uo C2 U 3. -C2 U 2 C3 U 5 -C3 U 4 C2 C3 U7 -C2 C3 U S

U2 -ClUa Uo ClUl CaUs -ClC3 U 7 -C3 U 4 ClC3 U 5

U3 -U2 Ul Uo C3 U 7 -C3 U 6 C3 U 5 -C3 U 4

(7)
U4 -CIU5 -C2 U S CIC2 U 7 Uo ClUl C2 U 2 -CIC2 U 3

U5 -U4 -CZ U 7 CZUs Ul Uo -C2 U 3 CZUz

Us CIU7 -U4 -CIU5 Uz ClU3 Uo -CIUl

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

H(Ui C ) = uol + Uk trk (8)

k=l

where 1 is the 2m x 2m unit matrix and the 2m x 2m (Clifford) matrices rk satisfy

(i and j=I,2, .. ,2m-l) (9)

It is thus possible to associate a Clifford algebra C(p, q) of degree p + q = 2m - 1

(with q - p being the signature of g) to the 2m-dimensional C-D algebra A(c). An
important property of the matrix H (u; c) is the following.

Property 1. The matrix H(u; c) belongs to R+ xO(2m) or CxO(m, m) according

to as the 2m x 2m metric
(10)
has the signature 2m or O. More specifically, we have

(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.

Lemma. The complex conjugation in a 2m-dimensjonal C-D algebra A(c) and

in its m-dimensional C-D subalgebras induces the existence of 2m - c5(m, 1) anti-
involutions on A(c).

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

where ck = 1 for k = 1,2, .. ,2m - 1. The notation

c = diag(l, cl, ,c2m-d (13)

shall be used below.

Definition 1. The mapping

T[l; e; c] : A( c) --7 A( c) : U f--+ x = U U (14)

is called a (right) Hurwitz transformation.

The Hurwitz transformations can be classified into several types according to the
various possibilities for the 2m x 2m matrix c.
u
Type AI. For = u: The transformations T[I; e; c], where c: is the unit matrix 1,
were called quasi Hurwitz transformations in ref. 10 and transformations of type Al
in ref. 12.
u
Type B 1 . For = j (u): The transformations T[l; c; c:J, where c: corresponds to an
arbitrary anti-involution j of A( e), were called Hurwitz transformations in ref. 10 and
transformations of type Bl in ref. 12.
u
Type C 1 . For -I- uor j(u): The transformations T[I;e;c], where c: corresponds
neither to the unit matrix 1 nor to an anti-involution j of A( e), were called pseudo-
Hurwitz transformations in ref. 10 and transformations of type C1 in ref. 12.

QUADRATIC TRANSFORMATIONS

The mapping (14) can be rewritten in matrix form as

R 2mX1 --7 R 2mx1 : U f--+ x = H(u; c) c: u (15)

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).

Property 2. The relation

(16)

is valid for all the Hurwitz transformations T[I; C; E].

We continue with some examples for 2m = 8, 4 and 2. The R 2m --+ R 2 m-n
quadratic transformations shall be given in explicit form only for 2m = 8. The cases
2m = 4 and 2m = 2 can be deduced from the case 2m = 8 by taking U4 = U5 =
U6 = U7 = 0 with C3 = 0 and U2 = U3 = U4 = U5 = U6 = U7 = 0 with C2 = C3 = 0,
respecti vely.
Type AI. The mapping T[I;CI,C2,C3;1] gives the generic R8 --+ R8 quadratic
transformation

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

f = diag(l, -1, 1, 1, 1, 1, -1, -1) (20)

gives the generic R 8 --+ R 5 quadratic transformation

X2 = 2( UOU2 + Cl Ul U3 + C3U4U6 - CI C3U5U7)

X3 = 2( UOU3 + UI U2 - C3U5U6 + C3U4U7)

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)

Transfor- Fibration Fiber L Lo Ll

mation

SI -+ {1} SI so(2) so(2,1)

R2 -+ R sp(4,R)
R -+ {I} R so(l,l) so(2,1)

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)

X4 = 2( UOU4 + Cl Ul Us - C2U2U6 + CI C2U3U7)

Xs = 2(UOUs + UI U4 + C2U3U6 - C2U2U7) (21)
with the property
X6 - C2X~ + CIC2X~ - + CIC3X~ =
C3X~
= (u~ - clui - C2U~ + CIC2U~ - C3U~ + CIC3U~ + C2C3U~ - CIC2C3U~)2 (22)
The matrix c in eq. (20) corresponds to an anti-involution of A(CI,C2,C3) of type
h,jz, ,h The other anti-involutions of type jl,jz"",h would lead to transfor-
mation formulae equivalent to eqs. (21) and (22).
The mapping T[1;CI,C2,C3;c] where

c = diag(l, -1, -1, -1, -1, -1, -1, -1) (23)

gives the generic R8 -+ R quadratic transformation

This transformation (which reads Xo = tu 1] u in a more compact form) corresponds

to the anti-involution jo of A( CI, C2, C3), i.e. to 11 = joe u) = u.

The geometrical classification of the R 2m -+ R m+ l transformations (2m = 8 and

4) associated to an anti-involution of type h,jz, ,jzm-l of A(c) leads to three

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

c = diag(l, -1, -1, -1, -1, -1, 1, 1) (25)

gives the generic RB -T R7 quadratic transformation

X3 = 2( -C3U4U7 + C3U5U6)
X4 = 2( CZUZU6 - Cl CZU3U7) X5 = 2(CZU2U7 - C2U3U6)

X6 = 2( UOU6 - Cl Ul U7) (26)

with the property

(27)

Other choices for c with I:r=1 ck = -3 or 5 lead to transformation formulae equiv-

alent to eqs. (26) and (27). Remark that the restriction of eq. (26) to the cases
2m = 4 and 2m = 2 does not lead to new transformations: The obtained quadratic
transformations are Hurwitz transformations of type B l . Furthermore, the (true)
transformations of type C l are equivalent to transformations of type Al for 2m = 2
and 2m = 4.
For 2m = 8, the R8 -T R7 transformations provide explicit realizations for (i)
the Hopffibration S7 -T CP 3 of compact fiber SI when Cl + C2 + C3 = -3 and for
(ii) its noncompact analogues, namely, R4 x S3 -T R4 X S2 of compact fiber Sl and
R4 x S3 -T R3 X S3 of noncompact fiber R when Cl + C2 + C3 =F -3.

159
DIFFERENTIAL ASPECTS OF QUADRATIC TRANSFORMATIONS

Differential and Lie-algebraic aspects of transformations of type Al and Bl were

studied in refs. 10 and 11. Here, we briefly discuss the situation for the R2m -r R 2m - n
quadratic transformations of type Bl with n = m - 1 + 8( m, 1) and 2m = 2, 4 and 8.
Let us consider the 2m one-forms defined by the column vector

n = 2H(u;c)cdu (28)

where the 2m x 2m matrix c is associated to a given anti-involution, of A( c), of the

type il,Jz,'" ,Jzm-l. Eq. (28) provides us with (i) 2m - n total differentials that
may be integrated to give the nonvanishing components Xj of x in eq. (15) and (ii)
n one-forms Wi that correspond to the vanishing components of x in eq. (15). As a
corollary of Property 1, we obtain

(29)

In view of the nonbijective character of the R 2m ---> R 2m-n transformation, we assume

that each one-form Wi satisfies the constraint condition Wi = O. The introduction of
these n conditions in eq. (29) makes it possible to connect the line element in R2m,
with the metric17, and a line element in R2m-n.
To each one-form Wi, we may associate a vector field Xi, defined in the symplectic
Lie algebra sp(4m.R), with an action of Wi on tXj, where r = t u17u, such that
Wi[tXj] = 8(i,j). Each vector field Xi has the property that Xi!/J = 0 for any
function!/J in C l (R2m-,,). The following theorems and definitions are interesting for
physical applications.

Theorem 1. The n = m - 1 + 8( m, 1) vector fields Xi span a Lie algebra Lo with

respect to the commutator law.

Definition 2. The subalgebra Lo of L = sp( 4m, R) is called a constraint Lie

algebra. Let us define the subalgebra Ll of L by

(30)

when Lo is one-dimensional and by

(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

Nonquadratic transformations may be obtained by replacing eq. (15) by

R 2mXl -t R 2mxl : u 1-+ x = H(u; c)N 5 u with NEZ - {1} (32)

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)

for the nonquadratic transformations. The quadratic and nonquadratic transforma-

tions are associated to Hurwitz transformations that we shall denote as T[N; c; 5]
with NEZ. They can be classified into transformations of type AN, BN and CN by
employing the.same character of distinction as for the transformations of type AI, Bl
and C l . We limit here our consideration to two specific examples: (1) transformations
of type AN with NEZ - {-1} and 2m = 2 and (2) transformations of type B-1
with 2m = 4.

Example 1. The mapping T[N; Cl; 1] defines the transformations u 1-+ x =

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

8 xoxo - c18xlXl = (N + 1)-2 (u~ - clui)-N (8uouo - cl8ulUl) (34)

which hold outside the domain {(uo,ud E R2 I u~ - clui = O}.

Example 2. The mapping T[-1; Cl, cZ; 5], where 5 = diag(1, -1, 1, 1), yields the
generic nonquadratic formulae

(35)

with pZ = u~ - Cl ui - C2U~ + Cl C2U~. The general property (33) can be particularized

as
(36)
Therefore, the nonquadratic transformation of type B-1 with 2m = 4 and Cl = C2 =
-1 corresponds to the R4 - t 53 Fock stereographic projection well known in quantum
mechanics.

161
APPLICATIONS

The applications of the Hurwitz transformations T[N; c; E] range from number

theory (Hurwitz problem, Pythagorean and Diophantine equations) to theoretical
physics (classical and quantum mechanics, gauge theory). We give below a few simple
examples of application of some Hurwitz transformations.

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

for 2m = 2, 4 and 8. As an interesting result, the solutions Wo:

eq. (37) are given by w = H(u; c)v.

Second, the three generic Diophantine equations

A2 - c1B z - C2CZ + Cl czD z - C3E2 + CI c3F2 + C2C3C2 - CI czc3Hz = [2 (38)

A2 - C2 Bz + CICZC z - C3 Dz + CIC3 Ez = FZ (39)

and
A Z - czB z + CIC z C 2 - C3Dz + CIC3Ez + czc3F2- CICZC3CZ = HZ (40)
admit solutions (A, B,' .. ,I) E Z9 ,(A,B,'" ,F) E Z6 and (A, B,, H) E Z8 cor-
responding to quadratic transformations of type AI, BI and C I , respectively (see
Table 2). In eqs. (38), (39) and (40), each Ci (i = 1,2,3) can take the values +1 or -1
(as in what precedes) in the case 2m = 8. Then, the solutions are quadratic functions
of Uo: E Z (a = 0,1"" , 7) given by eqs. (17) and (18) for (38), eqs. (21) and (22)
for (39) and eqs. (26) and (27) for (40). Other Diophantine equations are obtained
in the cases 2m = 4 and 2m = 2 by taking C3 = 0 and Cz = C3 = 0, respectively.
\Ve have reported in Table 2 some solutions for the Diophantine equations (38),
(39) and (40) and their particular cases. In Table 2, the solutions are (A, B, .. ,1) E
Z9, (A,B,,F) E Z6 and (A,B,,H) E Z8 for Ci = 1 with i = 1,2,3.
The two particular Diophantine equations inside round brackets admit solutions
(A, B, C, D, [) E Z5 and (A, B, C, F) E Z4 corresponding to U4 = Us = U6 = U7
and C3 = 0 while the Diophantine equation in square brackets admits solutions
(A, B, [) E Z3 corresponding to Uz = U3 = U4 = U5 == U6 == U7 and C2 == C3 = o.
For example, for CI == Cz == C3 - 1 == -1 we get from eq. (39) the Pythagorean
equation (see also Table 2)
(41)
with the solutions

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

A = u~ - 2u~ B =2uauz F = u~ + 2u~ ( 45)

are solutions of
( 46)
an equation which comes from eq. (41).

Table 2. Diophantine equations in zr for r = 3,4,5,6,8,9 with solutions

Equation Solution

AZ - clB z - czC z A = U5 + Clui + CZU~ - CICZU~ + C3U~

+ clczD z - C3Ez +
CIC3 Fz - CIC3U~ - CZC3U~ + CICZC3U~
+ c2 c 3 G 2 - CIC2C3Hz = [z B = 2UaUI C = 2uQuz D = 2UaU3
(A2 - cl B 2 - C2 C2 E= 2UQU4 F = 2UQU5 G = 2UQU6 H = 2UQU7
+ CIC2D2 = [2) 1= U5 - clui - C2U~ + CICZU~
[A2 - clB2 = [Z] - C3U~ + CIC3U~ + CZC3U~ - CICZC3U~

A = u5 - clui + C2U~ - CIC2U~ + C3U~

- + CZC3U~ - CI CZC3U~
Cl C3U~
A Z - czB Z + CICZC Z B = 2( UaU2 + CI UI U3 + C3U4U6 - CIC3U5U7)
- C3 DZ + CIC3 EZ = F Z C = 2( UOU3 + UI Uz - C3U5U6 + C3U4U7)
(AZ - czB Z + CICZ C 2 = FZ) D = 2( UQU4 + CI UI U5 - CZUZU6 + CI CZU3U7)
E = 2( UOU5 + UI U4 + C2U3U6 - CZUZU7)
F = u5 - clui - czu~ + CIC2U~
- C3U~ + CIC3U~ + C2C3U~ - CIC2C3U~

A = u6 - clui - czu~ + CICZU~ - C3U~

+ CIC3U~ - CZC3U~ + CICZC3U~
B = 2( -C3U4U6 + CIC3U5U7)
AZ - czB Z+ CICZC Z C = 2( -C3U4U7 + C3U5U6)
- C3DZ + CIC3EZ + cz c 3 F2 D = 2(CZU2U6 - CIC3U3U7)
- CI CZC3GZ = HZ E = 2(CZUZU7 - CZU3U6)
F = 2( UaU6 - CI UI U7 )
G = 2(UaU7 - UIU6)
H = u6 - clui - C2U~ + CIC2U~
- C3U~ + CIC3U; + CZC3U~ - CICZC3U~

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)

admit as (particular) solution A, Band C such that

(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

As a second illustration, we end up with a toy example concerning applications of

Hurwitz transformations to dynamical systems. Let us consider the two-dimensional
Schrodinger equation

(49)

where Z E Rand r = JX5 + xi. The application to eq. (49) of a transformation

T[ N; -1; 1] (of type AN) with NEZ - { -1 } leads to the partial differential equation

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)

so that the transformation of type AN allows to transform the R2 Schrodinger equa-

tion for the potential -Z(x~ +xi) -N/(N+I) and the energy E into the R2 Schrodinger
equation for the potential-(N + 1)2E(u5 +ui( and the energy (N + 1)2Z. (Note
that in such a transformation the rtles of the energy and the coupling constant are
interchanged.) The solutions for a E Z and NEZ - {-I} correspond to the cou-
ples (0, N) = (1,1), (3, -3) and (4, -2). (The solution (0,0) is trivial!) In other
words, the R2 Schrodinger equations for the potentiall/r (Coulomb potential), l/r 3
and l/r 4 are transformed into the R2 Schrodinger equations for the potentials p2
(harmonic oscillator potential), 1/ p6 and 1/ p4, respectively.
To close this paper, let us mention that there exist other applications to dynamical
systems. In classical mechanics, the Kustaanheimo-Stiefel transformation (a trans-
formation of type B l ) was used for the regularization of the Kepler problem [3].
In quantum mechanics, the latter transformation made it possible to transform the
Schrodinger equation for the three-dimensional hydrogen atom (in an electromagnetic

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

[1] Y. Choquet-Bruhat, C. DeWitt-Morette and M. Dillard-Bleick, Analysis, Mani-

folds and Physics, North-Holland, Amsterdam, 1982.
[2] Y. Choquet-Bruhat and C. DeWitt-Morette, Analysis, Manifolds and Physics.
Part II: 92 Applications, North-Holland, Amsterdam, 1989.
[3] P. Kustaanheimo and E. Stiefel, J. reine angew. Math. 218 (1965), 204.
[4] M. Boiteux, J. Math. Phys. 23 (1982),1311.
[5] M. Kibler and T. Negadi, Lett. Nuovo Cimento 39 (1984), 319.
[6] I. V. Polubarinov, On Application of Hopf Fiber Bundles in Quantum Theory,
Report E2-84-607, JINR, Dubna, 1984.
[7] T. Iwai, J. Math. Phys. 26 (1985), 885.
[8] A. C. Chen and M. Kibler, Phys. Rev. A31 (1985), 3960.
[9] M. Kibler, A. Ronveaux and T. Negadi, J. Math. Phys. 27 (1986), 154l.
[10] D. Lambert and M. Kibler, J. Phys. A: Math. Gen. 21 (1988),307.
[11] M. Kibler and P. Winternitz, J. Phys. A: Math. Gen. 21 (1988), 1787.
[12] M. Kibler and P. Labastie, in: Group Theoretical Methods in Physics, Y. Saint-
Aubin and L. Vinet (ed~.), World Scientific, Singapore, 1989.
[13] M. Hage Hassan and M. Kibler, in: Le probleme de factorisation de Hurwitz, A.
Ronveaux and D. Lambert (eds.), FUNDP, Namur, 1991.
[14] L. E. Dickson, Ann. Math. 20 (1919), 155.
[15] T. Ono, Variations on a Theme of Euler: Quadratic Forms, Elliptic Curves, and
Hopf Maps, Plenum Press, New York, 1994.
[16] M. Kibler, in: Proceedings oHhe IIIrd Catalan Days on Applied Mathematics,
J. Chavarriga and J. Gine (eds.), Institut d'Estudis Herdencs, Lleida, 1996.

165
QUANTIZATION OF SYSTEMS WITH CONSTRAINTS

John R. Klauder

Departments of Physics and Mathematics

University of Florida
Gainesville, FL 32611, USA

INTRODUCTION

The quantization of systems with constraints is of considerable importance in a

variety of applications. Let {Pj,qj}, 1 :s; j :s; J, denote a set of dynamical variables,
p a}, 1 :s; a :s; A :s; 2J, a set of Lagrange multipliers, and {<Pa (p, q)} a set of constraints.
Then the dynamics of a constrained system may be summarized in the form of an action
principle by means of the classical action (summation implied)

1= f[pd - H(p, q) - ),a<pa(P, q)] dt . (1)

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

{<Pa(P, q), <Pb(p, q)} = cab <Pc(p, q) ,

C (3)
{<Pa(p,q),H(p,q)} = hab <Pb(p,q) , (4)
then we are dealing with a system of first-class constraints. If the coefficients Cab C and
h} are constants, then it is a system of closed first-class constraints; if they are suitable
functions of the variables P and q, then it is a system of open first-class constraints. If
the first or both of the conditions in (3) and (4) fail, then the system involves second-
class constraints.
For first-class constraints it is sufficient to impose the constraints at the initial time
inasmuch as the equations of motion will ensure that the constraints are fulfilled at

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 167
all future times. Such an initial imposition of the constraints is called an initial value
equation. Furthermore, the Lagrange multipliers are not determined by the equations
of motion; rather they must be specified (a choice of "gauge") in order for a solution of
the dynamical equations to be given. For second-class constraints, on the other hand,
the Lagrange multipliers are determined by the equations of motion in such a way that
the constraints are satisfied for all time.
In the remainder of this section we briefly review standard quantization procedures
for systems with closed first-class constraints, both of the operator and path integral
variety, pointing out some problems along the way. In the following section we develop
our coherent state approach for closed first-class constraints, which are illustrated by
examples in the final section. For further details of such general systems, as well as a
discussion of open first-class and second-class constraints, see ref. [1].

Standard operator quantization

For a system of closed first-class constraints we assume (with 1i = 1) that

[<pa(P, Q), <Pb(P,Q)] = iCabc<pc(P,Q) , (5)

[<pa(P, Q), H(P, Q)] = ihab <p&(P, Q) , (6)

where <Pa and H denote self-adjoint constraint and Hamiltonian operators, respectively.
Following Dirac [2], we adopt the quantization prescription given by

iW(P, Q) = [W(P, Q), H(P, Q)] (7)

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

<Pa(P, Q)I7}!)phys = 0 (8)

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].

Standard path integral quantization

Faddeev [3] has given a path integral formulation in the case of closed first-class con-
straint systems as follows. The formal path integral

Jexp{iJ~[pjqj - H(p, q) - ),aYa(P, q)] dt} VpVq V)'

=Jexp{iJ~[pAj -H(p,q)]dt}5{Y(p,q)}VpVq (9)

may well encounter divergences in the remainingintegrals. Therefore, subsidiary condi-

tions in the form X"(p, q) = 0, 1 ::; a ::; A, are imposed picking out (ideally) one gauge

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)

This result may also be expressed as

J exp{ i J~[P;q*j - H*(p*, q*)] dt} Dp* Dq* , (11)

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.

COHERENT STATE PATH INTEGRAL

Canonical coherent states may be defined by the relation

(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

1= J Ip, q)(p, ql dj.l(p, q) , (13)

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

H(p, q) == (p, qIH(P, Q)lp, q) = (p, ql : H(P, Q) : Ip, q) (14)

which is related to the normal-ordered form as shown. If H denotes the quantum

Hamiltonian, then we shall adopt H(p, q) as the classical Hamiltonian. We note that in
the general case H(p,q) =f.H(p,q) but differ by terms which are O(fi). Although these
functions may be numerically different, in most cases the difference between these two
functions is qualitatively insignificant. As remarked below, however, there are analogous
cases where the qnalitative difference is quite significant. Lastly, we also note that an
important one-form is given by i(p, qldlp, q) = Pj dql.

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)

=M J exp{iJ[pjljj -H(p,q) - AarPa(p,q)]dt}'Dp'Dq. (15)

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)}.

Enforcing the quantum constraints

Let us next introduce the quantum analog of the initial value equation. For simplicity
we assume that the constraint operators generate a compact group; the case of a non-
compact group is implicitly discussed below (see ref. [1]). In that case

(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

J( pII ,qlilT e -iJ[H+>.a(t)\!>a]dt 1-'

p ,q-')(-'
p, -'IE
q I' p ,.q ') d/1 (-' -')
P ,q

lU' I'
- (" lilT e -iJ[H+,\a(t)\!>a]dt ""p,q ')
-p,q

= lim (p", q"I[II (e-icHe-ic.\l'\!>a)] E Ip', q')

z
= (p", q"le-iTHe-iTa\!>a E Ip', q')
= (p", qllle- iTH E Ip', q') , (19)

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 .

Reproducing kernel Hilbert spaces

The coherent state matrix elements of IE define a fundamental kernel

K(" " , ') =- ("

p,q;p,q p,q "IIEI' ')
p,q, (22)

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

'I/;(p,q) = LIYkK(P,q;Pk,qk) ' (25)

k=l

and the inner product of this function with itself has two equivalent forms given by
K

('I/;, '1/;) =L IYkIYIK(Pk, qk; pI, ql) (26)

k,l=l
= J'I/;(p, q)*'I/;(p, q) dJ-l(p, q) . (27)

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

K 1 (p";p') = K(p",c;p', C) , (28)

K 2(p"; p') = Jw(q")*w(q')K(p", q"; p', q') dq" dq' , (29)
K3(P", q"; p', q') = limK(p", q"; p', q') (30)

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

(p", q"llElp', q')

= 7['-1(2 r exp[-~(k _ p? + ik(q" - q') -
J-5
S
~(k - p')2] dk, (31)

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)

which has become a reproducing kernel that characterizes a one-dimensional Hilbert

space with every functional representative proportional to x(p) == exp( _p2 /2). This
one-dimensional Hilbert space representation also admits a local integral representation
for the inner product given by

(33)

This example is an important one inasmuch as it shows how a constraint operator

with a continuous spectrum is dealt with in the coherent state approach.

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)

with c a constant, which leads to the equations of motion

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)

which is defined in a fashion to yield

(p,q p,q{,
" "IIEI' '\ (39)

where

IE = IE( -8 < PI < 8)1E( -8 < P2 < 8) . (40)

In particular this leads to the fact that

(p", q"llElp', q')

= 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)

which clearly characterizes a particular representation of a one-dimensional Hilbert

space in which every vector is proportional to exp[-Hp~ +pDJ. This example is, of
course, related to the reduction example given earlier. Moreover, we can introduce a
local integral representation over the remaining p variables for the inner product if we
so desire.
Let us now turn attention to the second formulation of the problem by focusing (for
a different C(A)) on

M f exp[i!(PIql + P2q2 - AlPl - A2 ecQ'P2) dtJ DpDq DC(A) . (43)

This expression again leads (for a different IE) to

(p", q"llElp' ,q') , (44)

173
where in the present case the fully reduced form of this expression is proportional to

Jexp[-~(k2 p~)2 + ik2(q~ -q~) - ~(k2 - p~)2J

-
x exp[-!(kl - pn 2 + iklq~ - ~i'\lkl1
x exp[ -ixk1 - i'\2ecx k2 + iXKIJ
X
l' \ K I - tKlql
exp [-2'tlll . , - 12' ( Kl - PI, )2J
xdk 2 dk 1 dx dKI d'\l d'\2 . (45)

When normalized appropriately, this expression is evaluated as

(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

I = J (pq - '\p) dt . (47)

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

I> -1. (48)

Here the equations of motion include q = '\(1+ 1) Ipl"! which, along with the constraint
plpl"!= 0, may cause some difficulty in seeking a classical solution of the equations of
motion, e.g., if I < 0. vVhen I # 0, such constraints are said to be irregular. It is
clear from (9) that irregular constraints lead to considerable difficulty in conventional
phase-space path integral approaches.
Let us examine the question of irregular constraints from the point of view of a
coherent state, phase-space path integral quantization. For any, > -1, we first ob-
serve that the operator PIPI"! is well defined by means of its spectral decomposition.
Moreover, it follows that

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

(p", q"IElp', q') = M J exp[iJ(pq - '\plpl"!) dtJ VpVq VG"!(>.) , (50)

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

r/J(p, q) == (p, ql<l>(P, Q)lp, q) = (OI<l>(P + p, Q + q)IO) . (52)

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

r/J-y(p) = (Jrht 1 !2 j(k+p)lk+pl-Ye-k2!1idk

= ( 7rli t 1 !2 j klkl-Ye-(k- p )2/ li dk . (54)

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

A rough, but qualitatively correct expression for this behavior is given by

(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)

as A -+ 00, where ]{A is a suitable A-dependent parameter. If W = 0 is a point in the

discrete spectrum, as in the second example, then ]{A = 1 and the result is a projection
operator lE onto the subspace where TV = O. The case of a discrete spectrum is not
especially difficult and we choose to focus our attention on the first example for which
W = 0 is a point in the continuous spectrum. In that case it follows that

1~~ JiAT/21l' (pl,q"le-iTAP2/2Ip',q')

= lim JiAT/21l'
A~oo
jex [-t(k- pll )2
p

+ik(q" - q') - itT Ak2 - t(k - p')2] dk

= exp[-t(pll2 + p'2)]
, (63)
which is exactly the same result that we would find had we started with the usual
formulation with a Lagrange multiplier.
Let us now present a coherent state path integral for this version of doing constraint
systems. In particular we observe that

In this formal path integral we have included the term

(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

1= J[pq - ~A(p2 + ~1i)1 dt . (66)

When a term such as (65) arose as the coefficient of a Lagrange multiplier, we
concluded that it was unacceptable because it was impossible to satisfy p2 + tt/2 = 0 as
a constraint for real phase space variables. On the other hand, we now see such a term
arise as a coefficient of the large parameter A. Note that the extra 'It term does not
influence the equations of motion. In particular, the classical equations of motion are
q = Ap and p = 0 exactly as before. Thus the criterion that the classical solutions should
be independent of the large parameter A leads once again to the solutions p( t) = 0 and
q(t) = ct + q'. It is true that the "energy" given by A(p'2 + 'It/2)/2 = A'It/4 diverges as
A -+ 00, but that expression is not the real energy for any physical system. Thus we
conclude that the appearance of the extra term in the classical action does not interfere
with an appropriate solution of the classical equations of motion unlike the formulation
with a Lagrange multiplier.
Let us briefly raise the issue in the case of another even constraint, say p4 = 0 based
on the classical action

(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)

which implies that the proper factor in the classical theory is

(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

[1] J. R. Klauder, Coherent State Quantization of Constraint Systems, IHES/P/96/29,

quant-ph/9604033, Annals of Physics (in press); see also quant-ph/9607019 and
quant-ph/9607020.
[2] P. A. M. Dirac, Lectures on Quantum Mechanics, Belfer Graduate School of Science,
Yeshiva University, New York, 1964.
[3] L. D. Faddeev, Theor. Math. Phys. 1 (1970), 1.
[4] V. N. Gribov, Nuc!. Phys. B139 (197S), 1; 1. M. Singer, Commun. Math. Phys.
60 (197S), 7; see also J. Govaerts, Hamiltonian Quantisation and Constrained Dy-
namics, Leuven Notes in Mathematical and Theoretical Physics, Vo!' 4, Series B:
Theoretical Particle Physics, Leuven University Press, 1991.
[5] M. Henneaux and C. Teitelboim, Quantization of Gauge Systems, Princeton Uni-
versity Press, 1992.
[6] S. V. Shabanov, JINR Lecture Notes, Volume 54, Phase Space Structure in Gauge
Theories, Dubna (19S9) (in Russian); 1. V. Prokhorov and S. V. Shabanov, Sov.
Phys. Usp. 34 (1991), lOS.
[7] V. 1. Arnold, Mathematical Methods of Classical Mechanics, Springer, New York,
1975, 75ff and 91.

178
AUTOMORPHISMS AND DISCRETE FIBER
BUNDLES

Peter Kramer\ Uwe Beck\ and Shelomo 1. Ben-Abraham2

lInstitut fiir Theoretische Physik

Universitat Tiibingen
Germany
2Physics Department
Ben-Gurion University of the Negev
Beer Sheva
Israel

INTRODUCTION

The motivation of the present work is twofold. In an attempt to generalize no-

tions of crystallography for the description of systems with broken translational sym-
metry as for example quasicrystals, a scheme with non-commutative translations was
explored [1, 2]. This scheme is strictly discrete. It is based on free groups and their
automorphisms, and is linked to standard crystallography through two basic homo-
morphisms. In particular for the 17 space groups on the plane, 17 non-commutative
preimages were constructed.
Another type of symmetry breaking is associated with dislocations in crystals. A
systematic framework for dislocations was put forward among others by Kondo [3],
Bilby et al. [4], Kroner [5] and Ben-Abraham [6, 7]. This framework uses the con-
tinuum theory of solids and methods of differential geometry. Dislocations are de-
scribed by the concepts of connections and torsion. In a descriptive way, types of
dislocations are linked to the typical discrete symmetries of ideal crystals by Pond and
Hirth [8].
The present paper is an attempt to link concepts from these two fields on their ge-
ometric and algebraic level. Since we wish to maintain the discrete scheme throughout
we reformulate the notion of connection and torsion for this purpose. This is done by
adapting the concept of fiber bundles. We show that fiber bundles may be constructed
whose base space is one of the 17 planar space groups. These space groups are repre-
sented by their Cayley graphs. The non-commutative scheme appears in the fiber and
gives rise to an algebraic torsion specific for each space group.

Symmetries in Science IX. edited by Gruber and Ramek

Plenum Press, New York, 1997 179
THE FREE GROUP Fn

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

z = br;" ... b';:n . (5)

In the crystallographic context one considers a commutative translation group with n
generators which is isomorphic to zn.
Abelianization is the homomorphism

homi : Fn --+ zn, (6)

W --+ Z, mi = 2:.::: aij, 'Vi.

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

By construction this kernel is a normal subgroup of Fn.

We denote by K the commutator subgroup of Fn , i.e. the subgroup generated by
all the commutators. For any group G there are the well-known general relations

K(x,y):= xyx-1y-l, cK(a,b)c- 1 = K(ca,b)K(b,c), 'Vx,y,a,b,cE G. (8)

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)

where each of the conjugated factors K is an elementary commutator K = K(xtl, Xfl).

Upon applying homl to the reordered word, only the prefador before the commutators
contributes. If now w = u E ker(homl)' the prefactor must be unity and so this
word can be completely reordered into a product of elementary commutators, each
conjugated with some word from Fn. Since now ker(homl) E K, the two subgroups
must be identical D.
Example 2: We give in steps the reordering of a word u E F2 from ker(homl):

(10)

DISCRETE PRINCIPAL FIBER BUNDLES FROM GROUP /SUBGROUP

STRUCTURES

An excellent historical account of the concept of fiber bundles is given by Dieudonne

[11]. An exposition with applications in theoretical physics may be found in the mono-
graph by Curtis and Miller [12]. A fiber bundle (E, B, 7r, F) consists of a total space E,
a basis B with elements b, a projection 7r : E --+ B, and a fiber F = 7r- l (b), bE B. In
a principal fiber bundle, any fiber is a homogeneous space under the sharply transitive
action of a structural group H. There are a number of conditions from differential geom-
etry which cannot be maintained for discrete structures. A connection in a (principal)
fiber bundle is a prescription for lifting any path from the basis into the total space.
Let 'Y be a closed path in the basis starting from and returning to b E B. We define
the torsion along 'Y as the shift in the fiber obtained by following the path lifted from
'Y into E.
A particular type of principal bundle may be constructed from a group G and a
subgroup H < G: Take the total space as the group E = G and the bundle as the left
coset space B = G / H with the standard projection 7r : G --+ G / H. Then the fiber is
H, and H acts sharply transitive from the right within each fiber. Consider the left
action of G on G/H. In terms of left coset generators G/H = (ciH) this action is

(g, cih) --+ gCih = c/hh (11)

If for example 9 can be described by a set of generators, as will be assumed in what

follows, then the left action G x G --+ G provides
(i) a path in the basis from c;H to cjH,
(ii) a lifting of this path into the bundle by the left action h -7 hh in the bundle.
A related construction of torsion associated with the well-known left action G x G -7
G on a Lie group G is discussed by Goldberg (13], in terms of tensor calculus. From the
Killing metric he constructs a Riemannian metric on G. The connection is compatible
with this metric, has a symmetric Levi-Civita part and in addition a nonvanishing
torsion part linked to the commutator in the Lie group. In the terminology of Curtis
and Miller [12] this would be called an orthonormal frame bundle.

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 .

THE FREE GROUP F2 AS A PRINCIPAL FIBER BUNDLE

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)

The bundle will be

(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)

in the crystallographic notation. The graph r(Pl) yields a geometric representation of

the basis of the bundle. The fiber consists of copies of the commutator group K attached
to the vertices of the graph. The left action in F2 eq. (16) with respect to the graph
produces the new point (mI + mI, m2 + m2) and hence corresponds to a translation
in the graph or lattice r. Note that the left multiplication in the commutative group
PI lacks any information on the path followed between the initial and final points.
However, the full word 9 can be read as a path on r between the initial and final points,
and the full left action F2 x F2 --+ F2 yields a lifting of the path from the basis into
the bundle. The left action with an element u of the commutator group yields a closed
path on the graph r(pl). When we read the torsion eq. (17) from right to left, we find
that it describes exactly the closed path followed on r(Pl).

Figure 1. Left: A word from the commutator subgroup of the

free group F2 represented by a closed path. Right: The closed
loop is broken into two elementary squares.

Example 3: In Figure 1 we depict the path

(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.

CAYLEY GRAPHS FOR PLANAR SPACE GROUPS

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:

Group G Cayley graph r( G)

group element 9 vertex v
generator Xi directed edge of color i
left action with g, follow path on graph,
9 -+ 9 1 = 99, from v to VI,
9 = g(Xloo ,Xr) along edges 1 ... r
left action transitive f connected
relation on G, closed path on f,
R(Xl" ,Xr) along edges 1 ... r

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.

GROUPS OF AUTOMORPHISMS AND NON-COMMUTATIVE 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

<P:Yi =Yi(Xl,""xn), i= 1 ... n. (20)

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]

homz : Aut(Fn) -+ GL(n, Z) (23)

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

Yi = WXiW-I, i = 1 .. . n, wE Fn. (24)

Abelianization of these words yields the unit substitution matrix for any W E F n , and
so the free group embedded in this way falls into the kernel, Inn(Fn) < ker(hornz). We
clearly cannot use hornz if we wish to obtain full space groups as homomorphic images.
A way out is given, first for the case n = 2, by the construction of an embedding of
Aut(Fz) and a new homomorphism: The group Aut(Fz) can be embedded into Aut(F3)
in a nontrivial fashion [1] . The trivial embedding would let it act on the first two
generators (Xl, xz) of F3 only. The nontrivial embedding maintains this action on the
first two generators of F3, but at the same time transforms the generator X3 into a word
of the form Y3 = w( Xl, XZ)X3' The embedding construction resembles the embedding of a
space group in dimension 2 into a block triangular subgroup of the group GL(3, Z) [16].
Now we apply the homomorphism of type hom2 , but for the group Aut(F3), to the
embedded group Aut(Fz). Then it can be verified that the group Inn(Fz ) under this
homomorphism is mapped into a commutative translation group, represented by block
triangular matrices from G L(3, Z) with unit diagonal blocks.

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 )

(b) R1 = R R2 R3 = R' R4 = R2Y

(hom)-l (1zT2 (11 (12 T1(11
R (R1)2 (R2)2 (R3)2 (R4)2
ker
R (R1R2)2 (RzR3)2 (R3R4)2 (R4R1)2
ker K(Tl, T2- 1)
(30)

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.

FIBER BUNDLES FROM NON-COMMUTATIVE SPACE GROUPS

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
.... .... ....... .
. . ... .... .....
...... . . ......

Figure 3. Cayley graph of P2 generated by three involutions

SI, S2, S3' The torsion is carried by the hexagon.

Consider as the next example the fiber bundle

(E, B, 7r,F) = (P2, P2, hom2, JC) (33)
In the Cayley graph, all the relators appear as closed loops. Since the relators Rl have
the value e both in P2 and P2, the corresponding elementary loops are free of torsion.
All the elementary torsion is carried by the loops corresponding to the relators R 2 . In
Figures 2-7 we give the Cayley diagrams. The elementary loops which carry torsion
are marked by a small circle. For reasons of clearness, some of the loops are redrawn
on the right. Recall that undirected edges are short-hand for simple loops with two
vertices. Fortunately it turns out that, for all 17 constructions, these loops stay free of
torsion.
In Figure 2 we generate P2 by two translations X, Y and one involution T. The
elementary torsion is carried by the loops involving two copies of each translation gen-
erator. The loops corresponding to the relators T2, TXT X, TYTY stay free of torsion.

Figure 4. Cayley graph of Pg generated by two glide reflec-

tions P, Q. The torsion is carried by each rhombus.

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
. .--
....... " ...

.
...............~
.- - . ~

Figure 6. Cayley graph of P2mm generated by four reflec-

tions Rb R2, R 3 , R 4 The torsion is carried by one out of four
squares.

Figure 7. Cayley graph of P3 generated by two rotations

UI , U2 The torsion is carried by the hexagon. .

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

[1] P. Kramer, J. Phys. A28 (1995), 379.

[2] P. Kramer and J. Garcia-Escudero, in: Beyond Quasicrystals, F. Axel and D. Gratias
(eds.), Springer, Berlin, Heidelberg, 1995, 55.
[3] K. Kondo, in: Proc. 2nd Japan Nat!. Congress of App!. Mech., (1952), 4l.
[4] B. A. Bilby, R. Bullough, and E. Smith, Proc. Roy. Soc. London A231 (1955), 263.
[5] H. Kroner, Kontinuumstheorie der Versetzungen und Eigenspannungen, Erg. angew.
Math. 5, Springer, Berlin, Gottingen, Heidelberg, 1958, l.
[6] S. I. Ben-Abraham, in: Fundamental Aspects of Dislocation Theory, J. A. Simmons,
R. de Wit, and R. Bullough (eds.), Nat. Bur. Stand. (U.S.) Spec. Pub!. 317 II (1970),
943.

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

A. Leviatan 1 and R. Bijker 2

1 Racah Institute of Physics

The Hebrew University
Jerusalem 91904, Israel
2Instituto de Ciencias Nucleares
U.N.A.M., A.P. 70-543
04510 Mexico, D.F., Mexico

INTRODUCTION

In the usual description of baryons in terms of three constituents, the wave-function

is a product of a space part and an internal spin-flavor-color part. Algebraic methods
have been used extensively in the past to describe the internal part in terms of the
symmetry groups SUsj (6)cs>SUc (3) [1,2]. The difference between models lies in different
assumptions on the spatial dynamics of the three constituents. Quark potential models
in nonrelativistic [3] or relativized [4] forms emphasize the single-particle aspects of
quark dynamics [or which only a few low-lying configurations in the confining potential
contribute significantly to the eigenstates of the Hamiltonian. Excited baryons in this
description correspond to single-particle excitations of levels in the confining potential.
On the other hand, flux-tube models, soliton models as well as some regularities in
the observed spectra (e.g. linear Regge trajectories, parity doubling) hint that an
alternative, collective type of dynamics may playa role in the structure of baryons.
The aim of the present contribution is to report on an algebraic framework [5] which
encompasses both the single-particle and collective scenarios. The formalism is based
on a U(7) spectrum generating algebra whose bosonic realization and geometry were
discussed at length in a previous contribution to these proceedings [6]. The algebraic
formulation allows us to study a large class of models, all with the same spin-flavor-color
structure, but different types of spatial dynamics. Among the models is the familiar
harmonic oscillator quark model taken as a proto-type for single-particle dynamics,
and a collective model in which the baryon resonances are interpreted as rotations and
vibrations of an oblate-top shaped string with a distribution of charge magnetic mo-
ments. In what follows we report on an application of the model to the mass spectrum
of nonstrange baryons, derive expressions of form factors relevant for electromagnetic
and strong couplings [5]. We examine the effect of spin-flavor breaking and stretching

Symmetries in Science IX. edited by Gruber and Ramek

Plenum Press. New York, 1997 193
 A
d

u u

Figure 1. Collective model of baryons (the charge distribution

of the proton is shown as an example).

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-

potential model [10] where 81% of the wave function is a pure n =

cant spread should be compared to the structure of the ground state in the quark
configuration
and the 19% admixture of n = 2 components is induced by the hyperfine interac-
tion. The above analysis of the spectrum in the nonstrange sector, shows that mass-
es alone are not sufficient to distinguish between single-particle and collective forms
of dynamics. To do so require an examination of other observables (e.g. electro-
magnetic and strong couplings) which are more sensitive to the structure of wave-
functions.

COLLECTIVE FORM FACTORS

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)

where Am (m = 0, ) are Jacobi coordinates and (ko, k) is the four-momentum of the

absorbed quanta. The form factors of interest are proportional to the matrix elements
of these operators. In the algebraic approach, these operators are first mapped onto
the U(7) algebra and their matrix elements are evaluated in the wave-functions of
eq. (1). The algebraic images of the operators in eq. (4) and the calculation of matrix
elements were presented in [6J. Table I in [6] shows the resulting elementary form factors
corresponding to the case in which the charge and magnetization are concentrated at
the end points of the string of Fig. 1. Different types of collective models are specified
by a distribution of the charge, magnetic moment, etc., along the string. For the present
analysis we use the (normalized) distribution

(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

1112 912 712- 512- 312 112

- 1/2+ 312- 512+ 7/2- 9/2+

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

F(k) J d(3g((3) (if'fIUIif'i) ,

J d(3 g((3) (if'fITmlif'i) . (6)

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.

Elastic Form Factors of the Nucleon

Taking into account the overall symmetry of the nucleon wave function, the elastic
electric (E) and magnetic (M) collective form factors are given by

G~ 3 J d(3 g((3) (\II; MJ = 1/21 e3 U1W; MJ = 1/2) ,

G~ 3 J d(3 g((3) (\II; MJ = 1/21 P,3 e3 0"3,z U1\II; MJ = 1/2) (7)

197
Table 1. Collecti ve and harmonic oscillator form factors

Final state a ) :F(k) iJz(k)/m3 kOa iJ(k)/m3 k oa

[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

[56,2+]{0,0);0 !v15 [(1+;Ia2)2 J5 [3t7k2a2

-"21 5 ka{1+k2a2)3 y'"2 [ka{1+k-1 a )22 2

+ 2kt3 H(ka)] - 2k~a4 H(ka)] + 2k~a4 H(ka)]

Final stateb) (flU Ii) (fIt Ii) /m3 k of3 (f1'hli)/m3 k of3

[56,0+]0 e- k2 (32/6 ~ k(3 e- k2 (32 /6 0

[70,I-h -i.l k(3 e- k2 (32 /6

V3
i.l(1 _ iflf..) e- k2 (32/6
V3 3
'J"f-
=t=z 3
e _k2(32/6

[56,2+h __1_P(32 e- k2 (32/6 -2-k(3(1- k2(32) e-k2f32/6 =t=~k(3 e-k'(32/6

3y6 3y6 6

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)

where the corresponding magnetic moments are

2
/1p = /1 ; /1n= -"3/1 (10)

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,

(32 e-(3ja u /2a~ ,

(32e-(3/ad/2a~ . (11)

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]

A;; = 6 / f [ k(L, 0; S, vlJ, v) B - (L, 1; S, v - IIJ, v) A 1 ' (15)

where v = 1/2, 3/2 indicates the helicity. The orbit- and spin-flip amplitudes (A and
B, respectively) are given by

B .I d(3 g((3) (W j; MJ = vi/!3 e3 83,+ [r IWi; M~ = v-I)

A = .I d(3 g((3) (W j; MJ = vi J.l3 e3 1'+/ 931W i ; M~ = v - 1) . (16)

These observables correspond to an absorption process (I + B' --+ B) of a right-handed

photon with four-momentum (ko, k = ki). In eq. (16) IW i ) denotes the (space-spin-
flavor) wave function of the initial nucleon (B') with 28~2[56, 0+](0,0);0 and N = p, n,
and, similarly, IWj) that of the final baryon resonance (B). In general, the B and A
amplitudes of eq. (16) are proportional to the collective form factors F and ~h of eq. (6),
respectively. The breaking of spin-flavor symmetry has also influence on the helicity
amplitudes of eq. (15). They would now be given in terms of flavor-dependent collective
form factors Fu(k), Yu,+(k) and Fd(k), Qd,+(k), which depend on the size parameters,
au and ad, respectively.
When comparing with the experimental data one must still choose a reference frame
which determines the relation between the three-momentum k 2 and the four-momentum
Q2 = P _ k5. It is convenient to choose the equal momentum or Breit frame where

(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

0.0 0.2 0.4 0.6 0.8 1 .0

0 2 [ (GeV/c)2 ]

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

0.0 1.0 2.0 3.0 4.0

0 2 [ (GeV/c)2 )

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

Av( k) = (27r )3/21(2k o)1/2 [ (L, 0, S, vlJ, v) (0 Zo( k) 1

+ 2(L, 1, S, v ~ IIJ, v) (+ Z_ (k)

+~(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

0.0 1.0 2.0 3.0

Q 2 ( (GeV/C)2 ]

200

N(1 680)F'5

v = 312

'?
o
o ~~------------------------------------~~

100
v = 1/2

0.0 1.0 2.0 3.0

Q 2 [ (GeVfc)2)

Figure 6. Proton helicity amplitudes for excitation ofN(1520)D 13 (a factor of +i is sup-

pressed) and N(1680)F15 . The calculation with and without flavor breaking are shown by
dashed and solid lines, respectively.

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]

f(B -+ B' + M) = 27rPi _J2 ~ IAv(k)12 (21 )

2 +1L...-
v>o
where Pi is a phase space factor. For all resonances with the same value of (Vb V2), L P
this expression can be rewritten in a more transparent form in terms of only two ele-
mentary partial wave amplitudes Wi(k),

r(B -+ B' + M) = 27rPi (27r~32ko L cIIWI(k)1 2 . (22)

I=Ll
Here I is the relative orbital angular momentum between the final baryon and the
emitted meson. It takes the values I = L 1 (the value I = L is not allowed because
of parity conservation). For this set of resonances, the k dependence is contained in
the partial wave amplitudes Wl(k), while the dependence on the individual baryon
resonance is contained in the coefficients C[. In the algebraic method, the Wi (k) can be
obtained in closed form. For example, the corresponding Sand D elementary partial
wave amplitudes are

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
,

W2(k) = 3i {[gk - "61 hk1k(3 - 1

3hm3ko(3k (3 2 2} e- k 2 f32/6
, (24)

Use of eqs. (22)-(23) allows us to do a straightforward and systematic analysis of

the experimental data. The calculations are performed in the rest frame of the decaying
resonance, in which the relativistic expression for the phase space factor Pi as well as
for the momentum k of the emitted meson are retained. The expressions for k and Pi
are

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+

1700 3/2-} 1 1/2

1620 1/2-

124(144) 77(36) (4)

67(89) (29) (45)

1232 3/2+ 0 3/2

939 1/2+ 0 1/2

Figure 8. Strong decay widths for Ll * --+ N + 7r and Ll * --+ Ll + 7r decays of positive parity
resonances with L P = 2+ and negative parity resonances with L P = 1-. The theoretical
values are in parenthesis. All values in MeV.

SUMMARY AND CONCLUSIONS

In this article, we have exploited the algebraic approach to baryon structure to

analyze simultaneously elastic form factors and helicity amplitudes in photo- and elec-
troproduction and strong decay widths. The use of algebraic methods allows us to study
different situations, such as the harmonic oscillator quark model and the collective mod-
el, within the same framework. The logic of the method is that, by starting from the
charge and magnetization distribution of the ground state (assuming a dipole form to
the elastic form factor of the nucleon), one can obtain the transition form factors to
the excited states. In the 'collective' model, this procedure yields a power dependence
of all form factors (elastic and, inelastic) on Q2. For electromagnetic couplings we have
analyzed two aspects of hadronic structure: (i) the breaking of SUsj (6) symmetry, and
(ii) the stretching of hadrons with increasing excitation energy. We find that, whereas
the breaking of the spin-flavor symmetry hardly effects the helicity amplitudes, the
stretching of hadrons does have a noticeable influence.
The disagreement between experimental and theoretical elastic form factors and
helicity amplitudes in the low-Q2 region 0 S; Q2 S; 1 (GeV /C)2 may be due to coupling
of the photon to the meson cloud, (i. e. configurations of the type q3 - qij). Since such
configurations have much larger spatial extent than l, their effects are expected to
drop faster with momentum transfer Q2 than the constituent form factors. Also, since
meson exchange corrections contribute differently to different channels, this effect will
be state dependent.

208
Table 2. Strong decay widths for.6.* -+ N +7r and N* -+ N+7r in MeV. Experimental
values are from [23].

Resonance L f(th) f(exp)

Ref. [17] Ref. [20] Ref. [21] Present

.6.(1232)P33 0 70 121 108 116 1195

.6. (1950) F37 2 27 56 50 45 120 14
.6.(2420)H3.n 4 4 8 12 40 22
.6.(2950)K3 ,15 6 1 3 5 13 8

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

We have performed a calculation of the strong decay widths N* -+ N + 7r, N* -+

.6. + 7r, N* -+ N + 1], .6.* -+ N + 7r, .6.* -+ .6. + 7r and .6.* -+ .6. + 1] in a collective model of
baryons. The analysis of experimental data shows that, while the decays into 7r follow
the expected pattern, the decays into 1] have some unusual features. Our calculations
do not show any indication for a large 1] width, as is observed for the N(1535)Sll
resonance. The observed large 1] width indicates the presence of another configuration,
which is outside the present model space. This suggests, that in order to elucidate this
point, particular attention be paid at CEBAF to the N 1] channel.
In the present contribution we have focused the discussion to the nonstrange sector
(nucleon and delta). It will be of interest to extend the formalism to include strange
baryons as well.

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.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 211
 In section 6 we construct the unitary representations of Poincare group in the dis-
crete coordinate space. We recover the formulae for the momentum space by the use
of a discrete Fourier transform for non-periodic functions.

REFLECTION GROUPS

Let X be a metric space of dimension n of the type spherical sn, Euclidean En or

hyperbolic Hn. Let S be a side of an n-dimensional convex polyhedron P in X. The
reflection of X in the side S of P is the reflection of X in the hyperplane spanned by
S. The reflections of X in the sides of a finite sided n-dimensional convex polyhedron
P in X of finite volume generate a reflection group.
Let P be a finite-sided n-dimensional convex polyhedron in X of finite volume all of
whose dihedral angles are submultiple of Jr. Then the group generated by the reflections
of X in the sides of P is a discrete reflection group r with respect to the polyhedron P.
In order to construct a presentation for a discrete reflection group we take all the
sides {Sd of P and for each pair of indices, i, j, let k ij == Jr j{) (Si, Sj), where {) (Si, Sj)
is the angle between Si and Sj. Let F be the group generated by the sides Si and r
the group generated by the reflections on Si. When P has finitely many sides and finite
volume, the map 'lj; = G --> r is an isomorphism, where G is the quotient of F by the
normal closure of the words (SiSJkij. We call

a presentation of the discrete reflection group r.

A discrete reflection group r with respect to a finite-sided polyhedron P of finite
volume is isomorphic to a Coxeter group G, that is, an abstract group defined by a
group presentation of the form {Si, (SiSj)k ij }, where:

i) the indices i,j vary over some countable indexing set J;

ii) the exponent kij is either a positive integer or 00 for each i,j;

iii) kij = kji ;

iv) k ii = 1, for each i;

v) kij > 1 if i # j; and

vi) if kij = 00, the term (SiSj)= is omitted.

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)

If X = E3 we have the Euclidean triangle reflection groups with Coxeter diagram:

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

corresponding to the Coxeter graphs f n,!' (n = 2,3, ... ,9).

The groups fn,l have a very important connection with the Lorentian lattices
An or the set of points with integral components with respect to a Cartesian basis
{eo, e1, ... , en} equipped with the bilinear form

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.

A DIFFERENCE CALCULUS OF SEVERAL INDEPENDENT VARI-

ABLES

Given a function of one independent variable the forward and backward differences
are defined as

tlf(x) == f(x + tl~) - f(x) , '\If(x) == f(x) - f(x - tlx) .

Similarly, we can define the forward and backward promediate operator

~ 1 . - 1
tlf(x) == 2" {f(x + tlx) + f(x)} , '\If(x) == 2" {f(x - tlx) + f(x)}
Hence the difference or promediate of the product of two functions follows:

tl {f(x)g(x)} tlf(x)Lig(x) + Lif(x)tlg(x) , (1)

1
tlf(x)tlg(x) + 4tlf(x)tlg(x) .
~ ~
Li {f(x)g(x)} (2)

This calculus can b~ enlarged to functions of several independent variables. We use

the following definitions:

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:

tlf(x,y) tlxLiyf(x, y) + Lixtlyf(x, y) , (3)

- - 1
Lif(x,y) = tlxtlyf(x,y) + 4tlxtlyf(x,y). (4)

215
 Given a vectorial space vn over Z we can define a real-valued linear function over Z

f(u}=(w,u) UEV n (5)

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:

D.f(x,y) = ( D.t:f) D.x + (~~~Yf) D.y. (8)

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

D.x 1\ D.y = - D.y 1\ D.x .

With the help of this exterior product we can construct a second order discrete differ-
ential form or 2-form, namely

where (jOiP is an antisymmetric, tensor. Similarly we can define a discrete p-form in a

n-dimensional space (p < n)
1 .. .
v~ -- ,u
-~.
p.
. . AX'l
Z1 Z2 'ZpLl..
1\ Ll
A X '2 1\ Ll.
AX' p ,

where (jil i2 . ip is a totally antisymmetric tensor.

The dual of a p-form in a n-dimensional space is the (n-p)-form *a with components

where 5 is the n-dimensional Levy-Civitta totally antisymmetric tensor (5123 ... == 1).
EXTERIOR CALCULUS AND LORENTZ TRANSFORMATIONS

Given a I-form in a two-dimensional space

w = a(x, y)flx + b(x, y)fly

we can define the exterior difference, in the similar way as the exterior derivative,
namely

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,

w = a(x,y,z)fly /\ flz + b(x,y,z)flz /\ flx + c(x,y,z)flx /\ fly (11)

we can also define the exterior difference as

flw

(12)

The exterior derivative applied to the product of a O-form (scalar function f) and a
I-form (w = aflx + bfly) is

fl (fw) = Lif flw + flf /\ Liw , (13)

where Lif is expressed in (4) and Liw = Liaflx + Libfly.

The exterior difference of the product of two I-forms is easily obtained as

(14)
The exterior difference of the product of a p-form p and a q-form (J is

(15)

Finally for any p-form w we have

(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 ,

where J = JI't.xl' is the charge-current I-form.

Taking the exterior derivative of the last equation we get an other example of t. 2 = 0:

( 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:

{-VxVyVzV't (LixLiyLizt.t) + V'xVyVzVt (t.xLiyLizLit)

+VxV'yVzVt (Lixt.yLizLit) + VxVyV'zVt (LixLiyt.zLit)} (xyzt) = o. (20)

In order to compute the Lorentz transformation of the discrete differential forms,

we start with the coordinate-independent nature of I-form

(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)

transform covariantly like the coefficients WI" of (10).

The same technique can be applied to components of discrete p-forms.

UNITARY REPRESENTATIONS OF THE DISCRETE POINCARE

GROUP IN MOMENTUM SPACE

Let P + be the integral proper inhomogeneous Lorentz group or discrete Poincare

group, which is homomorphic to the semi direct product of the integral subgroups of
SL (2,C), and T4 be the transformation group on the Minkowski lattice. The multipli-
cation law for the Poincare group is

(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

U (a) = rr3 (1 + ~~EPI")al'

1 - -ZEP
, (27)
I" = 0 2 I"

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)

If we transform this equation to an arbitrary inertial system, we calculate

Q (P) = F (P) QF- 1 (P)

1j;(P) = F(P)1j; (p)
with

220
and ex = (~ ~), ~ = (~ ~I ) , we obtain

(ex. iJ + Iml ~) 1jJ() (P) = Po1jJ() (P) (35)

The transformation properties of the Dirac vectors are

for spacial rotations, and

U (a, A) 1jJ(+) (P) = rr3 (1 + ~~EP;,)"" (1 - ~~~. ~) 1jJ(+) (A -1 p) (37)

1'=0
1- -ZEP
2 I'
1 + -zv
2
.a

for pure Lorentz transformations.

The negative energy covariant vectors carry the same representation.

TRANSFORMATION OF THE COVARIANT FUNCTIONS IN CONFIG-

URATION SPACE

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

(jl E1, j2 E2, h E3, nT) :=0 (1, n ) ,

where E1,(:2,t3,T are small quantities in the space-time directions and j1,j2,j3,nare
integer numbers. We also define the difference operators

'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

provided the dispersion relation is satisfied

(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)

as in the continuous case. Multiplying this equation from the left by

we obtain the dispersion relation

(43)

The transformation properties of the wave functions in the configuration space are
given, as in the continuous case, as follows:

(44)

for spatial rotations of angle wand

(45)

for special Lorentz transformation of relative velocity it = vlv.

The connection of these transformations in the configuration space and the trans-
formation in momentum space is given via the Fourier transform .in the lattice, for
non-periodic functions, namely

ljJ(P) = "L.J
00 (1 + iiep)i
~.
1- -uP
ljJ(ej) (46)
j=-oo 2

where ljJ (j t) satisfy boundary conditions ljJ (ej) -7 0, when j -7 00 and

L
00

ljJ (fj) < 00. (47)

j=-oo

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

that satisfies lim SL (0) = 00 and

L-+oo

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

Lebedev Physical Institute

117924 Moscow, Russia
and
Osservatorio Astronomico di Capodimonte
1-80131 Napoli, Italy

INTRODUCTION

Standard quantum mechanics is based on notion of a complex wave function which

satisfies the Schrodinger equation [1]. The attempts to give classical-like interpreta-
tions of the wave function were done in [2-4]. It turned out that the new formulation
of standard quantum mechanics may be given in terms of classical probabilities for the
position [5, 6] based on symplectic tomography scheme [7, 8]. Recently, the energy levels
of the harmonic oscillator were discussed in frame of classical formulation of quantum
mechanics [9] as well as transition probabilities between the levels. Using marginal dis-
tributions for homodyne observable [10, 11] to describe the quantum states is based on
some relations of the density matrix to characteristic functions for the observable [12].
The marginal distribution for the position in ensemble of shifted, rotated, and scaled
reference frames in classical phase space of the system under study has been intro-
duced [7]. It was shown that this marginal distribution determines the quantum state
since the Wigner function is given by a Fourier component of the marginal distribution.
The invariant form for the connection of the marginal distribution for the position to
the density matrix was found [8] and the approach was extended to the system with
several degrees of freedom.
Classical formulation of quantum evolution was suggested and for the marginal dis-
tribution new quantum evolution equation was found [5], which is an alternative to the
time-dependent Schrodinger equation. This equation gives the classical-like description
of quantum evolution in terms of positive normalized distribution containing complete
information about the state of the system. Examples of free motion and some excited
states of the harmonic oscillator were also considered [5, 6]. The evolution of even and
odd coherent states [13] of a particle in a Paul trap was investigated [14] in frame of
the classical-like description [5-7]. The even and odd coherent states of a trapped ion

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 225
were discussed in recent papers [15, 16].
The aim of this work is to discuss the notion of a quantum state in the new for-
mulation of quantum mechanics. We review the classical-like description of transition
probabilities between stationary states (energy levels) of quantum systems and obtain
analogs of the orthogonality and the completeness relations. We show that, if the evo-
lution equation describing the dynamics of a quantum system is determined by the
imaginary part of the system potential energy considered as a function of a complex
coordinate, the energy states of the system are determined by the real part of the
potential energy. The energy levels of the harmonic oscillator are rederived in frame
of classical-like alternatives to Schrodinger evolution and stationary equations. The
new type of eigenvalue problem is formulated for the positive and normalized marginal
distributions.

NEW CONCEPT OF QUANTUM STATE

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

phase space is given.

It is worthy to note, that the information contained in the marghinal distribution
W (X, ft, v, 8) is overcomplete. To determine the quantum state completely, it is suffi-
cient to give the function for arguments with the constraints (8 = 0, ft2 + v 2 = 1) which

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

S (ft, v, 8) = - J dX 'W (X, ft, v, 8) In ['W (X, ft, v, 8)]. (5)

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.

ENERGY LEVELS AND QUANTUM DYNAMICS

As it was shown in [5], for systems with the Hamiltonians

(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

The measurable position is a cyclic variable for the evolution equation.

Let us rewrite the Schrodinger equation for the stationary state density matrix PE
of the quantum system with Hamiltonian (6)

lIPE = PElI = EPE (8)

in terms of the time-independent marginal distribution 'WE(X, ft, v, 8) of the squeezed

and rotated quadrature [7]. We have

The positive marginal distribution (eigendistribution) satisfies this eigenvalue equation

and also the equation

(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)

we reduce eq. (7) to the following one [5]

. {) {) 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

11!n(OS) (X, fJ" v, 8) -_ Wo

(OS)
(X, fJ" v, 8) _1_I Hn2 ( X- 8 ) , (13)
2nn. VfJ,2 2 +v
where the marginal distribution of the oscillator ground state is

[ (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

In this section, we express quantum transition probabilities in terms of an overlap

integral of classical-like marginal distributions describing the initial and final quantum
states.
If one has the initial pure state of a quantum system described by the marginal
distribution
Win = Wl (X, fJ" v, 8) (15)
and the final state of the quantum system described by the marginal distribution

Wf = 102 (X, fJ" v, 8) , (16)

the probability of the quantum transition P12 (1 ==} 2) may be obtained using the
known expression for the probability in terms of an overlap integral of the Wigner
functions Wi (q, p) and W 2 (q, p) of the initial and final states (see, for example,
review [17] )
(17)

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

J wp(X, {-t, v, 8) wp(Y, -{-t,

.
-v, 8) exp [z (X +Y - 28)]
df.idvdX dY
27r
= 1. (19)

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

J wn(X, {-t, v, 8) wm(Y, -{-t, -v, 8) exp [z (X

. +Y - 28)]
df.idvdX dY
27r
= 8mn , (20)

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

where the distribution w(wn) describes the white noise.

Thus, the classical marginal distributions describing the energy levels of quantum
systems are positive solutions to the system of equations (9) and (10) and these solutions
satisfy the orthogonality condition (20) and the analog of the completeness relation (22).
The distributions form interesting mathematical set which differs essentially from the
usual Hilbert space of states described by the normalized complex wave functions.

EXCITED STATES OF HARMONIC OSCILLATOR

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)

The marginal distribution is given by the integral

Wn(X, {-t, v) 7r- 5j2 T(n+2) (n!tl J [-l- :2 -

exp ik(X - {-tq - vp) - iPU]

<8l Hn (q +~) Hn (q -~) dkdqdudp. (25)

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

w",(X, p" 11) = exp (-Ial 2 ) "

L.J "
L.J aa ,wnm(X, p" v) , (29)
n=O m=O n.m.
where

Wnm(X, p" v) 7r- 5/ 2T(n/2+ m/2+ 2) (n! m!)-1/2

Q9 J [-l- :2 -
exp ik(X - p,q - vp) - iPU]

Q9 Hn (q+~) Hm (q-~) dkdqdudp. (30)

Then
(31)
The integral of (27) has the Gaussian form and for the marginal distribution of the
coherent state we obtain the expression

(32)

U sing the generating function for Hermite polynomials

(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

{~[ (:vr + (:~r] (8~) ~2 - ~(~2 + v (8~r} WE(X,~, v) 2

)

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)

for w = 1/2; M = 0; nr =n gives exactly the spectrum of the one-dimensional

oscillator
n = 0,1,2, .... (43)

231
 J
To find the marginal distribution, we have to calculate

wn(X, /L, v) -_ - 1 exp -

[P(/L2+ 2).]
+ zkX
V
Ln (P/L2+ P 2) V
dk. (44)
21l' 4 2

1:
In view of the integral

211l' exp ( - ~2 + ikX) Ln (~2) dk = 1l'-1/2 Tn (n!)-l exp (_X2) H~(X), (45)

one obtaines the marginal distribution (35).

It is worthy to remark, that the normalization condition for the marginal distribution
wn(X, /L, v) implies the condition for the Fourier component

J wn(k, /L, v) exp (ikX) dkdX = 21l'wn(k = 0, /L, v) = 1. (46)

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

(9 exp ( - ~2 _ X;2) Hn(X) Hm(X'). (47)

Since usnal creation and annihilation operators for two oscillators

(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:

a~p(X,X') -----+ G:q +i P) W(q,p);

a~'p(X, X') -----+ G :q -iP) W(q,p);

(53)
Xp(X,X') -----+
i
( q+"2op
a) W(q,p);
X'p(X,X') -----+ (q-~:p) W(q,p).

The operators a; b; at; bt of (51) satisfy the commutation relations

[a, b] = [a, btl = 0; (054)

and act on the function wnm(X' f.l, v) given by eq. (34) as follows

awnm(X, (1, v) vnWnm(X, IL, v);

at wnm(X, (1, v) Vn+1 W n +1m(X, il, v);
(55)
bWnm(X, f.l, v) VrnWnm-l(X, f.l, v);
bt wnm(X' Il, v) .Jm + 1 W nm +1(X, il, v).

Let us introduce the operators

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

n Wn-l (X, fl, v) ;

(57)

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

p (X, X', t) = W (X, t) W*(X', t). (59)

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

W (X, t) = J G (X, X', t) W (X') dX' . (60)

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)

J( (X, X', Y, V', t) = G (X, Y, t) G*(X', yl, t). (62)

Since the relation of the density matrix to the marginal distribution is known for any
time t (given by (59) and (60) ), it is possible to obtain

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(q,p) 217r J 'U!(X,jl,v)exp[-i(JLq+vp-X)]dJ-tdvdX; (64)

p(X, X') = 217r J w(Y,p,X-,X/)exp [i (y_pX~XI)] dpdY; (65)

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.

PATH INTEGRAL AND CLASSICAL PROPAGATOR

The path integral for the quantum propagator

G (X, X', t) = Jd[q] exp (i S[q]), (67)

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

6 (t2 - t l ) 6 (X - X') 6 (p - P') 6 (v - v') . (71)

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

II (X, Il, v, X', Il', v', tf, tin)

fn N-l

k=l
{II (Xk+l' Ilk+b Vk+1, Xk, Ilk, Vkj tk+l, tk) dXk dll k dvd , (73)

where the time interval tf-tin = NT j tk = tin +k T j k = 1, 2, ... , N. Taking in (73)

the limit T --+ 0 j N --+ 00 , one obtaines the expression for the classical propagator in
terms of the functional integral.

INTEGRALS OF THE MOTION AND CLASSICAL PROPAGATOR

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

Io(Y, yI) = I (Y, Y', t = 0). (77)

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

J1,r(t) wn(X, Il, v, t) = n wn(X, Il, v, t), (78)

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

We consider the system with the quadratic Hermitian Hamiltonian

(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 :

I (t) = A (t) Q + /:;.(t). (80)

Here the real symplectic 2 x 2-matrix A (t) and the real vector /:;. (t) satisfy the
equations
(81 )
and the initial conditions
A (0) = 1; /:;. (0) = o. (82)
As follows from (2) and from the property of theWigner function [17J, the classical
propagator is

II (X, fl, V, X', fl', v', t) = 8(X - X' +N A-1/:;')8(N' - N A-I), (83)

where the vectors Nand N' are

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)

there are two linear invariants [17, 18]

paCt) =p; qo Ct) =q- pt . (86)

It means, that /:;. (t) =0 and the symplectic 2 x 2-matrix is

(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

A (t) = ( co~ t sin t ) (90)

-SIll t cos t .

It means, that for the oscillator

N A-l(t) = (v cos t - f.l sin t, v sin t + f.l cos t). (91)

Consequently, the classical propagator of the harmonic oscillator 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)

Calculating the integral of (63), in view of delta-functions of (95), we obtain

f{
( , ') 1
X,X, Z, Z, t = (21f')2
J df.ldY
X -X'+f.lt

exp'[i ( Y - f.l - + X')

X -
2
- - i . Z - Z'
X - X' + f.lt
Z+
( Y - /-1 ~~-
2
Z')l (96)

Then integration over the variable Ygives delta-function

_1
21l'
J 8(1 _
dY --7 Z - ZI
X-X'+f.lt
)

In view of the relation

1
8 (f (f.l)) = 11'(f.lo) 18 (f.l- f.lo),
where the value Ito is obtained from the equation

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

J rr (f) (X , tt, v, X, "

tt ", v" , tt') "v" , X' , tt,, v,I t' , t 1 ) dX" dH"
z, rr(f) (X" , tt, r dH"
r

rr(f)(x, p" v, X', tt', v', t z , td . (99)

One can check, that the propagator (89) is the solution to the differential equation

arr _ tt i.. rr = 8 (tz - t l ) 8 (X - X') 8 (p, - tt') 8 (v - v'), (100)

at z av
which is the partial case of eq. (71) for zero potential energy function.

MEASUREMENTS IN QUANTUM AND CLASSICAL MECHANICS

We will review the discussion of quantum measurements done in [5, 6]. It is

known [19, 20], that quantum mechanics has trouble in the sense that it is incom-
plete and needs the notion of a classical device measuring quantum observables as an
important ingredient of the theory. Due to this, one accepts that there exist two worlds:
the classical one and the quantum one. In the classical world, the measurements of clas-
sical observables are produced by classical devices. In frame of standard theory, in the
quantum world the measurements of quantum observables are produced by classical de-
vices, too. Due to this, the theory of quantum measurements is considered as something
very specifically different from classical measurements.
It is phsycologically accepted, that to understand the physical meaning. of a mea-
surement in the classical world is much easier than to understand the physical meaning
of analogous measurement in the quantum world.
As it was pointed out in [5, 6], all the roots of difficulties of quantum measurements
are present in classical measurements, as well. Using the relations of the quantum states
in standard representation and in the classical one (described by classical distributions),
one could conclude that the complete information about a quantum state is obtained
from purely classical measurements of the position of a particle made by classical devices
in each reference frame of the ensamble o~ classical reference frames, which are scaled
and rotated in the classical phase space.
These measurements do not need of any quantum language, if we know how to
produce, in the classical world (using the notion of classical position and momentum),
reference frames in the classical phase space differing from each other by rotation and
scaling of the axis of the reference frame and how to measure only the position of
the particle from the viewpoint of these different reference frames. So, knowing how
to obtain the classical marginal distribution function w (X, tt, v), which depends on
the parameters tt; v, labeling each reference frame in the classical phase space, we
reconstruct the quantum density operator.

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

[1] E. Schrodinger, Ann. d. Physik, 79 (1926),489.

[2] L. De Broglie, Compt. rend., 183 (1926), 447; 184 (1927), 273; 185 (1927), 380.
[3] E. Madelung, Zeits. f. Physik, 40 (1926), 332.
[4] D. Bohm, Phys. Rev., 85 (1952), 166; 180.
[5] S. Mancini, V. 1. Man'ko, and P. Tombesi, Phys. Lett. A, 213 (1996), 1.
[6] S. Mancini, V. 1. Man'ko, and P. Tombesi, "Classical-like description of quantum
dynamics by means of symplectic tomography", Los Alamos Report No. quant-
ph/9609026; Found. Phys. (to appear, 1996).
[7] S. Mancini, V. I. Man'ko, and P. Tombesi, Quantum Semiclass. Opt., 7 (1995), 615.
[8] G. M. D'Ariano, S. Mancini, V. 1. Man'ko, and P. Tombesi, Quantum Semiclass.
Opt., 8 (1996), 1017.
[9] V. 1. Man'ko, "Energy levels of harmonic oscillator in classical-like formulation of
quantum mechanics," contribution to the Second Sakharov Conference (Moscow,
May 21-25, 1996); "Transition probabilities between energy levels in frame of clas-
sical approach," contribution to the Inaguration Conference of APCTP (Seoul, June
4-10, 1996), (to be published in the Proceedings of the Conferences by World Sci-
entific, Singapore, 1996); "Optical symplectic tomography and classical probability
instead of wave function in quantum mechanics," contribution to the XXI Intern.
Colloquium on Group Theoretical Methods in Physics (Goslar, July 15-20, 1996);
J. Russ. Laser Res., 17, No.6 (Plenum Press, New York, 1996).
[10] K. Vogel and H. Risken, Phys. Rev. A, 40 (1989), 2847.
[ll) D. T. Smithey, M. Beck, M. G. Raymer, and A. Faridani, Phys. Rev. Lett., 70
(1993), 1244.
[12) K. E. Cahill and R. J. Glauber, Phys. Rev., 177 (1969), 1882.
[13] V. V. Dodonov, 1. A. Malkin, and V. 1. Man'ko, Physica, 72 (1974), 597.

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

M. Moshinskyl and Yu. F. Smirnov

Instituto de Fisica, UNAM

Apartado Postal 20-364
01000 Mexico, D. F. Mexico

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.

SUPERMULTIPLET FORMULATION OF THE DIRAC OSCILLATOR

OF SPIN 1/2

Several years ago Moshinsky and Szczepaniak [3] introduced the concept of Dirac
oscillator through the equation

[a (p - iwr;3) + ;3]4> = E4>, (1)

with r, p the coordinates and momenta, w the frequency of the oscillator and

1 Member of EI Colegio Nacional

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 243
 (2)
where t1' is the vector of Pauli spin matrices.
The matrices a,(3 of dimension 4x4 can be converted into direct products of 2x2
matrices by introducing the definitions

j (3)

j = [ o1 0]
1 ,tl = 2"1 [01 1O] , t2 = 12"i
[0 -i] ,t3 = 2"1 [1
0 0 (4)

Clearly we have then a and (3 as the direct products

(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)

If we apply t3 to the ket I!T) we note from (4) that

t31!T) = TI!T) , T= !, (8)
There is another way though to write an operator that has the eiqenvalue T and it
is through the operator
(9)
as its action on I!T) will be

t( -1)!- It T ) = H-l)i- ltT ) = TltT),

t3 T (10)
as seen immediately from the fact that T = t. Thus besides the form of f3 given in
(5) we could also write it in the form

(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)

and the equation (15) becomes the matrix one

(17)

To express the matrix in (17) in a real form, we make a change

(18)

in the wave functions and obtain

(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

or E2 = 1 + 2w(N - j + ~). (21)

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

In ref. 1 we showed that a Dirac oscillator with spins up to (n/2), where n is an

arbitrary integer, satisfies the equation
n
L [au' (p - iwrB) + ,Bull/; = nEl/;, (22)
u=l

where on one hand

,Bu = 10 1 ... 0 I0,B 0 I ... 0 I, u = 1,2, ... n. (23)

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]

B = ,B 0 ,B . .. 0 ,B 0 ... 0,B. (24)

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

u=l u=l u=l

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

Laiu, L,Bu, i = 1,2,3, (27)

u=l u=l

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

Laiu = 4Ri1 , L,Bu = 2T3 , (28)

u=l u=l

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

B= i 0 (-1) 'i- L~=l 13u = i 0 (-1) 'i-To, (30)

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

H'lj; == 2i~{tRil [1-(-1)~-T3hi

+ t Ril [1 + (-1)~-T3Hi} 'lj;

+ 2T3'lj; = nE'lj;. (33)

Eq. (33) is the same as (22) but now expressed in the supermultiplet notation and
we shall proceed to discuss its solutions in the next section.

SOLUTION OF THE DIRAC OSCILLATOR OF ARBITRARY SPIN IN

THE SUPERMULTIPLET PICTURE

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

INZ; {h}siT;jm) = L(/fl,saljm)INZfl)l{h}jsatT). (38)

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', {h}y'S't'T',jmIR,\ . 11\Nl, {hhstT,jm)

= (-I(+s-jW(ll'ss'; Ij) [(21' + 1)(2s' + 1)]1/2
(tT, U\t'T') (N'['\\11\\Nl) ({ hh' s't'\\R\\ {h hst), (44)

where W is a Racah coefficient and < I > is Clebsch-Gordan one.

The reduced matrix of element of "l is given by [6]

(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)

When determining the matrix

II (N'l', {hh's't',jmIHI Nl, {hhsT,jm)11 , (47)

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

I ({ hh's't'llRl1 {h hst) II, (48)

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

Arnold-Sommerfeld-Institut fur Mathematische Physik

Technische Universitat Clausthal
D-38678 Clausthal-Zellerfeld, Germany

INTRODUCTION

The maximal group of space-time transformations which are automorphisms of the

solution variety of the free Schrodinger equation is the Schrodinger group. Its Lie al-
gebra is the (centrally extended) Schrodinger algebra s( n) in n space and one time
dimensions. In the first part of this article we present a method to reconstruct the
free Schrodinger operator from s(n). For this purpose we use a method based on the
determination of singular vectors of the Verma modules of the latter algebra. The
idea of using this method has been proposed for semisimple Lie algebras by Kostant
[1 J and others. It was only more recently that Dobrev gave a complete description
this programme [2]. As s(n) is not semisimple, the method cannot be applied di-
rectly here. However a closer inspection shows that a generalization to certain non-
semisimple Lie algebras is possible. If certain requirements are fulfilled such a gener-
alization gives in general a family of partial differential equations with invariant solu-
tion variety. Throughout this article we will call the corresponding partial differential
operators also invariant in this case. For s( n) this family contains the Schrodinger
operator.
In the second part of this article we will apply the generalized method to a q-
deformation of s( n) for n = 1 to derive a family of q-difference operators invariant
under this deformed algebra. Such q-deformations came up with the notion of gener-
alized symmetries and quantum groups and are considered as candidates for physically
relevant symmetries. Especially quantum groups lead to a renewed consideration of
q-difference operators, which have been introduced already about 100 years ago by
Jackson and others in a different context. A q-deformation of a partial differential
equation into a q-difference equation is not unique but depends on the principles that
were used to construct it. We use a deformation of s(l) in which some differential
operators in a given representation of s( 1) are changed into q-difference operators. It

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 251
turns out that the q-Schrodinger equation presented in this article is different from the
proposals that have been presented in the literature so far.

THE SCHRODINGER ALGEBRA

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)

i.e., the maximal group of space-time transformations

g: (t, x) --+ g(t, x) E IRt x 1R~

such that for each solution 1jJ of (1)

is also a solution. The transformations Tg are hence automorphisms of the solution

variety of (1). Their infinitesimal generators

X(t,x) = a(t,x)8t + bk(t,X)ak + c(t,x)

are the solutions of the operator equation:

[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:

g(t, x) (d,a,b,a,v,R) (t,x)

(3)
( d2 t+b dRx+vt+a)
1 + a(t + b)' 1 + a(t + b)
with real parameters d =I- 0, b, a, v
and a 50(3) matrix R.
We formulate the Lie-algebra s( n) -as essential part of this article-together with
its vector-field realization in slightly more general terms.

252
Definition and Basic Properties

Consider s(n), which is generated by Pt , Pi, D, Jik (= -Jki), D, G i and K, where

(i, j, k = 1, ... ,n), with the following non-vanishing commutation relations:

[Pt, D] = 2Pt , [Pt, Gd = Pi, [Pt, K] = D,

[Pi,D] = Pi, [Pi,K] = Gi , [D,G;] = Gi ,
[D, K] = 2K 1
(4)
[Pi, Jkl ] = DilPk - Dik P/ ,
[G i , Jik] = DikGj - DjkGi ,
[Jij, Jkd = Didjl + DjlJik - DilJik - Djdil .

Some of its subalgebras that are also interesting from a physical point of view are

the Euclidian- Lie algebra e( n) generated by: Jik, Ph Pi,

the Galileian-Lie algebra fl(n) generated by: Jik , Pt , Pi, Gi,

and the Euclidian Weyl-Lie algebra generated by: Jik, Gi , D,

sl(2) generated by PI, D, K.

As already mentioned, s( n) is not semisimple; it allows for any n exactly one central
extension realized by the additional relation

(5)

which gives the Schrodinger algebra.s( n).

Note that (5) is also an extension of the Galileian (sub)algebra fl(n). However, for
n = 1 or 2, fl(n) has n + 1 extensions (d. [4]).
Consider now representations of.s( n) through first order partial differential operators
(first order representations) on suitable function spaces on (t, x) like (2). One finds:

Pt Ot,
Pi Oi,
D 2tot +XiOi -d,
(6)
J ik XkOi - XiOk
Gi tOi + jJXi
K t20t + tXiOi + ~XiXi - td

where Ot == f and Oi == of). The arbitrary (real or complex) number d is called

conformal w~ight. (6) is the most general first-order representation of .s(n) up to
coordinate transformations, however, there exist e.g. more general representations on
vector-valued function spaces (d. [5]). For n = 3 the generators (6) can be transformed
into the form (2) if one chooses the special case d = 3/2.
The above elements are generators of the following space-time transformations (x E
~n):

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;

for /1 E IR and d = -n/2, s( n) is a symmetry algebra for the (n + 1)-dimensional

heat equation (or diffusion 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).

VERMA-MODULES AND SINGULAR VECTORS

Construction of Verma-modules

We present a method for the construction of partial differential operators invariant

under the Schrodinger algebra. We restrict ourselves to the case n = 1 in order to avoid
technical subtleties and to give a short description of the method. For this purpose
we generalize a construction developed by Dobrev [2] for semi simple Lie algebras. It is
based on the determination of singular vectors of Verma-modules over '5(1) analogously
to the corresponding procedure for semi simple Lie algebras (c.f., e.g., [12]). A more
detailed account of this construction can be found in [13].
For the construction of Verma modules it is necessary to have a triangular de-
composition or an analogue thereof. Notice that '5(1) admits the introduction of a
grading which for n = 1 is given by:
deg(Pt ) = -2, deg(Px ) = -1,
deg(D) = deg(/1) = 0, (7)
deg(G) = I, deg(K) = 2.

254
D acts as a grading operator, i.e., for any generator A of 5(1) we have

[D, A] = deg(A) A.

The grading (7) gives a triangular decomposition:

5(1) = ____ Ef)s(l)o

5(1)+ "-v-' Ef)s(I)-
____ . (8)
G,K D,p, Pt,Pz

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)

The Verma-module V d characterized uniquely (up to a factor) by d is then the 5(1)

module generated from Va by G k , ]{l, i.e., a vector space spanned by

(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 ,

D Vk,l = (k + 21 - d) Vk,l, (11 )

Px Vk,l = I Vk+l,l-l + {lk Vk-l,l ,

Pt Vk,l = l(k -1-1 - d) Vk,l-l + {lk(k:;l) Vk-2,1.

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

degv(Av) = deg(A) + degv(v)

holds.

Determination of singular vectors

For an analysis of the reducibility of V d it is useful to consider a singular vector

VB ofS(l). It is defined as a "second vacuum vector" of degree p > 0, i.e., some vector
v B ( =1= va) E V d with

D VB = (p - d) VB, Pt VB = Px Vs o. (12)

If such a vector exists then V d is reducible because

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

and checks the relations

Pt v P = Px v P = O.
The result is the following (for a proof see [13]):
The singular vectors of v d are (with an arbitrary factor aD 1= 0)
(13)

for pEN and d arbitrary if /l = 0;

v;: = aD f
1=0
(-2/l)1 0) Vp -21,1 (14)

for p even, d = ~ for jJ arbitrary.

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

which is a consequence of the factorization in (15) and of (14).

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:

For p even and d = this reduces to

(17)

In order to consider the corresponding partial differential equations in a function

space we need an explicit realization of the module ~ as a function space. For this
purpose we set
Va ---7 fa = 1

and

which is a 1-1 map. In special cases these basis vectors can be expressed using degenerate
hypergeometric polynomials:

fo,1

(18)

where (a h == a (a + 1) ... (a + b - 1) is the Pochhammer-polynomial. The other basis

vectors are more complicated polynomials in x and t for which we do not have such
convenient expressions.
Because of

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)

which therefore constitutes a family of5(1) invariant partial differential equations

(wi th d = Ef) . Setting f.1 = i W- the lowest (p = 2) member of this family is the usual
Schrodinger equation. For f.1 E lR it is the heat (or diffusion) equation.
Analogous results hold for s( n) with arbitrary n.

257
q-DEFORMATIONS

We present a q-deformation (in the sense explained in section 1) of :5(1), i.e., a

realization via q-difference and (partial) differential operators with q-Lie brackets and
usual Lie brackets. It will turn out to be possible to treat this deformation, i.e., a
q-algebra, in a completely analogous fashion as the un deformed one; consequently we
obtain a q-difference analogue to the free Schrodinger equation. For details of the
following construction we refer to [14].

A q-Schrodinger-algebra via q-assumptions

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:

so that with the q-numbers

a a
q'i - q-'i
[a]~ 1 _l.
q2 - q 2

the q-difference operators are

To construct a q-deformation of:5(1) we replace differential operators by q-difference

operators occuring in (6) in the following general way:

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).

2. The subalgebra structure should be preserved by the deformation and, in partic-

ular, the q-deformed s1(2) sub algebra generated by (the deformations of) D, J(
and G should coincide with the usual Drinfeld-Jimbo deformation Uq (s1(2)). Es-
pecially this means that D, which is the only element from :5(1) (and from s1(2)0),
remains undeformed.

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):

0, [Pt, D] 2Pi> Ptg - qgPt = Pi>

[Px,D] P x, pxg - q-lgpx = fl, (21 )

[D,g] = g, [D, JC] = 2JC.

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

- q-l [d]q t + q-~+d [~]q flX2 q2Nt -2Nx .

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 )

JC Vk,l = Vk,l+l, 9 Vkl = Vk+1,I,

DVk,1 =(k+21-d)vk,[,
Px Vk,[ = q d+l-l-k [I] q Vk+l,l-l +q 1-k
2 J1 [k]'q Vk-l,l , (23)
[k]~[k - 11~
P t Vkl = [l]q[k - 1-1 - dlq Vk,l-l + J1 [2]~ Vk-2,l

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)

with the q-binominal coefficient

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

With an analogous procedure as before we can now construct a realization of Lq=2

starting with Vo -> 1 by

(it is possible to find an analogue of (18) via q-degenerate hypergeometric functions, d.

[14]). Inserting the realization (22) into formula (24) yields a family of .5; (I)-invariant
q-difference operators, the "lowest" (p = 2) member is a q-analogue of the Schrodinger
operator. It is explicitly given as:

(25)

The q-difference equation

Sq'lj; =0
is a q-analogue of the free Schrodinger equation, which is its q -> 1 limit. It differs from
other q-Schrodinger equations given in the literature [17]. As special solutions of (25)
we have the basis vectors of the modules Lq=2 In special cases they are given as q-
analogues (d. [14] for the explicit form) of the functions given in (18), which are known
solutions of the free Schrodinger equation. More general solutions are yet unknown.

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

The free group on n generators, denoted byFn, is a characteristic subgroup of the

universal Coxeter group on 11 + 1 generators, which is denoted by W nH . Moreover
Aut(WnH ) is faithfully represented on Fn if 11 ;::: 2, which means that Aut(WnH ) is a
subgroup of Aut(Fn) for 11 ;::: 2.
This observation was the motivation for the presentation of Aut(F3 ) given by P.
Kramer [1] ; it is very likely that a similar presentation can be given also for arbitrary
n. The presentation given in [1] consists of 5 generators and 18 relations. There are 4
generators which are contained in Aut(W4 ), and 11 of the relations involve only those
4 generators. It turns out that those 4 generators and 11 relations actually form a
presentation of Aut(W4)'
This fact can be verified by applying a result of Fouxe-Rabinowitch on automor-
phisms of free products because Wn is the free product of n copies of Z2 (cf. [2], see also
[3]). There is another approach (d. [4]) to deduce a presentation of Aut(WnH ) which is
closer to the theory of Coxeter groups and which gives perhaps more information about
the structure of Aut(WnH ).
The purpose of the present paper is to give an idea how the methods of [4] work.
We will show how the action of Wn on its Cayley-graph (which is a labelled tree) can
be used to get information about Aut(Wn ). We will establish a set of generators of
Aut(Wn ) and give a proof for relations between Aut(Fn) and Aut(Wn+1 ) mentioned
above.

1 Supported by the Deutsche Forschungsgemeinschaft

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 263
Contents
A universal Coxeter system is a pair (W,5) where W is a group generated by a set 5
of involutions, such that the relations S2 = 1 form a presentation of W. If 151 = n then
W will be called the universal Coxetergroup on n generators and we denote it by W n .
In the following we fix a natural number n, and (W, 5) denotes the universal Coxeter
system on n generators S1, . .. , Sn.
Theorem A: Let 'P : W -> {+ 1, -I} denote the homomorphism which sends each
generator Si onto -1. Then

(i) we := Kern 'P is a characteristic subgroup of W.

(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 .

Given a permutation 7r E Sn, there exists a unique automorphism a" of W such

that oo 1f (si) = S1f(i) for all 1 :::; i :::; n. If 1 :::; i =I- j :::; n then there exists a unique
automorphism O"ij E Aut(W) such that O"ij( Sk) = Sk if 1 :::; k =I- j :::; nand O"ij( Sj) =
SiSjSi

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.

Groups acting on trees

The following proposition will be usedin the proof of Theorem A:

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.

Proof: This is an easy consequence of Theorem 4 in [5]. o

Let r be a tree of order n, where n is some natural number, and let I = {I, ... , n}.
A labelling of r is a mapping r : E ...... 1 such that r lEv is a bijection for each vertex
v E V.
We denote the group of label-preserving automorphisms of r by W. Here are some
elementary facts which follow immediately from the general theory of Coxeter groups
(see [6]), but which can also be verified at once in our very special situation:

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.

3. For W E Wand e E E we have E(W 0 O"(e) 0 w- 1 ) = w(e).

4. Two involutions 0"1, 0"2 E Ware conjugate in W ifand only if r( E( 0"1)) = r( E( 0"2)).
5. The group W acts regularly on V.

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

The following proposition will be used in the proof of Theorem B.

Proposition 3: Let F ~ E be a set of edges such that F = of. Put SF = {u(f) I f E

F} and WF = (SF). Then (WF' SF) is a universal Coxeter system. We have W = WF
if and only if F = E" for some vertex v EV.

Proof: This follows immediately from Proposition 3 in [7]. o

Convention
From now on we fix a natural number n and I denotes the set {I, 2, ... n}. The graph
r = (V, E) is a tree of order nand r : E ...... I is a labelling of r. We identify the Coxeter
system (W, S) as follows: The groupW is the group of label preserving automorphisms
of r and S = 0"( Eva) where Vo is a fixed vertex of r.

265
PROOF OF THEOREM A

Lemma 4: There exists a normal subgroup N of W such that WIN is isomorphic to

the elementary abelian group of order 2n.In particular, if W (resp. we) is generated
by a set X (resp. Y), then IXI;:::: n (resp. WI;:::: n -1).

Proof: Let i E I and define i : 5 -+ {+1,-1} by i(Si) = -1 and i(Sj) = 1

if j E 1\ {i}. By the universal property there exists a unique homomorphism <Pi :
W -+ {+1,-1} which extends k Let Ni denote the kernel of <p. Now N:= niE/Ni
has the required properties, which yields the first assertion. The second assertion is an
immediate consequence of the first. 0

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

Lemma 6: we is a characteristic subgroup of W.

Proof: We recall that we was defined as the kernel of the homomorphism \0 : W -+

{+ 1, -I} extendig the map sending each generator Si on -1. It follows by Remark 4 in
Section "Preliminaries" that \O(t) = -1 for each involution t E W. Thus we have that
\0 = \0 0 a for each automorphism a of W. Therefore each a E Aut(W) stabilizes the
kernel of 'P and hence we. 0

Lemma 7: Let n #- 2. An automorphism of W centralizing we is the identity.

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.

Proof: This is an immediate consequence of Proposition 3. o

LeIllIlla 9: Let F be a base, then T IF is a bijection between F and I.

Proof: Suppose that there is an i E I which is not contained in T(F). Define

;jJ : S --+ {+1, -I} by;jJ(Si) = -1 and ;jJ(Sj) = 1 if j E 1\ {i}. By the universal
property there exists a unique homomorphism > : W --+ {+ 1, -I} which extends ;jJ. It
follows immediately from Assertion 4. in Section "Preliminaries" that >( a(J)) = 1 for
all f E F. Thus a(F) is contained in the kernel of > which is a contradiction. 0

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.

Proof: This is immediate. o

LeIllIlla 11: Let i E I be such that fi {Xi, Yi} E Fe and put J {j I Ij E
F n H(Ji, Xi)}. Then D(F(i, J)) < D(F).

Proof: Let F' = F( i, K). If k, lEI \ J, then I~ = fk, II = ff and therefore

8(J~,f!) = 8(Jk,II)' If k,l E Ju {i}, then f~ = (a(Ji))(ik),ff = (a(Ji))(Jz) and as
a(f;) is an automorphism of r it follows also that 8(JL In = 8(Jk, fz). It remains
to consider the case where k E 1\ ({i} U J) and l E J. We apply Lemma 2 to
see that 8(J~, f!) < 8(Jk, fl)' As fi E Fe it follows that there is at least one pair
(k, l) E (I \ ({i} U J)) x J and therefore D(F') < D(F). 0

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:

1. The group H = (a" I JT E Sn) is a complement to Spe(W) in Aut(W).

2. If a E Aut(W), then c(a(S)) is a base.

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:

(aJ F is a root base.

(bJ D(F) = (O)i,jE!'

(cJ a E Inn(W).

Proof: That (a) is equivalent to (b) is obvious.

(a) implies (e): This is a consequence of Assertions 3. and 4. above.
(e) implies (a): Suppose that a corresponds to the conjugation with the element
w" E W. Then we apply Assertion 3. of Section "Preliminaries" to see that ( a( S)) =
w,,(Evo ) and as w" E Aut(r) we have w,,(Evo ) = EVI where VI = 11i,,(VO)' 0

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

Conclusion: Let a E Spe(W) and let F = t(a(S)). If D(F) = (O)i,jE[, then a E

Inn(W) by Lemma 12 and therefore a E (crij I 1 :::;c#- j :::; n) by Assertion 5. above.
If D(F) > (O)i,jEI then F is not a root base (by Lemma 12) and hence F #- of by
Lemma 8, which means that FO #- 0. Choose i, J as in Lemma l1, then it follows by
Lemmas l1 and 13 that D( t( a 0 criJ( S)) < D( t( a( S))). Thus it follows by induction
on D(t(a(S)) that a E (crij 11 :::; i #- j :::; n). Now Theorem B follows from Assertion
1. above.

REFERENCES

[1] P. Kramer, J. Phys. A: Math. Gen. 28 (1995), 379.

[2] D. I. Fouxe-Rabinovitch, Mat. Sb. 8 (1940), 265.
[3] N. D. Gilbert, Proc. London Math. Soc. (3) 54 (1987), 115.
[4] B. Miihlherr, Preprint, 1996.
[5] J. P. Serre, Trees, Springer, Berlin, Heidelberg, 1980.
[6] N. Bourbaki, Groupes et Algebres de Lie, Chapitres 4, 5 et 6, Hermann, Paris, 1968.
[7] J. Tits, J. Alg. 113 (1988), 346.

268
GENERALIZED QUANTUM MECHANICS AND
NONLINEAR GAUGE TRANSFORMATIONS

Peter Nattermann1

Institut fur Theoretische Physik

Technische U niversitat Clausthal
D-38678 Clausthal-Zellerfeld, Germany

INTRODUCTION

There have been several approaches to a nonlinear modification of orthodox quantum

mechanics and, in particular, of the Schrodinger equation. As the equations will appear
in this paper we cite the nonlinear Schrodinger equations proposed by Bialinycki-Birula
and Mycielski [1],

(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,

Their family of equations depends on a real quantum number D E JR obtained from

the classification of unitary representations of S(JR3 ) [6, 7], and five model parameters 2
Cj E JR,

1 aspn@pt.tu-clausthal.de
2 D' being; a constant with the dimensions of D, to make Cj dimensionless.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 269
For a further discussion of these equations we refer to [8].
Along with Weinberg's approach to a nonlinear modification of orthodox quantum
mechanics [9, 10] there have been many discussions on the physical implications of a
nonlinear quantum theory. In particular, Gisin [11] and Polchinski [12] pointed out
that - by an EPR-type experiment - a nonlinear evolution equation would give us
the opportunity of faster than light communications. Although this argument might
not be persuasive in a non-relativistic theory, the structural implications of this effect
led to Weinberg's sceptical attitude [13].
However, their argument is based on the notion of observables (as linear operators)
and (mixed) states (as density matrices) of orthodox, linear quantum mechanics. The
inconsistency of this notion for a nonlinear time evolution of pure states becomes most
evident for the time evolution of mixtures: Density matrices may be expanded in general
non-uniquely in terms of weights AJ') and eigenfunctions 1/;)'),

(4)

But evidently under a nonlinear evolution of the pure states 1/;j and 1/;; the two expansion
evolve differently,
(5)

and there is no consistently defined time-evolved density matrix Wt!

Rather than interpreting this inconsistency as a source for supraluminal communi-
cations we need a generalized concept of observables and mixtures for nonlinear time
evolutions.
A framework for such a generalization is due to Mielnik [14] and is based on the
observation that "what one observes in reality is a distinguished role of the position
measurement". The following section provides a brief review of Mielnik's Generalized
Quantum Alechanics along with two examples.
Similar considerations on the distinguished Tole of positional measurements have
recently led to the notion of nonlinear gauge transformation in nonlinear quantum
mechanics [15, 16, 17]. In the third section we merge these two ideas and introduce
generalized gauge transformations in generalized quantum mechanics.
A particular, st'riclly local class of generalized gauge transformations on L2(I~a, d3x)
is identified as the class of nonlinear gauge transformations previously introduced [15,
16, 17]. They are used in the fourth section to obtain linear quantum systems in
a nonlinear disguise; their gauge generalization (in analogy to minimal coupling in
orthodox quantum mechanics) leads to a unification of Bialinycki-Birula-Mycielski (1)
and Doebner-Goldin equations (2).
Using these transformations Lucke [18] proposed a generalized concept of nonlinear
observables slightly different from Mielnik's. The concept and its relation to generalized
quantum mechanics will be the content of the fifth section.

GENERALIZED QUANTUM MECHANICS

As indicated in the introduction, we are looking for a quantum theory capable of

nonlinear time evolutions of pure states. Mielnik's concept for generalized quantum
mechanics provides a framework for such a theory. It has three basic ingredients.
The first ingredient has to be a "model space" for the pure states. For a general
setting we will take a topological manifold ;r as the manifold of pure states. The

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,

Tt ,t' ,Ext : ::r ---; ::r. (6)

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 ,

o := {p 0 T Ip E !P, T E M} p.c. (9)

3 i.e. we allow for infinitesimal, shock-like evolutions, see [19].

4 i.e. again we allow for infinitesimal combinations.
5 Here we deviate from Mielnik [14] who took the lwear span for these combinations.

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.

Example 1: Orthodox quantum mechanics can be described in terms of generalized

quantum mechanics [14]. For a single particle in 1R3 we consider a linear quantum
system (:J-C, U(:J-C) , :P). Here the manifold of pure states is a separable Hilbert space in
a position representation,
(14)
Thus for all non-zero elements 1/; of:J-C 11/;(xW /111/;11 2 is a positional probability distri-
bution, i.e. the position observables are expectation values of characteristic functions

(1/;lxB1/;) IB ?j;(x)1/;(x) d3x

(15)
PB[1/;]:= 111/;112 = I~31/;(x)1/;(x) d3x .
The time evolutions are generated by linear Schrodinger equations

(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)

Example 2: A variation of the previous example of a single particle in ]R3 is due to

Haag and Bannier [23]. They consider a quantum system (9-C, M, 3J), where 9-C and
3J are as in (14) and (15), respectively. But for the time evolutions they include a
nonlinear term in the Schrodinger equation,

.
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

In orthodox quantum mechanics (on ]R3) a gauge transformation is a local unitary

operator De E Utoc
(25)
We may describe Utoc as the set of linear transformations that leave the position prob-
ability invariant. These gauge transformations are employed for introducing classical
electro-magnetic fields by replacing OtB and VB obtained in the Schrodinger equation
through (25) by more general scalar and vector potentials V and X, respectively, a
process denoted as gauge generalization in [17].
This notion of gauge transformations and gauge equivalence can naturally be adopt-
ed to the generalized setting of the previous section: We define two quantum systems
(!r{l) , M(1), 3J(1)) and (:r 2) , M(2l, 3J(2)) to be gauge equivalent, iff

273
 L 3>(1) and 3>(2) are positional observables over the same space M;

2. the external conditions C are the same,

3. and there is a homeomorphism N : :Jil) -> :Ji2) such that

(1)
PB(2) 0 N , Vp U) E 3>U) ,
PB
(1) N- o T(2)
1
t,t',Ext
N Vt, t' E ~, Ext E C.
Tt,t',Ext 0 ,

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.

GAUGE EQUIVALENCE ON L2(~3, d3x) AND GAUGE GENERALIZA-

TIONS

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

N[c1jJ] = c'(c)N[1jJ] , Vc E t. (27)

Applying this relation twice to c = Cl C2 leads to Cauchy's power equation on t,
(28)

with the continuous solution

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.

N[1jJ](x) = n(1jJ(x), x), (30)

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)

the semi-direct group structure being given by

(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:

i 12 at 1/;; = (- ;~ ~ + V) 1/;; - ~1tln 11/;;1 2 1/>;

h2 . 122 2
+ 4~t (iRd1/;] + 2R1 [1/;;)- 2R4 [1/;;)) 1/;; - 8~ (2R 2 [1/;;]- Rs [1/;;)) 1/;;
(37)
with Rj[1/;] as in (3).
Note, that these equations are formal generators of the nonlinear time evolution
[18,25]
lIt = N, 0 Vet) 0 N_, , (38)
where Vet) denotes the unitary time evolution of the linear Schrodinger equation (16).

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

VI (iRd7jJ] + ~R2[7jJ]-R3[1/']- ~R5[7jJ])

+ flIt (iR2[7jJ] + Rd7jJ]- R4[7jJ]) + fl2t (R2[7p]- 2Rs[7jJ]) + au In 17jJtl 2
(39)

that is invariant under gauge transformations NT In (39) we have used the expansion

(40)

and introduced new parameters Vj, flk, and al in an obvious way.

Further steps of gauge generalization are possible. For instance, if we distinguish real
and imaginary parts in (39) and close the resulting family of equations with respect to
the gauge group N we obtain a unified (and generalized) Bialynicki-Birula-Mycielski-
Doebner-Goldin equation
2 5

i f)~7jJt --,- flo V =i L VjtRj[7jJt] +L flktRk[7jJt] + au In 17jJtl 2 (41)

t j=l k=l

with real time-dependent parameters.

By construction - as in the linear theory - the generalized gauge transformations
are still automorphisms of the set of nonlinear Schriidinger equations obtained.

NONLINEAR OBSERVABLES

Having constructed the time evolutions of nonlinear quantum systems, we would

like to know the structure of the set of observables of these quantum systems.
To gain some insight we utilize the concept of generalized gauge transformations
and consider again a quantum system (1-f, JVC,:1 gauge equivalent to a linear quantum
system (1-f, U(1-f) , :1, i.e.1-f and:1> are as in (14) and (15), respectively, and the motion
group IS
(42)
where N is a (not necessarily strictly local) generalized gauge transformation on 1-f =
L 2(l~3, d 3x). However, we shall assume as in the previous section that N is norm-
preserving on 1-f = L2(l~?, d3x).
According to (19) of example 1 the observables of the linear quantum system
(1-f, U(1-f) , :1 are associated to projection operators E. For the nonlinear quantum
system (1-f, JVC,:1 the set of observables is then

Onl ':::' { E: 1-f -l- 1-f IE := No E 0 N- I , E E Proj (1-f) } (43)

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,

II(E 0 N- 1 0 N)[1/JJI12 liE [N[1/JJ] 112

(44)
111/J112 IIN[1/J] 112
The standard definition of observables in linear quantum theory, however, goes one step
further and associates to each observable A a self-adjoint operator, or equivalently (by
the spectral theorem) a projection-valued measure

(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)

It inherits the following properties from the linear case:

(i)
/I B f-+ /I (B)'= IIEA(B)[1/J]112 (47)
rV;' rV;' 11~)1I2

for all BE B(lR.), 1/J E 9i defines a probability measure on lR.;

(ii)
(48)

(iii)
(49)

If EA(B) may be interpreted as the probability of measuring a value of a physical

observable in B, we may call the generalized projection-valued measure a nonlinear
observable.
For linearizable quantum systems this generalized notion establishes the full equiv-
alence of the two descriptions [18]:

nonlinear theory linear theory

wave functions 1/J'=N[1/J] 1/J

time evolution lit = No V t 0 N- 1 V t = exp( -kHt)
observables { E A' = No EA 0 N- 1 } {EA}, A E Lsa(X)
position {XB} {XB}

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

I am grateful to the organizers of "Symmetry in Science IX" in Bregenz for their

invitation, support, and hospitality. Some of the results presented here have been
obtained in collaboration with H.-D. Doebner and G. A. Goldin; thanks are also due
to them as well as W. Liicke for valuable discussions on the subject.

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

Centro de Fisica Teorica

Universidade de Coimbra
P-3000-Coimbra, Portugal
and
Department of Physics
Kochi University
Kochi 780, Japan

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.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 281
 One of the most challenging problems in the current studies of nuclear physics is
to give a theory suitable for description of collective motions with large amplitudes in
both even and odd soft nuclei with strong collective correlations. For providing a gen-
eral microscopic means for a unified self-consistent description for Bose and Fermi type
collective excitations in such fermion systems, the new many-body theory has been pro-
posed by Fukutome, Yamamura and the present author standing on the SO(2N + 1) Lie
algebra of the fermion operators [3]. An induced representation of an SO(2N +1) group
has been obtained from a group extension of the SO(2N) Bogoliubov transformation
for fermions to a new canonical transformation group. We start with the fact that the
set of the fermion operators consisting of the creation-annihilation and the pair opera-
tors forms a larger Lie algebra, the Lie algebra of the SO(2N + 1) group. The fermion
Lie operators, when operated onto the integral representation of the SO(2N + 1) wave
function, are mapped into the regular representation of the SO(2N + 1) group and are
represented by Bose operators. The Bose images of the fermion Lie operators are ex-
pressed by the closed first order differential forms. The creation-annihilation operators
themselves as well as the pair operators are given by the finite Schwinger type boson
representation [4].
Embedding the SO(2N +1) group into an SO(2N +2) group and using SO(2N + 2)/
U(N + 1) coset variables [5], we have developed an extended TDHB theory. In that
theory, paired and unpaired modes are treated in an equal manner. A static extend-
ed HB theory on the coset space SO(2N + 2)/U(N + 1) is derived easily from the
extended TDHB theory. The extended HB theory applicable to both even and odd
fermion systems is a SCF theory with the same level of mean field approximation as
the usual HB theory for even fermion systems. We start from the Hamiltonian of the
fermion system which includes, however, the Lagrange multiplier terms to select out
the physical spinor subspace. The extended HB equation is written by the variables as
the respective representatives of the paired and the unpaired mode. The extended HB
eigenvalue equation can be solved with a method quite analogous to Baranger's two-
step diagonalization for the usual HB eigenvalue equation [6]. Unknown parameters
included in the Lagrange multiplier terms must be determined. The anti-commutators
of the fermion Lie operators expressed in the first order differential forms satisfy exactly
the anti-commutation relations, when they are operated onto the SO(2N + 1) HB wave
function. However, this fact makes no role to determine the parameters. A determi-
nation of the parameters is possible if we demand instead that the expectation values
of the anti-commutators by an approximate SO(2N + 1) HB wave function satisfy the
anti-commutation relations, i.e., the quasi anti-commutation relation approximation for
the fermions [7].
This paper is organized a follows. First, we recapitulate briefiy the induced represen-
tation of the SO(2N +1) canonical transformation group, embedding of the SO(2N +1)
group into an SO(2N + 2) one and introduction of the SO(2N + 2)/U(N + 1) coset
variables. We give a brief sketch of derivation of the extended TDHB equation from the
classical Euler- Lagrange equation of motion for the SO(2N + 2) / U( N + 1) coset variables
in the TDSCF. A static extended HB equation on the coset space SO(2N +2)/U(N +1)
is derived from the extended TDHB equation. In the third section, the extended HB
eigenvalue equation is solved with the two-step diagonalization method. The section
thereafter is devoted to a determination of the parameters in the Lagrange multiplier
terms by the quasi anti-commutation relation approximation for the fermions. Finally,
in the last section, we give some concluding remarks. Throughout this paper, we use
the summation convention over repeated indices unless we have a fear of misunder-
standing.

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

U(G)(c,c t , l/J2)Ut(G) = (c,cf, I/h)(z - p)G , (1)

G =- [~
y/V2 -y*/V2
~: -:;!f],
z
GtG = GGt = 1 , (2)

U(G)U(G') = U(GG') , U(G- 1) = U-1(G) = ut(G) , U(I) = 1 , (3)

where (c, ct , I/V2) is the (2N + I)-dimensional row vector (c",), (c~), I/V2) and A =
(Ai) and B = (B"'i) are the N x N matrices. The symbol * denotes the complex
conjugation. The p in the gauge factor z - p is defined as p = x",cl- x:c'" and satisfies
p2 = -x:x" = z2 - 1.
When z = 1, then G becomes to be essentially an SO(2N) matrix g. The HB
(SO(2N)) wave function Ig> is generated as Ig> = U(g)IO>, where 10> is the
vacuum satisfying c'" I 0> = O. Ig> is expressed as

Ig> = <0IU(g)10>exp(I/2q"ilc~cb)10>, (4)

<OIU(g)IO> =cfl~o(g)= [det(a)r/2 = [det(I_q*q)rl/4eiT/2, (5)

q = ba- 1 = _qT , (variables of the SO(2N)/U(N) coset space) (6)

where "det" represents determinant and the symbol T denotes the transposition. The
a = (an and b = (b"'i) are the N x N matrices and satisfy the orthonormalization
condition for the proper HB amplitudes [1]. On the other hand, the SO(2N + 1) wave
function I G> is generated as I G> = U (G) I 0> and can be expressed as [5,8]
1
I G> = < 0 IU(G) 10> (1+r"cl) exp(I/2 . q"j1 C l4) 10>, r", = -(x",+qo:iJX~) , (7)
I+z

<0 IU(G) 10> = cfl~o(G) = ) 1 ; z [det(I _ q*q)r 1/ 4 eiT / 2 . (8)

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

9 = [: ~:] , gtg = ggt = 1 , Q = BA- 1 = [_~T ~] = _QT , (10)

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

<E$ + !8a.(3)a = ROI(3 , <E,,(3)a = -K"(3 , < Ea.(3 >a = K~(3 ,

(11)
<COI>a = Kao - R';,o, <cl>a = K~o - R",o ,
where Rpq and Kpq are expressed in terms of the SO(2N + 2)jU(N + 1) coset variables
Qpq as

(12)
An expectation value of a two-body operator is given as

< EOI-Y E 5(3 >a = ROI(3R-y5 - Ra5R-y(3 - K:-y K6(3 . (13)

Let the Hamiltonian of the fermion system under consideration be

(14)
where the single particle Hamiltonian h,,(3 includes a chemical potential and

[af3b8l = -[a8bf3l = b8laf3l = [f3aI8I'l* (15)

are anti-symmetrized matrix elements of an interaction potential. The expectation
value of H by the I G> is given as

<H>a = h",(3<E$ + !8a.(3)a

+ ![af3b8J{ < E$ + !8",(3 >a< E? + !8-y5 >a + !< EOI-Y >a< E5(3 >a} . (16)
The images COl and cl of the fermion operators COl and c1, represented by the closed
first oder differential operators in terms of the SO(2N +1) variables qa(3 and r a, satifsy
the anti-commutation relations {cl, c(3} = 8,,(3, etc., when they are operated onto the
spinor subspace. Therefore, we use in the regular representation space the following
image of the Hamiltonian with the Lagrange multiplier terms to select out the spinor
subspace:

H ha(3(E$ + !8 (3) + Haf3b8lEOI-Y E5(3 + H' ,

01

H' !k",(3( {c~, cd - 8",(3) + ~l",(3{ cL cb} + ~l~(3{ c"" c(3} ,

(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

-~(1- R*)-1 8H>G + ~M,,<c~>G + ~M~<ca>G (1- R)-l

11, 8Q* '
i( )-1 8H>G+~Mo;<c~>G+~M~<cex>G ( *)-1
--I-R l-R (18)
11, 8Q '
in which the classical Hamiltonian function (16) accompanies an additional term ap-
pearing as a classical part of the Lagrange multiplier terms in (17). Eq. (16) and the
classical part are represented, respectively, as

+ 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

+ kM:{OpoQ"q + Qp"OOq + QpoQ",q - Qp"QOq}

z
+ 1iM",{QpoO"q + op"QOq + OpOO",q - Op"OOq} ) (20)

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

inq Fq + qF* + D + qD*q -rMtq - qM*rT -rMT + MrT ,

ini Fr + qD*r - qM* - (Mtr)r + M . (22)

1: 1:
Using the definitions (6) and (7) for the variables q and r, eq. (22) is rewritten as

[ifib - {F - z (xMt + Mxt}}b - {D - z (xMT - MxT) }a ]

+ {F* - 1: z (x* MT+ M*x T ) }a + {D* - 1: 0,

1:
- q [ina z (x* Mt - M*xt) }b] =

[inx - Fx + Dx* - 2zM - z {in!: + MT x* - Mtx}x]

+ q [inx* + F*x* - D*x + 2zM* - 1: z {in!: + MT x* - Mtx }x*] = 0, (23)

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.

A SOLUTION OF THE EXTENDED HARTEE-BOGOLIUBOV EIGEN-

VALUE EQUATION

We are now in a stage to solve the extended Hartree-Bogoliubov eigenvalue equation.

As a basis of the single-particle states, we use states of a spherical symmetric j-j
coupling shell model characterized by the set of quantum numbers ina, la,ja, rna}. We
denote it by the Greek letter a. The time-reversed single-particle state a is obtained
from a by changing the sign of rna' We use the phase factor Sa = (_l)ia- ma in the
time reversed quantities.
By introducing a new amplitude b defined by bai = sabai [9] and assuming the time
reversal symmetry F;(3 = sas(3Fafj, the extended HB eigenvalue equation is transformed

286
into

{ F:(3 - _l_(x~M(3 + M;x(3)}af

l+z
- {s(3D*0'(3- - _l_(x* (3 - M*0' S(3X~)
1 + Z 0' s(3M"- (3 }b(3',
- {F:(3 - _1_ (sO'xaS(3M5
l+z
+ sO'MaS(3X~) }b(3i
+ {sO'Da(3 - _l-(SO'Xa M(3 - SO'Ma X(3)}af (27)
l+z
-F:(3( - :;) - D~(3(0) + hM;z 0,

F:f3(sf3~) +sO'Daf3(- :;) +hsO'Ma z = O. (28)

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

fO'f3 + 2(1 ~ z) [a,8118J (X~<C8>G + <C~>GX8) ,

1 1
dO'(3 + 2(1 + z) 2 [a,I,88J (X-y<C8>G - <e-y>Gxo) , (29)

where fO'f3 and dO'(3 are the usual SO(2N) SCF parameters defined as

(Ea - >.) 80'(3 + [a,8118J P-yo, ft = f,

~ [a,I,88J (-KS-y) , J! = -d, (30)
with the SO(2N) density matrices PO'(3 and K",(3 given by

P",(3 <E~ + ~80'(3)g ,

(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

(~80'(3 - P~(3)X(3 + KO'(3X~ ,

(~80'(3 - POi(3)X~ + K~(3X(3 . (32)
For our aim of an easy seeking for a solution of the set of eqs. (27) and (28), we
solve the extended HB eigenvalue equation in the m-scheme representation with the
use of the two-step diagonalization method by Baranger [6J. Keeping the time reversal
invariance of the eigenvalue equation, we diagonalize in the first step the particle-hole
type interaction. For simplicity, we assume the interaction to be separable and neglect
the exchange effect. Then, the Hartree approximation leads to the eigenvalue equation

(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

a ,Oi ViW",i, (not summed over i), IUil2 + IVil2 = 1, (35)

for which we impose also the time reversal ansatz

s",t;a; = a*i S",tibCii b~i ,

Ui = u1 , Vi = v1 . (36)

We also introduce the following quantities associated with the unpaired mode ampli-
tudes

(37)

which have the time reversal properties

(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)

from which we have the formal solutions for Xi and Xi,

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

Ei = Je 2 + 1t:,.12 2z IMil 2+ IMil 2

, l+z Je;+1t:,.12
2z e(IMI2 -1M-12) - t:,.M*llf. - t:,.*M*M
+ 1 + z" Z e; + It:,. I; , ,z (45)

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

E. + . + ~2eiMtMi + t:,.MtMi + t:,.*MtMi

1 , e , l+z e;+1t:,.12
2" E _ ~ ei(MtMi - MtMi) - t:,.MtMi - t:,.*MtMi '
, l+z e;+1t:,.12
g _ . _ ~2eiMiMi - t:,.MiMi - t:,.*MTMi
~ ,e, 1+z e; + 1t:,.1 2
(46)
2E ~ei(MTMi-MtMi)-t:,.MtMi-t:,.*MtMi
,- 1+Z e; + 1t:,.1 2
We can also easily show that the above eigenvectors reduce to the usual HB eigenvectors
if we put z = 1 [ll]. From eq. (42), we can get the solutions for the unpaired mode
amplitudes and for z as

2 2ei1M;l2 + 1t:,.121Mi1 2- eit:,.MTAlIi - eit:,.*f.1tM;

Ix;! = 4z (e; + 1t:,.12)2 ,

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

ia{3 = (Ca - >.). 8",{3 L IV iI 2W",i WPi + [a,Bb8] L I iI 2W

V "{i W.5i ,
i i

X~<C8>G + <C~>GX8 = L {(1-l viI 2 -l v.i1 2 )xfxj + ~(UfVi +u;vi)xfXj

ij

+ ~(ujVj + ujvj)xf x j } W"{i W.5i , (48)

and the pairing potential (39) and the chemical potential are determined by the condi-
tions

Do ~ 9 ~ {l '- 1 ~ z ~ (l x il + IXiI 2 2 )} ufv; , (49)

N= R",,,, = ~ {1- 1 ~ Z ~ (IXiI2 + IXiI2)} IViI2 + 1; Z , (50)

where N is a nucleon number of the system. They contain unpaired mode effects and
reduce to the equations for the gap energy and the chemical potential of the HB theory
if we put z = 1 [11]. It should be noticed that the magnitude of the gap energy may
be decreased due to the presence of the unpaired mode effects.

DETERMINATION OF THE PARAMETERS k AND I IN THE LA-

GRANGE MULTIPLIER TERMS

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

eaU(9) 10> = (-r",+r"rec~-q",ecDU(G)IO>,

c~U(g)IO> -c~. U(G)I 0> . . (52)

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

(f;~o(G) = HcP~o(g) + Jl; Z cP~o(G)} , (54)

we can adopt an approximate SO(2N + 1) wave function expressed as

U(9)IO> ~ iP~o(G)(I+r",c~)exp(I/2qaiJC~cb)IO>

= ~(Jl;Z+I/Jl;Z)U(G)IO>. (55)

Operations of the anti-commutators of the differential fermion operators Ca and cl on

the approximate U( G) 10> of the U(9) 10> are made as follows:

(c~CiJ+CiJc~)U(9)IO> ~ (c~CiJ+CiJC~)U(G)IO>

~ ( Jl; z + 1/ Jl; Z ) DaiJU(G) 0> 1

+ ( J 1 ; Z - 1/ J 1 ; Z ) [(1 - 1 ! J DaiJ + ~ (1 ~ z)2 x:x~] . U(G) 1 0>

-~ (V 1 ;Z -1/ Vl;Z) I!Z [x~c~-x:CiJlU(G)IO>, (56)

(c~cb+cbc~)U(9)IO> ~ (c~cb+cbc~)U(G)IO> =

-~ (VI; Z - 3/ Jl; Z ) (1 ~ Z)2 (x:x1 + x~x~) . U(G)I 0 >

+~ (Jl;Z -1/ Jl;Z) l!z[x:c1+x1c~lU(G)IO>, (57)

(caCiJ + CiJCa)U(9) 10 > ~ (caciJ + CiJCa)U( G)I 0> =

- ~ ( Jl ; Z - 3/ Jl ; Z ) (1 ~ Z)2 (XaX~ + XiJX;) . U(G)I 0 >

- ~ (Jl ;z -1/ Jl ;z) 1!z[XaCiJ+X~Cal U(G)IO> , (58)

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

< 0 lut(9)( c~ciJ + ciJc~)U(9)1 0 > D"'iJ ,

< 0 IU t (9)( c~c1 + cbc~)U(9)1 0 > < 0 IU t (9)( caciJ + CiJ c",)U(9) I0 > o , (59)

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 ;

+ {II (21 ; Z) f C; Z- 3} .x~x~

{II C; Z) -I}. [x~<C(3)aT + <ca>a*x~]
{1/(21;Z)fC;Z -3) .~(x~x1+x~x~)
= {IIC;Z)-I}.[x~<c(3)at+x~<c,,>at],
{1/(21;Z)fC;Z -3) .~(XaX~+X(3X~)
= {II C; Z) -I} . [xa<cjJ>a + xjJ<c.,>a T] T (60)

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

2{I/C;Z) -lr [3 1;Z +I/C;Z) -2-4C;Z)IC;Z +1)] x.,k.,(3

1
.

+ c.,>ax:) x~k;(3 - c,,>a*x,,) x~k'Y(3 = 0 (not summed for a) ,

r
{1I (21; Z) f .{1I C; Z) - 1 C; Z- 3) . Ix.,12(x~k'Y(3 + x~l'Y(3)
1
.

- 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

gxO = -c~P+ = -P_c~ , P = HI (-It) ,

Eg = c~P_ , E~ = co:P+ , Eg = HP- - P+) , Eo:o = co:P- = P+co:, (62)
where n is the fermion number operator defined by n = c~c", and the P+ and P_ mean
the projection operators to the subspaces of even and odd fermion numbers, respectively.
The fermion SO(2N + 2) Lie operators, when operated onto certain functions on the
coset space SO(2N +2)/U(N +I), are mapped into the regular representation consisting
of functions on the coset space SO(2N + 2)/U(N + 1). The images of the fermion
SO(2N + 2) Lie operators are expressed by the SO(2N + 2)/U(N + 1) coset variables
and the closed first order differential forms over the SO(2N + 2)/U( N + 1) coset space.
Then, the images of the fermion SO(2N + 2) Lie operators e~ etc. can be represented
in terms of the SO(2N + 2)/U(N + 1) coset variables Qpq and the phase variable
T (= i/2 In[det(A*)/ det(A)]) of the U(N + 1) in a way quite analogous to the case of
the fermion SO(2N) Lie operators as

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,
~

C" = eo", - e~, c~ = e"'o - eg . (64)

By using the SO(2N + 2) Lie operators Epq, the expression (7) for the SO(2N + 1)
wave function I G> can be converted to a form quite similar to the SO(2N) wave
function I9 > as

(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)

(1 - R*)pu(l - R)vq - (1 - R*)pv(1 - R)uq ,

(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

Mariano A. del Olmo 1

Departamento de Ffsica Teorica

Universidad de Valladolid
E-47011, Valladolid, Spain

INTRODUCTION

The idea of symmetry is essential in our understanding of the physical world. By

symmetries of a physical system we understand a set of transformations that keep it
invariant, the knowledge of them gives a lot of information about the system. On
the other hand, the idea of "deformation" is very common in physics; very often it is
necessary to make (small) corrections depending on a parameter to a theory, in such
a way that its framework is preserved. All that is in the background of the interest
of the physics community for quantum groups, considering them as sets of "general-
ized" symmetries, in the sense of deformed objects (in a continuous way maintaining
many properties of them) of those that characterize the "undeformed" symmetries: Lie
algebras and Lie groups.
Quantum groups appeared originally in Physics in relation with quantization of
classical integrable systems by the method of quantum inverse scattering introduced
by Faddeev et aJ [1]. However, in a different context but closely linked to the concept
of symmetry mentioned above, many efforts have been devoted to study the quantum
algebras and groups related with the kinematical groups (Galilei, Poincare, etc ... ) in
the search of a better understanding of the space-time and its deformed or generalized
symmetries, perhaps related with the structure of the space-time at very small distances
and very high energies [2].
On the other hand, as is well known, Quantum Mechanics can be considered as a
"deformation" of the Classical (or Newtonian) Mechanics, such that in the limit when
the Plank constant goes to zero we recover the former theory. In the same way, Special
Relativity (whose kinematical group is Poincare) can be also seen as a "deformation"
of the Newtonian Mechanics (with the Galilei group as kinematical group), and in the
limit of the light velocity c going to 00 we recover Classical Mechanics. From a group-

1 olmo@cpd.uva.es

Symmetries in Science IX. edited by Gruber and Ramek

Plenum Press, New York, 1997 297
theoretical point of view this limit is equivalent to an Inonii-Wigner (IW) contraction
(from Poincare to Galilei) [3].
All that shows the interest for the study of deformations of the kinematical algebras
or groups. Nevertheless, this has not been an easy task. The pioneering works by
Drinfel'd [4] introduced q-deformed (complex) semisimple Lie algebras, and in some
sense, after the works by Drinfel'd [4] and Jimbo [5] the deformation of these al-
gebras can be considered a solved problem, but the kinematical algebras are inho-
mogeneous Lie algebras (i. e., IW contractions of semisimple ones) and there is not
a canonical procedure for their deformation. However, the method introduced by
Celeghini et ai. [6] in the early nineties has became the most successful procedure
for deforming these inhomogeneous algebras, by means of a generalized (IW) contrac-
tion from the corresponding semisimple algebras. In this way they obtained defor l
mat ions for the Euclidean algebra in two and three dimensions, and for Galilei and
Poincare in (2 + 1) dimensions. Other procedures were presented by Dobrev [7] and
Lukierski et ai. [8], these former introduced the so called ii;-Poincare in (3 + 1) di-
mensions. They started from the semisimple algebra 0(3,2) (real form of the com-
plex algebra sl( 4, C)) and by a contraction procedure and a non-linear transforma-
tion they obtained the ii;-Poincare. However, if one continues the contraction from
the ii;-Poincare does not obtain a deformed (3 + 1) Galilei but a q-Carroll algebra,
which can be considered as a Galilei algebra where the character absolute of time is
now played by the space [9]. In other cases deformation disappears after contrac-
tion.
A series of works [10, 11, 12] have presented a systematic and unified way of deform-
ing all the inhomogeneous algebras came from the pseudo-orthogonal algebras so(p, q),
i. e., related with the families Bz and Dz in the Cartan classification of (complex) Lie
algebras. This scheme is based in a family (Cayley-Klein algebras) of pseudo-orthogonal
groups and some of their IW contractions, which are the motion groups of the geome-
tries of real spaces with a projective metric [13), and uses the concept of quantum
algebra, i.e, a deformation of the (Hopf structure of the) enveloping algebra of a Lie
algebra [4]. In parallel with the non-deformed scheme, a global building appears for
the deformed counterparts of them in low dimensions and only for the inhomogeneous
algebras in high dimensions. Among these algebras we find kinematical ones like Galilei
and Poincare.
In this paper we would like to present a review of some results about deformation
of kinematical algebras and to show some possible applications, some of them are per-
tinent, in our opinion, for a description of the q-symmetries of the space-time. We will
consider two different points of view: firstly making use of the quantum algebras and
showing their relevance in the study of lattice physics, and afterwards in the context of
the non-commutative geometry and using quantum groups and their (dual) quantum
algebras for defining algebraic versions of the concept of infinitesimal transformation
and constructing quantum homogeneous spaces we shall solve a q-deformed Schrodinger
equation.
The paper is organized as follows. In Section 2 we present a brief review about
the classical scheme of (pseudo) orthogonal Cayley-Klein algebras. The deformed
counterpart of the CK algebras is presented in Section 3. A discussion about the
differential-difference equations obtained from the deformed Casimirs is presented in
Section 4. The non-commutative homogeneous spaces as configuration spaces of a
"q-deformed" Quantum Mechanics is the subject of Section 5. Some conclusions end
this paper.

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)

where i < j < k, i,j, k = 0,1, ... , Nand Wij = rr{=i+1Wt.

Note that when all the Wi are non-zero we obtain the algebrasso(p, q) (with p +
q = N + 1, p 2: q 2: 0), in particular so(N + 1) corresponds to the case with Wi =
1, Vi. The so-called quasi-simple or inhomogeneous algebras, i.e, those obtained from
so(p, q) through a sequence of IW contractions, appear when some Wi = O. It can be
easily proved that to make Wi = 0 is equivalent to perform an IW contraction (along a
determined subalgebra) with contraction parameter c = Iw;jI/2.
In particular, when WI = 0 the CK algebras SO(0,W2,,,,,WN)(N + 1) == iSO(W2, ... ,WN)(N)
can be realized as the algebras of affine groups of transformations on IR N. The above.
basis of generators can be renamed as B = {Fi := JOi , J ij ; i < j, i,j = 0,1, ... , N}, in
this way the role of the generators Fi and Jij as translations and rotations, respectively,
is emphasized. The Lie brackets of iso W2 ,,,,,WN(N + 1) are given by

(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)

where TN = (Pi; i = 1, ... , N) is an Abelian subgroup and SO(W2, ... ,WN)(N - 1) =

(Jij ; i,j = 1, ... , N) is an (N -I)-dimensional CK group.

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)

THE DEFORMED CK ALGEBRAS

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.

The N-dimensional affine CK deformed algebras

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

and is characterized by the following homomorphisms and commutation relations:

300
Coproduct:

~(X) = 1 @ X + X @ 1, X E {PN,J;j; i,j = 1, ... ,N -I},

Counit:
i,j = 1, ... ,N; (9)
Antipode:
')'(X) = _e(N-1)~PN X e-(N-1)~PN

-Pi, i = 1, ... ,N, if X = Pi,

(10)
= { -Jim' I,m = 1, ... ,N -1, if X = jim,
-JiN -wiN(N ~ IHPt, 1= 1, ... ,N -1, if X = J iN ;

Deformed commutators:

[JiN , P;j = ~ sinh(zPN), i = 1, ... , N - 1,

[JiN, JjN] = WjN{ Jij cosh(zPN) + ~(I::-:i WiNP.Wsij (11)
- I:~:~+1 w.NP.Wisj + I:::j~l wsNPsWijs)}, i < j,
where

i,j,k = 1, .. . ,N - 1. (12)

The rest of commutators are non-deformed and are given in (2).

The quantum analogue of the second order Casimir (4) is
N-1
C(q
W2,,WN
) = z42 2 PN)] 2 + ""'
[sinh(:' ~
WiN Pl. (13)
;=1

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]

Uz(500,W2"",WN(N + 1)) = U(50W2, ... 'WN(N)){3~" Uz(TN ), (14)

where Uz(TN) is a commutative but non co-commutative Hopf sub algebra of the affine
CK deformed Hopf algebra UZ (iso W2 "",WN(N)) generated by {Pi, i = 1, ... , N}, and
U(50 W2 "",WN(N)) is the non-commutative and co-commutative non-deformed CK Hopf
algebra spanned by {Jij; i < j, i, j = 1, ... , N}, hence, with commutation relations
given by (1) and primitive coproduct (i. e., !l(Jij) = lQ9Jij+JijQ91). This bicrossproduct
structure implements to the Hopf algebra level the structure of semi direct product (3)
at the algebra level of i50(W2"",WN) (N).

Quantum deformed kinematical algebras

We can recover the quantum kinematical algebras for any dimension (N ~ 2) giving
the appropriate values to the parameters W2, ... ,WN. However, in order to obtain the
kinematical algebras written in a suitable physical basis, i.e., in terms of the generators
of time-translations (H), space-translations (P), rotations (Jij) and boots (K) we need
to make an identification between these generators and those of the original basis {Jij }.
This kinematical assignation is

H = JOl , P = (J02,J03, ... ,JON), K = (J12 ,J13 , ... ,JIN ), (15)

and the remaining Jij , i, j = 2,3, ... , N, are the generators of the spatial rotations (the
so(N -1) algebra). In particular, in the 4D, i.e., for a (3+ 1)-space-time manifold, we
have 10 generators: generators of time-translations (H), space translations (PI, P2, P3),
rotations (J1, J 2, J 3) and boots (K1' K 2,!(3) related with the original ones by
H = J 01 , P = (J02 ' J03 , J04 ), K = (J12 , J13 , J14 ), J = (J34 , -J24 , J23 ). (16)
This assignation allows to identify the algebras isO(-1,+1,+1)(4) and iso(0,+1,+1)(4) with
the (3 + 1) Poincare algebra and the (3 + 1) Galilei algebra, respectively. Note that the
contraction W2 -+ 0 represents the nonrelativistic limit c -+ 00, and that after a suitable
dimensionalization of the generators and parameters we get that W2 = -1/ c2
Specializing in the general expressions (8)-(13) of the coproduct, counit, antipode,
commmutators and Casimirs for the 4D case, i.e., for (UZ isO(W2,W3,W4) (4)), the values
of the parameters w;'s we get the deformed Hopf structures of the (3 + 1) Poincare
and (3 + 1) Galilei algebras. Since in our approach z is real and in both cases the
exponent zP3 ought to be dimensionless and the dimension of P3 is (length)-l then
[z] =length. Remember that [H] =(timet\ [K;J =(velocityt 1 =time/length and the
J;'s are dimensionless.
We present the explicit form of the coproduct and deformed Lie brackets for this
(3 + 1) q-Poincare characterized by (W2,W3,W4) = (-1/c 2, +1, +1)

!l(X) = 1 Q9 X + X Q9 1, X E {P3, K 1, K 2, ls},

!l(Pi ) = e-~P3 Q9 Pi + Pi Q9 e~P3, i = 0,1,2.
!l(K3) = e-F3 Q9 K3 + K3 Q9 d + ~K1e-F3 Q9 PI - PI Q9 ~e~P3K1
P3
+F<2e-~P3 Q9 P2 - P2 Q9 ~e~P3 K 2,
(17)
!l(J2) = e-~P3 Q9 J 2 + J 2 Q9 e~P3 + ~Kle-~P3 Q9 H - H Q9 ~eF3 Kl
-~he-F3 Q9 P2 + P2 Q9 ~eF3J3'
!l(J1) = e-~P3 Q9 J 1 + J 1 Q9 e~P3 - ~K2e-fP3 Q9 H +H Q9 ~eF3 Kz
-~J3e-~P3 Q9 PI + PI Q9 ~dP3h;

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

where the remaining components of the deformed Pauli-Lubanski vector are

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

Ci = 1> [sinh(~P3)l2 + Pi + Pi,

q = Wi 2 + wt + [cosh(zP3) + 7(P{ + pn] Wi,
(23)

where

Wi = -Pdl - P2J2 - ~ sinh( zP3)J3, Wi = -P2 I<3 + ~ sinh(zP3)I<2,

(24)
Wj = Pl I<3 - ; sinh(zP3)I<1, WI == W4 = -PlI<z + P2I<l.

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

In the "classical" (or undeformed) theory of Lie group representations Casimirs

play an important role: they label the irreducible representations according with the
Schur lemma. On the other hand, let us consider the action of a kinematical Lie
group G on the space-time manifold M, which will be a G-homogeneous space, i. e.,
M == GIGo, with Go the isotropy group of a point (the origin for instance) of M. The
representations up to a factor of G on the Hilbert space of functions on M can be
induced from those of Go and are, in general, reducible. However, the Casimir provides
a differential equation characterizing the irreducible subspaces. This equation is the
motion equation for the elementary physical systems involved. In this way appears, for
example, the Schrodinger equation in the Galilean case, and it is possible to introduce
in a group theoretical manner the minimal electromagnetic coupling in many cases [21].
In the study of the deformed kinematical algebras we have found the deformed
counterpart of the Casimirs and also q-differential realizations of the generators acting
on (commutative) spaces of functions. However, in this case things are very different
from the above mentioned nondeformed situation.
Let us start analyzing the 2D case. We have seen in Section 2 that there are six
cases, those with W2 ::; 0, whose motion algebras are kinematical (oscillating Newton-
Hooke (WI = l,w2 = 0), Galilei (0,0), exponential Newton-Hooke (-1,0), Anti de
Sitter (1, -1), Poincare (0, -1/c2 ), de Sitter (-1, -1)). All of them present quantum
deformations [10], although here we are only presented the corresponding ones to WI =
given by (8)-(13), and in all the cases the deformation parameter has dimensions of

length.
There exists a differential realization of the generators H == J01 , P == J02 and
J{ == J 12 acting on commutative functions of 2(JR2), given by

H = -at, P = -ax, J{ = W2 x at - t sinh(zox)1 z, (W2::; 0) (25)

for all the q-algebras (Uz 50(Wl,W2)i Wz ::; 0).

In Ref. [22] it was showed that the deformed algebra q- Poincare(1 + 1) describes
the symmetry of a harmonic chain, whose lattice spacing is given by the deformation
parameter z. Let us consider the equation

(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

it is obtained from (26) the equation

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

Pi = ai, i =1,2, ... ,N,

J/ m= WlmXma/ - x/am, (30)
JIN = W/NXNa/ - X, sinh(zaN )/z, I; m = 1,2; ... , N - 1,

from which the 2D realization (25) is the specialization for N = 2 and WI = O.

Let us consider the Hilbert space of commutative functions "II1(x) E C2(JR N ). The
above differential realization (30) acting on these functions originates a discretization
on the XN coordinate since the behaviour of sinh( ZaN) / z .is like a symmetric discrete
derivative

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

can be considered the continuous version of the equation

N-l
\ ("II1(x, XN+ z) + "II1(x, XN - z) - 2"111(x, XN + L wiNa;"II1(X,XN) = c "ll1(x, XN),
z
z ;=1
(33)
which Can be seen as a (N -1,1) dimensional Db-E

L wiNalilij(x) = czilij(x).
N~1

:2 (ilij+l(X) + ili j _ 1 (x) - 2ili j (x + (34)

;=1 .

We have all a family of Db-E's, that by contraction, i. e., making Wi = 0 we pass

from one to another. An open problem is to study their solutions, and to see if it is
possible to relate them by contraction.
Another particular and interesting case corresponds to the (3 + 1) q- Poincare. A

305
differential realization (30) becomes now in coordinates (t, xl, X 2 , X 3 ):

H = -at, Pi = -Oxi, i = 1,2,3;

J{l = _~XlOt_tox" J{2 = _~X2 at - t ox2,
(35)
J{3 = _~x3 at - t ~ sinh(zox3), Jl = x3Ox2 - ~ sinh(zox3 ),
x2
J2 = x3 ax, - xl ~ sinh(zox3), J3 = x 2 ax, - x l Ox2,

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.

QUANTUM EQUATIONS ON q-HOMOGENEOUS SPACES

Another way to study q-deformed equations is from the perspective of quantum

groups, i. e., deformed Hopf algebras of polynomical functions on a Lie group. From
the beginning of the quantum group theory the concept of quantum space [35] appeared
in a natural way, and it is very closed to the non-commutative geometry [36].
We shall start remembering how the idea of quantum plane appears. Let us consider
the quantum group SLq(2) and a generic element of it

~)
T=(aed' (40)

whose entries generate SLq(2) and verify the commutation relations

ab= qba, ae=qea, eb = be, (41)

bd = qdb, ed = qde, ad - da =(q - q-l)bc.

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 ,

b.(al) = cos(O) @ al - sin(O) @ a2 + al @ 1,

b.(a2) = sin(O) @al + cos(O) @a2 + a2 @ 1 j (45)
7( al) = - cos( 0) al - sin( 0) a2 , 7( -.i) = ei ,

7(a2) = sin(O) al - cos(O) a2,

and vanishing counit. Supposing that the deformation parameter z is real, an involution
is defined by: ai = al ,a~ = a2 ,0* = 0 .
This q-group is the dual of the deformed enveloping algebra Uz ( e(2)) generated by
the unity and the three elemeJ;lts PI ,P2 ,J, whose algebraic structure is the follow-
ing
[J,Pl ] = (i/z)sinh(zP2 ),
b.Pl = e- ZP2 @ PI + PI @ eZP2 , b.P2 = P2 @ 1 + 1 @ P2 ,
(46)
b.J = e- ZP2 @ J + J @ eZP2 j
7(J) = -J - i z Pd2 j

with vanishing counity and involution: J* = J, P; = PI, P; = P2.

The duality pairing between Ez(2) and Uz e(2) is given by

with T = -i e- ZP2 (J - i(z/4) Pt), VI = -i e- ZP2 PI and V2 = -i P2.

By means of the left action of an element X of the quantized enveloping algebra on
a quantized function J defined by [39]

1(X)1 = (id (() Y) 0.6.1 = L 1(1) (X, 1(2) ( 48)

(I)

we can obtain a quantum homogeneous space as the invariant subalgebra and left coideal

Ex = {J E Ez(2) Il(X)J = O} . (49)

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

>.(X)J = (!'(Y) @ id) 0 b.J = Lb(X), J(l) J(2) , X E Uz e(2) (51)

(I)

allows to defined a q-deformed Schrodinger equation, which can be considered in this

context as the counterpart of the free Schrodinger equation, by means of the canonical
action of the Casimir of Uz e(2) on the q-homogeneous space.
The Casimir of Uz e(2) is

(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)

where X = x/z, X = -x/z and (X)n = rr~;:~(X + k) is the Pochammer symbol,

diagonalize the Casimir

(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

M. A. O. thanks to the Organizers of the Int. Symposium "Symmetries in Science

IX", Profs. B. Gruber and M. Ramek for their invitation to participate in this Sym-
posium. This work has been partially supported by DGES (project PB95-0719) from
Spain.

REFERENCES

[1] 1. D. Faddeev, E. K. Sklyanin, Sov. Phys. Docl. 23 (1978), 902; L. D. Faddeev,

Sov. Sci. Rev. Math. Phys. C1 (1981), 107.
[2] G. Domokos, S. Kovesi-Domokos, J. Phys. G27 (1994), 7805.
[3] E. Inonii and E. P. Wigner, Proc. Natl. Acad. Sci. U. S. 39 (1953), 510.
[4] V. G. Drinfel'd, Sov. Math. Dokl. 27 (1983), 68; Quantum Groups, ICM Proceed-
ings Berkeley, 1986, 798.
[5] M. Jimbo, Lett. Math. Phys. 10 (1985), 63; 11 (1986),247.
[6] E. Celeghini, R. Giachetti, E. Sorace and M. Tarlini, in: Lecture Notes in Math.
1510, Springer, Berlin ,1992, 22l.
[7] V. K. Dobrev, J. Phys. A: Math. Gen., 26 (1993), 1317.
[8] J. Lukierski and A. H. Nowicky, Phys. Lett. B279 (1992), 299.
[9] H. Bacry and J. M. Levy-Leblond, J. Math. Phys. 9 (1968), 1605.
[10] A. Ballesteros, F. J. Herranz, M. A. del Olmo and M. Santander, J. Phys. A: Math.
Gen. 26 (1993), 5801; J. Phys. A: Math. Gen., 27 (1994), 1283.
[11] A. Ballesteros, F. J. Herranz, M. A. del Olmo and M. Santander, J. Math. Phys.
35 (1994), 4928.
[12] A. Ballesteros, F. J. Herranz, M. A. del Olmo and M. Santander, Lett. in Math.
Phys. 33 (1995), 273.
[13] M. Santander, F. J. Herranz and M. A. del Olmo, Kinematics and homogeneous
spaces for symmetrical contractions of in: Anales de Ffsica, Monograffas, 1.1,
CIEMAT /RSEF, Madrid, 1993, 455.
[14J M. de Montigny and J. Patera, J. Phys. A: Math. Gen. 24 (1991), 525; R. J.
Moody and J. Patera, J. Phys. A: Math. Gen. 24 (1991), 2227.

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

Institut fiir Theoretische Physik A

TU Clausthal
D-38678 Clausthal-Zellerfeld, Germany

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

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New York, 1997 313
 /1

Figure 1. Identification of the 2-dimensional torus.

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.

THE n-DIMENSIONAL TORUS Tn

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)

Taking the (trivial) complex line bundle over Rn and identifying

(Xl, ... ,Xn,Z)'" (Xl, ... ,xn,z)+(mlal, ... ,mnan,z) zEC (4)

we obtain a complex line bundle 17 over Tn 17 = (E,Rn/"',pr,C,(.,.)) (E total space,

pr : E --+ Rn /"', C fiber, (.,.) scalar product on C).
It will be convenient to view sections in the complex line bundle as functions on Rn
that are invariant under the periodicity conditions (1). Concerning the integrability
of the solutions, we will only have to assure local integrability on Rn in order to have
integrability on the torus.

Flat Connections and the Lifting of the Laplacian on Tn

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).

If \7 is not flat, the function h will be defined on the set of paths on R n. As h is a

mapping on the covering space, it has to satisfy the conditions

h( x + ajej) - h( x) = CXj ej = (0, ... , 1, ... 0), CXj E R, j = 1, ... ,n,

(6)
x = (Xl, ... ,Xn).
Furthermore, h becomes unique by the condition h(O) = O.
We will consider sections

CTp(x)=exp(ih(x)) X=(Xl, ... ,X n ), (7)

which are locally parallel with respect to the connection \7 on a suitable subset of Tn.
The relation for the classification of flat complex line bundles in (3) and the functions
h are discussed below.
To define a Laplacian on the complex line bundle 17 we furnish first Rn /'" with a
Riemannian metric 9 and get a Laplace Beltrami operator 6. 9 on Rn /'" defined by

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

~;,(To) = ( l:,g ~) . ao. (10)

If a connection V is given on T/, a natural choice of ao is given by 'Vao = O.

If we choose the trivial metric gij = Oij, we obtain the lifted Laplace labeled by
functions h that satisfy (6):

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

The domain of l:,hg is given by

dom(l:,~) = {a E L2(T/,djJ)lda absolutely continuous} (12)

(The absolute continuity of the da assures, that the second derivative exists almost
everywhere. )
On the I-[ilbert space 1i:= L2(T/,djJ) together with the usual inner product, l:,~ is
a symmetric operator whose closure is self-adjoint. In the following, we often use that
L2(T/, djJ) is isomorphic to Lfoc(Rn, djJ), the space of locally square integrable functions,
i. e. functions, that are square integrable on all compact subsets of Rn.

Equivalence of Quantization Mappings

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 .

Diagonalization of the Lifted Laplacian over Tn

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

which yields by separation

n

fj(xj) = yI-E(j)xj, E = .LE(j). (18)

j=l

From the restriction of Rn to Rn I~, the translational symmetry of Tn (see (4) we have

yI-.E(J") = 21Tmj - Qj mj E Z, j = 1, ... ,n. (19)

aj

Hence, the eigenvalue problem has, for m = (m1, . .. , m n) E zn. the solutions

spec (6";) = {Em = _ t

J=l
(21Tm~"- a
J
j )2, m = (m1, ... ,m n ) E zn, (20)
Qj fixed in [0, 21T) }

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.

THE KLEIN BOTTLE K2

Topology of K2

K2 is non-orient able and modeled in a two-dimensional plane R2 with the identifi-

cations
(a1 + XI,a2 - X2)
(22)
(Xl, a2 + X2) .
The fundamental domain of the Klein bottle is given in Figure 2. As R 2 is a universal
covering space for K2, the first fundamental group given as
(23)
is a non-Abelian group. 7rl(K2) is generated by the two elements '/'1 and '/'2. In order to
calculate its group of characters, we take a homorphism [ : 7rl (K2) -+ Sl and observe
that
[( '/'2)[( '/'1)[('!'2)[( '/'11)
= [('!'1)[('!'lt 1[('!'2)2 (24)
[(,!,d 2 .
As '/'2'/'1'/'2'/'11 is the neutral element e in 7rl(K2), it must be mapped to 1. Hence

l('!'l) E [0,27r)
(25)
[('!'2) E {1} ~ Z2

318
and the group of characters of K2 is

(26)

Flat Connections and the Lifting of the Laplacian on K2

71";(K2) classifies line bundles with flat connections. Elements in Z2 give rise to two
non-isomorphic line bundles TJ given by the identifications

TJ : (Xl, X2, z) (Xl + all az - X2, z)

(27)
(XI,XZ'Z) (Xl,X2+ aZ,Z) .

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.

Diagonalization of the Laplacian overK2

As indicated in Figure 2, the two-dimensional torus is a two-fold covering of the

Klein bottle. Hence solutions of the eigenvalue equation in L2( TJ, df.t) associated with
(29) will be solutions ofthe eigenvalue problem for TZ on L2(TJ(T 2), df.t) that satisfy the
additional symmetry of K2 given by (22).
With az = 0 we get from (21) with ml, mz E Z

(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)

The eigenvalues m = (mI, m2) E Z2 are

(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.

THE PROJECTIVE SPHERE RP2

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

(R3 X R3 \ D) /~ (D diagonal, ~ interchange of the particles)

R3 x (R3/~ ) (centre of mass x relative coordinates)
(34)
'" R 3 X R+ x (5 2 / ~ ) (radial and angular coordinates)

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)

Sections in TJ correspond to functions on R" with

a( XI, X2) a( al+ Xl, a2 - X2)

(38)
a( Xl, X2) a(al - Xl, a2 + X2)'

Diagonalization of the Laplacian on RP2

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)

The corresponding eigenvalues on TJ are, (ml' m2) E Z2,

(41)

RP2 as a Subset of 8 2

As mentioned at the beginning of this section, an alternative construction of RP2

is given by the identification of diametral points in S2. As the natural metric differs on

321
both constructions, the Laplacian is also different. The two complex line bundles T/
on Rp2 are given by the identifications

(x,z)~(-x,z) XE5 2 ,ZEC (42)

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

We have considered Laplacians on line bundles over different 2-dimensional mani-

folds. In order to compare the results, we rename the perimeters of the Klein bottle by
bj , its quantization parameter by (31 and the perimeters of the projective sphere by Cj.
The eigenvalues on the manifolds are then given by

M eigenvalues quantization parameter

T2 _ C7rm~l- a1 ) 2 _ C7rm~ - a2 ) 2 aj E [0,27r)

2

f{2 - ( 7rm~: (31 y _ C:m2y (31 E [0, 27r),

2

RP2 - (7r:1y - e~2y

2

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

S. Szpikowski, L. Pr6chniak, and W. Berej

Institute of Physics
M. Curie Sklodowska University
20-031 Lublin, Poland

INTRODUCTION

The idea of application of the supersymmetry to nuclei is based on the assumption

that in ground and low excited states of nuclei valence nucleons are mostly coupled to
the static pairs which with unpaired nucleons, one or two, form a system for the super-
symmetry application. Such a picture is, from the very beginning, an approximation
because pairs of nucleons can be only approximately treated as bosons.
The generators for the supersymmetry unitary transformations are composed of
the products of second order creation and annihilation operators (fermion and boson).
Hence, the supermultiplet is defined by the N = const, where N is the total number of
fermions and bosons.
We have already considered several supermultiplets of nuclei from the sd-shell [1].
The three or four parameter model has been able to describe about a hundred excited
states for a given supermultiplet consisting of even-even, odd-odd, and even-odd nuclei.
The B(E2) transition probabilities has been also treated within the same model [2].
The present application of the supersymmetry idea is the following. Suppose we
have firmly fixed the supersymmetric parameters for the well known nuclei with their
ground and excited energies and angular momenta. Suppose then that to the same su-
permultiplet belongs also an exotic hypothetical nucleus for which we want to calculate,
with already fixed supersymmetric parameters, the binding energy to check whether or
not the hypothetical nucleus exists.
The main technical problem to be solved is connected with the fact that excited
energies have been taken relative to the ground state energy, and hence, all of the
constant contributions to the energy for a given nucleus have been put into the Ho,

1 The work is partly supported by Polish Committee for Scientific Research (KBN) under Contract
No.2 P03B 19909.

Symmetries in Science IX, edited by Gruber and Ramek

Plenum Press, New Yark, 1997 325
where
H = Ho + Hsuper .
However, while comparing the binding energies of the different nuclei of a given super-
multiplet, the structure of theHo must be carefully treated. We will see the problem
in the following section.

THE SUPERMULTIPLET N =5
The supermultiplet which has been considered in [3] has the following members:

odd~odd nucleus itNa13 (2 fermions + 3 bosons)

even-odd nuclei i~Mg13 and i~Ah2 (1 fermion + 4 bosons)
even-even nuclei i~Mg14 and i~Si12 (5 bosons)

The constructed Hamiltonian consists of several second order Casimir invariants of

the relevant groups

H = Ho + AC2[SOB(5)] + BC2[SOf(3)] + aC2 [S00(3)] +

+ ;3C2[SO~F(3)] + ,C2[SU(2)] + 8C2[SU:'(2)] +
+ EC2 [SUr(2)] (1)

with the energy for excited states calculation

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.

Table 1. The Hamiltonian parameters for N = 5 supermultiplet (in MeV)

A B ,8 llao - ,

0.19 0.12 0.08 0.82

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

E~ = Eo + aN] + bNb + ET(T + 1), (3)

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:

Go =0 and hence 'Y = -0.82. (4)

Hence, we get

E = Eo + aNj + bNb + AT(7" + 3) + BL(L + 1)

+ (3J(J + 1) + 'YTj(Tj + 1) + T(T + 1). (5)
The new parameters which have to be adjusted to the binding energy calculation are
a, b, and .
Let us consider separately the members of the N = 5 supermultiplet:
~iNaI3: Nf = 2, Nb = 3,7" = L = 0, J = 4, Tj = T = 1
and hence
El = Eo + 2a + 3b + 2 - 0.04 (6)
Nj = 1, Nb = 4, 7" = L = 0, J = 5/2, Tj = T = 1/2
E3 = Eo + a + 4b + 0.75 + 0.08 (7)
26M
12 g14 ( or 14
26S' ).
112 . Nj = 0, Nb = 5, 7" = L = J= Tj = 0, T = 1
E5 = Eo + 5b + 2 . (8)
The binding energies of the mirror nuclei are, of course, different. But in our su-
persymmetric calculation we do not take into account the Coulomb part of binding
energies. Hence, the energies in (6, 7, '8) are understood as EB - Eo, where EB is the
total binding energy and Eo is the Coulomb contributions. The Coulomb energy is
calculated according to a standard shell model approach [4]. To adjust the supersym-
metric parameters we have to put in the (6-8) the experimental binding energies [5]
without the Coulomb contributions. Both are displayed in Table 2.

Table 2. Experimental binding energy and Coulomb contribution Eo

~608 ~iNal3 ~~MgI3 ~~AI12 ~~Mg14 ~~Si12

Eo El E2 E3 E4 Es
(EB)exp -127.62' -193.52 -205.59 -200.53 ~216.68 -206.05
Eo 18.29 30.26 35.10 40.18 35.10 45.77
Ei ==EB -Eo -145.91 -223.78 -240.69 -240.71 -251.78 -251.82

Let us notice, that for mirror nuclei the EB - Eo is practioally the same. We
precisely fix the values:

E2 = E3 = -240.70 E4 = Es = -251.80 (9)

Solving the equations (6-8) with the energies of Table 2 and (9) we get

a = -8.12 b = -22.14 = 2.41 (in MeV) (10)

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.

(EB - Eo)exp (EB - Eo )mean field (EB - EO)supersym

for 24 0 for 26 0 for 26 0

-187.17 -191.03 -184.31

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

flb == bN - 1- bN = 0.60 (MeV) (12)

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).

Table 3. The parameters band f for supermultiplets N = 2,3,4,5,6 (in MeV)

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)exp -217.75 -217.78 -233.36 -187.14 -233.09 -266.73 -282.31 -

(EB - Ec)th -217.61 -217.61 -232.07 -183.87 -234.15 -267.83 -282.35 -181.13

Even better reproduction of experimental binding energies we could get if we allow

for the. small changes in the parameter c, the strength of the isotopic term, which shows
the minimum value around the middle of the shell sd ~ Figure l.
To conclude, let us gather, with the slowly changing parameter c, the calculated
binding energies of oxygen isotopes A=24, 26 and 28 and compare them with experi-
mental energy for 24 0 and also with earlier theoretical predictions [5, 6, 7, 8, 9].

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.

Table 5. The predicted binding energies of oxygen isotopes A = 24, A = 26 and

A = 28 in different calculations versus the experimental energy for A = 24

0
24 26 0 280
Eexp -168.48 ? ?
Etheor (this work) -168 -166 -163
Etheor [5] - -168.43 -

Etheor [6] -170.46 -172.94 -177.40

Etheor [7] -165.39 -167.85 -166.01
Etheor [8,9] -168.48 -169.66 -168.88

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

Institut fiir Theoretische Physik (A)

TU Clausthal
D-38678 Clausthal-Zellerfeld, Germany

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)

Symmetries in Science IX. edited by Gruber and Ramek

Plenum Press. New York. 1997 331
where the quantity a can be, e. g., a number or a operator like N z. In particular, [NzJ
is given by a difference of shift operators qNz which act on functions f(z) as

(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.

QUANTUM KINEMATICS AND q-QUANTUM KINEMATICS ON Sl

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.

The Undeformed Case

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

[Q (f) 1/>] (</;) f( </;)1/>( </;)

(8)
[P(X)1/>](</;) [-iX + (-~i + D) (div X) + w(X)] 1/>(</;).
They are characterized by the quantum number D E R and by a closed one-form w
on 81, w = Bd</; with B E [0,1), i. e., by two numbers. Different values of D and

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

in which z := exp(irp), fn = !-n and Xn = X-n hold, then (8) is given by

n=-oo
(10)

n=-oo

with

zn (z~ + ~ + 0)
dz 2
(ll)

If 0 E R is a fixed constant we have

[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.

The Deformed Case

The derivation of a q-deformed quantum Borel kinematics is based on a q-deforma-

tion of the kinematical algebra 5(5 1 ), in terms of which (10) is given. In particular, we
are looking for operators In which are q-deformations of Ln, i. e., the zOz are "replaced"
by q-difference operators and, which satisfy an algebraic relation that in the limit q --+ 1
reduces to the Witt algebra. Since the process of q-deformation, even with the above
restriction, is not unique, we make the following two additional q-assumpiions:
(ql) The operators In should contain no additional derivatives.
(q2) The operators In should contain a minimal number of extra parameters.
The reason for q-assumption (q 1) is to prevent that the q-Schrodinger equation
derived later contains derivatives (or derivatives of higher order) than the undeformed

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

(f(),g()) = 2~ 121r /()g()d<P (16)

in L(SI, d) to a q-integral S inverting the q-derivative. We will discuss this aspect

in a forthcoming paper. Here we will concentrate on the algebraic structure, which is
independent on the function space, and furthermore on SJv. Here the Hilbert space is
finite dimensional and no integrals appear.
The following object satisfies the requirements (ql) and (q2):

r(j,B) '= zm [j (Nz + rg. + B)] (17)

J..- m . ~j]'

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)

otherwise. As required, these identities reduce to (13)-(15) for q = 1.

Another possible q-deformation would be [6]

r(j,r,) _ m [j (Nz + rg. )]'f [r N z ]

J..- m - Z [jh [r]' (21)

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)

which can be expressed as a commutator as follows:

[ (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

However, the projection part of the operator Pj(X) in Tn

(25)

formally contains a derivative (because of the factor n)

p(X)i
{Tn}
= i (Nz) f
n=-oo
X"DTn. (26)

Since we are aiming at a theory in which derivatives N z occur exclusively as q-derivatives

[NzJ, it is consequent to make the following q-assumption:

(q3) The integer n should be replaced by [nJ.

This leads to

pg(X)i
{Tn}
= i [NzJ f
n=-oo
Xn DT" =i f
n=-oo
Xn[nJDTn . (27)

The q-quantum Borel kinematics on Sl is then given by

00

L
n=-oo
iSn (= QU))
00
(28)
PJ(X)
n=-oo

DYNAMICS

The evolution equation on 8 1

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.)

H1f; + G[1f;,1f;]1f; (32)

fJ1f; + G[1f;,1f;]1f;,
where H is a linear operator, which will later be interpreted as Hamiltonian, and
G[1f;,1f;] == ReG[1f;,1f;] + iImG[1f;,1f;] a nonlinear function of ~6, 1f; and their derivatives
(possibly also on t and 8), one obtains the following nonlinear Schrodinger equation
(Doebner-Goldin equation) [8, 9]:

which contains a nonlinear term proportional to iD and an arbitrary nonlinear real part
which cannot be derived from (29) without further assumptions.

The q-Analogue of the Fokker Planck Type of Equation on Sir

Now we are interested, as mentioned above, in the construction of a dynamics on
the discrete configuration space Slv given by the N-th roots of unity, the operators (28)
have to be restricted to Slv, i. e.,
N-l

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

1f;(1) = ~ 1f;nl zr with zr = exp C~inl) (35)

and the scalar product is given by

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

D t (1/J( t, z), Q~(f)1/J(t, z) d = ( 1/J(t, z), P/,d(Xgrad q f )1/J(t,z) ) d ' (38)

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

atp = r;J{ ([Nzl [j(Nz2-0)]~) (q-<I NZ+<2j INZ/")1j;)

V
- ([NzJ Nz2+ 0)] 1j; ) ( q'INz+<2j(Nz2-e) ~)
_ ([NzJ [j~] q<2j(NZ2_e)~) (q-<3~Nz-'INz1j;)
- ([Nzl [hi8] q(2).~)
1/J (q42
2 iN z+'I"" -1/J)
Z
(40)

_ ([Nzlq<2jINZ2-e)~) ([j~z] q-<iNZ1j;) q<3j~

+ ([NzJq'd~1/J) ([j~z] q'INz~) qqj~ }

~ D ([NzJ2~1j;) ,
where qaNz acts as a shift-operator on functions fez), i. e.,

(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

(q5) 103 = -104 = t2

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,

iotw0 = (Hg1fJ)(50) + (G~[0,1/J]1/J)(R0) (44)

-iot 0w = (HN)(Sw)+(G~[0,w]0)(RW).

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)}.

Consequences of the q-Fokker Planck Equation - Linear Part

The defining relation for the linear part is thus given by

2 {(H1W1)(52W1 - 5 1 Wz)+ (HZW2)(51W2 - 5 Wl) Z

+ (H2W1)(51W1 + 521/J2) + (H1W2)(51WI + 5 2 W2)} (46)

(V1/J2)(WWI) - (V1/Jt}(W1/J2),
where the operators

(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

2 {(VI)(S21 - SI2) + (V2)(SI1 + S22)}

= (V2)(WI) - (VI)(W2) (49)

from which one can conclude that

2(SI1 + S22) (50)

2(S21 ~ SI2). (51)

Hence,

W
(52)
2
o. (53)

With this, the linear part of the Schrodinger equation is given by

(Ht) (S1/!) ((HI + iH2)) ((SI + iS2)1/!)

= (V) (1f1/!) . (54)

Case b: HI = 0, H2 = V
With

2{(V2)(SI2 - S2t) + (VI)(SII + S22)}

(V2)(WI) - (Vt)(W2) (55)

one has

2(SI2 - S21) (56)

2(SI1 + S22) , (57)

which implies

o (58)
W
(59)
2

It leads to the same Schrodinger equation (54) as case a.

Case c: HI = V, H2 = V
Here, the defining relation is

2{ (VI)(S21 - SI2 + SII + S22)

+ (V2)(SI2 - S21 + SII + S22)} (60)
= (V2)(Wd - (V1)(W2)

which leads to
(61)

339
With

((V + iV)'I/J) (!f + i!f))

(V'I/J) (1f) (62)

one again obtains (54).

Hence one obtains in all cases the linear part (j E 2Z, 1':1, 1':2 = 1)

(Ht'I/J)(S) '= ([Nzl [j~Z ]'I/J ) ~(qflNz + q-f1Nz)qf2j~, (63)

which leads for all j in the limit q ~ 1 to the Hamiltonian - ~ d~2 obtained in the
undefortned case. 2

Consequences of the q-Fokker Planck Equation - Nonlinear Part

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

- :iI':2~h('l/JII' - "'I/J') + O(h2)

Dp" + O(h)
{(c + d)('I/J~ + 'l/JD + h {(er + d6)('l/Jl'I/J~ + 'l/J2'I/J;) (65)
+(aa + b(8)(?,b2?,b~ - .,pll/J~)} + O(h2)}
{(a + b)(l/Ji + 'l/Ji) + h {(aa + b(3)(l/Jl.,p~ + 'l/J2'I/J~)
+(er + d6)(1/J1l/J~ - 'l/J2'I/JD} + O(h2)}
With G1 and G 2 from (64) and (65) the nonlinear term in the q-Schrodinger equation
is given by
(66)
In particular, (65) shows that the behaviour of the defining relation for G1 and G2 in
the limit q ~ 1 depends on the coefficients a, b, c, d, a, (3, , and 6 which define the
shift R.
'JLhe following cases have to be treated independently:

Case a: R = Rl and aa + b(3 =f. 0

Since this case implies c = d = 0, the defining relations for Bl and B2 are given by
(compare with (65))

Bl {h( aa + b(3)( 'l/J21/J~ -l/Jll/J;) + O( h2)}

B2 = {(a+b)(l/Ji+'l/Ji)+h{(aa+b(3)(l/Jl'I/J~+'l/J2l/J;) (67)
+(er + d6)(1f!1'I/J; - 'l/J2l/JD} + O(h2)} .

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

-C2j (1/;//;{;' - ;{;//1/;')

(70)
8( aa + b(3) (1/;1/;' - 1/;1/;')
Dp//
G2 = (71)
2p(a + b)
Since one has R ---'> (a + b) in the limit q ---'> 1, the nonlinearity is given by
((GI + iG 2 )1/;)(Rl + iR 2 );{; (72)
-c2j(a + b) (1/;//;{;' - ;{;//1/;') .Dp//
(73)
8(aa + b(3) (1/;1/;' -1/;1/;') + z2P
In particular, the imaginary part coincides with the one which was found without q-
deformation. The real part depends on the choice of the parameters a, b, a and (3. It
is interesting that in the special case R = 5, i. e., if the shifts for the linearity 5 and
for the nonlinearity R in (44) coincide, the real part is independent on the coefficients.
In particular, one has (aa + b(3) = c2j (a + b) and with this the real part
(;{;//1/;' -1/;//;{;')
(74)
8(1/;1/;' -1/;1/;')

Case b: R = Rl and aa + b(3 = 0

It is clear that either a =f. 0 or b =f. O. Otherwise one would have R = which

would correspond to a linear ansatz for the evolution equation in (44). One can thus
assume w.l.o.g. that a =f. 0, which implies that a = and hence R = a is a trivial shift
proportional to a. With this Bl = 0 and the identity (64) reduces to

(75)

In particular, the real part is no longer restricted by it and remains undetermined as in

the un deformed approach. The imaginary part, however, can be calculated from (75)
and gives in the limit q ---'> 1 the usual result.
This case contains also the situation, where no shift is present, i. e., a = 1. The above
discussion shows that the absence of a shift yields to a loss of information. However,
for q-deformed models, this is not generic.

Case c: R = R2 and c, + do =f. 0

This case is similar to case a. Here one has

(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

((GI + iG 2 )1/;)(R I + iR 2 ){;

-c2j(c+d) (1/;//{;'-{;"1/;') .Dp"
8(q + dEi) (1/;1/;' -1/;1/;') + z2"P (80)

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.

Case d: R = R2 and q + dEi = 0

This case is similar to case b. With c I' 0 and "( = 0 one has

(81 )

i. e., G 2 remains undetermined. However, since R2 -+ ic for q -+ 1, one again obtains

the usual imaginary nonlinear functional in the limit q -+ 1. The lack of a non-trivial
shift leads again to the fact that the real part of the nonlinearity remains undetermined.

Case e: R I , R2 I' 0
In this case both BI and B2 are of order 0(1) and one has

(c + d)p + O(h) (82)

(a+b)p+O(h). (83)

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.

With this, the nonlinear part (for () =I 0) is given by:

(iot'IjJ) '0 = ([Nzl [j~z] 'IjJ) (51'0)-

i [j~] {q-2~B (i[NzJq-2~Nz'IjJ) + q2~B (i[Nzlq'2~NZ~)) q-2jNz} (51'0) (93)

+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

q-quantum mechanics on 51 has been obtained via a q-deformaljon of the underlying

kinematical structure. In particular, the process of discretization 5j. of 51 was linked
to a q-deformation of the kinematical algebra S (5j. ).
The limit q --4 1 of the derived q-Schrodinger equation is of independent interest: it
yields the Doebner Goldin equation, which was also derived in the undeformed theory,
but with a fixed form for the real part. It is an interesting feature that the real part
depends on the shift R in the ansatz for. the q-Schrodinger equation.
A generalization of the formalism to other configuration spaces will be the issue of
further investigations.

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

[1] T. Kobayashi and T. Suzuki, Phys. Lett. B317 (1993), 359.

[2] M. Gerstenhaber, Ann. Math. 79 (1964), 59.
[3] V. K. Dobrev, H.-D. Doebner, and C. Mrugalla, J. Phys. A: Math. Gen. 29 (1996),
5909.
[4] B. Angermann, H.-D. Doebner, and J. Tolar, in: Lecture Notes in Mathematics
1037, Springer, Berlih, 1983.
[5] H.-D. Doebner and J. Tolar, Ann. Phys. Leipzig 47 (1990), 116.
[6] R. Kemmoku and S. Saito, J. Phys. A: Math. Gen. 29 (1996),4141.
[7] H.-D. Doebner and J. D. Hennig, in: Symmetry in Science VIII, B. Gruber (ed.),
Plenum, New York, 1995, 85.
[8] H.-D. Doebner and G. A. Goldin, Phys. Lett. A162 (1992), 397.
[9] H.-D. Doebner and G. A. Goldin, J. Phys. A: Math. Gen. 27 (1994), 1771.
[10] V. K. Dobrev, H.-D. Doebner, and R. Twarock, A discrete, nonlinear q-Schrodinger
equation via Borel quantization and q-deformation of the Witt algebra. AS 1'- TPA
19/96, INRNE-TH 12/96, submitted to Phys. Rev A.

344
LIST OF PARTICIPANTS

Engin Arik v. K. Dobrev

Department of Physics Institute of Nuclear Research and
Bogazi<;i University Nuclear Energy
TR~80815 istanbul Bulgarian Academy of Sciences
Turkey 72 Tsarigradsko Chaussee
BG~ 1784 Sofia
Bulgaria
Metin Arik
Department of Physics
Bogazi<;i University
H.-D. Doebner
TR~80815 istanbul
Institut fiir Theoretische Physik
Turkey
Technische Universitiit Clausthal
LeibnizstraBe 10
R. Bijker D~38678 Clausthal- Zellerfeld
Instituto de Ciencias Nucleares Germany
U.N.A.M., A.P. 70-543
04510 Mexico, D.F.
Mexico Ismail H. Duru
Mathematics Department
Ernst Binz University of Trakya
Fakultiit fiir Mathematik und Infor- Edirne
matik Turkey
Universitiit Mannheim and
D~68131 Mannheim Physics Department
Germany TUBITAK, Marmara Research Cen-
tre
P.K. 21
Matthias Brunner
TR~41470 Gebze
Universitiit Konstanz
Turkey
EggerstraBe 2
D~ 78464 Konstanz
Germany
Huseyin Erbil
Fisik Bolumu
J6szef Cseh Fen Fakiiltesi
Institute of Nuclear Research Ege Universitesi
Hungarian Academy of Sciences TR~35100 Bornova-Izmir
H~4001 Debrecen POB 51
Turkey
Hungary

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

John R. Klauder Hans-Jiirgen Mann

Departments of Physics and Mathe-
matics
University of Florida
Gainesville, FL 32611
USA
Institut fiir Theoretische Physik
Technische Universitiit Clausthal
LeibnizstraBe 10
D-38678 Clausthal-Zellerfeld
Germany

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
Arnold Sommerfeld Institut
Technische Universitiit Clausthal
Leibnizstrafie 10
D-38678 Clausthal-Zellerfeld
Germany
Christoph Schulte
Institut fur Theoretische Physik
Technische Universitiit Clausthal
Leibnizstrafie 10
D-38678 Clausthal-Zellerfeld
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

Abelianization, 180-185 Bose condensation, 281

Algebra Bose type excitation, 281-293
Cayley-Dickson, 153-155 Boson realization, 126, 133
Cayley-Klein, 298-301 Boson-fermion realizations, 117, 129
Euclidian, 298, 302-303 Bosonic quantization, 21
Galilei, 298, 302-303 Bosons, 14
Grassmann, 120-122 Bragg spectrum, 139, 147
Heeke, 57 Broken translational symmetry, 179
Kac-Moody, 48, 118 Bruhat order, see Coxeter group
Lie, 84, 281-293,334 Bundles
Euclidian, 253 fiber 153, 179-190
Galilean, 253 line, 313-323
inhomogeneous, 298
q-deformation, 1 CH 4, 107-113
semisimple,47-69, 117,298-300 Cartan-Weyl basis, 48, 119
simple, 48-69 Cartan-Weyl generators, 51
Poincare, 298,302-307 Cayley graphs, 179-191
Schrodinger, 251-253, 332 Cayley-Dickson algebra, 153-155
W-,48 Cayley-Klein algebra, 298 301
Weyl-Lie Chevalley basis, 52
Euclidian, 253 Chevalley generators, 51
enveloping, 298, 308 Clebsch-Gordan coefficients, 103
universal, 119 Clifford matrices, 155
kinematical, 269, 298, 302 Coherent state, 118,124, 168-174
quantum, 48 path integral, 169-177
Alphabet, 140-145 Commutative group, 183
Angular momentum, 14 Constraint Lie algebra, 160
Anti-involution, 156 Coon-Yu-Baker oscillator, 6
Aperiodic structures, 140 Couplings, Darling-Dennison type, 109
Bargmann representation, 126 Coxeter graph, 212-214
Baryon Coxeter group, 141-144,212-214,263-268
resonances, 17-19 Bruhat order, 57
structure, 193-209
Basis Danzer tiling, 144-148
Cartan-Weyl, 48 Darling-Dennison type couplings, 109
Chevalley, 52 De-Sitter universe, 4
Be3, 106 Decomposition, Robinson, 141-142
Be4, 102-105 Diophantine equations, 162-163
Bernstein-Gel'fand-Gel'fand example, 51-77 Dirac oscillator, 243-249
Bialynicki-Birula-Mycielski equations, Doebner-Goldin equations, 83-91,254,
275-276 270~276, 344
Bogoliubov transformation, 282 Drinfel'd-Jimbo group, 47-48
Borel kinematics, 333-344 Drinfel'd-Jimbo scheme, 300
Borel quantization, 83-91, 313, 331-344 Dynamic supersymmetry, 44
Borel subgroup, 69-70 Dynamic symmetry, 37-44

351
Dynamics
Riemannian, 86
quantum, 86
Dyson realizations, 117
Ehrenfest relation, 87-88. 313
Einstein equation, 25
Enveloping algebra, 298, 308
Equation of motion, 281-295
Euclidian Lie algebra, 253
Euclidian Weyl- Lie algebra, 253
Euclidian algebra, 298, 302-303
Euclidian group, 300
Exotic nuclei, 325-330
Factor group, 182
Factor-module, 48
Faddeev-Popov determinant, 169
Fermi type excitation, 281-293
Fermion
Lie operators. 285-294
system, 281-293
Fiber hllndles, 153, 179-190
Fibonacci sequences, 140
Fokker-Planck equation, 336-343
Form factor, 18-19
of nucleons, 195-1\)\)
Free group, 182-185
Giirsey Radicati form, 195
Galilean Lie algebra, 253
Galilei algebra, 298, 302-303
Galilei group, 84, 300
Gauge transformation, 270-275
Generators
Cartan-Weyl,51
Chevalley, 51
Gleason's theorem, 273
Golden section, 140
Grassmann algebra, 120-122
Grassmann envelope, 121-122
Grassmann variable, 169
Greens function, 94-96, 234-238
Hi, 106
Hadron, 203, 205
Hamiltonian functional, 83
Hamiltonian operator, time-dependent, 170
Harmonic oscillator, 38,93,100-104,165,
226-240, 254
quark model, 206
Hartree-Bogoliubov equation, 281-293
Hecke algebra. 57
Heisenberg ferromagnet, 305
Heisenberg-Weyl superalgebra, 126
Helicity amplitudes, 200-203
Hodge star operator, 32
Hopf fibration, 153-159
Hopf structure, 300-302
Hurwitz theorem, 162
352
Hurwitz transformations, 156-164
Icosahedral symmetry, 144
Inhomogeneous Lie algebra, 298
Integral linear group, 185
Jacobi vector, 194
Kac-Moody algebra, 48, 118
Kazhdan-Lustig polynomials, 47-79
Kepler problem, 164
Kepler-like potential, 94-96
Killing vectorfields, 84
Kinematical algebra, 269, 298, 302
Kinematical groups, 298
Klein bottle, 31~:-323
Klein-Gordon equation, 25
Kustaanheimo-Stiefel transformation, 164
Langrange multipliers, 3, 167-175,284-292
Levi-Civita transformations, 157
Lie algebra, 84, 251-260, 281-293, 3:14
constraint, 160
inhomogeneous, 298
q-deformation, 1
semisimple, 47-69, 298-300
simple, 48 69
symplectic, 160
Lie bialgebra, 300-301
Lie group, 181
Lie operators, fermion, 285-294
Lie superalgebra 117-,118, 126
Line bundle, 31::-323
Lorentz group, 82, 219
Lorentz manifold, 27
Lorentz metric, 25-28
Lorentz representation, 71-76
Lorentz transformation, 215-222
Lowest weight module 49-69
Magnon,306
Maxwell equation, 75, 218
Maxwell field, 75
Methane, 107-113
Minkowski lattice, 219
Minkowski space-time, 71,218
Molecular dynamical group, 107
Molecular spectroscopy, 99-114
Morse oscillator, 100-104
Morse potential, 93
Motion group, ~m-272
Nat, 106
Non-commutative space groups, 184-190
Nonbijective transformations, 153-165
Nuclear states. 41
Nucleon
resonances, 19
form factor, 195-199
0(3), 41
Oblate symmetric top, 13
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
U(7), 9, 11, 209
Universal enveloping algebra, 119
Usui operator, 126-132
Vector coherent state theory, 118
Verma module, 47-68,117,119,254-260
Vibrational spectra, 99-114
354
Vibron model, 9, 39, 40, 100
Von Neumann equation, 82, 90
W-algebra,48
Weyl equations, 75
Wigner function, 226-233
Wigner's spin-isospin group, 39
WigIier-Eckhart theorem, 248