Spinors, Twistors, Clifford Algebras and Quantum Deformations

Fundamental Theories of Physics

An International Book Series on The Fundamental Theories of Physics:

Their Clarification, Development and Application

Editor: ALWYN V AN DER MERWE

University of Denver, USA

Editorial Advisory Board:

ASIM BARUT, University of Colorado , US.A.
BRIAN D, JOSEPHSON, University of Cambridge, UK.
CLIVE KILMISTER, University of London, UK.
GUNTER LUDWIG, Philipps-Universittit, Marburg, Germany
NATHAN ROSEN, Israel Institute of Technology, Israel
MENDEL SACHS, State University of New York at Buffalo, US.A.
ABDUS SALAM, International Centre for Theoretical Physics, Trieste, Italy
HANS-JURGEN TREDER, Zentralinstitutfur Astrophysik der Akademie der
Wissenschaften, Germany

Volume 52
Spinors, Twistors,
Clifford Algebras
and Quantum Deformations
Proceedings of the Second Max Born Symposium
held near Wroclaw, Poland, September 1992

edited by

Zbigniew Oziewicz
Bernard Jancewicz
and
Andrzej Borowiec
Institute of Theoretical Physics,
University ofWroclaw,
Wroclaw, Poland

SPRINGER SCIENCE+BUSINESS MEDIA,B.V.

Library of Congress Cataloging-in-Publication Data
Sp,nars, tWlstars, Cllffard algebras, and quantum defarmatlans
edlted by Zblgnlew OZleW1CZ, Bernard JanceW1CZ, and AndrzeJ
BaraWlec.
p. cm. -- (Fundamental thearles of phYS1CS v. 52)
ISBN 978-94-010-4753-1 ISBN 978-94-011-1719-7 (eBook)
DOI 10.1007/978-94-011-1719-7
1. Splnor analysls--Congresses. 2. TWlstar theary--Congresses.
3. Cllffard algebras--Congresses .. 4. Mathematlcal phYS1CS-
-Cangresses. 5. Rzewuskl, Jan, 1916- I. OZlewlcz, Zblgnlew.
II. Jancewlcz. Bernard. III. Barawlec, AndrzeJ, 1950-
IV. Ser les.
aC20.7.S65S67 1993
530.1'5257--dc20 93-7413

ISBN 978-94-010-4753-1

Printed on acid-free paper

AH Rights Reserved
© 1993 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1993
Softcover reprint ofthe hardcover Ist edition 1993
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner.
This volume is dedicated to Professor Jan Rzewuski) pioneer and teacher)
on the occasion of his 75-th birthday
 TABLE OF CONTENTS

Foreword
Zbigniew Oziewicz xiii

Photograph of Professor Jan Rzewuski XIX

Homage to Professor Jan Rzewuski

Jan Lopuszariski xxi

Homage to Professor Jan Rzewuski

Jan Mozrzymas xxiii

Photograph of participants xxiv

SPINORS
Structure of matrix manifolds and a particle model
Jan Rzewuski 3
Complex stuctures and the Elie Cartan approach to the theory of spinors
Michel Dubois- Violette 17

Spin structures on hypersurfaces

and the spectrum of the Dirac operator on spheres
A ndrzej Trautman 25
Algebraic construction of spin structures on homogeneous spaces
Vladimir Lyakhovsky 31
The Kummer configuration and the geometry of Majorana spinors
Gary W. Gibbons 39
viii TABLE OF CONTENTS

Pauli-Kofink identities and pure spinors

Helmut Urbantke 53
General covariance and spinors
Ludwik Dgbrowski 61

Tensored division algebras: origin of geometry, spinors and symmetry

Geoffrey Dixon 67

G-Structure for hypermanifolds

Andrzej Borowiec 75

Towards a unification of "everything" with gravity

Jerzy Rayski 81

Generalized Fierz identitites and

the superselection rule for geometric multispinors
William M. Pezzaglia, jr. 91
Electrons, photons and spinors in the Pauli algebra
William Baylis 97

TWISTORS
Twistors and supersymmetry
Dmitrij Volkov 109

A twistor-like description of D=10 superstrings

and D=l1 supermembranes
Igor Bandos and Aleksandr Zheltukhin 121
Born's reciprocity in the conformal domain
Arkadiusz Jadczyk 129
Self-dual Einstein supermanifolds and supertwistor theory
Sergey Merkulov 141
An approach to the construction of coherent states for massless particles
Andrzej Karpio 147
TABLE OF CONTENTS ix

CLIFFORD ALGEBRAS
What is a bivector ?
Pertti Lounesto 153

Monogenic forms on manifolds

Vladimir Soucek 159

On invertibility of the Clifford algebra elements with disjoint supports

Pavel Semenov 167

Clifford algebras and algebraic structure of fundamental fermions

Wojciech Krolikowski 171
Clifford algebra of two-forms, conformal structures and field equations
Ingemar Bengtsson 183
Dirac form of Maxwell equation; ~n-graded algebras
Jaime Keller 189

Travelling waves within the Clifford algebra

Bernard Jancewicz 197

Hamiltonian mechanics with geometric calculus

David Hestenes 203

Grassmann mechanics, multivector derivatives and geometric algebra

Chris Doran, Anthony Lasenby and Steve Gull 215

Intrinsic non-invariant forms of Dirac equation

Josep Manel Parra 227

2-spinors, twistors and supersymmetry in the space-time algebra

Anthony Laseby, Chris Doran and Steve Gull 233
x TABLE OF CONTENTS

QUANTUM DEFORMATIONS
Quantized Minkowski space
Julius Wess, Bruno Zumino, Oleg Ogievetsky and W.B. Schmidke 249

D = 4 Quantum Poincare algebra and finite difference time deri'Vative

Jerzy Lukierski, Anatol Nowicki and Henri Ruegg 257
Quantum Lorentz group and q-deformed Clifford algebra
Ursula Carow- Watamura 267
Isotropic q-Lorentz group
Jakub Rembielinski 277
Lorentz algebra and twists
Ralf Engeldinger, Michael Schlieker and Wolfgang Weich 281

On a noncommutative extension of electrodynamics

John Madore 285
Bicovariant differential calculus and q-deformation of gauge theory
Satoshi Watamum 299

Cyclic paragrassmann representations for covariant quantum algebras

Alexey Isaev 309
Hecke symmetries and braided Lie algebras
Dmitri Gurevich 317
Anyonic quantum groups
Shahn Majid 327
On S-Lie-Cartan pairs
Wladyslaw Marcinek 337
New real forms of Uq(G)
kn&k~ ~
TABLE OF CONTENTS xi

RELATED TOPICS
~3-graded structures
Richard Kerner 349
On Jordan block form
Garret Sobczyk 357
Unified theory of spin and angular momentum
Leopold Halpern 365
Noetherian symmetries in particle mechanics and classical field theory
Dan Grigore 371
Lienard- Wiechert Yang- Mills fields
I<. Paul Tod 379
The twist prescription in the topological Yang-Mills theory
Sorin Marculescu 391
On symmetry properties of classical Lagrange functions
under rotations
Peter Stichel and Jan Lopuszmiski 403
Tunnelling of neutral particle with spin 1/2 through magnetic field
Mijat Mijatovic, G. Ivanovski, B. Veljanoski and G. Apostolovska 413

List of participants 421

FOREWORD
ZBIGNIEW OZIEWICZ
University of Wroclaw, Poland

December 1992

The First Max Born Symposium in Theoretical and Mathematical Phy-

sics, organized by the University of Wrodaw, was held in September 1991
with the intent that it would become an annual event. It is the outgrowth
of the annual Seminars organized jointly since 1972 with the University of
Leipzig. The name of the Symposia was proposed by Professor Jan Lopu-
szanski. Max Born, an outstanding German theoretical physicist, was born
in 1883 in Breslau (the German name of Wrodaw) and educated here.
The Second Max Born Symposium was held during the four days 24-
27 September 1992 in an old Sobotka Castle 30 km west of Wrodaw. The
Sobotka Castle was built in the eleventh century. The dates engraved on
the walls of the Castle are 1024, 1140, and at the last rebuilding, 1885. The
castle served as a cloister until the end of the sixteenth century.
The Second Max Born Symposium is dedicated to Professor Jan Rze-
wuski. Professor Rzewuski was born in 1916, earning his doctoral degree
in 1948 and his habilitation in 1950 at the University of Warsaw. He was
a professor at Copernicus University in Torun* until 1952 when he was
forced by the communist regime to move to the University of Wrodaw**. In
Wrodaw, Professor Rzewuski founded the Institute of Theoretical Physics
and became its first director. He was forced to resign from the positions
of Dean of Faculty and Director of the Institute after March 1968 when
he supported the demands of protesting students. During and after Martial
Law in Poland, over the years 1982-1988, Professor Rzewuski supported the
fight for independence. ***
In 1958 Rzewuski started to propagate the idea that the spinor space is
more fundamental than Minkowski space-time and that classical and quan-
tum field theories needed to be formulated in complex spinor space rather
than in real Minkowski space. Later, Professor Rzewuski independently dis-
covered the notion of Penrose's twistor and the Penrose transform. In the
* The Stefan Batory University in Vilna, founded in 1570, was relocated in TorUli in
1945.
** Jan Kazimierz University in Lwow was moved in 1945 to Wroclaw.
H* When I was in prison in 1982, and again in 1984, Professor Rzewuski bravely worked

for my release on his bail. To give bail required an extraordinary amount of courage and
a good heart.

Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, xiii-xviii.
© 1993 Kluwer Academic Publishers.
xiv ZBIGNIEW OZIEWICZ

1980's, Professor Rzewuski investigated the submanifolds of Hom (<En ,<Em) of

fixed rank. This approach, presented in Professor Rzewuski's article in the
first Chapter of this volume, generalizes the Penrose transform to arbitrary
dimension.
The subject of the Second Max Born Symposium Spinors, Twistors and
Clifford Algebras reflects the domains in which Professor Rzewuski's contri-
butions are notable. The subject has been extended by Quantum Deforma-
tions.
The lectures in mathematics and theoretical physics attracted 65 partici-
pants from 20 countries. During the four days there were 54 lectures with 38
in parallel evening sessions. The lectures at the Symposium were grouped
according to subject. In this volume we have again grouped together related
contributions.
The Symposium was opened by a welcoming speech and two speeches
in honor of Professor Rzewuski. The welcoming speech, not included in this
volume, was delivered by the Director of the Institute of Theoretical Physics,
Professor Jerzy Lukierski.
The first plenary lecture was given by Michel Dubois-Violette. He
explored the identification of the Clifford algebra C£2£ of even dimensional
euclidean space with the algebra of the fermionic anticommutation relations
by means of an isometric complex structure. We learn that neither Cartan's
terminology (simple spinors) nor Chevalley's (pure spinors) are appropriate
in this case because these spinors are Fock states.
The first Chapter on SPINORS contains also papers on spin structures,
one by Andrzej Trautman who unfortunately was unable to participate
because of the tragic death of his son, and the other by Vladimir Lyakhov-
sky who also delivered a second lecture about multiparametric deformations
of quantum groups (i.e. Hopf algebras). This second lecture by Lyakhovsky
is published elsewhere. A paper related to spin structures is presented also
by Ludwik D<}browski.
Gary Gibbons describes the geometry of the Majorana spinors in terms
of the real projective space.
The second chapter, devoted to TWISTORS, starts with the plenary lec-
ture by Dmitri Volkov, one of the pioneers of graded symmetries. Supersy-
mmetry (LZ2-graded Lie algebras) was introduced by Volkov and Akulov in
1973 and independently by Julius Wess and Bruno Zumino in 1974. I should
add that professor Jan Lopuszanski "almost" discovered supersymmetry in
1971 and made significant contributions to this domain later on. The or-
ganizers were proud to have gathered together at the Sobotka castle the
pioneers of supersymmetry, professors Dmitri Volkov, Julius Wess and
Jan Lopuszanski. Dmitri Volkov showed in his lecture how twistors are
related to supersymmetry. Volkov's colaborator, Aleksandr Zheltukhin
describes superstrings and supermembranes in terms of twistors.
FOREWORD xv

Twistor space is a U2,2-module, a four dimensional <V-space with a hermi-

tian form and with the automorphism group U2 ,2. The sequence of nested
subspaces of the twistor space is called a flag. An G L-orbit of a flag is called a
twistor flag space. Anatol Odzijewicz from Bialystok, Poland, has since 1979
been considering twistor flag spaces as phase spaces. Arkadiusz Jadczyk
in a plenary lecture identified the reltitivistic conformal phase space with the
symmetric homogeneous space of the automorphism group U2 ,2 of twistors.
We learn that such a homogeneous space carries a quantum geometry with
inclusion of Planck's constant, which to me is mysterious. Jadczyk's lecture
also contains a discussion of Max Born's scientific works in historical per-
spective. He showed transparences with citations from Max Born's papers
related to Born's reciprocity principle (q ~ -p,p ~ q invariance). Born's
principle is presented by Jadczyk in a new light.
We invited Professor Albert Crumeyrolle from Toulouse, France,
as the key speaker to the session on CLIFFORD ALGEBRAS. With deep
sadness we were forced to announce the sudden death of our friend and
colleague. Professor Crumeyrolle was one of the major contributors to the
theory of Clifford algebras and spinor structures, including oo-dimensional
symplectic Clifford algebras and symplectic spinors, invented by him in 1975.
Professor Crumeyrolle was born in 1919 and died in June 1992.
Vladimir Soucek presented his recent monograph on Clifford analysis
written jointly with Delanghe and Sommen.
Professor Wojciech Krolikowski from Warsaw was unable to partici-
pate in person; however, he was kind enough to send us his lecture in which
he explains how a sequence of Clifford algebras leads to the existence of three
and only three families of fundamental fermions with a mass spectrum for
charged leptons.
David Hestenes and Garret Sobczyk are the authors of the mono-
graph, Clifford Algebra to Geometric Calculus, published by Kluwer in 1984*
soon after Garret Sobczyk was expelled from Poland**. This monograph is
a polemic assault against the Cartan's calculus of differential forms.
David Hestenes in a plenary lecture Reconcilling Cliffor'd and Grass-
mann, stressed again that "the modern calculus of differential forms is a
step backward" in comparison with geometric calculus, i.e. "calculus with
the structure of Clifford algebra". The differential forms were known since
Pfaff (1815). The calculus of differential forms was completly developed
by Ellie Cartan around 1900 and was based on Grassmann's exterior al-
gebra invented in 1844. Hestenes's and Sobczyk's geometric calculus is more

* Second printing in 1987.

** Numerous members of our faculty were arrested during Martial Law in Poland. Among
others, Garret Sobczyk was forced to live in the underground for two months, was arrested
by the SB (Polish KGB) in February 1983 and then brutally expelled from Poland to
Helsinki.
 xvi ZBIGNIEW OZIEWICZ

known as the (Pseudo )Riemannian Differential Geometry. This differential

geometry include the theory of the Dirac operator and utilize completely
the Cartan's calculus for riemannian structures. The riemannian differen-
tial geometry was developed among other by Dirac, Hodge, Kodaira and de
Rham. Hestenes's polemic lecture of 36 pages is published in the Journal of
Mathematical Physics (1993) and is not included in the present volume. We
include another paper by Hestenes on Hamiltonian mechanics in terms of or-
thogonal Clifford algebra. Hestenes' approach is presented more extensively
in his monograph New Foundations for Classical Mechanics (fourth printing
1992). Again this approach is controversial because it requires that the phase
space manifold be endowed with a riemannian structure, whereas no natural
riemannian structure seems to exist. Any choice of scalar product*** gives a
rather obscure identification of vector fields with differential forms.
Spinor and twistor spaces are minimal one sided ideals in Clifford al-
gebras. This approach has a long history. Invented by Elie Cartan (pure
spinors), presented by Marcel Riesz at the Mathematical Congress in Stock-
holm in 1946, reinvented and used by Claude Chevalley in his book in 1954,
used by Atiyah, Bott and Shapiro in 1964, and by Penrose in 1967 who
utilized Witt's decomposition. Since 1974, Albert Crumeyrolle made these
ideas more popular by utilizing Witt's splitting for descriptions of spinors
and twistors*.
David Hestenes in 1975 associated the name "spinor" with Clifford sub-
algebra of even multivectors. Even subalgebra is not an ideal. The trou-
ble is that spinor and twist or spaces are not just linear spaces; they are
linear spaces with structure tensors, namely, bilinear or hermitian forms.
SU2-spinors are members of the two-dimensional oriented <E-space with a
hermitian form and with the automorhism group SU2 . Whereas the ba-
sis independent identification of spinors as ideals is compatible with these
structure forms (inheriting these structures from Clifford algebra), Hestenes'
frame-dependent "identification" of spinors with even subalgebras seems to
be no more than the identification of linear spaces with the same dimension.
Josep Parra recognized the difference between Dirac spinors and Hestenes's
missleading "spinors". Unfortunately most of audience was not able to fo-
llows his reasoning**.
*** Hestenes' choice is that the tangent space splits into the direct sum of the two or-
thogonal spaces of equal dimensions in such a way that the directions of the position and
momenta are orthogonal.
* Twistors as a minimal ideal in the complexified Dirac-Clifford algebra (or alternatively
as a minimal ideal in the real anti- De Sitter-Clifford algebra) was explored by Professor
Jan Rzewuski in a joint paper with Ablamowicz and myself in 1982.
** The young Cambridge group, Anthony Lasenby, Chris Doran and Steve Gull, pre-
sented a paper about how to generalize Hestenes' basis-dependent linear isomorphisms
from, what they call 2-spmors (SL 2 (a')-spinors, which are members of a symplectic space),
and from twistors to the subalgebra of the even multivectors of real Clifford algebra of
Minkowski's space-time. These linear isomorphisms are not convincing because they are
FOREWORD xvii

The next chapter is about one of the central topics of this Symposium:
QUANTUM DEFORMATIONS. In his plenary lecture Julius Wess gave
an introduction to quantum groups also known as noncommutative Hopf al-
gebras. He presented a new geometric framework based on the algebra gene-
rated by noncommutative spacetime "coordinates". This leads to a discrete
spacetime described by eigenvalue equations of operator-valued spacetime
"coordinates". J erzy Lukierski in his plenary lecture presented a nonline-
ar quantum deformation of the Poincare algebra and pointed out that such
deformations lead to the field equations with finite difference time deriva-
tives.
Ursula Carow-Watamura explained the construction of the quantum
Lorentz group, the quantum Minkowski space and the q-deformed Dirac
"(- matrices.
Differential calculus for noncommutative Hopf algebras has been elabo-
rated by Woronowicz since 1979. A calculus for associative rings, which does
not have the Hopf algebra structure, has been considered by Alain Connes,
by Michel Dubois-Violette (since 1988), by Julius Wess and Bruno Zumino
in 1990 and by many others. John Madore applies this noncommutative
calculus to electrodynamics and Satoshi Watamura applies the bicovariant
differential calculus in quantum deformations of gauge theories.
Braided Lie algebras were presented by Dmitri Gurevich who invented
this generalization.
Shahn Majid delivered two lectures, one of which is included in these
Proceedings. Majid considers Hopf algebras with a braided structure on the
tensor product. In this way he obtain a generalization of supersymmetry
(supergroups and superalgebras).
The last chapter contains several important and interesting lectures which
do not fit into any of the previous chapters. One of the most interesting
plenary lectures was delivered by Richard Kerner on .z:;'3-graded algebras.
Leopold Halpern was for four years, 1956-1959, an assistant of Erwin
Schrodinger at the University of Vienna and for eleven years, 1974-1984,
an assistant of Paul Dirac at Florida State University. Halpern claims that
every great physical theory contains an equally great absurdity "that no rea-
sonable person can believe in it". Halpern has been proposing a way to avoid
the absurdity in Einstein's theory of gravity by introducing a Lagrangian
nonlinear in curvature. Halpern has proposed also a Kaluza-Klein gauge the-
ory of gravity based on the anti- De Sitter universe 50(3,2)/50(3,1). In
his lecture Halpern considered spin in Einstein's theory of gravitation and
explained his philosophy that an absurdity is unavoidable (and not obvious)
in any "good" physical theory.
Multisymplectic geometry in classical field theory was initiated by De-
frame-dependent and need "specific Clifford elements allowed on the right and not allowed
on the left" .
xviii ZBIGNIEW OZIEWICZ

decker in 1953 and was developed in Warsaw by Wlodzimierz Tulczyjew

around 1968, and since 1972 by Jerzy Kijowski, Krzysztof Gaw~dzki and
Wiktor Szczyrba. In this geometrical approach, the presymplectic differen-
tial form of classical mechanics is replaced by a symplectic differential form
of degree (2+m) for an m-dimensional classical membrane theory, so that the
zero-dimensional membrane corresponds to mechanics. A differential form
of degree (2+m) is called symplectic if
it is closed, which assures the existence of a variational formulation,
and in particular the existence of the local action,
it is regular, assuring that the maximal integral submanifolds of the
ideal generated by the appropriate (l+m)-forms are exactly (l+m)-
dimensional.
In the last chapter, Dan Radu Grigore gives an overview of the symplectic
formulation of the Lagrangian formalism, following only his own papers and
those by Krupka and Betounes. The paper deals with the Lagrange-Souriau
differential form which does not seem to be regular in the above sense.
Instead, it vanishes on the bivector fields spanned by the integrable vertical
distribution and satisfies another condition. These requirements are needed
for the existence of a local Lagrangian density and are just a fixing of the
local symplectic potential.
I have limited my remarks mostly to contributions which provoked the
lively discussions during the Symposium. The only complaints of partici-
pants were about the overcrowded programme.

Organizers and Editors:

The Symposium was organized by Zbigniew Oziewicz (Chairman), Andrzej
Borowiec and Bernard Jancewicz with the great help of Professor Jerzy
Lukierski.

Acknowledgements
The Editors would like to thank Mrs. Anna Jadczyk and to Dr. Krzysztof
Rapcewicz from University of Wrodaw for all their help and assistance du-
ring the preparation of these Proceedings.
The English of the Foreword has been corrected by Dr. Garret Sobczyk.
The Editors are grateful to Ms Margaret Deignan and Ms Anneke Pot
from Kluwer Academic Publishers for all of her kind help in the publishing
of these Proceedings.
Professor Jan Rzewuski and his wife Alicja at front of their house in Wrodaw
(1989).
 Homage to Professor Jan Rzewuski

by Professor Jan Lopuszanski

Most honorable Professor Rzewuski,

Dear JaS,

It is a pleasant opportunity to celebrate your jubilee in this beautiful

scenery: a gathering of the physical community from all over the world in an
old castle amid the wooded hills of Sobotka. Outside - a nice Indian Summer.
Our Indian Summer, dear Jas, yours and mine, the Indian Summer of our
lives is also quite nice.
Poland has became finally a free country again. The economy, although
still ailing, is slowly improving. We both are healthy and able to follow
actively the exciting developments in contemporary physics. We have loving
partners in life whom we love too. So we may look with confidence to the
future.
As mentioned already your scientific activities are still very vita1. This is
testified by the main issue of this conference: spinors and twistors, as well
as by your book "Introduction to Quantum Theory", published recently.
The exposition of the subject is clear, straightforward and elegant and the
approach is modern.
The theory of spinorial spaces was the main topic of your numerous pub-
lications for many, many years. I do not go wrong claiming that you were the
founder of this direction in physics. You were the first to emphasize that the
spin or space is more fundamental in physics than that of space-time concept.
Dear Jas, you did also a pioneering work in the theory of non-local field
theory, classical and quantum, as well as in the functional approach to quan-
tum field theory, in particular to the theory of the scattering operator, the
so called S-Matrix.
Your work was numerous and well received by experts. Your papers
were frequently quoted. Your work found also followers among the younger
generation of physicists who have been developing your ideas.

Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, xxi-xxii.
© 1993 Kluwer Academic Publishers.
xxii

This year is also the 40-th anniversary of your arrival in Wrodaw. You
were the founder of our Institute and your merit is that this Institute was
and is still thriving.
It is also the 25-th anniversary of your becoming a member of the Polish
Academy of Sciences.
Passing to problems linked closer with daily life, I would like to sketch a
picture of this deep thinking scientist as a man. I shall rely on own expe-
rience gained during many years of collaboration. I would like to stress the
extraordinary moral uprightness and personal charm of Professor Rzewuski.
He is a type of man qualified coloquially as a manly type; tall, strongly built
and deft, excellent skier and swimmer; his behaviour and conduct is charac-
terized by self-control, an restraint and quickness of decision. In relations
with other people he is straightforward, sensitive and of high personal cul-
ture, showing a deep wisdom concerning human nature and life. He likes
music and is a connoisseur in this field of arts. As a superior he rarely makes
use of the prerogatives of power. He is a good organizer who accurately
distinguishes among important and insignificient issues. In his work and in
every activity he is exact, persistent and careful. He is courageous and firm,
if needed. This was testified during the war when he fought as a voluntary
soldier in the Warsaw uprising as well as in 1968 when he dared to oppose
the totalitarian communist regime as Dean of our Faculty.

Jas, have a nice time in Sobotka!

 Homage to Professor Jan Rzewuski

by Professor Jan Mozrzymas

Dear Guests and Participants,

I have known Professor Jan Rzewuski since 1957 which means from the
third year of my studies at the University of Wroclaw. I have been, from
this time, under the impression of his personality; he has always impressed
me as a physicist, a scholarly teacher and a human being. But now, after
thirty five years of our mutual acquaintance, I would like to say that the
most inspirational and meaningful for me was his behaviour during March
1968. It was the period of the anti-Jewish campaign unleashed by the ruling
powers. Professor Rzewuski was at that time the dean of our faculty and, as
it turned out, our faculty was the only one that expressed official opposition
to this campaign.
In the years 1981-1984, when I was the dean of the faculty and in the
years 1984-1987 during which I was the rector of the University of Wro-
claw, we lived through Martial Law which was as distressing as March of
1968. Throughout these six years of work first as dean and later as rector,
and, especially, in the most difficult situations, I tried to follow the example
that Professor Rzewuski provided so many years before. Today on this most
solemn occasion, I would like to take the opportunity to thank Professor
Rzewuski for all that he has done for me and, in particular, for this example.

Jan Mozrzymas

Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, xxiii.
© 1993 Kluwer Academic Publishers.
 .... {1I.,.. 14.. 3-1... ~.
t0~r,Qn'~,:fo1~1Brnn~~~~~')f
;'~ L ) - \ \ ' - \ , 1z
'/4s'"\::J4,~ro~~J3~1~~{\(".Il J, ., ~ ~~ ,15"""';
~ J L~4ft/ 'JSitfl 8
fO ,.0, ~f
4~~ ~ ~)6
PARTICIPANTS on the PHOTO

1. Jerzy Cislo, 2. Anthony Lasenby, 3. Helmut Urbantke, 4. Steve Gull, 5.

Chris Doran, 6. David Hestenes, 7. Jerzy Hanckowiak, 8. William Baylis,
8a. Jerzy Rozanski, 9. Cezary Juszczak. 10. Wladyslaw Marcinek, 11, Prze-
myslaw Siemion, 12. Josep Manel Parra, 13. Leopold Halpern, 14. Shahn
Majid, 15. Aleksey Isaev, 16. Ireneusz Tobijaszewski, 17. Bernard Jancewicz,
18. Ziemowit Popowicz, 19. Marek Mozrzymas, 19a. Jan Milewski, 20. Mijat
Mijatovic, 21. Andrzej Borowiec, 22. Jan Sobczyk, 23. Aleksandr Zheltukhin,
24. Zbigniew Oziewicz, 24a. Viktor Abramov, 25. Wojciech Kopczynski, 26.
K. Paul Tod, 27. Gary Gibbons, 28. William Pezzaglia, 29. Pertti Lounesto,
30. John Madore, 31. Garret Sobczyk, 32. Jaime Keller, 33. Ralf Grunewald
or Uwe Semmelmann, 34. Jerzy Lukierski, 35. Julius Wess, 36. Jakub Rem-
bielinski, 37. Jan Lopuszallski, 38. Dan Radu Grigore, 38a. Michel Dubois-
Violette, 39. Krzysztof Rapcewicz, 40. Richard Kerner, 41. Dmitrij Volkov,
42. Vladimir Soucek, 43. Sorin Marculescu, 44. Valeriy Tolstoy, 45. Ursula
Carow- Watamura, 46. Michael Schlieker, 47. Satoshi Watamura, 48. Sergey
Merkulov, 49. Vladimir Lyakhovski, 50. Mrs. Susan Dixon, 51. Geoffrey
Dixon, 52. Dmitri Gurevich.
SPINORS
STRUCTURE OF MATRIX MANIFOLDS
AND A PARTICLE MODEL*)
JAN RZEWUSKI
Institute of Theoretical Physics,
University of Wroclaw, pl. Maxa Borna 9, 50-204 Wroclaw,
Poland

Abstract. The decomposition of matrix manifolds into homogeneous spaces of direct

products of certain groups is described. The results are applied to derivation of the internal
structure of SU(2, 2) x SU(m) and P4 x SU(m) invariant particle models.

1. Introduction

The mathematical description of physical laws is based on the observed sym-

metries and the underlying geometry. An example is the Poincare symmetry
P4 = T4 5D SO(3,1) and the underlying space-time M4 which is one of the
homogeneous spaces of P4 namely

(1.1 )

This fact inspired some physicists (cf. e.g. [1], [2]) to investigate also other
homogeneous spaces of the Poincare group

/ ,..... P4 SO(3,1) Hi C SO(3,1)

P4 Hi = SO(3,1) (1.2)
Hi i = 1,2, ...

with respect to their applicability in physics. E.g. the local coordinates on

SO(3, 1)/ Hi can be considered as internal degrees offreedom of a relativistic
particle.
In this paper we wish to combine an old idea [3] of describing the par-
ticle structure in complex space rather than in Minkowski space with the
investigation of homogeneous spaces of the whole physical symmetry group
consisting of external as well as internal symmetries [4]. We shall assume,
in accordance with experiment, that the physical symmetry is the direct
product SU(2, 2) x SUe m) or P4 X Sue m) C SU(2,2) X SUe m) of external
conformal or Poincare and internal SU(m) symmetryl where P4 is now to
be considered as a subgroup of SU(2, 2) .
• Supported by Polish Committee of Science, Grant No 2-2419-92-03
1 We keep m arbitrary to cover such possibilities as SU(3), SU(3) x SU(2) x U(l) etc.
3
Z. Oziewicz et al. (eds.). Spinors, Twistors, Clifford Algebras and Quantum Deformations, 3-16.
© 1993 Kluwer Academic Publishers.
4 J. RZEWUSKI

The natural representation space for a direct product G L( n,~) X

GL(m,~) is a complex matrix manifold ~nm. In the case of SU(2, 2)x SU(m)
it will be ~4m. In this space both internal and external symmetries have a
common geometrical basis (in contradistinction to space-time where only
external symmetries are geometrized).
We shall consider homogeneous manifolds of SU(2,2) X SUe m) and
P4 X SU(m) in ~4m and show that there exists one and only one such man-
ifold which admits a unique and consistent projection on the compactified
complex Minkowski space. In the case of the smaller P4 X SU(m) symmetry
we arrive at the homogeneous manifold

P4 X SU(m) ~ P4 SO(3,1)xSU(m)
(1.3)
SO(2) X SU(m - 2) SO(3,1) X SO(2) X SU(m - 2)·

It is seem that the particle structure in this model is described by the m-

independent 5-dimensional real manifold SO(3, 1)/SO(2) and the manifold
SU(m)/SU(m - 2) depending on the kind of internal symmetry.
The structure of homogeneous submanifolds of ~nm can be investigated,
up to a certain stage, without additional difficulties in the case of arbitrary
nand m (Section 2). It provides the theory of n complex m-vectors (or m n-
vectors) subject to certain invariant conditions and generalizes in a certain
sense the theory of spinors, bispinors, twistors etc. to arbitrary dimensions.
In the case when the symmetry is the direct product of more than two groups
one has to generalize to matrix manifolds of matrices with more than two
indices [5]. One can also consider supermatrices being representation spaces
of direct products of supergroups and their decomposition into homogeneous
structures [6].
In the case of sets of vector fields the general theory provides a classifi-
cation of all possible invariant constraints.
In Section 3 we derive the internal structure in the SU(2, 2) x SUe m) and
P4 X SU(m) invariant particle models which follows uniquely from the above
mentioned assumptions. This structure is described in terms of homogeneous
spaces (cf. e.g. (1.3)) and it remains to describe invariant dynamics and
invariant differential operators in these spaces. This task will be the subject
of a separate publication.
The present brief report contains only the main features of the theory.
Proofs and more details will be published elsewhere (cf. however also [4] and
[7]).

2. Matrix Manifolds
Let us consider the set ~nm of all complex n X m matrices. The elements of
this set may be viewed as n complex m-vectors (or m complex n-vectors)
STRUCTURE OF MATRIX MANIFOLD 5

or as homomorphisms Hom(C m ~ Cn ) of the vector space C m into en (or

vice-versa).
e
The set mn ~ Hom(C m ~ en) decomposes in a natural way into sub-
manifolds oin,m) of matrices of equal rank

Oknm ) := {M E cnm : rankM = k} , (2.1 )

M = {maa} a c
= 1, ... ,n E nm ~ Hom(C m ~ en) ,
0: = 1, ... ,m
min(n,m)
cnm = U oin,m), O k(n,m) n O(n,m)
l
_ {; O(n,m)
- Ukl k . (2.2)
k=O

A matrix of rank k is characterized by the fact that all determinants

malal

m (0: 1 , ... ,0:1 ) := det [ : (2.3)

a1, ... , al

m alal

of order higher than k vanish and that there exists at least one subdetermi-
nant of order k different from zero

(2.4)

Equation (2.4) determines a coordinate neighbourhood for the manifold

Okn,m). There are G)(';:) such neighbourhoods according to the (~) possi-
bilities to choose k rows out of n rows out and to the (';:) possibilities to
choose k columns out of m columns.
Let us choose on oin,m) a neighborhood corresponding to the square
submatrix

ma:.lal

J(= [ det J( i 0 (2.5)

m akal
6 J. RZEWUSKI

and denote the complementary matrices by A, B, Y.

Y=

manO'm
(2.6)
In the neighborhood
m (l, ... ,k) ;;i 0
(2.7)
1, ... ,k
the picture is particularly simple

(2.8)

According to well known relations from linear algebra, we have

A = aK Y = aB
(2.9)
B = Kb Y = Ab
where

a = {a~:/} a'= al, ... , an and b= {b~:,} a , = al, ... ,ak

a" = ak+l, ... ,an a" = ak+l, ... , am
(2.10)
are k X (n - k) and k x (m - k) matrices resp.
Due to invertibility of K (det K ;;i 0) we obtain from (2.9)

Y = aKb = AK- l B (2.11)

providing two natural coordinate systems on oin,m) corresponding to the

neighbourhood det K ;;i O.
• The index sets are ordered in the sense that al < a2 < ... < a. ;
ak+l < ak+2 < ... < an and similarly for the a's.
STRUCTURE OF MATRIX MANIFOLD 7

The coordinates a and b playa particularly important role because of their

invariance properties. The coordinates a are the same for all columns and
the coordinates b are the same for all rows. The first (second) are invariant
with respect to arbitrary transformations of columns (rows).
It is seen from (2.11) that

di'l1l¢Oin,m) = ken + m - k). (2.12)

The space with lowest dimension, for k = 0, is the point mae> = 0, a =
1, ... , n, k = 1, ... , m. The next complex dimension is already n + m - l.
Each space ofn,m) with I < k lies in the boundary of oin,m) in the sense
dn,m) C O(n,m) 1< k (2.13)
I k'

where the "bar" denotes closure in the topology induced on oin,m) from the

°
natural topology in c(n,m).
The manifolds ofn,m) all have elements arbitrary close to := O~n,m) and
form a flag of manifolds [7] in the sense that oin,m) c oi~;,). All closed
06
orbits meet at the point 0 = n ,m) and their tangent spaces at this point
form a flag of spaces in the usual sense.
The homomorphism M E oin,m) admits a canonical decomposition

Cm ~ em / Ker M ~ 1m M --S Cm . (2.14)

the kernel J( er M (the image 1m M) consisting of an (m - k )-dimensional
(n-dimensional) hyperplane in Cm (C n ). To this canonical decomposition
there corresponds a fibering (oin,m) , G:_ k X G"k, 11"0) of oin,m) where

11"0 : O(n,m) -----> Gm X Gn

k m-k k
(2.15)
M -----> J(er M X 1m M

is a projection on the base G:_ k X G'k and the fibre is homeomorphic with
GL(k,C). The G:_ k and G'k are Grassmann manifolds parametrized by the
coordinates a~:, and b~:, resp. and consisting of all (m - k )-dimensional
planes in C m and all k-dimensional planes in C n . Two other fiberings are
possible
O(n,m) ----->
11"1 : k G'k

M -----+ ImM,
(2.16)
O(n,m)
11"2 : k -----+ G:_ k

M -----+ J(er M
8 J. RZEWUSKI

as illustrated on the following graph:

o(n,m)
k

(2.17)

(id,O) (O,id)
Gk

The group theoretical structure of oin,m) follows from the observation

that ~nm is the representation space of the group GL(n,~) X GL(m,~):

M ....... M' = gMh- 1 , 9 E GL(n,~), h E GL(m,~)

according to the commuting diagram

~n_. _______M_________
(2.18)
g h

~n_. _______M_'________
The manifolds oin,m) are orbits of GL(n,~) X GL(m,~) [8]

dn,m) _ GL(n,~) X GL(m,~)

k - H(n,m)
(2.19)
k

For the point Mo =( oI 0)

11k h. "1
(n m) t e Isotropy group IS eaSI y
E Ok'

calculated to be

(2.20)
STRUCTURE OF MATRIX MANIFOLD 9

where 91 E GL(k,C), 92 E ck(n-k), 93 E GL(n - k,C), h2 E c(m-k)k,

h3 E GL(m - k,C).
Another group theoretical description of oin,m) is obtained if we represent
the complex Grassmann manifolds in the base as homogeneous spaces
U(n) D< GL(n,C)
G nk D<
- (2.21)
U(k) XU(n - k) Hi:
where

Hf = (~) (2.22)

and 91 E GL(k,C), 9k E ck(n-k), 9k E GL(n - k,C).

It is important to note that the Grassmann manifolds G'k and G:_ k ofthe
base in the fibre bundle oin,m) are invariant with respect to 1n X GL(m,C)
and GL(n,C) X 1m resp. This follows immediately from the remark after
formula (2.11) stating that the coefficients a (b) of the linear combinations
A = aK (B = Kb) do not depend on the columns (rows) of the matrix M.
Up to now we have considered the general symmetry GL(n,C) X
GL(m,C). In physical applications we mostly have to do with symme-
tries restricted by the existence of metric. We shall consider here only
the case when the general symmetry is reduced to the direct product
SU(n-p,p) X SUe m-q, q) defined by the invariant hermitean metric tensors
F1 and F2 . In this case there appear real invariants
(2.23)

and the manifolds oin,m) decompose into SU(n - p) X SU(m - q)-invariant

sub manifolds oin,m) determined by the invariant equations

(2.24)
It can be shown that only the first k invariants are independent (cf. e.g.
[4]) so that we have to do with a k-parametric family ('" = {"'1, "'2, ... ,"'k})
of homogeneous spaces of the group SU(n-p,p) X SU(m-q, q). Analytically
these manifolds are obtained by introducing (2.11) into (2.24).
To simplify the notation we extend the matrices a~'11 and b~:, (cf. (2.10))
by the unit matrices a~', 6~'" b~: = 6~;, so that relation (2.11) can be
extended to
M = a K b. (2.25)
Introducing now (2.25) into (2.24) we obtain

In = tr(F2 b*ICa*F1aKbt = tr(hK*hKt (2.26)

\0 J. RZEWUSKI

where
it=a*F1 a, (2.27)
are the metrics induced from the metrics F1 and F2 on the columns and
rows of the k X k matrix J(.
The induced metrics it and h are functions of a and b resp. Their sig-
nature is determined by the roots of the secular eqs.

det (Ii - .H) =0 i = 1,2 . (2.28)

Not all eigenvalues can appear in the induced metrics. The number of posi-
tive (negative) roots can not exceed the number of positive (negative) signs
in the original metric. If the original signature in (;n is (n - p, p) then the
admissible signatures on k-dimensional planes in (;n are (k -l(k), l(k)) with
the obvious relations

k -l(k) 5: n - p, l(k) 5: p, l(k) 5: k, l(k) 2 °

or, jointly,

lmin(k) = max{O,k + p - n} 5: l(k) 5: min{p,k} = lmax(k). (2.29)

In this way the Grassmann manifolds G k and G~_k are decomposed

into domains corresponding to different induced metrics divided by borders
determined by zeros of various multiplicity of the secular equation (2.28).
The zeros correspond to degenerate metrics, the number of zeros in the
metric equaling the multiplicity of the 0 root. This structure of G~_k and
G'k can be lifted by the inverse of one of the projections 11"0,11"1,11"2 to ot~m)
(d. (2.17)). Details can be found in [7].

3. The Model
To construct a particle model one has to derive the structure of the space of
internal parameters. The derivation is based on two plausible assumptions:
1) The physical symmetry group is represented by the direct product
SU(2, 2) X SUe m) or its subgroup P4 X SUe m). SU(2, 2), the covering group
of the conformal group, or its Poincare subgroup P4 , are supposed to describe
the external, SUe m) the internal symmetries in accordance with experimen-
tal evidence. External symmetries are represented by SU(2,2) or one of its
subgroups in order to have a common geometrical basis ((;4m) for both ex-
ternal and internal symmetries. It is not necessary, so far, to specify m. One
can think e.g. of SU(3) or SU(3) X SU(2) X U(l) < SU(6).
2) The external and internal parameters of the particle are represented
by local coordinates of an invariant homogeneous submanifold of the linear
representation space C4m of SU(2, 2) X SUe m). This manifold has to satisfy
STRUCTURE OF MATRIX MANIFOLD 11

the following correspondence principle: It must admit a projection on the

Minkowski space-time which is unique and consistent with the symmetry.
It is easy to show that there exists one and only one such submanifold of
<c 4m •
To find the manifold satisfying the above conditions we use decomposition
(2.2)
(3.1 )
and the fiberings (2.17)2. It is seen immediately that the only submanifold
containing G~ is O~4,m) with the local trivialization G~ X GL(2,<c) X G2'. It is
well known that G~ is isomorphic with the compactified complex Minkowski
space ~, the isomorphism being given by the well known relations

(3.2)
a =
where a is a 2 X 2 complex matrix its entries being Grassmann coordinates of
the two-dimensional hyperplane in <e 4 • The dimensional parameter A with
dimension of length has to be introduced in a relation connecting the com-
plex vector zJ.L = x J.L +iyJ.L with the dimensionless ratios a = AJ(-l. According
to the remark after formula (2.11) the coordinates a~:, do not depend on the
selection of columns in k which proves uniqueness of the projection 11"1. To
prove consistency with the group we have to derive the transformation prop-
erties of the coordinates zJ.L induced by SU(2,2) transformations of the rows
in M by the intermediary of the Grassmann coordinates a = AJ(-l.
If d,pJ.L' kJ.L and mJ.LV are the generators of dilatations, translations, special
conformed transformations and rotations in <e4, then the induced infinitesi-
mal transformations of the zJ.L are

(3.3)

The proof of (3.3) can be found in [4] and in the complete version of this
report.
2 The case m = 2 corresponds to the Penrose model [9]. In this case C(4,2) = 0~4,2) U
0\4,2) U 0~4,2) and the internal symmetry is restricted to SU(2).
12 J. RZEWUSKI

It is seen from (3.3) that dilatations d and rotations mj1.V are linear trans-
formations and, therefore, act in the same way on the real and imaginary
parts of the complex vector zj1. = xj1. + iyw Special conformal transformation
are non-linear and, therefore, they mix xj1. and yj1. according to

(3.4)
kj1.Y>' = 2igj1.>' xvqv - 2i (xj1.y>, + Yj1.x>.)
Also translations are not linear and it follows from (3.8) that the real part
xj1. transforms like a vector Pj1.x>, = -igj1.>' whereas the imaginary part is
translation ally invariant Pj1.Y>. = O.
The transformation properties (3.3) prove that the condition of consis-
tency of the projection (3.2) with the group SU(2,2) is satisfied for the
complex vector zj1.'
The real and imaginary parts of zj1. = xj1. + iyj1. transform like vectors
with respect to rotations and dilatations. The fact that Yj1. is invariant un-
der translations and x j1. transforms like a vector suggests the interpretation
of xj1. as the local coordinates of the centre of mass and of Yj1. as the rel-
ative coordinates with respect to the centre of mass. This interpretation
corresponds to Yukawas idea of bilocal theory [10), [11].
Let us go over to the calculation of invariants of the theory. According
to (2.26) In = tr (12K* ItKt, n = 1,2 on O~4,m). The metric of the group
SU(m) is necessarily F2 = nm • The invariant form of the group SU(2,2)
must be chosen in accordances with the representation of the generators of
SU(2,2) in 1[:4. It is shown in [4] that we must take

Fl = - (~2 ~2) (3.5)

The metric It induced from (3.5) on 1[:2 is

(3.6)

With the help of the isomorphism (3.2) we can express a~;, through zj1. and
obtain
(3.7)
and

(3.8)
STRUCTURE OF MATRIX MANIFOLD 13

where
22m
TJ1. = - L L L m~a (o-J1.)"b mba· (3.9)
a=1 b=1 a=1

Details of the derivation are published in [4].

Instead of II and h we can use, equivalently, the invariants YJ1.TJ1. and
YJ1. y J1.T vT V and describe the decomposition of O~4,m) into a two-parameter
family of submanifolds by the two SU(2, 2) X SUe m) invariant equations

(3.10)

In the case when we restrict the external symmetry to the Poincare group
P4 another invariant appears, namely the Poincare invariant YJ1.yJ1. (d. (3.3)).
A further decomposition of O~4,m) takes place into a three-parameter family
of submanifolds ot~m) described by the equations

(3.11)

Let us consider now the decomposition of the Grassmann manifold into

domains of different induced metrics. According to (2.28) the induced sig-
natures are determined by the roots of the secular equation for the induced
metric. In our case (d. (3.6), (3.7))

(3.12)

is a 2 X 2 matrix with the two eigenvalues

(~tr It) 2 - det It

(3.13)
4
tT It = -I Yo·

According to the general scheme (2.29), we have the following domains corre-
sponding to the admissible induced metrics: (++), (+- ), (- -), (+0), (0- ),
14 J. RZEWUSKI

(0,0). (cf. also [9])

(+ +), Yo < 0, YILy IL < 0, Al > 0, A2 > °

(+ 0), Yo < 0, YILy IL = 0, Al = A2 > °
0,

(+ -), Yo § 0, YILy IL > 0, Al < A2 > °

0,

A2 = °
(3.14)
(0 -), Yo> 0, YILy IL = 0, Al < 0,

(- -), Yo> 0, YILy IL < 0, Al < A2 < °

0,

(0O), Yo = 0, YILy IL = 0, Al = 0, A2 = °.

It is seen that the classification of domains and metrics depends entirely

on the character of the fourvector YIL and does not depend on Xw
The invariant conditions (3.10), (3.11) contain two translation ally invari-
ant Minkowski fourvectors YJ.I and rw The variables xJ.I do not enter and the
variables J( and B (mala, a' = 1,2, 0: = 1, ... ,m) enter by the intermediary
of the vector rJ.l (cf. (3.9)). Solving (3.19) for y we obtain

y2 _ (ry + C12) Cl
(3.15)
r20 rJ.lrJ.l

or, after diagonalization,

rILrl"
(YD2 + (y~)2
70
(y~)2 =
CI - Ci2
rJ.lrJ.l
(3.16)

But from (3.9) we easily derive that the vector r IL is time-like and points
towards the future (rJ.lr IL < °, °).
ro > Equation(3.16) represents therefore
an ellipsoid which is real when CI - Ci2 < 0. In the case of conformal symme-
try the axes are functions of rILrJ.l and roo In the case of Poincare symmetry
rJ.lrJ.l = -C22, C22 > 0, Cl = CllCn and the ellipsoid (3.16) depends only on
the component ro, the condition for the reality being

det C = Cll C22 - Ci2 < °. (3.17)

It is important to have also a coordinate free description of the spaces

determined by equations (3.10) or (3.11). We restrict ourselves here to the
(easier) discussion of the Poincare invariant case (3.11). Consider the point
STRUCTURE OF MATRIX MANIFOLD 15

~alcn ~I" a' = 1,2, a = 1, ... ,m, 11 = 0,1,2,3, satisfying conditions

a = 2,3, ... ,m,

a = 3,4, ... ,m, (3.18)

o 0
Y1 = Y2 = o.
and
22m
~i = - L: L: L: :it :a (iTi)ab mba = 0
a=1 b=1 a=1
22m
~o = - L: L: L: ~ ~a (iTo)ab ~ba = .jC22
(3.19)
a=1 b=1 a=1
o 0
Yo TO = -C12, del c < 0

One easily convinces oneselve that this point satisfies conditions (3.11) and
that the isotropy group of this point is SO(2) x SU(m - 2).
Moreover, every point satisfying (3.11) can be reached from the point ~, ~
(3.18-19)) by a proper transformation of SO(3, 1) x SU(m). The remaining
coordinates x I' are unrestricted and we have, therefore,

O(4,m) C>! P4 SO(3, 1) x SU(m)

x (3.20)
2,c - SO(3,1) SO(2) x SU(m - 2)

where P4/S0(3,1)) stands for the real (external) Minkowski space-time

parametrized by the coordinates x I' = Re ZI' and c denotes the three real
parameters Cik satisfying det C < o.

The internal space can be considered as the direct product of a five-

dimensional outer internal space SO(3,1)/SO(2) parametrized by the co-
ordinates YI' and Til subject to conditions (3.11) and an inner internal
space SU(m)jSU(m - 2) parametrized by the coordinates malen a' = 1,2,
a = 1, ... , m subject to conditions TI' = const, 11 = 0,1,2,3.
The domains described by the admissible metrics (+ +), (+ - ), (- -) (of
(3,14)) are represented by the inside of the future (- -) and past (+ +) light
cones and by the outside (+ -) of the light cone. The degenerate metrics
(+ 0) and (0 -) are represented by the past and future light cones. The
metric (0,0) corresponds to the point YI' = o. (cf. also [9]).
16 J. RZEWUSKI

The internal space in the model described here is necessarily not compact.
It consists of the non-compact outer internal spaces SO(3, 1)/SO(2) which is
topologically equivalent to the direct product S2 X H3 of a two-dimensional
sphere and a three-dimensional hyperboloid and of the compact inner inter-
nal space SU(m)/SU(m - 2) topologically equivalent to S2m-l X S2m-3.
For physical interpretation it remains to find the representations of
SU(2, 2)/ SU(m) and P4 X SU(m) in the corresponding homogeneous man-
ifolds. We shall present the results in a separate publication.

References
[1] Finkelstein D., Phys. Rev. 100 (1955), 924.
[2] Bacry H., Kihlberg A., J. Math. Phys. 10 (1969), 2132.
[3] Rzewuski J., Bull. Acad. Polon. Sci. 6 (1958), 261, 335;
7 (1959), 571; 8 (1960), 777, 783; Nuovo Cimento, 5 (1958), 94; Acta Phys. Polon. 17
(1958),417; 18 (1959), 549.
[4] Rzewuski J., Reports on Math. Phys. 22 (1985), 235.
[5] Rzewuski J., Reports on Math. Phys. 28 (1989), 167.
[6] Rzewuski J., Reports on Math. Phys. 29 (1991), 321.
[7] Kocik J., Rzewuski J., in Symmetries in Science II, Ed. B. Gruber, R. Lenczewski,
Plenum Press, 1986.
[8] Crumeyrolle A., Reports on Math. Phys., 28 (1989), 27.
[9] Penrose R., Ann. Phys. 10 (1960), 171;
J. Math. Phys. 8 (1967), 345; "Quantum Theory and Structure of Space-Time",
Miinchen 1975.
[10] Yukawa H., Phys. Rev. 77 (1950), 219; 80 (1950), 1074; 91 (1953),415.
[11] Rayski J., Nuovo Cimento 2 (1955), 255.
COMPLEX STRUCTURES AND THE ELIE
CARTAN APPROACH TO THE THEORY OF
SPINORS
MICHEL DUBOIS-VIOLETTE
Laboratoire de Physique Theorique et Hautes Energies,
Biitiment 211, Universite Paris XI,
91405 ORSA Y Cedex, France

Abstract. Each isometric complex structure on a U-dimensional euclidean space E corre-

sponds to an identification of the Clifford algebra of E with the canonical anticommutation
relation algebra for £ ( fermionic) degrees of freedom. The simple spinors in the terminol-
ogy of E. Cartan or the pure spinors in the one of C. Chevalley are the associated vacua.
The corresponding states are the Fock states (i.e. pure free states), therefore, none of the
above terminologies is very good.

1. Introduction

In this lecture, we will discuss complex structures and spinors on euclidean

space. This is an extension of the algebraic part of a work [1] describing
a sort of generalization of Penrose and Atiyah-Ward transformations in 2£
dimension. We shall not describe this work here, refering to [1], but concen-
trate the lecture upon the notion of simple spinor of E. Cartan [2] (or pure
spinor in the terminology of C. Chevalley [3]). Many points of this lecture
are well known facts and, in some sense, this may be considered as an in-
troductory review. The notations used here are standard, let us just point
out that by an euclidean space we mean a real vector space with a positive
scalar product and by a Hilbert space we mean a complex Hilbert space.

2. Isometric Complex Structures

2.1. NOTATIONS
Let E be an oriented 2l-dimensional euclidean space (E ':::' JR2C) with a scalar
product denoted by (.,.). The dual space E* of E is also, in a canonical way,
an euclidean space and we again denote its scalar product by (.,.). On the
complexified space E; = E* @ <C of E* one may extend the scalar product
of E* in two different ways: Either one extends it by bilinearity and the
corresponding bilinear form will again be denoted by (.,.) or one extends it
by sesquilinearity and the corresponding sesquilinear form will be denoted
17
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 17-23.
© 1993 Kluwer Academic Publishers.
18 MICHEL DUBOIS-VIOLETTE

by (ele ). As for any complexified vector space, there is a canonical complex

conjugation w t-7 W on E*, (an antilinear involution), and the connection
between the two scalar products is given by:

2.2. ISOMETRIC COMPLEX STRUCTURES OR HILBERTIAN STRUCTURES

Let 11( E) be the set of isometric complex structures on E or, which is the
same, the set of orthogonal antisymmetric endormorphisms of E, i.e.

11(E) = {J E End(E)IJ E O(E) and J2 = -ll} =

= {J E End(E)IJ E O(E) and (X, JY) = -(J X, Y), 'v'X, Y E E}
Let J E 11( E) and define

(x + iy)V = xV + yJV, 'v'(x + iy) E ~, 'v'V E E

and
(XIY)J = (X, Y) - i(X, JY), 'v'X,Y E E.
Equipped with the above structure, E is a f-dimensional Hilbert space which
we denote by EJ. For a basis (el, ... , et) of the complex vector space EJ,
(el, . .. ,et, J el, ... ,Je() is a basis of E the orientation of which is inde-
pendent of (el' ... ,e() but only depends on J. Accordingly, 11( E) splits in
two pieces: 11(E) = 11+(E) U 11_(E). The orthogonal group O(E) acts
transitively on 11(E) and the subgroup SO(E) of orientation preserving or-
thogonal transformations acts transitively on 11+(E) and on 11_(E).
Thus one has 11(E) ~ O(E)jU(EJ) and 11+(E) ~ SO(E)jU(EJ) ~ 11_(E)
where U(EJ) is the unitary group of EJ for a fixed J E 11(E) (i.e. U(EJ) ~
U(~()). We equip 11(E),11+(E) and 1L(E) with the corresponding mani-
fold structure. In particular, dimR 11(E) = dimR11±(E) = f(U - 1) - f2 =
f(f - 1).

2.3. IDENTIFICATION OF DUAL SPACES

The dual Hilbert space of EJ can be identified with the Hilbert subspace
A1,0 Ej of E; defined by

A 1 ,OEj = {w E E~lw 0 J = iw}.

One verifies easily that A1 ,0 Ej is maximal isotropic in E~ for (e, e) or, which
is the same, that A1,0 Ej is orthogonal to its conjugate Al,o Ej = A0,1 Ej in
E~ for (ele) (Le. E~ is the hilbertian direct sum A 1 ,0 Ej EEl AO,1 Ej).
Conversely if F C E~ is a maximal isotropic subspace for (e, e), then there
COMPLEX STRUCTURE AND SPINORS 19

is a unique J E 1t(E) such that F = AI,O Ej. It follows that 1t(E) identifies
with a complex algebraic submanifold of the grassmannian Gt(E~) of £-
dimensional subspaces of E~, (Gt(E~) ~ G i ,2i«(;)). In particular, 1t(E) is a
compact Kahler manifold of complex dimension i(t~l) and its Kahler metric
is given by ds 2 = ~tr( (dP}O)2) where p}O is the hermitian projector of E;
on AI,o Ej. Notice that one has p}O = pJ,1 = 11- pl,O.
Furthermore AI,OEj is the fibre at J E 1t(E) of a holomorphic hermitian
vector bundle of rank £ over 1t(E) which we denote by AI,o E*.
Finally notice that one has the hilbertian sum identifications

Ak E; = EB AT,S Ej, tlJ E 1t(E)

T+s=k

where AT,SEj = AT(AI,OEj) 0 AS(AI,OEj), (here the tensor product is over

(;).We denote by p;,s the corresponding hermitian projectors.

2.4. EXAMPLES

One has 1t+(lR?) = {I+}, 1i+(JR4 ) = (;pl, 1t+(JR6 ) = (;p3 and, as will be
shown below, 1i+(JR2i ) C (;p2 l - 1 _1 but the inclusion is strict for £ ~ 4 as
it follows by comparison of the dimensions.

2.5. HODGE DUALITY

On AE* there is a linear involution, *, defined by *(w l /I. •.• /I. w P ) =

w p +1 /I. ••• /I. w U for any positively oriented orthonormal basis (wI, ... ,w2t).
One extends this involution by linearity to AE;. One has the following
lemma.

Lemma. Let n n n
be an element of Ai E~. Then one has + it * = 0, (resp.
n - it * n = 0), if and only if pJ,in = 0, tlJ E 1t+(E), (resp. tlJ E 1i_(E)).
For £ = 2 (i.e. in dimension 4), this is the basic algebraic lemma for the
Penrose-Atiyah-Ward transformation.

3. The Clifford algebra as C.A.R. algebra

3.1. DEFINITION

We define the Clifford algebra Cliff(E*) to be the complex associative *-

algebra with a unit 11 generated by the following relations
20 MICHEL DUBOIS-VIOLETTE

The ,),(w),w E E*, are hermitian generators and')' : E* -. Cliff(E*) is an

injective IR-linear mapping. One extends,), as a (>linear mapping, ')' : E~ -.
Cliff(E*), by setting ,),(w) = ,),(w)*.

3.2. COMPLEX STRUCTURES AND THE C.A.R. ALGEBRA

Let J E 1i(E) be given. The algebra Cliff(E*) is generated by the ')'(w) with
w E A1,o Ej and their adjoints ,),(w)* = ,),(w). In terms of these generators
the relations read

These are the defining relations of the algebra of canonical anticommu tation
relations (C.A.R. algebra) for C (fermionic) degrees of freedom. Thus each
J E 1i(E) corresponds to an identification of the Clifford algebra with the
C.A.R. algebra. Furthermore, the action of the orthogonal group O(E) on
1i( E) corresponds to the Bogolioubov transformations. One has, as well
known, Cliff(E*) ~ M2t(q:;).

4. Spinors and Complex Structures

4.1. DEFINITION

We define a space of spinors associated to E to be a Hilbert space S carrying

an irreducible *-representation of Cliff(E*). The spinors are the elements of
S. Since Cliff(E*) is isomorphic to M 2€(q:;), S is isomorphic to q:;2t and the
representation is an isomorphism. We shall identify Cliff(E*) with the image
of this representation.

4.2. THE SIMPLE SPINORS OF E. CARTAN

Let 'ljJ E S with 'ljJ :j:. 0 and set I,p = {w E E;I')'(w)'ljJ = o}. If WI and W2
are in I,p, one has [,),(wt},,),(wz)]+'ljJ = 2(WI,WZ)'ljJ = 0, so I,p is an isotropic
subspace of E~ for (.,.).
If I,p is maximal isotropic, Le. if dim(I,p) = £, then 'ljJ is called a simple spinor
by E. Cartan [2] or a pure spin or by C. Cheval ley [3]. We denote by F the
set of these spinors and by P(F) the corresponding algebraic submanifold
of P(S) = q:;p2t-I, (Le. P(F) is the set of directions of simple spinors).
For 'ljJ E F,I,p = h1/J V>' E q:;\ {O}, so the maximal isotropic subspace I,p of
E; does only depend on the direction ['ljJ] E P(F) of 'ljJ. On the other hand
we know that there is a unique J E 1i(E) such that I1/J = A1,o Ej. It follows
that one has a mapping of P( F) in 1i( E) which is in fact an isomorphism
of complex manifolds. In the following, we shall identify these manifolds,
writing P(F) = 1i(E).
COMPLEX STRUCTURE AND SPINORS 21

4.3. THE NATURAL LINE BUNDLE

The restriction to P(F) = 1t(E) of the tautological bundle of peS) is a
holomorphic hermitian vector bundle of rank one, L, over 1t(E). One has
L = F u {the zero section}.
As holomorphic hermitian line bundles over 1t(E), one has the following
isomorphisms, see in [1]: Ae,oE* ~ L0L and A l(l:;l)T*1t(E) = L® 2(1'-1).

4.4. SEMI-SPINORS AND SIMPLE SPINORS

To the action of SO(E) on E* corresponds a linear representation of its cov-
ering Spin(E) in S. Under this representation, S splits into two irreductible
components S = S+ EEl S_ with dimS+ = dimS_ = 21'-1. The elements of S+
and S_ are called semi-spinors. One the other hand P(F) = 1t(E) splits into
two transitive parts under the action of SO(E), 1t(E) = 1t+(E) U 1t_(E).
It follows that F = F+ U F_ with F± = F n S± and (with an eventual
relabelling in the ±) P(F±) = 1t±(E). In other words F consists of semi-
spinors. It turns out that for f ~ 3 all non vanishing semi-spinors are in F
(i.e. F± = S±\{O}) but for f;::: 4 the inclusions F± c S±\{O} are strict
inclusions. For f ;::: 4 1t+(E) is no more a projective space.

5. Fock States and Simple Spinors

5.1. STATES ON ALGEBRAS
Let A be an associative complex *-algebra with a unit :Ii. We recall that a
state on A is a linear form 4> on A such that 4>( X* X) ;::: 0, 'It X E A and
4>(:Ii) = 1. The set of all states on A is a convex subset of the dual space A*
of A. The extreme points of this convex subset (i.e. which are not convex
combinations of two distinct states) are called pure states. To the states on
A correspond cyclic *-representations of A in Hilbert space via the G.N.S.
construction; pure states correspond then to irreducible representations.
Coming back to the case A = Cliff(E*), we see that to each spinor 'lj; i' 0
corresponds a state X 1-+ (iJ~lt) (its direction) which is a pure state leading
to an irreducible, or simple, representation. This is why the terminology of
C. Chevalley or E. Cartan to denote the elements of F is somehow mislead-
ing. What characterizes the elements of F is that the corresponding states
(Le. elements of P(F) = 1t(E)) are Fock states or free states on Cliff(E*)
(see below); thus the name Fock spinors or free spinors would be better.

5.2. FOCK STATES ON THE CLIFFORD ALGEBRA

First of all it is clear from above that the elements of F are all possible vacua
corresponding to the identifications of Cliff(E*) with the C.A.R. algebra. It
22 MICHEL DUBOIS-VIOLETTE

is well known that given a vacuum, the vacuum expectation values factorize
and only depend on the "two-point functions" i.e. on the vacuum expectation
values of ,(wlh(w2) for Wi E E*, (this is the very property of the free states).
More precisely, a Fock state, (see for instance [4]), on Cliff(E*) is a pure state
</> satisfying the following (Q.F.) property:

k
for Wi E E*, (where ~ means omission of the kth term). From (Q.F.) one sees
that </> is determined by the </>( ,(WI h(W2)) = h(WI, W2) +iO'(WI, W2), Wi E E*,
where hand 0' are real bilinear forms. The defining relations of Cliff(E*)
implie that h(Wl'WZ) + h(W2,wI) = 2(Wl,W2) and O'(wl,w2) + O'(W2, wI) = O.
The positivity of </> is equivalent to </>(,(Wl + iW2h(wl - iW2)) ~ 0 which is
equivalent to h(WI,W2) = (Wl,W2) and O'(w},wz) = (AWI,W2) = -(WI, AW2)
with II A II::; 1. By polar decomposition, A = JIAI with J E 1i(E) and
IAI ~ 0 (II IAI II~ 1). Then, </> is pure if and only if IAI = 1. Therefore,
</> is a Fock state iff. it satisfies (Q.F.) and </>(/(WI),(W2)) = (Wl,W2) +
i(Jwl,W2), 't/Wi E E*, with J E 1i(E). Thus, the Fock states are parame-
trized by 1i(E) = P(F) and, in fact, the set of Fock states is P(F); indeed
if'¢ E F is such that It/; = A1,0 Ej then one has

('¢I,(wI),(W2)'¢) .
II '¢ 112 = (WI, W2) + l(JWI, W2), 't/Wi E E*

and (Q.F.) is satisfied.

6. Sp~nors and Fock Space Constructions

The standard construction of the Fock space for the C.A.R. algebra implies
that, for each J, S is isomorphic to
EBAO,nEj.
n

However, there is the vacuum, namely an element of LJ, which is hidden

here.
In fact, one has an isomorphism it> of hermitian vector bundles over 1i(E)
from
EBAO,n E* ® L
n

onto the trivial bundle with fibre equal to S, such that

't/W E AO,lEj and 't/rp E EBAO,nEj ® LJ.

n
COMPLEX STRUCTURE AND SPINORS 23

More precisely one has the following

which gives the identification of semi-spinors.

7. Bundles of Complex Structures

Let M be a 2f-dimensional oriented riemannian manifold. The tangent space
Tx(M) at x E M is an oriented 2f-dimensional euclidean space so one can
consider the complex manifold H(Tx(M)) as above. H(Tx(M)) is the fiber
at x E M of a bundle H(T(M)) on M which we call the bundle ofisometric
complex structures over M. This bundle is associated to the orthonormal
frame bundle so there is a natural connection on it coming from the Levi-
Civita connection of M. On H(T(M)), there is a natural almost complex
structure defined by the following construction. Let J x E H(Tx(M)), then
by horizontal lift, J x defines a complex structure on the tangent horizontal
subspace at J x ; on the other hand the tangent vertical subspace at J x is the
tangent space to the complex manifold H(Tx( M)) so it is naturally a com-
plex vector space, so by taking the direct sum one has a complex structure
on the tangent space to H(T(M)) at J x and finally, H(T(M)) becomes an
almost complex manifold. It is easy to show that the almost complex mani-
fold H(T(M)) only depends on the conformal structure of M. In particular,
the almost complex structure of H(T( M)) is integrable, i.e. H(T( M)) is a
complex manifold, whenever M is conform ally flat. The Penrose and the
Atiyah-Ward transformations are obtained, in the four-dimensional case, by
lifting to H(T(M)) various objects living on M (see in [1)).
Let us end this lecture by noticing that the complex manifold H(T(SU))
identifies with the complex manifold H(]R2C+2) of isometric complex struc-
tures on the euclidean space ]R2C+2, [1]. So, in particular, by restriction to
the positively oriented complex structures one has H+(T(S4)) = H+(]R6) =
(;p3.

References
[1] DUBOIS-VIOLETTE M. : 1980, "Structures complexes au-dessus des varietes, ap-
plications", Seminaire math. E.N.S. in "Mathematique et Physique", L. Boutet de
Montvel, A. Douady and J.L. Verdier, eds. Progress in Mathematics vol. 37, 1-42.
[2] CARTAN, E. : 1938, Lec;ons sur la theorie des spineurs I,ll, Hermann et Cie.
[3] CHEVALLEY, C. : 1954, "The algebraic theory of spinors", Columbia University Press,
Morningside Heights, New York.
[4] MANUCEAU, J. : 1970, "Etude algebrique des etats quasi-libres; A. Etats quasi-
libres des fermions". In Cargese lectures in Physics vol. 4, D. Kastler ed., Gordon and
Breach.
SPIN STRUCTURES ON HYPERSURFACES
AND THE SPECTRUM OF THE DIRAC
OPERATOR ON SPHERES*
ANDRZEJ TRAUTMAN
Institute of Theoretical Physics, Warsaw University,
Hoza 69, 00-681 Warsaw, Poland.

Abstract. Recent results on pin structures on hypersurfaces in spin manifolds are re-
viewed. A new form of the Dirac operator is used to compute its spectrum on n-dimensional
spheres. This constrbution is based on two papers by the author, where details and proofs
can be found (Ref.4 and 5).

1. This research has been motivated by, and can be summarized in, the
following observations:
(i) In odd dimensions, it is appropriate to use the twisted adjoint represen-
tation p : Pine n) -+ O( n) to find a cover of the full orthogonal group O( n)
which extends the standard homomorphism Spine n) -+ SOC n). Here p is
given by p(a)v = a(a)va-l, where v ERn, a E Pin(n) C Cl(n) and a is the
grading (main) automorphism of the Clifford algebra Cl(n) [1]. Using the
twisted representation leads to modifying the Dirac operator [2].
(ii) The bundles of "Dirac spinors" over even-dimensional spheres are trivial
[3]; this observation generalizes to hypersurfaces in Rn+l: every such hy-
persurface, even if it is non-orient able, admits a pin structure with a trivial
bundle of Dirac (n even) or Pauli (n odd) spinors [4]
(iii) The spectrum and the eigenfunctions of the Laplace operator d on the
n-dimensional unit sphere Sn are easily obtained from the formula
n+l
L 8 2 /8xf = r- 2 d + r- n 8/8r(r n 8/or) (1)
;=1

This formula generalizes to a foliation of R n + 1 by hypersurfaces and extends

to the Dirac operator, allowing a simple computation of the Dirac spectrum
of n-spheres [5] .

• This research was supported in part by the Polish Committee for Scientific Research
under grant No.2-0430-9101

25
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 25-29.
© 1993 Kluwer Academic Publishers.
26 ANDRZEJ TRAUTMAN

2. Consider the vector space R n with the standard scalar product (u Iv) and
the associated positive-definite quadratic form (u I u), where u = (uJ.L) ERn,
J-L = 1, ... , n. The corresponding Clifford algebra CI( n) contains REB R n,
one has

uv+vu=-2(ulv), where U,VER n , (2)

and uv is the Clifford product of u and v. Let (eJ.L) be the canonical frame
in R n so that u = uJ.LeJ.L for every u ERn; similarly, (ei), i = 1, ... , n + 1, is
the canonical frame in R n+!. The group Pine n) is defined as the subset of
CI( n) consisting of products of all finite sequences of unit vectors.
Let CI(n) = Clo(n) EB Ch(n) be the decomposition of CI(n) defining its
n
Z2 grading so that Spine n) = Pine n) Clo( n). Let a = ao + al be the
corresponding decomposition of a E CI( n). The map h : CI( n) -+ Clo( n + 1)
given by a -+ ao + al en+! is an isomorphism of algebras with units. By
restriction, it defines the commutative diagram of group homomorphisms

Pin(n) ~ Spin(n + 1)
p ! ! p (3)
O(n) !! SO(n + 1)

where the horizontal (resp., vertical) arrows are injective (resp., surjective).
For every n, there is a representation , of CI( n) and a representation
" of CI( n + 1) in the same complex vector space S. The representation "
extends, in the sense that, = " 0 h. One puts

,i = ,'(ei) i = 1, ... ,n + 1. (4)

and defines the helicity automorphism r = (-1)n(n-l)/4,n2 ... 'n so that
r 2 = 1. Note that ,(el-') = 'J.L'n+! and ,(eJ.Lell) = 'J.L'II. For n = 2m, , is
the Dirac representation in a complex vector space of dimension 2m and "
is one of two Pauli representations, characterized, say, by ,n+l = .j - 1r.
For n = 2m - 1, " is the Dirac representation, whereas , is a faithful
representation that decomposes into two irreducible Pauli representations.
This terminology generalizes the one used by physicists in dimensions 3 and
4.
3. Consider now an n-dimensional pin manifold M, i.e. a Riemannian man-
ifold with a pin structure
(5)

where P is the O(n)-bundle of all orthonormal frames on M so that a(q) =

(aJ.L(q)), q E Q, is an orthonormal frame at i(q) = 11" 0 a(q) E M and
i : Q -+ M is a Pin(n)-bundle such that a 0 6(a) = 6(p(a)) 0 a, where 6(a)
is the (right) translation by a E Pin(n) of elements of Q. The Levi-Civita
SPIN STRUCTURES ON HYPERSURFACES 27

connection on M defines a "spin" connection on the pin-bundle Q -> M

which can be described by giving on Q a collection of n horizontal vector
fields V/L (Jl = 1, ... ,n) such that, for every q E Q, one has T q 1i'(V/L(q)) =
a/L(q).
By restriction, one has the representation -y: Pine n) -> GL( S) and one
defines a spinor field on M, with its pin structure (5), as a map 'Ij; : Q -> S,
equivariant with respect to the action on Pine n), 'Ij; 0 b( a) = "'(( a-I) 0 'Ij;.
Alternatively, and equivalently, a spinor field can be described as a section
of the bundle E -> M, associated with Q -> M by the representation "'(.
The Dirac operator V = "'(/LV /L transforms spinor fields into spinor fields.
4. A hypersurface M in an (n + I)-dimensional connected Riemannian
manifold M' is an n-manifold M with an immersion f : M -> M'. The
metric tensor on M' induces a Riemannian metric on M. If M' is orient able
and P' is its bundle of orthonormal frames of coherent orientation, then the
bundle P of all orthonormal frames on M can be identified with the set
{(x,p) E M X P':p = (Pi),i = 1, ... ,n+ 1 wherepis a frame at f(x) such
that Pn+l is orthogonal to Txf(TxM) C Tf(x)M'}.

The group O( n) acts in P via H. Assume now that M' has a spin structure
Q' ~ P' -> M'; a spin-structure on Mis (5), where Q -> P is the Z2-bundle
induced [6] from Q' -> P' by the map F: P -> P',F(x,p) = p, i.e.

Q = {(p,q) E P X Q' : F(p) = a'(q)}.

As an example illustrating this construction, one can mention the embedding
of real projective spaces, RP n ->RP n +l . Since RP 4m +3 is a spin manifold,
there is a pin structure on RP 4m+2 [7].
Immersions of M, which are differentiably homotopic one to another, give
rise to equivalent pin structures on M, but otherwise not, in general. For
example, the "identity" and the "square" immersions of SI in R2 give rise
to the non-trivial and the trivial spin structures on the circle, respectively.
Assume now that the spin structure on M' is trivial, i.e. there exists a
map 9 : Q' -> Spin(n + 1) such that g(qa) = g(q)a for every q E Q' and a E
Spine n + 1). The pin structure on the hypersurface M need not be trivial, but
the bundle E -> M of spinors, associated by "'( with Q -> M, is isomorphic
to the direct product M X S. /
Indeed, the bundle E can be identified with the set of equivalence classes
of the form [(p,q,¢»], where (p,q,¢» E P X Q' X S, F(p) = a'(q) and
[(p,q,¢»] = [(p',q',¢>')] iff there is a E Pin(n) such thatp' = pp(a), q' = qh(a)
and ¢> = "'((a)¢>'. The map [(p,q,¢»]-> (-rr(p),,'(g(q))¢» trivializes E. For ex-
ample, if M is a hypersurface in Rn+l, then its bundle of Dirac or Pauli
spinors is trivial. Since RP 3 =SO(3) has a trivial spin bundle, the bundle
of two-component "Dirac" spinors on RP 2 is also trivial. In general, the
28 ANDRZEJ TRAUTMAN

bundles of Weyl (half) spinors on even-dimensional hypersurfaces in Rn+l

are not trivial (example: even-dimensional spheres).
5. Let f : M -+ M' be an embedding (i.e. injective immersion) of the
hypersurface M in the manifold M' with a trivial spin structure Q' -+ P' -+
M'. The maps P -+ P' and Q -+ Q' are then also injective and the extension
Q" of the Pine n )-bundle Q to the group Spine n) is also trivial. A spinor fiels
'IjJ: Q -+ S extends to a map 'IjJ": Q" -+ S such that 'IjJ"(qa) = ,'(a-1)'IjJ"(q)
for every q E Q" and a ESpin( n +1). Instead of working with 'IjJ, one can now
take a global section s of the trivial bundle Q" -+ M and the composition
\Ii = 'IjJ" 0 s : M -+ S as an equivalent way of describing the spinor field. One
defines the Dirac operator D acting on \Ii by the formula
D\Ii = (\l'IjJ)" 0 s. (6)
6. The above considerations are particularly useful and simple when M is
an orientable hypersurface embedded in Rn+!. This being so, let (Xi) be
the unit normal vector field on M and let (xi) be the Cartesian coordinates
in R n+!. Each of the n( n + 1) /2 vector fields
Xij = Xi)i - Xi OJ , where Oi = O/OXi, 1 ~ i <j ~ n,
is tangent to M. Introducing the notation
0ij = hilj - ,j/i )/2, X= X I'i, div X = OiXi,
so that

Oij = Oij + ,ilj,

one can write (6) as
1 ..
D\Ii = 2X(o'JXij-divX)\Ii. (7)

The right side of (7) is invariant with respect to the replacement of X by

-X and one can show that the assumption of orient ability of M is irrelevant.
Assume now that R n +1 is foliated by a family of hypersurfaces so that
the field X of unit normals is defined over an open subset of Rn+!. The
identity
(8)
leads to a decomposition of the Dirac operator ,iOi on Rn+l into parts
tangential and transverse to the foliation,
,iOi = D + X(%r + ~divX), (9)
where a/or = Xi Oi is the derivative
in the "radial" direction, transverse
to the foliation. There is an analogous formula for the Laplace operator
SPIN STRUCTURES ON HYPERSURFACES 29

[4]. Since the operator D anticommutes with X and X2 = -1, if W is an

eigenfunction of D, then (l+X)W is an eigenfunction of XD with the same
eigenvalue. Therefore, for M orientable, it is enough to consider the spectrum
of the latter operator.
7. As a simple application, consider the spectrum of the Dirac operator on
the unit sphere Sn. The space Rn+l with its origin removed is foliated by
the spheres r = v(xi + ... + x;+l) = const. so that Xi = xi /r,the vector
fields Xij are generators of rotations, divX = n/r and equation (9) gives

(10)
Let cp:Rn +1 -+ S be a spinor-valued harmonic polynomial of degree 1 +

,i
1, where 1 = 0,1, ... The polynomial W = ('''/Oi)CP is of degree 1 and is
annihilated by the Dirac operator Oi . Therefore, on the unit sphere r = 1,
one has

XDw = (I + n/2)W and XD XW = -(I + n/2)XW. (11)

and the spectrum of the Dirac operator on Sn, for n > 1, is the set of
all numbers of the form ±(l + n/2), where I = 0,1,2, .... There is a gap
°
of length nand is never an eigenvalue, this being a simple consequence
of the celebrated Lichnerowicz theorem [8]. For n = 1, there are two spin
structures. The previous formula applies to the non-trivial structure; for the
trivial one, the spectrum is Z.

References
[1] M.F. Atiyah, R. Bott and A. Shapiro, Clifford modules, Topology, 3 Supp!. 1(1964)3-38
[2] A. Trautman, The Dirac operator on hypersurjaces in Euclidean space, Trieste Seminar
on Spinors, Letter 13 (10 April 1991).
[3] S. Gutt, Killing spinors on spheres and projective spaces, in: "Spinors in Physics and
Geometry" (Proc. Conf. Trieste, 11-13 Sept. 1986) ed. by A. Trautman and G. Furlan,
World Scientific, Singapore, 1988.
[4] A.Trautman, Spinors and the Dirac operator on hypersurjaces. 1. General Theory,
J.Math.Phys. (in print).
[5] A.Trautman and E. Winkowska, Spinors and the Dirac operator on hypersurjaces.II.
The spheres as an example (in preparation).
[6] S.Kobayashi and K. Nomizu, Foundations oj Differential Geometry, vol.1, Interscience,
New York, 1963.
[7] 1. Dabrowski and A.Trautman, Spinor structures on spheres and projective spaces,
J.Math.Phys. 27(1986)2022-2028.
[8] A.Lichnerowicz, Spineurs harmoniques, C.R. Acad.Sci.Paris A-B 257(1963)7-9.
ALGEBRAIC CONSTRUCTION OF SPIN
STRUCTURES ON HOMOGENEOUS SPACES

VLADIMIR LYAKHOVSKY
Theoretical Department, Institute of Physics, St.Petersburg University, St. Petersburg,
198904, Russia

Abstract. We consider the manyfold B to be the homogeneous space G/ H, where G

and H are the reductive pair of compact Lie groups. The spin structures on B can be
costructed by purely algebraic methods based on the lattice theory.The efficiency of the
proposed algorithm is demonstrated on examples.

1. Introduction

The necessary and sufficient condition of existence of spin structures on the

oriented Riemanian manifold M is the trivialization of its second Schtiffel-
Uitney class (Borel,1959). In practice the problem of evaluating this object
is very difficult. It can be totally solved only for the manifolds with quite
simple topology or of small dimensions (Avis,1979jPetry,1984)
In multidimensional models for elementary particles interactions it is
highly important to know the total classification of spinor fields on the ho-
mogeneous spaces of the type B ~ G / H where G and H are the symmetry
groups. Thus we have the main fibre bundle (H, G, B) with the structure
group H and the base B. Let (SO(n), E, B) be the fibre bundle of the or-
thonormal frames over the oriented Riemann space B ( n = dimB ). For
the subgroup K C SO(n) and K' - the lift of K in Spin(n) the bundle
(SO(n),E,B) has the so called reduction p : (K,R,B) -+ (SO(n),E,B).
The spin structure also has the reduction p' to (K',R',B). In the corre-
sponding part of the strict homotopic sequence 11"1 (1() ~ 11"1 (R) -+ 11"1 (B)
the group 1I"1(K) contains the subgroup 1I"1(K').One can prove that for con-
nected space R' the criterium (Baum,1981) must be generalized as follows:
the spin structure on B exists iff Ker( 7) C 11"1 (K') and N C 11"1 (R)j here
the subgroup N is normal in 1I"1(R) and has the properties 1I"1(R)/N ~ Z2 ,
Nn7(11"1(1()) = 7(11"1(1(')). The disconnected R' has the form R X Z2 and
appear only in the case when 1(' ~ K X Z2. So the trivial spin structure
always exists when 1(' is disconnected.
In this report we propose the algebraic method of the explicit construction
of spin structure on homogeneous spaces B ~ G / H for reductive pair of
connected compact Lie groups G and H. It is based on the lattice theory
31
Z. Oziewicz et al, (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 31-38.
© 1993 Kluwer Academic Publishers.
32 V.D. LYAKHOVSKY

for Lie groups. We shall demonstrate its efficiency on examples that have
physical applications.

2. The group structure revised

Consider the fibre bundle of orthonormal frames (SO(n),E,B) on the ori-
entable Riemann manifold B : SO (n) ---+ E ---+ B, with n = dim B and the
universal covering w : Spine n) ---+ SOC n) . The existence of the spin structure
on B means that the bundle morphism ~ ,

~: (Spin(n),E',B) ---+ (SO(n),E,B)

exists and has the following restrictions: ~lB = id, ~ lSpin = w.

In our case B ~ GIH and (G,H) is the reductive pair of the connected
compact Lie groups. Let g and h be the corresponding Lie algebras and
Vg , Vh - their vector spaces. We shall use the direct sum decomposition

(1)
where Vd is the space of the representation D of the group H induced by
the adjoint representation of G,

Ad(G)!H ~ Ad(H) EB D(H). (2)

For its kernel and image the following notations will be used K == ImD C
SO(n), N == KerD C H . The important object for us is the reduction p of
the main fibre bundle

p: (K,R,B) ----+ (SO(n),E,B),

where R is the factorspace GIN. For the morphism p the spin structure on
B means the existence of such a morphism 'f/ and a reduction p' that the
following diagram is commutative (Dabrovsky,1986)

(Spin( n), E', B) --L (SO(n),E,B)

/1 p1 (3)
1)
(K',R',B) ----+ (K,R,B)
Here K' is the coimage of K for w : K' == Coim w( K) C Spine n).
First let us show that N is normal in G and thus R is the factor group
(just as K ~ HIN ). Suppose that No ,the connected component of N ,is
a nontrivial Lie group. Its algebra no commutes with the space Vd and is an
ideal in h. So no is an ideal in g (see the decomposition (1)). For connected
group G this means that its subgroup No is normal. Thus in the definition
of the factor space B we can use the factor groups GINo and HI No.
ALGEBRAIC CONSTRUCTION OF SPIN STRUCTURES ON HOMOGENEOUS SPACES 33

Having this in mind consider the case of discreat group N. It is cen-

tral in H because H is connected: N C Z(H). Consider the morphisms
Ad(H), Ad(GhH, D and their kernels: Ker(Ad(H)) = Z(H), Ker(Ad(G)!H) C
Z(G), KerD = N. The relation (2) imposes the condition

Ker(Ad(G)!H) = Z(H) n N = N. (4)

The result is that N is the central subgroup of G : N C Z( G) .

We have proved that in the definition of the factorspace B we can factorize
out the kernel of the representation D and reformulate B in terms of the
factor groups R == G/N and 1( == H/N ,

B = G/H ~ R/K. (5)

3. Necessary and sufficient conditions

Let the spin structure (Tf,p') exist (see the diagram (3)). The morphism
Tf : R' - - R is the twofold covering of the group R. So R' is a group. If
R' is connected it belongs to the class of locally isomorphic groups with Lie
algebra r and can be realized as a factor of the universal covering group ii,
R' ~ ii/ N'. Consider the coimages Nand 1( of the groups e and 1( in R.
Then the following prism is commutative

Fili l.
R

K~l~
'N'l~R'~
R

1(' 1(

The conclusion is: when the spin structure exists and R' is connected,
i~ the }ernel N of the universal covering ii -- R there exists a subgroup
N' C N such that
R' ~ ii/N', 1(' ~ 1(/N' and Z2 ~ N/N'. (6)
Note that the covering 1( __ 1( is not universal.
When R' is not connected it is equivalent to the direct product R' ~
R X Z2 and for 1(' the same is true: 1(' ~ 1( X Z2. So the spin structure
(I( X Z2,R X Z2,B) ....!!...... (1(,R,B) is trivial.
34 V.D. LYAKHOVSKY

When the existence of the spin structure is in question one can still put
the problem whether the bundle (K', R', B) and morphism 'fJ exist with K'
as a coimage of w. Now it is clear that to solve this problem one must find
the universal covering group R and the corresponding kernel N containing
a subgroup N' such that the commutative diagram (Fig. 1) holds. In the
case of disconnected K' ~ K X Z2 the disconnected R' ~ R X Z2 plays the
necessary role and the trivial spin structure on B exists. The number of
inequivalent spin structures is just the number of inequivalent subgroups N'
with the described properties (7) plus the trivial one in case of disconnected
K' .

4. Algebraic construction
The problem is how to construct the subgroups N'. It is sufficient to consider
only the front triangle of the diagram (Fig. 1), all the necessary kernels play
there. Now we shall expose the algebraic algorithm which solvs the problem.
It is based on the lattice theory (Loos,1985;Adams1979).
The group K must not be connected. Let 10 be it~c0I!!P0nen~ of ~it
and consider the corresponding subgroups of Nand N': No == N n Ko ,
N~ == N' n Ko. The group K can obviously be written as the factor group

Ko/ No. Let Kb be the connected component of K' and K-the universal
covering for Ko, Kb and K. Then together with the fron!...triangle in Fig.1
we obtain th~rectangular commutative diagram, where Q is the kernel of
the covering K ---> Kb.

Fig. 2.

Diagram 2-a Diagram 2-b

L L ~

Ko K A(Ko) • A(K)

N'0

K'0
X K
p N'0

A(Kb)
X p

A(K)
ALGEBRAIC CONSTRUCTION OF SPIN STRUCTURES ON HOMOGENEOUS SPACES 35

Now the group No is fixed by the factorization

(7)
The group K is compact and connected so one can look for its maximal
torus T(K). Construct all its coimages, contained in the Diagram 2-a . The
obtained abelian groups T(Ko),T(K),T(Kb) together with the initial torus
T(K) form the commutative diagram that isjust a copy of2-a. Now take the
Lie algebras t of these groups and consider the corresponding unit lattices
A == exp-l(l) ct. Once again one obtaines the diagram induced by the
Diagram 2-a .
On the diagram 2-b all the morphisms are injections and~the discreat ~
groups indicated on the diagram are the corresponding factors: p';:;j A(K)/ A(K)
, etc. The lattices A(K) and A(K) are known. It is easy to find A(Ko) . Con-
sider the maximal tori T(R) and T(R) and the corresponding unit lattices
A(R) and A(R) The space Vt(K) of algebra t(K) is the subspace of Vt(R) .
The necessary lattice is obtained by the interseption:
(8)

Now the groups Land No are fixed:

L ';:;j A(Ko)/ A(K), (9)

No ';:;j A(K)/ A(Ko).
So the only object that was not yet defined is the lattice A(Kb) . It
can be identified examining the structure of the corresponding sublattice in
A(Spin( n)) . For the semisimple groups K and R we propose an easy way
to solve this remaining problem.
Consider the complexification Dc of the exact representation D : K ---+
SO(n) and the corresponding complex algebras kc and rc . The decomposi-
tion of the type ( 1 ) is still valid. Vrc ';:;j Vkc EB Vd C . Take the projection {I}
of the root system {I r} of rc of the subspace kc and eliminate the subsys-
tem {I k} (the roots of kc ). The obtained set of vectors {,B} == {I} \ {I k}
is the weight diagram of the representation DC. Let {,B} + be the subset of
positive weights.
When for b E T(K) the rotation D(b) is lifted to Spin(n) we obtain the
product of the operators (Adams,1979) :
exp ± 7r,B( b)e;e;; (10)
where e;, e;; are the Clifford algebra generators corresponding to the plane
V.8EBV-.8. Thus D(b) can enter the lattice A(Kb) only when the sum of ,B(b)
for all the positive weights is even.
36 V.D. LYAKHOVSKY

The lattice A(](b) thus constructed,

A(](b) = {b E A(]() I L f3(b) = 2Z}, (11)

i3E{i3}+

makes it possible to calculate the group Q= A(](b)1 A(]() and together with
--.. -I --. -. -I
L defined earlier ( 9 ) one can finally calculate No =Q I L . Knowing No
and the relations between Rand ]( we can easily reestablish N', obtain the
groups ](' and R' (see ( 6 ) ) and thus conclude the construction of the fibre
bundle that define the spin structure.
First one must find the group ](b =]( I Q . This almost immediately
gives the structure group ](' .

](' ;: :; ](b X Z2 ~ ]( X Z2 for A(](b) = A(](), (12)

](' ;: :; ](b for A(]()/A(](b) ~ Z2.

Then one must search the subgroups N' in N with the properties N' n ](0 =
Nb, NIN' = Z2. Every class of equivalent subgroups N' defines the spin
structure (R', ](', B) with R' ~ RI N'. In case of disconnected 1(' (see ( 12
) ) the trivial spin structure with R' ~ R X Z2 also exists.

5. Examples
5.1 Let us start with the factorspace where it is quite difficult to use the or-
dinary topological methods. Consider the space B ~ GI H ~ SO(5)1 SO(3)sp,
where the subgroup SO(3)sp is the image of the special injection. For this
injection the fundamental representations of SO(5) treated as the repre-
sentations of its SO(3)sp subgroup remain irreducible (Lyakhovsky,1986):
(5)180(3) = (5) , (4)180(3) = (4) .
The 7-dimensional space Vd in the decomposition (1) is also irreducible.
Here the center Z(G) is trivial. So we have R = SO(5), 1( = SO(3)sp
-
and, obviously, R = Spin(5), 1(= SU(2). The groups Nand D coinside:
~ ~

N ~D= Z2 . Now it is necessary to draw the lattices of the groups under

consideration in the standard e-frame (see Fig. 3 ).
Now it is easy to check that the sum 'L,f3i(b) = 6.A(b). is even for every
element of A(SO(3)sp) . This means that A(1(b) ~ A(1() . So the group ](' is
disconnected: 1(' = SO(3) X Z2, the group Q coinsides with p : Q ~ p = Z2'
~ ~ ~

So in this case only the trivial spin structure with R' ~ SO(5) X Z2 exists.
5.2 For regular injections H ----+ G the proposed algorithm gives the results
valid for the whole ansemble of similar factorspaces.
Let G = SU(n) and H = (SU(p) X SU(q) X U(l))/Zu ,where p+ q = n
and u is the minimal proportional to p and q. Here all the necessary lattices
ALGEBRAIC CONSTRUCTION OF SPIN STRUCTURES ON HOMOGENEOUS SPACES 37

Fig. 3.

b {.} - the central lattice

A}(SO(5)) ~ A(SO(5))

{ * } - the dual root lattice

Ao(SO(5)) ~ A(Spin(5))

{ + } - the central lattice

A}(SO(3)sp) ~ A(SO(3)sp)

b - the dual root of SO(3)sp

are well known (Adams,1979), so we shall expose only the final result. One
must consider separately three types of factorspaces.
There is no spin structure when numbers p and q have different pairity.
When both p and q are even the only possible spin structure is the
trivial one with the total space R' ~ R X Z2 = (SU(n)/Zn) X Z2 .
For p and q odd the unic nontrivial spin structure exist with
R' = SU(n) / Zn/2,
K' = (SU(p) X SU(q) X U(l)) / Zu / Zn/2.
5.3 In multidimensional quantum gauge theories the study of symmetry
breaking due to the nontrivial topological configurations leads to the in-
vestigation of the model spaces of the type M4 X sm / Zp (Hosotani,1983).
In these theories the spinor fields on sm / Zp can be treated as the Zp
-invariant spinor fields on sm. The spinor fields as the global sections of
the bundle associated to the pricipal Spin(m)-fibre bundle must retain the
38 V.D. LYAKHOVSKY

initial symmetry. It is known that for m = 2n + 1 the SO(2n + 2) sym-

metry is broken to SU(n + 1) when the factorization by Zp (p f; 2)
is produced (Lyakhovsky,1991). So we are forsed to consider sm
as the
SU(n + 1) jSU(n) factorspace rather then the SO(2n + 2)jSO(2n + 1) .
Let m = 3 and take the triple (e, SU(2), S3) as the initial bundle. Here
we obviously have K = e and R = SU(2) ::::: R. Thus the group Ko = e
-,
is trivial. The diagram (2) shows that No and Kb must also be trivial and
Kb ::::: Ko = e. So we have only the trivial spin structure: K'::::: Z2 and
R' ::::: SU(2) X Z2' The analysis of the injection R' --+ SOC 4) shows that K'
is the diagonal subgroup of R'. As a result on the space (SU(2) X Z2)jZf ag
one must study the spinor fields 'i[f corresponding to the exact 2-dimensional
representation of zf ag . The harmonic expansion for the components 'i[f j
of 'i[f will contain only those representations of SU(2) X Z2 that have the
exact I-dimensional reduction to zf ag . These representations D(l)(SU(2))0
B(Zfag) will contain the exact subrepresentations B for the integer 1 and
the trivial for the halfinteger ones. As a result the basis of the harmonic
expansion for 'i[f j is formed by the matrix elements of the full spectrum of
representations of SU(2).

Acknowledgements
It is the pleasure for the author to express his gratitude for warm hospitality
of the scientists of the Institute of Theoretical Physics in Wroclaw.

References
Borel, A. and Hirzebruch, F.: 1959, Amer. Journ. of Math. 81, 315
Avis, S.J.and Isham, C.J.: 1979, in Quantum Field Theory and Fibre Bundles in General
Space-Time, ed(s)., Recent Developements in Gravitation, Levy M.,Deser S., Plenum
Press:London
Dabrovsky, 1. and Trautman, A.: 1986, Journ. Math. Phys. 27, 2022
Petry, H.R.: 1984, Spin structures on Lorentz manifolds, ISAS-44/84/EP: Triest
Baum, H.: 1981, Spin-Structuren und Dirac-Operatoren iiber Pseudoriemannscher Man-
nigfaltigkeiten, Teubner:Leipzig
Loos, 0.: 1985, Symmetric Spaces, Science:Moscow
Adams, J.: 1979, Lectures on Lie Groups, Science:Moscow
Lyakhovsky, V.D. and Vassilevich, D.V.: 1986, Teor. Matern. Phys. 66(3), 350
Hosotani, Y.: 1983, Phys. Lett. B126, 309
Lyakhovsky, V.D. and Shtikov, N.N.: 1991, Yadernaya Physica 54(8), 595
THE KUMMER CONFIGURATION AND THE
GEOMETRY OF MAJORANA SPINORS
GARY W. GIBBONS
DAMTP
University of Cambridge
Cambridge CBS 9EW
UK

Abstract. In this article I show how the properties of Majorana spinors in four space-
time dimensions may be understood in terms of the real projective geometry of ordinary
three-dimensional space. They may be viewed as points in projective space equipped with
a linear line congruence. The discrete group generated by the i-matrices may be viewed
as the automorphism group of Kummer's configuration 166. As an application of line ge-
ometry which I develop I show how the skies of events of 2 + I-dimensional Minkowski
spacetime correspond to the lines of a linear line complex in projective three space.

1. Introduction
The real Clifford algebra Cliff(l, 3) generated by the relations:

16=-1,1;=1, i=1,2,3 (1.1 )

is isomorphic to the algebra R( 4) of real four by four matrices and therefore

admits a representation in which the gamma matrices 111 are real four by four
matrices and act on a real four dimensional vector space whose elements are
called Majorana spinors (see e.g. Dabrowski, 1988). In this representation
the charge conjugation matrix C, which satisfies CIIlC- 1 = -lilt, 15 and
10 may be taken to be anti-symmetric and the Ii to be symmetric.
Viewed projectively one may think of Majorana spinors as points in real
projective space P3(R). This fact allows one to relate the projective geometry
of ordinary 3-space and spinor algebra. One aspect of this relationship is
that one may identify the 32 element group G 32 generated by the gamma
matrices as the 2-fold cover of the automorphism group G 16 of Kummer's
self-dual configuration 166 consisting of 16 points and 16 planes in P3 (R)
.mch that every plane contains 6 points and 6 planes pass through every
point (Hudson, 1905). This fact was known to Eddington (Zariski 1932;
Eddington,1935,1936 ) and others in the 30's: it arose in his "Fundamental
Theory". The numerological properties of what he called "E-numbers" are
in fact just the properties of this Clifford algebra. This talk is intended as
39
Z. Oziewicz et al. (eds. J, Spinors, Twistors, Clifford Algebras and Quantum Defonnations, 39-52.
© 1993 Kluwer Academic Publishers.
40 G.W. GIBBONS

an expository account of these matters. !vI uch of the material is extremely

ancient if not especially well known nowadays and no particular originality is
claimed, indeed, much of what I have to say lIlay be found in the interesting
book by Paerls but not in quite the way I shall describe it. The reader may
decide for him or herself whether or not the material should remain in the
decent obscurity in which it has been left hitherto. My own motivation was
not to resurrect Eddington's ghost but rather to develop some geometric
intuition for Majorana spinors analogous to that one has for two-component
Weyl spinors, considered projectively as points on the complex projective
line PI (C) which has been so successfully exploited by Roger Penrose.
Of course the action of the group GIG may be extended complex projective
space P3 ( C) but something is lost ill the wa.y of visualizability and moreover
the limitation to real numbers serves to illuminate the differences between
the Clifford algebra Cliff(1,3) defined by (1.1) and Cliff(3,1) defined by:

,6 = 1, ,l = -1, i = 1,2,3, (1.2)

which as a real algebra is isomorphic with the algebra of two by two quater-
nion valued matrices H(2). Now the group r of invertible elements of the
Clifford algebra Cliff(1,3) defined by (1.1) is isomorphic to GL(4,R) which
acts on P3 ( R) as P S( 4, R) the natural group associated to the Projective
Geometry of ordinary three dimensional space. This group is isomorphic
to P SO(3, 3) the group of linear transformations of six-dimensional space
preserving a metric of signature (3,:3), I and is the hasis of Plucker's Line
Geometry in which lines in ordinary space are associated to null rays in
R 3 ,3. By contrast the Clifford algebra Cliff(3,1) defined by (1.2) leads to
the group P SO( 4,2) associated to the Conformal geometry of Minkowski
spacetime. The analogue of Plucker's construction is Lie's Sphere Geometry
in which spheres in ordinary space are associated to null rays in R 4 ,2. Lie
realised that there is no distinction between Line Geometry and Sphere Ge-
ometry if one works over the complex numbers and this idea is at the heart
of Penrose's Twistor Theory. The passage between these two view points is
essentially no more than one of endowing the real fOllr dimensional vector
space of Majorana spinors with a complex structure which allows one to

itly the complex structure is given by the ,-matrix ,5

identify it with the complex two-dimensional space of Weyl spinors. Explic-
whose square is minus
one no matter which signature is chosen. Nevertheless for some purposes,
supersymmetry and supergravity for example, it is much more convenient
to work over the reals and in particular it is often most useful to use Majo-
rana spinors. Indeed it has recently been pointed out by DeWitt and Carlip
1 A metric of signature (p,p) is sometimes called neutraL I prefer the adjective
"Kleinian" by analogy with "Riemannian" for signatnre (p,O) or (D,p) and "Lorentzian"
for signature (p,l) or (l,p). The flat model spaces I like (.0 call Plucker space, Euclidean
space and Minkowski space repectively.
THE KUMMER CONFIGURATION AND THE GEOMETRY OF MAJORANA SPINORS 41

that the under certain global conditions the choice of signature may, in a
certain sence, actually have have physical consequences. For these reasons I
shall restrict myself to real projective geometry.
While working over this material I became aware of various other appli-
cations of the projective geometry to be described below to physics. One in-
volves linear line complexes to describe the causal structure of 3-dimensional
Minkowski spacetime and will be sketched below. Others relate to the ge-
ometry of De-Sitter and Anti-De-Sitter spacetime, the Petrov classification
of curvature tensors and even to mechanics.

2. Projective Geometry
The fix notation and terminology it will be useful to recollect some ele-
mentary geometrical ideas. Points p in P3 ( R) corresponds to rays in in
some real four dimensional 4-dimensional vector space V with homoge-
neous co-ordinates pO ,a = 0,1,2,3. Planes 11' in P3 ( R) correspond to rays
in the dual vector space V* and also have four homogeneous co-ordinates
11'0 ,a = 0, 1,2,3. The point ]J lies in the plane 11' if and only if:

(2.1)

Linear maps from V to V are called collineations. They induce projective

transformations of P3 (R) taking points to points, lines to lines and planes
to planes. In homogeneous co-ordinates they have components: La {3 Lin-
ear maps from V to V* are called co'rrelations. They induce interchanges of
points and planes in P3(R). In homogeneous co-ordinates they have compo-
nents: La{3. A correlation is contact preserving if it preserves the property
of a point lying in a plane. Analytically:

(2.2)

For such correlations it follows that La(3 mllst either be symmetric or anti-
symmetric.
If La{3 is symmetric the correlation is called a polarity or a Legendre
transformation and the plane 1I'(]J)" = L,,{3:/·(3 associated to a point x contains
the point x and is tangent to the quadric:

(2.3)
The point p and the plane 11' are said to be pole and polar respectively to the
quadric. It is clear that there is an equivalence beween quadrics, polarities
and real symmetric four by four matrices.
If on the other hand LO'{3is skew symmetric the correlation is called a null
correlation. It then takes a point 1) to a plane 7r(p) passing through the point.
It is clear that there is an equivalence beween null correlations and real four
42 G.W. GIBBONS

by four skew-symmetric matrices. If the null correlation is non-singular (i.e.

invertible) it also defines a symplectic structure on the vector space V which
descends to a contact structure on P3( R). The word "symplectic" was coined
by Weyl to replace the confusing use of the word "complex" applied to the
group associated to a "linear line complex". This will be defined shortly.
Two points p, q define a unique line 1 pa.ssing through them which corre-
sponds to a simple bi-vector IC'i(3 in V II V

(2.4)

The space of bi-vectors V II V (i.e. skew-symmetric second rank tensors)

is six-dimensional and carries a natural Kleinian metric of signature (3,3)
given by the alternating symbol fC'if3ll v . The simple bi-vectors lC'if3 satisfy:

l af3 I IlV f a f3llv =0 (2.5)

and therefore constitute the four-dimensional manifold of null rays in R 3 ,3

which has topology (52 X 5 2)/ ± 1. the set of lines inherits a conformal
structure because two lines 11 and 12 intersect if and only if :

"21/af3Zllv
1 2 fC'i(3Il = gAB '1 2
V
- LA IB -
-
0, (2.6)

where A = 1,2,3,4,5,6 and [1 = 1° 1 etc, and gAB has signature (3,3). In

what follows, we shall employ the convention that lower indices on bi-vectors
have been lower using the alternating symbol. We are now in a position to
return to line complexes which are 3-dimensional families of lines in P3 (R)
("complex" is here being used to denote a collection of lines which is "plaited
together"). A linear line complex is the intersection of a hyperplane in R 3 ,3
with the Plucker Quadric and thus is of the form:

C'C'if3 /C'i!-1 - 0
- . (2.7)

Clearly associated with every null correlation is a lineal' line complex and
conversely and both are associated to a direction in R 3 ,3. A singular line
complex,sometimes called special is associated with a. null ray in R3,3 and
consists of those lines which intersect a fixed lim'.
A non-singular or non-special linear lin(' complex has the property that
every line of the complex passing through a fixed point 1) ill P3 (R) lies in a
fixed plane 1I"(p) containing p and conversely every line of the complex lying
in a given plane 11" passes through a fixed point p( 11") in the given plane 11".
The point pa and plane 11" a are related by

(2.8)
THE KUMMER CONFIGURATION AND THE GEOMETRY OF MAJORANA SPINORS 43

The point and plane are sometimes referred to as pole and polar repectively.
A more detailed description will be given later. The subgroup of SO(3,3)
leaving invariant the 6-vector CA is the Anti-De-Sitter group SO(3, 2) which
is double covered by what is usually called, following Weyl's suggestion, the
real symplectic group Sp( 4, R). We shall return line complexes later when
we look at causality in 3-dimensional Minkowski spacetime.
A quadratic line complex is the intersection of a quadric in R 3 ,3 with the
Plucker Quadric and thus is of the form:

(2.9)

It has the property that the lines passing through a point p in P3(R) lie on a
quadratic cone. Points at which this cone degenerates to a pair of planes are
called singular points and they lie on the singular surface of the quadratic
line complex. These surfaces are called Kummer surfaces and turn out to be
quartic surfaces. Over the complex numbers they possess 16 double points
or nodes and 16 singular tangent planes or tropes. The nodes and tropes
make up a Kummer conjigumtion. An example of a real Kummer surface
which arises in physics is Fresnel's Wave Surface. The generic quadratic
line complex determines a symmetric tensor Q AB on R 3 ,3 which may be
diagonalized over the reals with respect the Kleinian metric gAB to give a
privileged orthonormal sextad for R 3 ,3. This will be used later.
Finally we recall the a 2-dimensional family of lines is called a line con-
gruence and that a linear line congruence is the intersection of the Plucker
quadric with 4-dimensional plane in R 3 ,3.

3. Spinors in Six Kleinian Dimensions

One may if one wishes pass to the projective space P5 (R) and view the space
of lines in P3 ( R) as the Plucker quadric in P5 ( R) but in what follows it is
more useful to remain in R 3 ,3. Indeed the key idea (going back at least as
far as Cartan (see Paerls)) is to consider the Clifford algebra Cliff(3,3) of
R 3 ,3. This is isomorphic to the algebra of real eight by eight matrices R(8)
and is generated by six real eight by eight matrices fA satisfying:

(3.1 )

which act on an 8-dimensional real vector space 5 whose elements q,aare

"Majorana spinors for 80(3,3)". Since the 50(3,3)-invariant volume ele-
ment f7 == foflf2f3f 4f5 has square unity, the real vector space of Ma-
jorana spinors splits as the direct sum of two real four dimensional vector
spaces which may naturally be regarded as the duals of one-another. In fact
one vector space summand maybe taken as the original vector space V, rays
in which are associated with points in projective space P3 ( R) and the other
44 G.W. GIBBONS

as its dual V*, rays in which correspond to planes in P3(R), i.e. :

5 = V EB V*. (3.2)

In other words one may think of a Majorana spinor for 50(3,3), 111 a , as a
pair (pa, 7r a) consisting of a point pa and a plane 7ra in ordinary (projective)
3-space. Explicitly:
(3.3)

The 50(3,3) invariant Dirac adjoint spinor 111 a is given by

111 = (7ra pa). (3.4)

So the point lies in the plane if and only if:

111111 = o. (3.5)

With these notational conventions we represent points x A in R 3 ,3, i.e. bivec-

tors x a {3 as elements of the Clifford algebra Cliff(3, 3) in terms of the eight
by eight dimensional matrices:

(3.6)

As a check the reader may wish to verify that use of the identity:

(3.7)

yield the basic identity:

(3.8)

J.
where:
II = (8~{3 82 (3.9)

Since the 50(3,3) invariant product f7 acts on the space of point-plane

pairs as
f7 W = ( -7r{3
pO: ). (3.10)

from a geometrical point of view it has no effect. On the other hand R 3 ,3

acts on the space 5 of point-plane pairs as null correlations. One has the
following
THE KUMMER CONFIGURATION AND THE GEOMETRY OF MAJORANA SPINORS 4S

Proposition. If
¢'I! = 0,
then the 6-vector x A must be lightlike, tlle bi-vector x°{3 must be simple and
represents a line in P3(R), moreover if

the point pO lies in the plane 7r 0 and the line x°{3 lies ill the plane 7r 0 and
passes through tlle point pO.

4. Reflection Groups and Their Covers

If we choose an orthonormal basis for R 3 ,3, for example that invariantlyas-
sociated to a generic quadratic line complex, we may consider the group
of reflections with respect to these basis vectors. Each reflection is an in-
volution and all reflections commute so it is an abelian group G32 with
26 elements. According to the usual procedure in Clifford algebra we con-
struct a non-abelian double cover G64 of this this group by multiplying
the six I-matrices fA. The group G64 acts on the eight dimensional spinor
space S. The commutator group of G64 consists of the elements ±1L and
its quotient by the commutator is of course just the abelian group of re-
flections G 32 • Projectively speaking the signs are irrelevant and it is G 32
which interests us. As mentioned above the six I-matrices fA act as six null
collineations. There are 6 X 5/2! non-trivial ways of multiplying together
two I-matrices. The resulting 15 matrices f[AfB] commute with f7 and act
on P3 (R) as collineations. There are 6 X 5 X 4/3! = 20 non-trivial products
f[AfBfC] which anticommute with f7 and act as correlations but since the
off-diagonal blocks are symmetric rather than anti-symmetric like the off-
diagonal blocks of the 6 fA we obtain polarities with associated quadrics.
Proceeding in the same way one obtains 15 collineations f[A f B f c f D] and 6
correlations f[AfBfCfDfE]' Of course there is just one product of the form
f[AfBfCfDfEfF] and this just f 7. It would seem therefore that we arrive
at 6 + 20 + 6 = 32 correlations and (including the unit element 1 and f7)
1 + 15 + 15 + 1 = 32 collineations. However these will not all be distinct
because 'I! and f 7 'I! correspond to the same geometrical point-plane pair
and so projectively speaking we have 16 collineations and 16 correlations,
including the identity. The 16 correlations split into 10 polarities and 6 null
correlations depending upon whether the off-diagonal blocks are symmetri-
calor anti-symmetrical. Associated with these are 10 quadrics and 6 linear
line complexes. Since one of the symmetrical matrices is the identity matrix,
one of these quadrics has no real points.
46 G.W. GIBBONS

The 16 correlations form a necessarily invariant subgroup U 16 of G32 , the

quotient G 3 2/G 16 having the effect of interchanging points and planes, i.e. of
projective duality. The abelian group G16 of correlations acting on the space
P3 ( R) of projective Majorana spinors has a non-abelian 32 element double
cover G32 which is just that group generated by multiplication of the four
by four I-matrices hO'/d in (1.1). It is this group which so pre-occupied
Eddington and which he eventually learnt was the automorphism group of
the Kummer configuration which we are now in a position to describe.
Let us start with an arbitrary point ZF and act with the group G16 of
collineations. A simple argument based on the fact that I-matrices have
square ±1 allows one to deduce that the group and Gl6 acts effectively on
P3 ( R) so we shall obtain 15 additional distinct points making up a 16 points
orbit of G16 in P3(R). These 16 points are the points of Kummer's config-
uration. Now let us acting with the 6 null correlations associated with the
six I-matrices r A will give 6 planes passing the original point q"'. Acting
with the 15 non-identical collineations on one of these planes we obtain 15
other planes making 16 in all. These are the planes of Kummer's configura-
tion. By the commutativity and the duality properties of the group that this
exhausts the set of planes in the configuration and moreover not only does
each point of the configuration lie in 6 of the planes of the configuration but
every plane of the configuration contains 6 of the points of the configuration.

5. An Explicit Basis
To make this explicit it is convenient to adopt basis for V. Geometrically a
basis for V determines four distinct points in P3 ( R) which form as the ver-
tices of a "tetrahedron of reference". The faces of the tetrahedron determine
a basis for V*. The six edges of the tetrahedron determine a real null sextad
for V 1\ V falling into two triples {if, nf} such that each vector is lightlike
in the Klein metric and the only non-vanishing iuner products are between
opposite edges: if gABnp = bij. The tetrahedron of reference is left invariant
by the tetrahedral group which is the three-dimensional abelian subgroup of
correlations corresponding to boosts in the three 2-planes spanned by {if
and nf}. the orbit in P3(R) of a one parameter subgroup of the tetrahedral
group is called a W-curve and the orbit of a two-parmeter subgroup is called
a W-surface. Interestingly, it has recently been suggested that the shapes of
growing buds are well described by W-curves.
With respect to the given tetrahedron of reference a null correlation de-
termines a second tetrahedron of reference each face of which (possibly ex-
tended) contains one of the original vertices. Dually each vertex of the sec-
ond tetrahredron lies on one of the planes of the original tetrahedron. In
this way one obtains eight points and eight planes making up the self-dual
Moebius configuration 84 , When we need to we may take the four points of
THE KUMMER CONFIGURATION AND THE GEOMETRY OF MAJORANA SPINORS 47

our tetrahedron of reference to be the origin and the intersection of the three
co-ordinate axes with the plane at infinity. The faces of the tetrahedron are
then the co-ordinate plane and the plane at infll1ity.
Algebraically we are now allowd ignore the distinction between co- and
contra-variant indices and to decompose the space of bi-vectors 1\2(V) into
the orthogonal direct sum of self dual and antiself dual 2-forms: 1\2(V) =
I\~(V) EB I\:'(V). Two mutually commuting bases {pi} and {Ai}, i = 1,2,3,
for self-dual and anti-selfdual 2-forms respectively may be found generating
two copies of the quarternion algebra Cliff(2):

(5.1)

and
(5.2)
with
(5.3)
One then has:
( O. pi) (5.4)
-p' 0
and
(5.5)

The ri+ have square plus one and the ri- have square minus one.
In this basis the collineations and the correlations consist of 1, pi ,AJ and
pi Aj • These are easily seen to generate under multiplication a 32 element
subgroup (;32 of SOC 4) isomorphic to D4·D4 where D4 is the binary dihedral
group (which is isomorphic to the multiplicative group of unit quaternions).
To see that this group is just the same group that one gets if one multiplies
together the usual 4-dimensional,-matrices it suffices to note that one may
take
(5.6)

(5.7)
It follows that
(5.8)
and we may take as the charge conjugation matrix:

(5.9)
48 G.W. GIBBONS

p
An explicit example is provided by setting:

p'~
-1
0
0
1
0
0
-1
0
0
I) C
o
~ ,p
2

= ~1
0
0
0
0
-1
1
0
0
0
~) "- C
o 'p
o
- 0
0
-1
0
0
0
0
0
0
-1 D
(5.10)
and
o o
~) A2 = (~1 o°°
-1 -1
~ ~) (~0 ~ ~I)
Al -- ( 0 o 1 o o
-1 o 0 ' A3 =
o 0' o
-1 o o o 0 -1 o o 0 o
(5.11)
Although not the same as those used by Eddington the reader will have no
difficulty in verifying that these {'-matrices do indeed satisfy the conditions
of his celebrated competition for Caliban's puzzle column in the Christmas
1936 copy of New Statesman and Nation (Eddington, 1936,1937) in which
three boys and two girls visit a zoo in which the labels on the cages of four
pairs, male and female, of animals with known names have unfortunately
been lost. For every animal, the Tove for example, John supposes that the
animal he supposes to be Mr Tove is the animal he suppose to be Mr Tove
while Mary supposes Mr Tove to be the animal she supposes to be Mrs Tove.
The same is true for all the boys and all the girls. Moreover the animal which
John supposes to be the animal which Mary suposses to be Mr Tove is the
animal which John supposes to be Mrs Tove and the same is true for all
pairs of children.
The permutations that arise from mistaken identities are represented by
the 5 ,-matrices, the boys coresponding to the {'i and the girls to,o and {'5.
The animal species are associated to points equi-distant from the origin along
four orthogonal axes in four dimensional euclidean space. To get the group
of Kummer configuration, G l 6 one simply ignores the sex of the animals.
The six null correlations are given by G, G{'5 and G{'nl'. The ten quadrics
are given by G{'J.l and G{'[P{'v]. The sixteen collin cations are of course 1, ,I',
,5, {'[I'{'v] and {'5{'1'.
To use this representation to construct an example of a Kummer config-
uration in P3(R) we start with the plane (1,1,1,0), i.e.
x+y+z=o
which passes through the origin and which is perpcndicular to one of the
four body-diagonals ofthe cube whose vertices are (±, ±, ±, 1) with all eight
combinations of signs and act with pi and Aj . One obtains six points lying
in the plane at the vertices of a regular hexagon:
(0,1, -1, 1), (-1,0, -1, 1), (1, -1. 0,1),
THE KUMMER CONFIGURATION AND THE GEOMETRY OF MAJORANA SPINORS 49

and
(0,-1,1,1), (0,1,0,1), (-1,1,0,1).
The vertices of this planar hexagon comprise half of the 12 mid-points of the
cube. These twelve points together with four points at lying on intersection
of the four body-diagonals of the cube with the plane at infinity comprise
the sixteen points of Kummer's configuration. The sixteen planes consist of
the four planes passing through the origin perpendicular to the four body-
diagonals of the cube, each of which contains six points arranged at the
vertices of a regular hexagon, together with twelve other planes. Each of
these other twelve planes contains 4 of the mid-points of the cube arranged
in a rectangle and is parallel to a pair of body-diagonals. The twelve planes
therefore also contain two points at infinity. Thus each plane belonging to
the configuration contains six points and it is not difficult to see the truth of
the dual proposition that each point lies on six planes of the configuration.
An example of a real Kummer surface in P3 (R) having the maximum
complement of 16 nodes and 16 tropes making up the above Kummer con-
figuration is given in affine coordinates by:

X4 + y4 + Z4 + 1 _ y2 z2 _ Z2 x 2 _ x 2 y2 _ x2 _ y2 _ z2 = o.
6. Causal Struct ures and Linear line Complexes
The conformal compactification of any flat spacetime RP,q, RP,q is obtained
by considering the space of null rays through the origin in RP+l,q+l. Two
points in RP,q are null separated if and only if the rays have vanishing in-
ner product. One recovers Rp,q as those null rays which intersect a null
hyperplane which does not contain the origin. Four-dimensional Minkowski
spacetime is just a special case and we obtain in this way the standard iso-
morphism between the conformal group and P SOC 4,2). If we had started
from flat four-dimensional space time with the ultra-hyperbolic signature
(2,2) we would have obtained PSO(3,3). The fact that the conformal ge-
omtry of R2,2 coincides with the projective geometry of P3(R) is of course
the Plucker correspondence.
From the point of view of Clifford algebras it is clear that p + q + n = 4
spacetime dimensions is a special case since the dimension of the conformal
group (n + 2)( n + 1)/2 is only equal to one less than the dimension of the
group invertible elements of the Clifford algebra Cliff(p, q), 2n - 1 . This is
what enabled us to bring in projective geometry. However one can also use
projective geometry in lower dimensions and in this section we shall do so
in 2+ 1 spacetime dimensions.
The topologically and metrically Rp,q == (SP x sq) / ± with its product
metric. In the Lorentzian case q = 1 each null geodesic is projects to a great
circle on Sp and we can thus identify the space N of null geodesics as the
50 G.W. GIBBONS

bundle of unit tangent vectors of SP. In the case p = 2 the space N of null
geodesics may thus be identifed with real projective space P3(R).
In this case the conformal group is the Anti-ODe-Sitter group SO(3,2)
which is double covered by Sp( 4, R). The obviolls idea now is to exploit
the fact that the conformal compactification of R 2 ,1 lIlay be obtained as a
restriction of the conformal compactification of R 2 ,2 and then to use the
Plucker correspondence. Thus two points in R2,1 will he null separated if

(i) two lines in P3 (R) intersect

but

(ii) the lines considered as simple bi-vectors 10l{3 are orthogonal to a fixed
bi-vector C OIOI ' i.e.

C 0I{3 ZO:{3 =.0 (6.1)

In other words the lines in P3 ( R) must beloH·g to a non-singular linear
line complex.
It follows that the light cones of spacetinw poi n ts :r ill R2,1 considered
as the skies, or set of null geodesics sky( x) passing through the event x
in may be identified as lines belonging to the linea.r line complex. Since
points of P3 (R) may be identified with null geodesics in R2,1 one may encode
the causal structure of R2,1 into the projective properties of the linear line
complex. For example if two lines [1 = sky(xt) and l2 = sky(x2) belonging
to the complex intersect in a point pin P:3(R) then p represents the common
null geodesic generator of two light con('s of the two spacetime points Xl and
X2 joining these two events. The plane 11(])) containillg all the lines passing
through the point p corresponds to all light cones in Il2,1 sharing p as a
generator. Since we wish to investigate t he possible rela.tion between linking
and causality it will be useful to have a descrptioll of the lines belonging
to the complex. One such description has been provided by Woods (Woods,
1922).
Choosing affine co-ordinates x, y, z such that the contact form:

(6.2)

the lines are found to be tangents to helices drawn on cylinders whose axes
coincide with the z-axis and whose pitch is 211(1.2 when' A is the radius of
the cylinder. The lines of the complex passing through points on the z axis
(which itself does not belong to the C"Olllplex) lie in planes perpendicular to
the the z-axis. It seems appropriate to recall at this point that the word com-
plex is derived from two Latin words for "plait('(l together" and the word
sym-plectic is derived from two Greek words bearing the same meaning.
Clearly each line is linked just once with every other one. Thus linking alone
THE KUMMER CONFIGURATION AND THE GEOMETHY OF MAJORANA SPINORS 51

is not sufficient to determine whether the events ill R2 •1 associated to two

skies are spacelike or timelike separated. This conclusion is consistent with
that of Robert Low, who discuses tIll' case of R 2 •l (Low, 1990).
To compare it is convenient to abandon the use of affine coordinates and
adopt the model of projective space as a solid ball in R3 = {x = (x,y,z)}
with opposite points of the boundary identified. Using stereograllic rather
than central projection one finds that:

1 - 1'2
xC< ex (-2-'x) (6.3)

The polar plane through a point x is:

(6.4)

This becomes a sphere:

(6.5)
If the poles lie on the vertical axis x = (0,0, z) then polar planes are:

22 1_ L2 1_12
x +y +(z- "2(z- 5")) = 4(z+ i)' (6.6)

This gives a family of spheres passing through the equator of the unit ball,
i.e. z = 0, x 2 = y2 = 1. The lines of the romplex belonging to this family of
planes are great circles passing through the vertical axis. One may identify
the generators of I as points on the equator of the boundary sphere, i.e.
x 2 + y2 = 1, z = 0, and I itself as the set ofits generators, I = sky(}).

7. Majorana Spinors and Linear Line Congruences

We have seen above that endowing P3 ( R) with a linear line complex, Cc<(3 re-
duces SO(3, 3) to SO(3, 2). The introduction of a second linear line complex,
Ci5 which is timelike in the Klein metric will fed lice further to the Lorentz
group SO(3, 1). The lines common to both determine a linear line congru-
ence. One line of the congruence passes through each point p of P3(R), this
line is the intersection of the two polar planes of]J determined by the two
null correlations C and Ci5' The associated Lo]'cntziall 4-plane through the
origin in R 3 ,3 is left invariant by SO(2) X 80(;3,2), where the 80(2) factor
corresponds to cMral rotations in the 2-dimensional normal space spanned
by C and Ci5. The infinitesimal generator of this 80(2) is of course just
the involutive colIineation i5' Acting on P3(R) the rhiral rotations move
the points along the lines but leave the lines of th~ congruence itself invari-
ant. We thus obtain a fibring of P3 ( R) by Pd R) and it is not difficult to
see that this is the quotient of the standard IJopj fibration of 3 by Sl by
52 G.W. ·GIBBONS

the antipodal map. Because as mentioned in tilt' introduction if we pass to

complex co-ordinates for V using {5 as a complex structure we obtain two-
component Weyl spinors and now chiral rotations a.ct by phasing.

References
L. Dllbrowski, Group Actions on Spinors Bibliopolis (1988)
A. S. Eddington, New Pathways in Science Cambridge University Press p.271 (1935)
A. S. Eddington, Relativity Theory of Electrons and Protuns Cambridge University Press
1'.36 (1936)
A. S. Eddington, The New Statesman alld Nation Dec J 9 1936 p 1044, Jan 9 1937 PI'.
62-64.
R. W. H. T. Hudson, Kummer's Quartic Surface Cambridge University Press
1905reprinted 1990
R. J. Low, Classical and Quantum Gravity 7 177-187 (1~190)
E. R. Paerls, Representations of the Lorenth G'Ollp alld l'roject.ive Geometry
F. S. Woods, Higher Geometry Ginn and COIllPlLlIY (1922) reprinted by Dover 1961
O. Zariski, American Journal of Mathematics 54 466-4711 (1932)
PAULI-KOFINK IDENTITIES AND PURE
SPINORS
HELMUT URBANTKE
Institut fUr Theoretische Physik,
Universitiit Wien
Austria

Abstract. A machinery producing identities between the bilinear covariants of spinors,

devised by Pauli and Kofink, is extended to the n-dimensional case and applied to pure
spinors.

1. Introduction
Space-times of dimension higher than four have found their way into at-
tempts of establishing a unified theory of all interactions, notably through
the Kaluza-Klein construction and its generalizations to include non-abelian
gauge fields and supersymmetry multiplets. This made it necessary to con-
sider spinors in arbitrary dimensions. Some special dimensions > 4 where
spinors are considered in more detail are 6 (twistor theory (Penrose 1986);
Calabi-Yau manifolds (Candelas 1985)), 7 (parallelized 7-sphere (Englert
1983)), 8, 10, 11 (extended supergravities - see e.g. Julia (1982)); we also
wish to mention the description of classical strings without differential con-
straints as given by Hughston (1987). In all these works certain identities
playa basic role. As probably well-known to the practitioneers, these can
be derived according to a scheme first devised by Pauli (1935) and slightly
extended by Kofink (1937, 1940); it is probably better known as Fierz rear-
rangements (Pietschmann 1983). For dimensions ~ 7, identities of this type
are also related to the theory of pure spinors (purity conditions and purity
syzygies; see Cart an (1966); Chevalley (1954); Hughston (1987); Budinich
(1989)).
In this note we describe the Pauli-Kofink type approach to some of these
identities in a unified manner (Sect. 4) after some preparatory material in
Sect. 2 and 3. Application to pure spinors are given in Sect. 5. All consider-
ations are restricted to the complex domain, for simplicity.

2. Clifford Algebra and Spinors. Completeness. Semispinors

As a generalization and complexification of ordinary Minkowski vector space,
we consider a complex vector space V, dim V = n finite, together with a
53
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 53-60.
© 1993 Kluwer Academic Publishers.
54 H.URBANTKE

non-degenerate symmetric bilinear form g. The Clifford algebra for V, 9 is

the complex associative algebra with unity generated by V in which the
relations xy + yx = g(x, y) . 1 hold for all x, y E V. Here iuxtaposition of
vectors means algebra multiplication, whereas the scalar product of x, y will
always be written g(x,y), 1 is the unit element of the algebra. By spinors we
mean the elements of a complex spinor space S which carries an irreducible
representation 1 of the Clifford algebra by linear operators. The set of these
operators is thus generated by the operators I( e,.) = I,. satisfying

(2.1 )

where {e,. E V} is an orthonormal basis, g(e,.,e ll ) = g,.11 = 8,.11' The indices

appearing in (2.1) can also be considered as "abstract indices". Similarly, we
shall have occasion to use indices in spin space as well, either to be thought
of as referring to some basis in S, or as abstract indices. For any subset
A = {AI, ... ,AIAI} of the set N = {l, ... ,n}, where the elements Ai are
ordered according to 1 ::; Al < A2 < ... < AlAI::; n, we define

10 := idS, (2.2)

where [... J indicates antisymmetrization and JAJ is the cardinality of A. By

(2.1), the IA linearly span the algebra of operators generated by the I>"
With these notations, we can now state the basic properties of the spin
representation of the Clifford algebra; but we have to distinguish the cases
n = even and n = odd. (For a fuller treatment, see, e.g. Penrose (1986),
Budinich (1988).)
Case n = even = 2m. All irreducible representations of the Clifford
algebra are equivalent and faithful; dimS = 2m; tr fA = 0 for all A :10; the
fA are linearly independent and span the whole set End S oflinear operators
on S - in more detail, any F E End S may be expanded

F == rm L (tr fAI FhA. (2.3 even)

ACN

This formula is the basis of the Fierz rearrangements and for the derivation
of Pauli- Kofink type identities (see Sect. 4).
The element IN satisfies

(2.4)

The spaces of semispinors (= half spinors = Weyl spinors = chiral spinors)

S± are defined as the projections

S .- idS
±.-
±2imlN S
. (2.5)
PAULI-KOFINK IDENTITIES AND PURE SPINORS 55

They are invariant under the even part of the Clifford algebra, but get
interchanged by the odd part, as we have

(2.6)

Case n = odd =
2m + 1. The irreducible representations of the Clifford
algebra fall into two equivalence classes, none of them faithful, distinguished
by fN = ±imids; for both classes the dimension of spin space is again 2 m;
tr fA = 0 for all A i- 0, N. The fA form an overcomplete system, i.e., we
have

F =T m- 1 L (tr fAl FhA =T m L (tr fAl FhA. (2.3 odd)

AcN AcN,IAI~m

3. The Fundamental Bilinear Form

If x 1--+ f( x) is an irreducible representation of the Clifford algebra on the
spin space S, then
(3.1 )
is, up to a sign, the transpose representation, acting irreducibly on the dual
space S*. Since in the even case n = 2m there is just one equivalence class of
irreducible representations, there exists a nonsingular intertwiner B : S -+
S*, i.e.
(3.2)
Defining
(3.3)
it can be shown that

(3.4)

The sign in (3.1) has been chosen in such a way that the same properties
for B, including its existence, also hold in the odd case.
In the even case, we can define chiral projections in S* by taking the
transposes of the chiral projections in S. Then B maps S± to S± or S~
according to m = even ("splitting case") or m = odd ("mixing case"):

(3.5)

In terms of indices, if we use small Latin upper indices for elements of

S, lower indices for S*, B appears as Bab, which is either symmetric or
antisymmetric (d. (3.4)). We can use Bab as a "spinor metric" to move
indices, with the usual care for signs in the antisymmetric case. For n even,
we can adapt the index notation to the chiral decomposition and will use
56 H.URBANTKE

capital Latin indices, undotted for S+, S:t, dotted for S_, S:':.: then Bab
defines B AB, B AB in the splitting case, B AB' B AB in the mixing case. So we
have separate semispinor metrics in the splitting case, but an equivalence of
semispinors of one chirality and dual semispinors of the opposite chirality in
the mixing case. (See Penrose (1986) for details.)

4. Fierz Rearrangements and Pauli-Kofink Type Identities

Pick four spinors c.p, 'IjJ,w,x E S and two operators M, L on S: we then
can form, e.g., the invariants obtained from evaluating B M 'IjJ E S* on c.p,
BLX E S* on w, written usually as

( 4.1)

A Fierz rearmngement of the product of these,

(4.2)

is an expression that results when the indicated operator F C e := 'ljJcwd Bde

(of rank one) is expanded according to (2.3), resulting, e.g., in the even n
case in

(c.pTBM'IjJ)(WTBLX) == T m L (wTB'"(;,.l'IjJ)(c.pTBM'YALX). (4.3)

AcN

Of course, one could have included the factor M, or L, or both, in the

definition of F to arrive at other versions. This procedure has played a role
in the discussion of possible interaction terms in weak interactions (Fierz
1937). In concrete examples, it is wise to observe the symmetry properties
(3.4) of B'YA = BA if, e.g., w = 'IjJ in (4.3), and also the chiral properties
of B, M, L, /A if chiral spinors are involved. One can decompose (4.3)
with respect to these two aspects to obtain a group of identities with a
considerably smaller number of terms on the right hand side of each.
Generalized Pauli-Kofink type identities are a special case where M and
L are replaced by /[/-11 .•. //-Ip]' /[Vl ···/Vq]' and contractions over some of
the indices Jli, Vj are taken (Pauli (1935); Kofink (1937, 1940); in this work
c.p, w, X are all assumed to be related to 'IjJ; the scalar identities (i.e. all Jli, Vj
contracted) are also called Fierz identities (Fierz 1937) and have been worked
out in n dimensions by Case (1955)). A more recent application with n = 6,
M = /[/-I/A] , L = /[A/v], 'IjJ = X = (Majorana) spinor, c.p = w = /N'IjJ,
where (4.3) yields _('ljJT B'IjJ)28~, showing that a (resp. covariantly constant)
Majorana spinor field on a 6-dimensional Riemannian manifold defines an
almost complex (resp. complex) structure, is given in (Candelas 1985).
PAULI-KOFINK IDENTITIES AND PURE SPINORS 57

The type of identities about which we want to go into more detail here
arises if we take M = L =,I-', 'I-'
and w = 'Ij;. Then only those A where
B,A = BA is symmetric will contribute, Le., according to (3.4), IAI = m, m+
1; m - 3, m + 4; .... If further cp, 'Ij;, w have a definite chirality, then for (4.3)
to be non-trivial cp and X must have chirality (-lr+1 relative to 'Ij;, cf.
(3.5), and we will have only contributions from A satisfying m + IAI = even.
These two conditions together then require m -IAI to be a multiple of four.
We also can easily prove that (for even or odd n)

( 4.4)

This is the second row ofthe 'Fierz matrix' and shows, in particular, that for
even n = 2m the 'middle' terms IAI = m in the sum (4.3) drop out. Finally,
for odd n quite generally and for even n in the chiral case the sum (4.3)
may be restricted to IAI ::; m, but must then be taken twice: this is because
if A, A' are (ordered) complements in N, they differ by a factor ±1'N which
is a multiple of ids in the odd n case and acts as such on chiral spinors in
the even case, while the numerical factors (4.4) are the same for A, A' if n
is odd, opposite if n is even but this is then compensated by an additional
(-1) that remains from the relative chirality (_I)m+1 between 'Ij; and X.
The identity in question thus becomes, in the even case, with chiral cp,
'Ij;, X and relative chirality ( -1 )m+1 between 'Ij; and cp, X:

(_2)m-4(cpT B'I-''Ij;)('Ij;T B,I-'X) =L CA('Ij;T B'Al'lj;)(cpT B'AX)

A

IAI = m - 4, m - 8,... (?: 0), CA = 8 if IAI = m - 48. (4.5 even)

In the odd case, we have

2m (cpT B,I-' 'Ij;) ( 'lj;T B,I-'X) = ~) -1)IAI(n - 21AI)( 'lj;T B'Al'lj;)( cpT B,AX)
A

IAI = m,m - 3,m - 4,m - 7,m - 8, ... (?: 0). (4.5 odd)
Here the coefficient of the highest terms, IAI = m, is (-1 and we may r,
get rid of these terms by subtracting the corresponding expansion of the
expression (-lr(cp T B'Ij;)('Ij;T BX), resulting in

(_2)m-3{(cpTB'I-''¢)('¢TB,I-'X) _ (_l)m(cpTB'¢)('Ij;TBxn =

= LCA(,¢TB,A1'lj;)(cpTB,AX), IAI = m - 3, m - 4, m - 7, m - 8, ...

A
(4.6)
CA ={ 8 for IAI = m - 48 +1
-8 for IAI = m - 48.
58 H.URBANTKE

Before applying (4.5), (4.6) to pure spinors (Sect. 5), we shall discuss
them in a few special cases. For n = 4 and n = 6 the sum on the right
is empty; for n = 4, one gets a formula well-known from two-component
spinor algebra with a well-known geometric content (Penrose (1984); for
n = 6, using semispinor indices, (3.4) tells us that the quantities ,tJ. AB are
antisymmetric, so that our identity (4.5) says that

(4.7)
are totally antisymmetric. It is well-known that such an object is the basic
structure oftwistor algebra (Penrose 1986). - n = 8 is the first instance in the
even case where we have a nonvanishing r.h. side in (4.5); dim V = 2m = 8
is distinguished among all dimensions by the fact that 8± are of the same
dimension i. 24 = 8 as V. Further, it here happens for the first time in the
even case that B is both symmetric and splitting, thus defining quadratic
forms in 8+ and 8_; together with 9 on V, we have altogether three spaces
with quadratic forms and of the same dimension 8. Eqs. (3.4,5) tell us that
B,tJ. is symmetric-mixing, so we have ,tJ. AC = 1tJ. CA' and the content of our
identity (4.5) may be written, after raising one index

1tJ. ABI tJ..DC + 1tJ. ACI tJ..DB -_ 2B BCUiJ

.A ( 4.8)
plus an analogous equation with dotted and undotted indices interchanged.
If the defining relation (2.1) is also split into its semispinor versions, one
recognizes, on comparison with the relations just obtained, that, just as
(8+,8_) are semispinor spaces for (V, g), (V, 8+) are semispinor spaces for
(8_, Bls_) and (8_, V) are semispinor spaces for (8+, Bls+). This is one
version of the 'principle of triality' (cf. Cartan (1966), Chevalley (1954),
Penrose (1986)). - For n = 10, the right hand side of (4.5) contains only
terms with IAI = 1; B is antisymmetric and mixing, B,tJ. is symmetric and
splitting. Identity (4.5) then can be rewritten semispinorially
(4.9)

where ( ... ) indicates symmetrization. These relations playa significant role

in Hughston (1987), where they are called 'purity syzygies' for reasons to be
explained in the next section.

5. Application to Pure 8pinors

In this section all spinors are to be chiral if n = 2m is even. For a given
nonzero spinor 'I/J, consider the associated subspaces of V defined by
Vt/; {x= xJLeJLlxtJ. = <pTBItJ.'I/J,<p E 8}
Nt/; V~ = {xlxJLItJ.'I/J = o}.
PAULI-KOFINK IDENTITIES AND PURE SPINORS 59

If N", =I {O}, it is totally null, as follows from 0 = ,(x)2'1j; = g(x,x)'Ij;, and

then N", C N = Vi i1. = V",; thus V", is null. (Recall the terminology: a
subspace W C V is called r-fold isotropic or null iff dim(W n W1.) = r,
totally null iff W C W 1.. Recall further that the maximum dimension of
totally null subspaces is m and that the null cone of (V, g) carries two
~m( m - 1) parameter families of maximal totally null subspaces if n = 2m
and one ~m(m + 1) parameter family of such if n = 2m + 1.)
A spinor 'Ij; is called pure iff N", is maximal, i.e., m-dimensional and thus
V", is of dimension m resp. m + 1 in the even resp. odd case; in the former
case it follows that N", = V",.
Considering first the even case, identity (4.5) shows immediately - be-
cause of V", = N", - that 'Ij; (chiral!) is pure iff 'lj;T BA 'Ij; = 0 for IAI =
m - 4, m - 8, .... The identity goes, however, somewhat beyond this well-
known fact if m > 4, as it yields some 'purity syzygies', i.e., identical re-
lations satisfied by the 'lj;T B A'Ij; which explain the discrepancy in the corre-
spondence between the number of (linearly independent!) purity conditions
just obtained and the dimension of the two families of totally null m-spaces
in V: they are obtained by putting <p = 'Ij; in (4.5), leaving X arbitrary. More
of these syzygies are obtained if we apply the general Pauli-Kofink machin-
ery mentioned above. It seems difficult, in the general case, to give a detailed
count of independent syzygies to make up for the discrepancy mentioned -
more knowledge from invariant theory is required, at least.
Considering the odd case, identity (4.5) immediately shows that the pu-
rity conditions 'lj;T B A'Ij; = 0, IAI = m - 3, m - 4, ... are sufficient to guarantee
dim N", = m, because according to it the spinors <p mapped onto N", = Vi
by M", : <p I--t <pT BI-''Ij; then form a subspace 'lj;1. of S given by <pT B'Ij; = 0,
i.e. having codimension 1; thus

- d'1m S/."1.
1- 'P
-- S/ ker M",
d'lm.',-L/k = d imV", / N",=n-2 d imN",.
'P er M",

Concerning necessity, we were not able, on the odd case, to get along without
resorting to the rank results of Veblen (1955) or the recursive procedure of
Budinich (1989); both of these arguments work, however, equally well with-
out the benefit of our identity, whose main advantage thus seems to be to
provide a short sufficiency proof. (Note that Budinich (1989) aims at provid-
ing simple arguments in the theory of pure spinors, but refers to Chevalley
(1954) for sufficiency - not to Cartan (1966), whose proof contains a point
that remains cryptic to the present author.) Again there is a discrepancy
in the correspondence between the number of (linearly independent) purity
conditions and the dimension of the family of totally null m-spaces in V,
if m > 3, to be explained in terms of syzygies, and we do not yet have a
systematics for that.
60 H.URBANTKE

Acknowledgements
I am indebted to L.P. Hughston and A. Trautman who know everything
that has been said in this paper.

References
Budinich, P. and Trautman, A.: 1988, 'The Spinorial Chessboard', Springer: Berlin.
Budinich, P. and Trautman, A.: 1989, J. Math. Phys. 30, 2125.
Candelas, Ph., Horowitz, G., Strominger, A. and Witten, E.: 1985, Nucl. Phys. B258, 46.
Cartan, E.: 1966, 'The Theory of Spinors', M.l. T. Press: Cambridge.
Case, K.M.: 1955, Phys. Rev. 97, 810.
Chevalley, C.: 1954, 'The Algebraic Theory of Spinors', Columbia U.P.: New York.
Englert, F., Room an , M., and Spindel, Ph.:1983, Phys. Lett. 130B, 50.
Fierz, M.: 1937, Z. Physik 104, 553.
Hughston, L.P. and Shaw, W.T.: 1987, Proc. R. Soc. Lond. A414, 423.
Julia, B.: 1982, 'Marcel Grossmann Meeting on General Relativity', R. Ruffini, editor,
North-Holland: Amsterdam, 79.
Kofink, W.: 1937, Ann. Phys. (Leipzig) 30, 91.
Kofink, W.: 1940, Ann. Phys. (Leipzig), 38, 426.
Pauli, W.: 1935, 'Zeeman-Verhandelingen', M. NijhoJJ: s'Gravenhage.
Penrose, R. and Rindler, W.: 1984, 'Spinors and Space-Time', Vol. 1, Cambridge U.P.:
Cambridge.
Pietschmann, H.V.R.: 1983, 'Formulae and Results in Weak Interactions and Derivations',
Springer: Wien.
Veblen, O. and v. Neumann, J.: 1955, 'Geometry of Complex Domains', The Institute of
Advanced Study: Princeton.
GENERAL COVARIANCE AND SPINORS
LUDWIK D~BROWSKI
SIS SA , Strada Costiera 11,
34014 Trieste, Italy

Spinors play an important role in general relativity. Besides describing

the matter fields, they have been employed for example to classify 'types'
of Weyl tensor, in the E > 0 theorem of Witten and other instances. In
this note we shall discuss the issue of general covariance in gravity with
spinors. This is usually regarded as a statement that Lie derivative of the
Lagrangean L, with respect to arbitrary vector field X, is zero ax L = O.
To check this condition, one does not apparently need the Lie derivative of
spinors. Namely, if L is a scalar (density) function of spinors, one may sub-
stitute the Lie derivative of L with a (fiducial) covariant derivative \7 X of
L and, using the Leibnitz rule, apply \7 X to spinors, which is a well defined
operation. However, similarly to gauge theories, where one insists on the
group action of gauge transformations on the gauge potentials, it should be
equally useful to have the general covariance as a well defined group action
of diffeomorphisms on the configuration space of gravity with spinors. Con-
cerning this point one encounters two quite opposite claims. It is a poular
opinion in the literature on gravity and supergravity, that no problem arises
as spinors transform as 'scalars' uder coordinate changes and as 'spinors'
under (local) vierbein rotations. Instead, differential geometers often claim
that there is no satisfactory action of diffeomorphisms on spinors. Our aim
is to explain why both these opinions are essentially right. In order to do so,
we shall briefly recall:
1. various kinds of spinors
2. 'no-go' theorem for some space-times
3. nonuniquness problem
4. reformulation of the problem in terms of GL(4).
We refer to [D<}browski, 1986] for more details and the list of relevant refer-
ences.
1. Let G be a chosen structure group under which tensors transform, e.g.
G = SOo(3, 1) in the Lorentz space-time. The type of a spinor is specified by
a spinorial (projective) representation R of G in a linear space V. For conve-
nience, this is commonly viewed as a true (nonprojective) representation of
a double covering of G, i.e. of a group G equipped with a 2:1 homomorphism
61
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 61-65.
© 1993 Kluwer Academic Publishers.
62 LUDWIK D4BROWSKI

ponto G

Z2
1
G
p 1 '\.R
G .!!:.... L(V). (1)
If G = SOo(3, 1), G = Spino(3, 1) == SL 2 (C), and typically one choses the
Weyl (chiral) representation in V = C 2 or the Dirac representation in V =
C 4 , etc. These standard data are best defined in terms of Clifford algebras
and are also well known for other dimensions and signatures. However, one
often needs a bigger group G, such as the (Cartesian) product of SOo(3, 1)
times one of the following groups: Z2 X Z2 (reflections), R* (conformal),
U(l) (electric charge), SU(2) (nonabelian gauge), or yet a bigger group. For
a chosen G there may be several possibilities for G and for p, some of them
(but not all) defined by means of Clifford algebra (c.f. [D'}browski, 1986] for
details). Let us mention that G = G L( 4) and its (two) nontrivial coverings,
though not directly applicable as a structure group (demanding spinorial
representations to be finite dimensional), will be employed in the sequel.
In order to pass to manifolds, the above data are usually slightly rein-
terpreted as follows. Let Fg be the space of orthonormal frames E = {Ea}
in R 3 ,1, with respect to some chosen metric g. Let T(E) be the components
of a tensor T in the frame E. They transfrom as T(E) = R(g) T(E'), with
E' = Eg == Ebg't, for 9 E G. Thus one can regard a tensor as an R-equvariant
map, E ~ T(E), from F to the space of components. Analogously, for
spinors, one makes use of an (abstract) space Fg of 'spinor frames', together
with a definite 2:1 (equivariant) assignment 1] between spinor and orthonor-
mal frames (G acts freely and transitively on Fg and 1] intertwines this ac-
tion and the action of G on Fg). Next, spinors are regarded as G-equivariant
functions 'IjJ on Fg

pI
G<-+ (2)
2. The content of 1. can be globalized [Haefliger, 1956],[Milnor, 1963] by
taking in (2) Fg to be the bundle of (space and time oriented) orthonor-
mal frames defined with respect to a metric tensor g. One can also take a
(related) bigger principal G-bundle over the manifold M. Next, Fg is some
principal G-bundle over M, and TJ - an equivariant 2:1 bundle homomor-
phism. The spinor fields of type R are just functions from Fg to V, which
GENERAL COVARIANCE AND SPINORS 63

are equivariant with respect to the (free and transitive on fibers) action of
G. It is well known, though surprising, that such a spin structure (Fg, '1]) not
always exist. For G = SOo(3, 1), besides the orient ability and existence of
Lorentzian metric, a topological condition for a manifold is vanishing of the
second Z2 Stiefel-Whitney class. This is clearly a global problem (locally one
always have spinors), which can be visualised e.g. by a paradox with a one-
parameter family of parallel transports along closed paths in CPl [Geroch,
1970]. For another choice of G, the obstruction is generally weaker and may
completely disappear.
3. Another known 'complication' is that if it exist, spin structure may be
not unique, cf. [Isham, 1978]. Remind that two spin-structures cr;,
rO and
(Fg, '1]) are equivalent iff there exist a bundle isomorphism (3 such that
-,
Fg
f3
-+ Fg
r/ ! ! 1]
id
Fg -+ Fg (3)
commutes. It can be seen that the number of inequivalent spin structures
equals the number of classes in H l (M,Z2) ~ HOM(rrl,Z2).
Now, assuming that two inequivalent spin-structures are isomorphic as
bundles (which is a typical case), we mention some aspects of the inequiva-
lence between (Fg, '1]') and (Fg , 'IJ). The first one concerns the spin connection,
defined as a pull back to Fg of the Levi-Civita connection on M (composed
with the isomorphism of Lie algebras of SOo(3, 1) and Spino(3, 1)). In our
case we have two different spin connections r' =: 1]'*r f:. 1]*r =: r, though
locally they are equivalent just by a Lorentz gauge transformation. This
yields diffrent covariant derivatives and consequently different Dirac oper-
ators. As an alternative, one can perform (locally) a gauge tranformation,
i.e. pass to another gauge eE,1]), such that 1]'(E') = 1](E). Then, the (lo-
cal) expressions for the covariant derivatives coincide and the same holds
for Dirac operators, but the (anti-) posed periodicity conditions (along the
loops in rrt) are different. Altogether, it is clear that inequivalence leads,
in general, to different spectra of Dirac operator, positive eigenspaces and
second quantization.
4. Now we pass to the question of diffeomorphisms. We have seen that
the spinor fields are subordinated to metrics; i.e one first needs g, then a
spin structure and finally the spinors. Therefore, the configuration space W
of spinor fields coupled to gravity has the structure of (infinite dimensional)
vector bundle over the space M X ~, where M is the space of metrics
and ~ is the (discrete) set of spin-structures. The fiber r 9 over g, is the
space of spinor fields defined as above. Now, to implement the action of a
diffeomorphism j, one observes that j transforms metrics by a pull back,
g' = 1* g, and maps r g to r g'. It may also change a spin structure. In order
64 LUDWIK D4BROWSKI

to compare spin-structures related to different metrics, and then to define

the action of diffeomorphisms on I:, sort of a canonical isomorphism between
the bundles of orthonormal frames for different metrics would be completely
sufficient. This may be the case if M has some additional structure (e.g. the
complex structure). Related interesting claims were also presented in [Binz
and Pferschy, 1983], [Bourguignon and Gauduchon, 1992] and [Hennig and
Jadczyk, 1987], but the author is not aware yet of a completely conclusive
proof in the case of Lorentzian signature. Our solution uses a reformulation
of the defini,tion of spin-structure and spinor fields [Dg.browski and Percacci,
1986]. First define a (nontrivial) double covering of GL+( 4) by the following
commutative diagram
Spino(3,1)
pl
SOo(3, 1) '-+ (4)
Associated with (4) there is a reformulated definition of spin structure, i.e.
a double covering of the bundle of all (oriented) linear frames on M.

Fg '-+ F ~ V
TJg 1 1 TJ
Fg '-+ F. (5)
This is really nothing but a reformulation: the existence conditions are iden-
tical (for oriented Lorentzian M) as one easily passes from one definition to
the other, with the help of a metric or by the associated bundle method.
Also the equivalence of spin structures is preserved. Nevertheless it gives us
a possibility to define Fg and Fgl to be equivalent when they originate from
the same F. Note that for a given metric g, spinor fields are functions on
F, which are supported only on F g , and which are equivariant with respect
to the subgroup Spino(3, 1) of GL+( 4). Note also that GL+( 4) is used here
merely as a tool and we don't need its representations.
Now, given a diffeomorphism f and some (reformulated) spin-structure
1
(F, TJ) we ask for a lift of f (and of its derivative T f)
7 F
F' -+
TJ' ! ! TJ
Tf
F -+ F
1 1
f (6)
M -+ M.
It has been shown that such a lift always exists for precisely one (F', TJ'). We
use this fact to define (F', TJ') to be the result of transformation of (F, TJ) by
GENERAL COVARIANCE AND SPINORS 65

f (in fact this defines an action of diffeomorphisms on II, with some nice
properties ).
Finally, we can define the transformation of 'ljJ by f as

'ljJ' = 'ljJ 07. (7)

It is easily seen that 'ljJ' is a spinor field for the metric J*g and that (6)
defines an action of (a double cover) of diffeomorphisms on spinor fields.
The formula (6) has a rather simple meaning. With repect to appropriately
chosen local spin frames E' in F' and E in F, such that E = 7(E'), the
local components of 'ljJ' and 'ljJ are equal; thus we essentially recover the
statement "spinors transform as 'scalars' uder coordinate changes", except
one important detail that it is not sufficient to refer to the usual linear
frames, as is clear from the famous 'sign' ambiguity.
We close with some remarks. The most satisfactory result, the action of
diffeomorphisms on the space of spin-structures, was succesfully applied to
two-dimensional oriented and nonoriented surfaces with or without bound-
ary [DC}browski and Percacci, 1987]. However, some complications arise in
the Lorentzian case for bigger structure groups and also for diffeomorphisms,
which do not preserve the orientation. The reason is the subtlety that GL(4)
contains those coverings of 0(3,1) (and even of SO(3, 1)), which are not
defined in terms of Clifford algebras. This difficulty should be possible to
overcome. A more important issue is that the action of diffeomorphisms on
spinor fields is not a representation in a fixed space of spinor fields. As a
consequence, given a one-parameter subgroup {f} of diffeomorphisms, the
notion of a Lie derivative of a spinor fields is still not defined. It is this fact
behind the statement that there is no satisfactory action of diffeomorphisms
on spinors. A remedy would be a possibility to compare the spaces of spinor
fields at least along paths {J* g} in M. For that purpose, a sort of a (nat-
ural ?) connection on the infinite dimensional vector bundle (configuration
space) W would be helpful.

References
E. Binz and R. Pferschy, 1983, C.R.Math.Rep.Acad.Sci. Canada V269
J-P. Bourguignon and P. Gauduchon, 1992, Commun.Math.Phys.144581
L. D~browski, 1986, Group Actions on Spinors, Naples Lecture Notes Bibliopolis, Naples
L. D~browski and R. Percacci, 1986, Commun.Math.Phys.l06691
L. D~browski and R. Percacci, 1987, J.Math.Phys. 29580
R. Geroch, 1970, J.Math.Phys. 91739
A. Haefliger, 1956, C.R.Acad.Sci. Paris 243558
A. Hennig and A. Jadczyk, 1987, Clausthal T. U. preprint
C.J. Isham, 1978, Proc.Roy.Soc. London 362A383
J. Milnor, 1963, Enseign.Math. 9198
TENSORED DIVISION ALGEBRAS: ORIGIN
OF GEOMETRY, SPINORS AND SYMMETRY
GEOFFREY DIXON
Department of Physics
Brandeis University
Waltham, MA 02254
USA

and

Department of Mathematics
University of Massachusetts
Boston, MA 02125
USA

Abstract. A summary of a decade's work on the tensored division algebras is presented,

emphasizing the remarkably complete and exact correspondence of this mathematics to
the design of our physical reality. The meaning of this correspondence is explored.

1. Playing with Blocks

Clifford Algebras Lie Groups for

Rl,o Internal Symmetries
Ro,l I U{l} ~ SO{2} I
R 2 ,o SO(3)

R 2 ,2 I SU{21 I
I Rl,3 I I 3'U{31 I
Ro,4 SU(4)

I Rl,9'? I Sp(2)

Nature has been too kind to us. Had our mathematics one Clifford algebra
corresponding to a single geometry and acting on a solitary spinor space,
67
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 67-74.
© 1993 Kluwer Academic Publishers.
68 GEOFFREY DIXON

and did the spinors help describe particles that fell into the only multiplets
of a lone internal symmetry, we would have little difficulty explaining our
universe. It is the way it is, we would say, because it can not be other than
it is.
Instead we are awash in Clifford algebras and their spinors, Lie groups
and their multiplets. Out of this surfeit a select few play fundamental roles in
the design of physical reality. Why? Larger, encompassing Clifford algebras,
groups, or supersymmetries and string theories, offer the hope of combining
several unexplained features under a single unexplained banner. But our
only hope presently of achieving such a unification is to stumble upon it.
Nature presumably did not stumble upon Her design. It is my belief that
what we see of this design is merely the surface, that beneath the surface
there is a Truth that functions in the absence of our attempts to perceive
it, in the absence of us altogether. Within this Truth are the conditions
underlying viable reality and existence, and they are stringent enough to
exclude any type of existence other than our own. Flatland is not viable.
With mathematical physics we have gained some understanding, but
mathematics and physics are still viewed as different sciences. I believe that
to go further we must develop a new instrument to help us see below the
surface, a hypermathematics, which would bring mathematics and physics
to a common focus. Within this there would be a hypervariational princi-
ple with which we might explain the selectness of our physical reality. At
the same time, I believe that Truth, and our symbolic representation of it,
however encompassing, are distinct.
It is not my intention here to develop this hypermathematics, merely to
open a door onto a path that may eventually lead in the right direction.

Division Algebras

As indicated above, there are only four real normed division algebras, the
reals, the complexes, the quaternions, and the octonions. Three are already
employed in physics. The reals give us the results of measurements, quantum
mechanics is done over the complex field, and the Pauli algebra, P, without
which we could not describe fermions, is isomorphic to R ® C ® Q. On
the mathematical end, in Lie group theory, the sequences of orthogonal,
unitary, symplectic, and special classical Lie groups arise, respectively, from
the algebras R, C, Q, and O. In topology the parallelizable spheres, Sl, S3,
TENSORED DIVISION ALGEBRAS 69

and 57 arise from C, Q, and

arises from R).
° (SO, the null sphere, trivially parallelizable,

The sequence of division algebras is unique, finite, and generative in both

mathematics and physics. Wouldn't it be great if it were in fact an integral
part of the design of our universe. I believe that it is. I have demonstrated
that it is [1], which tends to confirm me in my belief.
This belief has been shared by many other physicists and mathematicians.

or °
Many attempts have been made to make a quantum mechanics based on Q
instead of C. Because C is a subalgebra of Q, and Q is a subalgebra
of 0, it seems superficially reasonable that in making the jump to Q or
0, we may discard the smaller algebras, which are in any case present as
subalgebras. In my opinion this is an example of confusing Truth with the
symbolic tools we use to explore Truth. The definition of an algebra did not
miraculously appear on a stone tablet; we made it up. And Nature is not
obliged to order reality to fit our prejudices.
The sum of all the properties associated with each of the division algebras
exceeds those included in the definition of an algebra. Let KL and KR be
the adjoint algebras of left and right actions of the algebra K on itself, K =
R, C, Q, or 0. Then, for example,

CL and CR are the same algebra, both isomorphic to C,

QL and QR are distinct algebras, both isomorphic to Q,
0L and OR are the same algebra, both isomorphic to R(8),

where R(8) is the algebra of 8 x 8 real matrices. (I have found the natural oc-
tonion products more useful than those based on defining ° as an extension
of Q. In particular, given a basis e a , a=1, ... ,7, for the hypercomplex part of
0, and given that distinct ea anticommute, that ea(eaeb) = (ebea)ea = -eb,
that e~ = -1, then the rest of the multiplication table is determined by
the equation e a e a+1 = e a+5, indices modulo 7, from 1 to 7. The ease of
this multiplication is enhanced by the following property: if eaeb = e c then
e(2a)e(2b) = e(2c). So, ele2 = e6 yields, via this index doubling property,
e2 e4 = e5.) My point, if it is not already plain, is that the algebraic inclusion
property is irrelevant. These algebras stand on their own. I am suggesting
that the physical relevance table for division algebras should look like this:
70 GEOFFREY DIXON

Division Algebras

That is, the algebra

H=R®C®Q®O
must form part of the algebraic design of reality. I assume it underlies the
Clifford algebra/spinor, Lie group/multiplet structure of our universe, and
with physical reality as a guide (as little as possible), I shall outline the way
in which this may be seen to be true.

2. Playing with H
The individual division algebras may be associated with Clifford algebras in
the following ways:
CL ~ R o,!,
QL ~ R O,2,
0L ~ R O,6.
(Note that C, Q and ° are 2k -dimensional, k=1,2,3, and their left adjoint
algebras are isomorphic to the Clifford algebras of spaces of dimension k!.
This may be accidental, but there is a more natural way of extending the
sequence than the Cayley-Dickson prescription, and I suggest further work
in that direction to resolve this interesting dimensionality question.)
The spinor spaces of these Clifford algebras are just C, Q and 0, the
object spaces of CL, QL and 0L. Of the three only QL does not act effec-
tively. QR provides an internal degree of freedom. The group of elements of
QR of unit norm is SU(2).
Two physically relevant isomorphisms are
(1)
and
(2)
this last being the complexification ofthe Clifford algebra of (1,3)-Minkowski
space. This is the Dirac algebra. The spinors OfPL(2), namely p2 (the space
of 2xl matrices over P) are more complicated than those of C( 4) in having
TENSORED DIVISION ALGEBRAS 71

the QR internal freedom. Each PL(2) spinor contains a pair of ordinary

Dirac spinors.
The isomorphisms (1) are just mathematics, but those in (2) are arrived
at only by following the lead of physics, which we know rests on the Dirac
algebra. I will follow this lead a bit further with the following isomorphisms:

(3)

and

(4)

There are, of course, indications in other branches of theoretical physics

that Rl,9 is a physically relevant geometry. Because the complexification
of its Clifford algebra is to H what the Dirac algebra is to P, I believe it
necessarily is, although the role of the extra six dimensions is not yet clear.
Step back to just H and its adjoints for a moment. H is not a division
algebra (it is not even alternative), and its identity admits nontrivial reso-
lutions into four orthogonal primitive idempotents. There are at least three
(I suspect no more) inequivalent resolutions of this identity, but only one
(into elements ~m,m = 0,1,2,3,3 ~m~n = 8mn~m (orthogonality), and
~o + ~l + ~2 + ~3 = 1) which satisfies the associativity conditions

~m(X ~n) = (~mX)~n'

~m(~nX) = (~m~n)X, (5)
(X ~m)~n = X(~m~n)
for all X E H.
These associativity conditions allow us to consistently define the compo-
nents ~mX ~n of elements X E H with respect to the ~m' Of particular
interest are the diagonal components

(6)

where Xm E C. (That is, in a mathematical sense, with respect to the

resolution {~m}' H is a complex algeb~a. Therefore, if H is indeed part of
the design of reality, it is not surprising that physics is inherently complex
as well.)
It is possible to design a group of H-automorphisms Mm, m = 0,1,2,3,:3
Mm(~o) = ~m (Mo = identity map). With their help, and the help of
Hermitian conjugation (X ---+ xt), we can define the real part of the trace
of X E H:
72 GEOFFREY DIXON

(note: ~tn ~m). Likewise, if X, Y E H, we may construct their inner

product:
1
< X, Y >= 8" 2: ([Mn«X ~m)t(y ~m))] + [same]t}. (7)
mn

(= real part of xty or yt X).

The ~m in (7) may be replaced by more general rm satisfying
r~r n = {jmn~n (no sum), (8)
and an associativity relation like (5),
r~(Xr n) = (r~X)r n (9)
for all X E H. The symmetry of (8) and (9) is somewhat complicated, but
on each of the r m it boils down to
U(2) X U(3) (10)
(in fact, they are each U(3)-invariant, but not U(2) invariant, and that SU(2)
is chiral, SU(3) nonchiral, is eventually seen to result from this U(2)jU(3)
distinction) .
I now let physics take the lead and define 'l/J E H which transforms with
respect to the symmetry (10) like r6
(any other rtn
would have done as
well). Since 'l/J is a general element of H, and SU(3) is a subgroup of the
invariance group of 0 (ie., G2 ), 'l/J is not U(3) invariant. With respect to
SU(3) it transforms as
1 EB 3 EB I EB 3. (11)
In fact, with respect to the symmetry (10), 'l/J transforms exactly like a
family plus antifamily, including all correct quantum numbers, of lefthanded
leptoquark Weyls spinors (ie., the individual invariant vector components are
2-dimensional over C).
The ~m can be expressed as follows:
~o = A+P+,
~1 = A_P+, (12)
~2 = A+P_,
~3 = A_P_,

where A± are SU(2) eigenvector projectors, and P± are SU(3) multiplet pro-
jectors (the fact that the ~m break 'l/J down to the vector level with respect
to SU(2), and to the multiplet level with respect to SU(3), is eventually
seen to be the explanation for why SU(2) breaks and SU(3) is exact). In
particular, p+ 'l/J transforms under SU(3) like 1 EB 3, and p_ 'l/J like i EB 3. So
the operation of P± from the left projects the matter jantimatter half of 'l/J.
TENSORED DNISION ALGEBRAS 73

What about righthanded leptoquark Weyl spinors? The symmetry (10)

is carried by the Mm from ro to the other r m' In fact, since the r mare U(3)
invariant, only U(2) is nontrivially carried. Therefore each of the elements
p+"pr m have the same U(3) charges as p+ "ptl m , but altered U(2) charges
(in fact, they are SU(2) singlets). They have precisely the charges observed
on righthanded leptoquark spinors (p+ from the left assures us that this is
matter; the antimatter half is treated similarly).
The particle assignments are

lefthanded righthanded assignment

p+ "ptl o p+"pr o neutrino
p+ "ptl l p+"pr l (-1 )-lepton (13)
p+ "ptl z p+"pr z (2/3)-quark
p+ "ptl3 p+"pr 3 (-1/3)-quark

Left- and righthanded assignments can be simultaneously manifested in the

larger context of HL(2) and its spinor space H2. The details are developed
elsewhere [1].
There is more one can derive from the mathematics, including a natural
weak mixing and spontaneous symmetry breaking, but I want to conclude
this section with a look beyond the Standard Model. Most ofthe early devel-
opment was based on HZ-fields "iII considered functions of (1,3)-Minkowski
space. I recently looked at what would happen if this were expanded to a
functional dependence on (1,9)-space. The extra six dimensions carry SU(3)
charges. I have demonstrated that inorder for the (1,9)-Dirac operator to
reduce to the (1,3)-Dirac operator, the lepton fields above must be indepen-
dent of the extra six dimensions, and the quark field may depend only upon
parameters carrying SU(3) charges parallel to those carried by the fields
themselves. The reason this is essential is phenomenological. With respect
to (1,9)-Minkowski space, matter and antimatter are indistinguishable, and
a Lagrangian of the form

C l ,9 =< "iII, 9'1,9"iI1 >

=< p+"iII, 9'1,3p+"iII > + < p_"iIl, 9'1,3P- W > (14)
+ < p+"iII, 9'O,6p-"iII > + < p- W, 9'O,6P+ W >,
where the (1,9)-Dirac operator !h,9 is constructed on HL(2), gives rise to
un mediated (or space mediated) matter into antimatter transitions via the
last two terms above. Such transitions have not been observed. They dis-
appear from the mathematics if the lepto-quark functional dependencies on
the extra six dimensions are arranged as outlined above (ie., the last two
terms in (14) are identically zero in this case).
74 GEOFFREY DIXON

3. If the Shoe Fits, Wear It

Division Algebras
IRICIQIOI II

es

For some reason Nature sprouts only certain geometries and symmetries
and multiplets and no others. A finite number are chosen, infinitely many
rejected. There must exist a kernel - a seed - which guides Nature in its
very select development. H does this, simply, elegantly, and at a mathemat-
ical level below geometry and symmetry, using algebraic objects of great
and generative importance to many branches of our mathematics. As I see
it, while it is necessary now to use physics to guide the researcher in the
development of this mathematical application, ultimately other generative
mathematical ideas would be included, and physics could then be derived
from pure mathematics, a mathematics in many respects unlike any we have
yet seen. This would take us down a little deeper, a little closer to whatever
ultimate Truths generate reality. Quite frankly I do not imagine these truths
are totally accessible to us. But no matter, this just means there'll always
be records to break. In the meantime, whether it be based on strings or
twistors or ought else, no theory not based on H has a chance of succeeding.
Nature is the way it is, I would say, because it can not be other than it is.

References

[1] In a series of papers -

Geoffrey Dixon: II Nuovo Cim 105B 349(1990).
Geoffrey Dixon: Phys. Lett. 152B 343(1985).
Geoffrey Dixon: Phys. Rev. D 29 1276(1984).
Geoffrey Dixon: Phys. Rev. D 28 833(1983).
- I developed most of the features outlined in this paper. However, as these
ideas evolved some notions supplanted others (in particular, there is no
mention of adjoint algebras in the earlier papers). I am presently at work
on a book in which a decade of mathematical and physical research will be
presented as a unified and coherent whole.
G - STRUCTURE FOR HYPERMANIFOLD

ANDRZEJ BOROWIEC
Institute of Theoretical Physics
University of Wroclaw
pI. Maxa Borna 9, 50-204 Wroclaw,
Poland, e-mail: [email protected]

Abstract. This paper extends the earlier version (Borowiec 1989). It points out the fact
that Riemannian geometry plays a very exceptional role among geometries represented
by a symmetric tensor 9JLl ... JLn of degree n. In particular, we question the claim made by
Holm (1986) that there exist torsion free Christoffel coefficients on a hyperspin manifold.
we develope a G-structure formalism for hypermanifolds.

Key words: G-structure, tensor concomitant, connection, chronometric, hypermanifold

1. Introduction

Recently David Finkelstein (1986) proposed a new idea in physics - a hy-

perspin theory. While Riemannian geometry is based on a second-rank
symmetric metric tensor g/-LV, he suggested replacing it by an n- symmet-
ric form g/-Ll ... /-Ln called a chronometric. A manifold M equipped with such
O-deformable tensor will be called a hypermanifold. The name hyperspin
manifold was reserved for a particular kind of hypermanifold. C. Holm (1986)
discussed the geometry of hyperspin manifolds in a complete analogy with
the metric geometry. Our purpose is to explain why some important con-
cepts of Riemannian geometry work differently in the chronometric case. In
particular, we will criticize some oversimplified statements formulated by C.
Holm.
We will confine ourselves to the case of n = 3 (i.e. a cubic metric). A
generalization from 3 to n is evident. A flat space with a cubic metric and
the construction of the classical and quantum mechanics of a particle moving
in it has been recently considered by Yamaleev (1989).
Let TJ = (TJijk) be a symmetric 3-rd rank covariant tensor on a linear
space V = JRN, i.e. TJ E V* 8 V* 8 V* where 8 denotes a symmetric tensor
product. Assume that TJ is nondegenerate in the following sense: the mapping
(also denoted by TJ)

(1)

* Dedicated to Prof. Jan Rzewuski

Supported by KBN under Grant PB # 2 2419 03
75
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 75-79.
© 1993 Kluwer Academic Publishers.
76 ANDRZEJ BOROWIEC

is injective (we use the Einstein summation convention). The Riemannian

case is an exception because (1) becomes an isomorphism of V onto V*.
Decomposing V* 0 V* = L" EEl L' into a direct sum, where L" = 1m T/,
one can define (depending of the choice of L') the linear mapping called an
inverse or dual chronometric ii : V* 0 V* --+ V, by

iilL' = 0 -I
T/L'=T/
-1
(2)
By definition ii = (T/ijk) is any solution of the eq.

(3)

Notice that in general T/ki j will be symmetric only in the last two indices.
Even if there exists a completely symmetric solution of (3) it is nonunique.
In the following we will fix some dual tensor ii.
It is convenient to introduce the tensor .x~ = T/mijT/mkl. Then (3) guar-
antees the following properties of .x : V* 0 V* - - V* 0 V*

.xijkl .xmn
kl - .xij
- mn
(4)
\ ij ",mkl _ ",mij \ ij '" .. _ '"
/lkl'l - '1 , /lkl'lm~] - 'Imkl

This means that .x is a projection operator on a subspace L" along the

direction L". Immediately one obtains that the conditions

(5)

as sufficient and necessary for an element (Vi]) E V* 0 V* to be a value of

some vector (v m = T/mi j vij ) E V under the action of (1).

2. Tensorial G-structures
Let LM denote the frame bundle over an N-dimensional paracompact man-
ifold M. LM is a principal GL(lR,N)-bundle. Consider a tensor field t E
v
r (Tf M) of type (p, q) on M. Let t: LM --+ Tf (lRN) denote the corre-
sponding equivariant mapping (see e.g. Gancarzewicz 1987). A tensor t is
v
called O-deformable if GL(IR, N) acts transitively on t(LM) C Tf(IR N ).
v .
Let T E t(LM) and G(T) C GL(lR, N) be the Isotropy subgroup of T. 0-
deformability of t is equivalent to the existence of a subbundle P( T) of LM
on which t is constant and equal to T. T is called a canonical form of t. P( T)
v
is a G(T)-structure on M. If T' is an another element in t (LM) then P(T)
and P( T') are isomorphic. Notice that there is no canonical isomorphism
between P(T) and P(T').
G - STRUCTURE FOR HYPERMANIFOLD 77

A tensor field s E r (T;' M) is called a concomitant of i E r (T: M) if

there exists a GL(JR, N)-equivariant map f :t (LM) -----* ~(LM) such that
~ = fo t. By definition each equivariant map produces some concomitant.
If i is O-deformable and s is a concomitant of i then ~ is constant on P( T)
and G(T) C G(f(T)) (Zajtz 1985).
A given linear connection r on M is said to be i-connection if \1i = O.
The existence of a t-connection is equivalent to the O-deformability of i.
A G(t)-structure P( T) is said to be integrable if every point x E M has a
coordinate chart (xJ.L) such that the (local) cross section (8/ 8xl, ... ,8/ 8x N )
of LM is a (local) cross section of P( T). An integrable G-structure is locally
flat and it admits a torsion free connection (Kobayashi 1972).
Kobayashi and Nagano (1965) have found all subgroup G of GL(JR, N)
which satisfy the following condition: every G-structure P on a manifold M
admits a torsionfree connection. It should be noticed that the group G(T))
does not belong to this class, where T) is taken from the previous Section.

Example.
Each (pseudo- )-Riemannian metric on manifold M is O-deformable. The
metric 9 = (gJ.Lv) and its inverse g = (gJ.LV) are mutually concomitant to each
other. The Riemannian connection is a unique torsionfree g-connection on
M.

3. Hypermanifolds

Let M be an N-dimensional manifold with given G(T))-structure, where

7]= V* 8 V* 8 V*. Let 9 = (gJ.LVA) denote the corresponding tensor field
on M. According to the hyperspin philosophy one has to assume that there
exists a tensor concomitant g = (gJ.LVA) of 9 such that

(6)

The resulting structure will be called a hypermanifold. It is also convenient

to introduce a tensor concomitant h by

(7)

The algebraic relations (3), (4) and (5) remain valid for tensor fields g, g,
and h.
Let r J.L~ be the connection coefficients for some g-connection on M. From
\1g = 0 one gets
(8)
 78 ANDRZEJ BOROWIEC

where r J.LV>'p = r J.L~go:>.p. Holm (1986) has found a "unique tor",ionfree" solu-
tion of (8) of the form

1 1
DJ.Lv>,p = "3 (0J.L9v>.p + Ov9J.L>'p) - 6" (Op9>.J.Lv + O>.9pJ.Lv) (9)

Indeed, (9) is an algebraic solution of (8) and is symmetric in the first pair
of indices. Unfortunately the DJ.Lv>,p do not transform into themselves under
the gauge transformations, so they depend on the choice of coordinates. In
particular (9) does not yield a global object. The next remark is that the
solutions (9) are not connection coefficients at all. To see this let us observe
that DJ.Lv>,p does not satisfy the constraints (5), i.e.

(10)

and hence there do not exist coefficients r J.L~ such that DJ.Lvo:(3 = r J.L~9>.o:(3.
For a paracompact M, a 9-connection always exists, but the problem of
finding of a canonical 9-connection of M remains open. If it even exists it
will be in general with torsion.

4. Particle Trajectories
It follows from the above considerations that one cannot use of the geodesic
principle of general relativity in order to obtain particle trajectories in a
space-time M with a geometry represented by the tensor field 9.
Let 89 = 9' - 9 b e a symmetric tensor of type (0,
for matter regularly concentrated on a curve K c M can be written in the

J
form
(aJ.Lv>'89J.Lv>,) ds = 0 (11)
K

where the xJ.L's are a coordinate system on M and K respectively, and aJ.Lv>,
denotes a density on K with values in symmetric tensors of type (0,3). 89J.Lv>,
is of the form LY9J.Lv>, with Y being any vector field on M. Then after some
long calculations and using the methods developed by Jadczyk (1983), one
finds
d (J.L v ) _ 1 J.L v~ _ (12)
-d v v 9J.Lv>' - --v v u>.9J.Lvp - 0
s 3
where vJ.L = d::
are velocity components. It is remarkable that the same
equations can be derived from the variational principle for the functional
b
J \!9J.Lv>,vJ.Lv v v>' ds.
a
As an accidental result we should noticed that (12) turns out to be
"geodesic motion" for Holm's connection.
G - STRUCTURE FOR HYPERMANIFOLD 79

Acknow ledgment
I am grateful to A. Jadczyk for valuable suggestions.

Note added
After completing this work, the author has become aware of an interesting
paper by Urbantke (1989) in which the historical remarks on the "space
problem of Weyl" are also contained.

References
Borowiec, A., (1989) Int. Jour. Theor. Phys. 28 (10) 1229
Finkelstein, D., (1986), Phys. Rev. Lett. 56 (15) 1532
Finkelstein, D., Finkelstein, S.R., Holm, C. (1986), Int. Journ. Theor. Phys. 25 (4) 441
Holm, C. (1986) Int. Journ. Theor. Phys. 25 (ll) 1209
Jadczyk, A. (1983) Annales de L'Institut Henri Poincare - Section A, XXXVIII (2) 99
Gancarzewicz, J. (1987) "Differential Geometry" PWN, Warsaw (in Polish)
Kobayashi, S. (1972), "Transformation Groups in Differential Geometry", Springer - Ver-
lag, Berlin
Kobayashi, S., Nagano, T., (1965) J. Math. Soc. Japan, 17 (1) 84
Souriau, J.M. (1974), Annales de l'Institute Henri Poincare - Section A, vol. XX, 967
Yamaleev, R.S. (1989), "On Geometric Form in Three Dimensional Space with Cubic
Metric", Communications, JINR P5 - 89 - 269
Yamaleev, R.S. (1989), JINR Rapid Communications, 1 (34) 50
Urbantke, H., (1989), Int. Journ. Theor. Phys. 28 (10) 1233
Zajtz, A. (1985), Colloquium Mathematicum XLIX (2) 232
TOWARDS A UNIFICATION OF
"EVERYTHING" WITH GRAVITY
JERZY RAYSKI
Institute of Physics
Jagellonian University, Krakow, Poland

Abstract. Combining the ideas of gauge interactions with a global supersymmetry a uni-
fied model in six dimensions is built up step by step starting with a single generation of
leptons and ending with three generations of leptons and coloured quarks forming a super-
multiplet characterized by a most general extension N = 8. The puzzle of supersymmetric
partners like gravitino, photino, s-leptons and s-quarks is seen in a new light.

Let us begin with a simple model of one generation of leptons interact-

ing electro-weakly and gravitionally. Consider a set of fields with the same
number of fermionic and bosonic degrees of freedom consisting of one tensor
field, two Rarita-Schwinger fields, four vector fields, six Weyl fields and six
scalars. All these fields are at first massless two-component fields except for
one-component scalars. This set may be split either into (N = 1)- super-
multiplets or into extended (N = 2)-supermultiplets ( see the tables 1 and
2 where the rows denote fields with spins 2,3/2,1,1/2 and 0 ).
table 1

m
1

[~
1 1
+ 1 +3x 1
1
+3x

table 2

m
1 1
2 2
4 1
6
+3x
6

81
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 81-89.
© 1993 Kluwer Academic Publishers.
82 JERZY RAYSKI

The first column to the right of the table 1 is interpretable as graviton

and gravitino. The second represents photon and photino whose spin is un-
expectedly not smaller but higher by 1/2 from unity. The triplet appearing
in the third column denotes three vector fields W± and ZO intermediating
weak interactions and is combined with a triplet of their supersymmetric
partners with spins 1/2 to be called W-ino and Z-ino. A triplet of Weyl
fields appearing in the last column may be interpreted as a triplet of weakly
interacting leptons, possibly eR, eL and vI. Two of them viz. eR, eL may
be fused into a 4-component Dirac field but the third Weyl field has no
partner with an oppposite helicity and, consequently, exhibits a chiral char-
acter of the weak interactions. Also the number of Z-ino is unity, i.e. is an
odd number and, together with ZO must violate parity conservation of weak
coupling.
The next problem is that of spontaneous symmetry breaking. Three of six
scalars appearing in the tables 1 and 2 should be swallowed up by the triplet
of vector fields appearing in the third row of table 1 endowing them with big
masses in agreement with experimental evidence. It confirms our previous
guess that these fields are W± and Zoo Similarly two of the Weyl fields, viz.
W- ino's have also to be swallowed up by the two spin 3/2 fields endowing
them with very big masses which fact explains why corresponding particles
could not be observed yet. The table 3 shows the results of spontaneous
symmetry breakdown.
table 3

where "prime" denotes "heavy". It is seen that besides a triplet ofleptons:

electron and left-handed neutrino one more Weyl field has been playing a
twofold role: it is a Z-ino which at the same time may be regarded as a right-
handed neutrino. It is excluded from weak interactions but it is involved into
a supersymmetric interaction within a local super-doublet with two scalars.
According to the presence of a triplet and a singlet of vector fields, the
symmetry of gauge interactions is G = U(1) x SU(2). The above sketched
model of a lepton generation appears much simpler and intelligible if looked
upon from a six-dimensional viewpoint. Let us assume that spacetime is
six-dimensional with topology of M4 x S2 or AdS X S2 where S2 is a two-
dimensional spherical surface. Its radius is assumed to be extremely small.
TOWARDS A UNIFICATION OF "EVERYTHING" WITH GRAVITY 83

The metric field is

(1)

with M,N = 0,1, ... ,5; J.L,v = 0, ... ,3; and e,7] = 4,5. The mixed metric
tensor components 91-'e appearing as if components of two fourvectors assume
the form [4]
3
9/1-e = LA~J(~ (2)
all

where J(~ are the Killing vectors of a sphere. The fields A~ are to be identi-
fied just with Wi and Z~ exhibiting the symmetry of a sphere being a fun-
damental representation of SU(2). From the point of view of Minkowskian
observer the components 9('1) look like scalars.
Three fourvectors A~ have been incorporated intrinsically into the six-
dimensional metric field 9MN in agreement with Kaluza's original assump-
tion however the electromagnetic field is not interpretable as a constituent
of generalized metric but denotes the first four components of a six-vector:

(3)

It follows that the common view that all vector fields have a metrical
origin may be regarded as a prejudice. The extra components of the six-
vector V, form something like a "tail" and look like scalars for macroscopic
observers. They may be identified with two apparent scalars appearing in
the last row of the table 3 so that the number of genuine scalars in our
scheme reduces to a singlet a Goldstone scalar.
Inasmuch as the number of independent components of the generalized
metric field is eleven ( two 9/1-1/, six A~ and three 9('1) ) while that of massIes
Rarita-Schwinger field components in D = 6 is twelve whereas the numbers
of components of massIes vector fields as well as of Weyl spinors in D = 6 is
four the table 1 may be rewritten and simplified enormously, see the table 4
table 4

which justifies ex post our initial choice of the multiplet appearing to the
left hand sides of the tables 1, ... ,4.
Three generations of leptons.
84 JERZY RAYSKI

In order to perform a transition to a triplet of leptonic generations we

proceed as follows: Assume a (reducible) supermultiplet involving 1 tensor
field, 4 Rarita-Schwinger spinors, 12 vector fields, 24 Weyl spinors and 30
scalars. This supermultiplet consists of 56 fermionic and 56 bosonic degrees
of freedom and splits into irreducible supermultiplets characterized by the
following indices of extension: once N = 4, six times N = 2 and eight times
N = 1 as seen from the table 5

table 5

Applying Higgs mechanism for spontaneous symmetry breaking eleven

from twelve vector fields acquire considerable masses by swallowing up eleven
scalar fields. Similarly four Rarita-Schwinger fields become very heavy too

l
by swallowing up four of the Weyl spinors as is to be seen from the table 6
table 6

11 ;;b 1
4 4'
12 1 + 3' + 8'
24 12 + 8
30 18 + 1 D=6

According to the appearance of a set of 1 + 3' + 8' vector fields in the

third row of the middle column of the table 6 it is seen that the gauge group
is
GI = U(I) X SU(2) X SU(3) (4)
due to the fact that the octet of heavy vector fields means a fundamental
representation of the SU(3) group. The particles forming this octet may be
called para-gluons. They must possess a considerable mass value in order to
prevent a quick decay of higher into lower generations.
The adequacy of the SU(3) symmetry group is confirmed also by a dis-
cussion of the set of fields appearing in the fourth column of the table 6.
Their number 24 splits naturally into 12 + 8 + 4 wherefrom four have to be
swallowed up by the Rarita-Schwinger fields, further eight are also related
to Rarita-Schwinger fields inasmuch as they form their "tails" if going over
to a six-dimensional description so that finally we are left with only 12 two-
component (or six fourcomponent) Weyl spinors in four (or six) dimensions.
These numbers are factorizable by the factor 3, viz. 12 = 2 X 2 x 3 which
TOWARDS A UNIFICATION OF "EVERYTHING" WITH GRAVITY 85

means that the corresponding spinor fields form triplets. Thus the number 12
of Weyl spinors denotes nothing else but three generations of leptons. They
include also right-handed neutrina although the latter do not participate in
weak interactions.
The 18 scalars appearing in the last row of the mirldle column of the table
6 mean "tails", i.e. additional components of the 8 + 1 six-vectors so that
finally we are left with only one single genuine scalar a Goldstone boson if
regarding and interpreting the multiplet from a 6-dimensional viewpoint.
Three generations of coloured quarks
Let us consider now a supermultiplet consisting of 96 bosonic and 96
fermionic degrees of freedom, viz. 1 tensor, 6 Rarita- Schwinger spinors (spin
3/2), 20 vectors, 42 Weyl spinors and 54 scalars. It forms a (reducible)
supermultiplet splitting into the following irreducible constituents (see table
7)
table 7

1
6
20
42
54

The irreducible constituents are: a single (N = 6)-extension and a quartet

of (N = 4)-extensions with the highest spin values 2 and 1 respectively.
The number of vector fields splits as follows: 20 = 1 + 3 + 2 X 8 and is
compatible with a symmetry group of gauge interactions

Gqg = U(I) X SU(2)L X SU(3)g x SU(3)c (5)

where one of the two octets is related to the symmetry group of three gener-
ations, the other with the group of colour. In order to prevent a quick decay
of higher into lower generations the para-gluons must be massive which may
be achieved by a (generalized) Higgs mechanism. Three vector fields repre-
senting the group SU(2)L together with eight representing the group SU(3)g
have to swallow up eleven scalars while the six spin 3/2 fermions must swal-
low up six Weyl spinors endowing the respective particles with high masses
(see the table 8)
86 JERZY RAYSKI

table 8

[ 1+ ~~+ 8,1
8+1 D=6

From the fourth row of the middle column it is seen that the number
of Weyl fields is 36 interpretable as three generations of coloured quarks
according to the splitting 36 = 2 X 2 X 3 X 3 where 2 X 2 denotes a doublet
of helicities times a doublet of charm or flavour whereas 3 X 3 denotes a
triplet of generations times a triplet of colour. It should be stressed that the
octet of lepton and quark generations must be different one from the other
in order to prevent decay of quarks and hadrons into leptons. In order to
account for the masses ofleptons the SU(3) symmetries of generations must
be (slightly) broken but the mechanisms of these breakdowns are not yet
clear.
The problem of a N = 8 extension
It is suggestive to assume that the set of all fundamental fields and particle
types existing in Nature should form the most general, irreducible super-
multiplet characterized by the index of extension N = 8. This supermultiplet
includes 1 tensor, 8 Rarita-Schwinger fields, 28 vectors, 56 Weyl fields and
70 scalars. In view of the splitting 28 = 1 + 3 + 8 X 3 it could be supposed
that the symmetry group of gauge interactions were
G = U(l) X SU(2) X SU(3) X SU(3) X SU(3) (6)
of rank 8. However, it seems impossible to "put quarks and leptons into one
basket". Inasmuch as strong gauge interactions are not universal and quarks
interact with leptons only via universal gravitational and electro-weak cou-
plings we assume that the Lagrangian splits into a leptonic and quarkonic
parts interacting only via subgroups (4) and (5) of the most general possible
group (6) of gauge interactions of ranks 4 and 6 respectively. The generality
of the scheme will be preserved only in so far that all fields appearing within
the (N = 8)-multiplet partake either in the leptonic or quarkonic parts of
the Lagrangian
(7)
The lepton and quark parts £/ and £q involve the interaction terms with
bosons whereas the bosonic part £b denotes a sum of interaction-free bosonic
fields. The gauge groups of interactions with leptons and quarks in £/ and
£q are the groups (4) and (5) respectively. If all interactions are of a gauge
type and gravitational ones and of Yukawa-Higgs type then writing down a
Lagrangian (7) is rather a matter of standard techniques.
TOWARDS A UNIFICATION OF "EVERYTHING" WITH GRAVITY 87

Assume that the triplet W± ZO as well as both para-gluonic octets become

massive as visualized by the table 9
table 9

[~]
The large masses of the two octets of para-gluons prevent a possibility
of a quick decay of higher into lower generations. The appearance of two
different octets of para-gluons for leptons and quarks assures a lack of decay
of hadrons into leptons.
The reduction of the number of Weyl spinors from 56 to 48 is just suf-
ficient and necessary to interpret the remainig 48 as three generations of
leptons and quarks. To see this let us perform the following splitting

48 = 12 + 36 = 2 X 2 x 3 + 2 x 2 x 3 X 3

where 12 denotes the number of leptons and 36 that of quarks. As before

2 X 2 means two helicities times a doublet of charm (or flavour). One of the
triplets accounts for the three generations of leptons or quarks while the
second triplet accounts for the colour of quarks. Nevertheless, the multiplet
N = 8 is not simply a sum of formerly discussed lepton and quark multiplets
because of different roles of universal electro-weakly-gravitational and spe-
cific SU(3) interactions. Adding simply the schemes would mean doubling
the gravitational field and the number of fields representing the W±, ZO
bosons. Instead, we may perform a decomposition according to the table 10
where the first column to the right hand side of this table
88 JERZY RAYSKI

table 10

2
8
12 + 2
16

denotes the leptonic sector, the third describes the quarkonic sector while
the middle column denotes their common part, i.e. the terms joining leptonic
with quarkonic worlds via universal gravitational, electro-weak and Yukawa-
like interactions.
The additional numbers 2 and 6 of Weyl spinors from the fourth rows
of the table 10 are to be swallowed up by the corresponding spin-3/2 fields
endowing them with masses by means of a mechanism of spontaneous sym-
metry breaking. The numbers of scalars are explicable if reinterpreting the
supermultiplet from a six-dimensional viewpoint. The 16 and 32 scalars ap-
pearing in the first and third column are the "tails" of the corresponding
six-vectors whereas the number of 22 = 1 + 3 + 2 + 16 scalars from the mid-
dle row denotes respectively a Goldstone boson, a triplet of metric tensor
components 9f,TJ' the "tail" of the six-vector of electromagnetic potentials,
and a set of further 16 apparent scalars which will be swallowed up by the
two octets of para-gluons endowing them with masses.
If reinterpreted from a four-dimensional to a six-dimensional viewpoint
the (N 8)-dimensional extended supermultiplet assumes the following
form
table 11

D=6

(still prior to the spontaneous symmetry breakdown).

Let us notice the following remarkable circumstance: the numbers of Weyl
spinors viewed from a six-dimensional perspective are odd, viz. 5 and 15 in
the fourth row of the table 11 if decomposed into leptonic and quarkonic
sectors, i.e. into singlets and triplets. This explains why parity conservation
must be violated by weak interactions. Weyl spinors cannot be fused into
Dirac spinors in D = 6.
Concluding remarks
Our tables may be regarded as analogues of the Mendelejev table of
chemical elements, but this time applied to elementary particles. Similarly
TOWARDS A UNIFICATION OF "EVERYTHING" WITH GRAVITY 89

as the original table of Mendelejev exhibited some signs of periodicity (and

therefore was called "periodic table of elements") also here we notice signs
of periodicity since the rows denote bosonic and fermionic fields alternately.
Moreoveer, similarly as in the case of Mendelejev tables there appear also
here empty places to be filled up by some expected but not yet discovered
elements (particle types).
The table 11 reveals also a possibility of another gauge symmetry viz.
U(1)xSU(2)xSU(5). This possibility should be worked out also. The fact
that our scheme fits so well and smoothly into a six-dimensional framework
shows decisively the inadequacy of the concept of superstrings, the latter
requiring at least a lO-dimensional space.
In spite of the fact that now the problem of writing down explicitly a
Lagrangian for unified theory is only a matter of standard techniques, the
above sketched assumptions cannot be regarded yet as a full unification
because they do not predict all possible relations between coupling constants,
provided they are indeed constants if viewed upon from a point of view of a
cosmological time scale.
A possible objection that the above models may turn out to be mathe-
matically inconsistent (because of non-renormalizable couplings, etc.) is to
be refuted since obviously it does not apply only to such or similar endavours
of unification but equally well to any contemporary quantum formalism of
gravitation. Probably such objections cannot be avoided unless a profound
modification of the concept of general relativity will be achieved.

References
[1] J.Rayski, Acta Phys. Polonica 27, 89 (1965)
[2] J.Rayski, in Unified Field Theories in more than 4 dimensions,
ed. by V. De Sabbata and E.Schmutzer, World Sc Singapore, 1982
[3] J. Rayski and J. M. Rayskijr, Nuovo Gim 103 A, 1729 (1990)
[4] E. Witten, Nuclear Physics B 186, 412 (1981)
GENERALIZED FIERZ IDENTITIES AND THE
SUPERSELECTION RULE FOR GEOMETRIC
MULTISPINORS

WILLIAM M. PEZZAGLIA JR.

Department of Physics and Astronomy
San Francisco State University
1600 Holloway Avenue
San Francisco, California 94132
U.S.A.

Abstract. The inverse problem, to reconstruct the general multi vector wave function from
the observable quadratic densities, is solved for 3D geometric algebra. It is found that op-
erators which are applied to the right side of the wave function must be considered, and
the standard Fierz identities do not necessarily hold except in restricted situations, cor-
responding to the spin-isospin superselection rule. The Greider idempotent and Hestenes
quaterionic spinors are included as extreme cases of a single superselection parameter.

Key words: fierz - multi vector - superselection - spinors

1. Introduction

In a recent paper Crawford [1] explored the inverse problem of Dirac

bispinor algebra, to reconstruct the wave function from the observable quadratic
densities. Other authors 2 ,3 have presented parallel developments for multi-
vector quantum mechanics in which column spinors are replaced by Clif-
ford algebra aggregates. However, these expositions have only considered
restricted cases (e.g. minimal ideals) for which the multi vector analogies of
the observable bispinor densities obey the standard Fierz 1 ,2,4 identities.
Previously we have proposed 5 ,6 a more general multivector wave func-
tion in which all the geometric degrees of freedom are used. To obtain the
complete set of observable multispinor densities one needs to augment the
standard sinistral 6 operators (applied to the left side of the wave function)
with new dextral 6 (right-side applied) operators, and also bilateral 6 oper-
ators (multivectors applied on both sides of the wave function). It is found
that if and only if the multivector wave function is restricted will the multi-
spinor densities obey the standard Fierz identities. In this paper we propose
to solve the inverse problem for the general unrestricted multivector wave
function of the 8-element 3D geometric algebra C(2), i.e. the Pauli algebra.

91
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 91-96.
© 1993 Kluwer Academic Publishers.
92 WILLIAM M. PEZZAGLIA JR.

2. The Algebra of Standard Pauli Spinor Densities

In non-relativistic quantum mechanics, the electron is represented by a
two-component Pauli spinor. The endomorphism algebra (module structure
on spinors) is C(2), i.e. two by two complex matrices. This Clifford algebra
has as its basis the 8 element group generated by 3 mutually anticommuting
basis vectors, {O"j,O"k} = 20jk for (j,k = 1,2,3), and where i = 0"10"20"3.
As operators, their "bilinear expectation values" yield real densities which
are interpreted to be the projection of the spin along the j-th spatial axis,
Sj =< '0!O"j!'0 >.
The 4 densities {p, Sd, where p =< '0!'0 >, are invariant with respect
to the phase parameter of the spinor. Hence they satisfy a single constraint
equation which can be derived by substituting the projection operator into
the square of the normalization. The magnitude of the spin is found to be
constrained by the Fierz 1 ,2,4 identity,

(1)

The spinor can be reconstructed in terms of a U(2) unitary rotation matrix,

U( A, (}, </>, a) = exp( ia/2) exp( i0"3</>/2) exp( i0"2(} /2) exp( i0"3A/2), (2)

where ((}, </» are the orientation angles of the spin and (A, a) do not con-
tribute in the bilinear form O"jSj{(},</» = pU0"3Ut. Choosing a starting
spinor to be the "plus" eigenstate of 0"3, the wave function can be expressed,
'0(p, (}, </>, (3) = pU!1J >, where the net unobservable phase is f3 = A + a.

3. The Algebra of Geometric Multispinors

We consider an unrestricted multivector wave function 5 which has the
same 8 degrees of freedom as the Clifford group,

(3)

where {a, b, c, d} are complex coefficients. Note each column of the matrix
is a minimal left ideal of the algebra and will hence behave like a column
spinor 8 for all sinistral 6 (left-sided ) operations. Each row of the matrix
is a minimal right ideal of the algebra and will behave like a row isospinor
for all dextral 6 (right-sided) operations. Hence the complete solution can
be interpreted6 as an isospin doublet (of spinors) coupled by now-allowable
dextral application of the Pauli operators.
GENERALIZED FIERZ IDENTITIES AND THE SUPER SELECTION RULE ... 93

3.1. MULTIVECTOR DENSITIES

A complete set of 16 generalized quadratic forms are defined in terms of

the matrix trace (i.e. half the scalar part of the Clifford multi vector )9,
(4a)
Sj = Tr('ljJtaj'ljJ) = Tr('ljJ'ljJtaj), (j=1,2,3), (4b)
Tj = Tr('ljJaj'ljJt) = Tr('ljJt'ljJaj), (j = 1,2,3), (4c)
Rjk = Tr('ljJtaj1/Jak) = Tr(1/Jak1/J t aj),
(j,k=1,2,3). (4d)
They are interpreted to be the probability, spin, isospin and bilateml densi-
ties respectively. From these we can construct the multi vector densities,
1/J1/J t = (p + ak Sk )/2, (5a)
1/J t 1/J = (p + akTk) /2, (5b)
1/Jtaj 1/J = (Sj + Rjkak)/2, (5c)
1/Jak1/Jt = (Tk + a j Rjk)/2. (5d)

3.2. GENERALIZED FIERZ IDENTITIES

The 16 densities are all independent of the phase parameter, hence must
satisfy 9 constraint equations. In general these identities have the form,
Tr[ ('ljJ t a a'ljJ)a,e( 1/J t a" 1/J )a6] = Tr[ ( 1/Ja,e'ljJ t)a,,( 'ljJa61/J t)a a] (6a)
where the indices can take on values 0 through 3, and ao = 1. The parenthesis
indicate where one inserts eqs. (5abcd). It can be shown from these relations
that the bilateral density eq. (4d) contains all the other densities. Further,
we find that the magnitudes of the spin and isospin are equal, but that eq.
(1) is no longer valid,
(6b)

3.3. INTERPRETATION OF THE BILATERAL DENSITY

Counting degrees of freedom,we see that there is one free internal "hidden
variable" contained in Rjk which does not affect the other densities. To
gain some insight as to the nature of this parameter we consider the class
of unitary (hence p invariant) transformations that will leave the densities
{sj, Tj} invariant, but modify the bilateral density.
The special case sinistral operator, U(>') = exp[iajS j /(2ISI)], will leave
the spin invariant (as well as the isospin) as it corresponds to a rotation
about the spin axis by angle >.. The bilateral density will be modified by
this transformation, hence we should be able to parametrize Rjk in terms of
the densities {p, sj, Tj} and a bilateral phase angle >..
94 WILLIAM M. PEZZAGLIA JR.

4. Inverse Theorem and Superselection Rule

We assert that the projection operator for the multi vector wave function
has the bilateral form, "p = (p"p + Sk(1k"p + "p(1kTk + (1j"p(1k Rjk)/( 4p).

4.1. INVERSE THEOREM

The multi vector wave function can be reconstructed from the observable
densities by a applying the projection operator to an arbitrary starting so-
lution TJ, and renormalizing. Hence we assert,

where a is a phase factor and TJ is an arbitrary starting multivector subject

only to the normalized trace constraint Tr( TJt TJ) = 1.
The normalization factor is most directly determined by requiring the
reconstructed wave function to reproduce the probability density eq. (4a),

where identities have been used to reduce the quadradic terms to linear ones
in terms of the observable densities. This construction will fail if eq. (7b)
yields zero, in which case a different starting solution should be used.

4.2. SPECIAL CLASSES OF SOLUTIONS

In order to insure a scalar norm, Hestenes[3,9] proposed a unitary or

quaternionic solution which has 5 parameters,

where unitary matrix U(>',B,</>,a) is given by eq. (2). The alternate quater-
nionic Cayley-Klein components {r, Bj} are all real numbers, subject to
constraint r2 + B2 = 1. Only 4 parameters are however needed to describe
an electron, hence Hestenes (arbitrarily?) sets the parameter a to zero.
This unitary class of solutions is synonymous with zero magnitude spin
and isospin as defined by eqs. (4bc). The bilateral density eq. (4d) is pro-
portional (by a factor of p) to the 0(3) rotation matrix R(>.,e,</» asso-
ciated with the U(2) matrix U(>',B,</>,a). This allows Hestenes to make
an alternate definition of a "spin" vector in terms of the bilateral density,
Sj = Rj3 = Tr("pt(1j"p(13) = ~Tr(U(13Ut(1j)p. It is easily verified that Rjk
is invariant with respect to the>. parameter of the unitary matrix, allowing
Hestenes to reinterpret it as quantum phase, and dextrally applied i(13 as
the quantum phase generator (replacing the usual commuting i).
GENERALIZED FIERZ IDENTITIES AND THE SUPER SELECTION RULE ... 95

In contrast, Greider[7] proposed an idempotent spinor which has the al-

gebraic form of the projection operator,

where the magnitude of the spin is subject to the standard Fierz constraint
of eq. (1). This makes the determinant zero, hence the wave function is of
the "singular class" distinctly different from the "unitary class" discussed
above. There are only 4 degrees of freedom, exactly that needed to describe
a single Pauli spinor (i.e. isospin is everywhere parallel to spin).
Isospin degrees of freedom can be re-introduced by applying a dextrad
rotation operator to eq. (10). Equivalently, consider the following factorized
idempotent form,

_ (1+(73) t
'ljJ - vIP U(A,Os,1>s,a) 2 U (A,OT,1>T,-a), (lOa)

= eia (p + Sk(7k)(p + Tj(7j) [4p(p2 + SkTk)r~, (lOb)

where the singularity constraint eq. (1) still holds. The angles {Os,1>s} give
the orientation of the spin, while {OT, 1>T} that of the isospin. Our solution
has 6 degrees of freedom, exactly that which is needed to describe an isospin
doublet of Pauli spinors (i.e. two particle generations in the family). The
net phase f3 = A + a shows A is indistinguishable from parameter a, hence
Rjk = pSjTk/1S12, has no A dependence.

4.3. SUPERSELECTION PARAMETER

Our multispinor solution is subject to a spin-isospin superselection rule lO •

While certain linear combinations are allowable, the superposition of "spin
& isospin up" with "spin & isospin down" would yield a "forbidden" uni-
tary class solution with zero spin and isospin. Equivalently such a state is
inaccessible by any spinlisospin rotation from a "spin & isospin up" state.
Mathematically this constraint manifests as requiring the determinant of
our wave function to be zero.
Consider a new superselection parameter 8, defined: lSI = pcos8,

where nk(Os, 1>s) = Sk/ISI. For 8 = 0 the wave function becomes a Greider[7]
idempotent with zero determinant, and when the spin vanishes in the limit
of 8 = 7r 12, the solution is of the Hestenes[3,9] quaternionic form. Note the
bilateral phase A is independent of the ordinary imaginary phase a for 8 > O.
96 WILLIAM M. PEZZAGLIA JR.

The remaining two isospin degrees of freedom can be reintroduced as

before by a dextrad rotation operator. A complete parameterization of the
general 8 degree of freedom solution can be expressed in polar form,
_ (1 + eiou3) t
lI!(p,a,)..,o,Os,¢S,(h,¢T) - Vi> U()..,Os,¢s,a) 2 U ()..,OT,¢T,-a).
(12)

5. Summary
We have solved the inverse problem for the completely general eight de-
gree of freedom wave function of 3D geometric space. Our results are more
general than other treatments in that a more complete set of quadratic
multispinor densities is introduced which includes sinistral, dextral and bi-
lateral operations. The 16 densities satisfy generalized Fierz-type identities.
The new bilateml density is found to contain one new independent "hidden"
variable which does not affect the more familiar probability, spin and isospin
densities. It is an open question as to whether this quantity can be physically
measured, or is unobservable like the overall quantum phase parameter.
The standard Fierz identities (for column spinors) are found not to hold
except for a restricted singular class of wave functions. This appears to be a
manifestation of the spin-isospin superselection rule, and may be the critical
constraint which classifies the solution as being a fermionic particle. A con-
tinuous superselection parameter is introduced for which the singular class
of solutions (which includes the Greider idempotent form) is one extreme
case; the Hestenes quaternionic spinor form is at the other extreme.
Extending the work to 4D Minkowski space with a 16 degree of freedom
wave function we will find 136 quadratic forms, which obey 121 generalized
identities. One or more new "hidden" variables will be found, and the stan-
dard Fierz identities will not be valid except for a restricted wavefunction,
corresponding to the charge superselection rule of bispinors.

References
lCrawford, J.P.: 1985, J. Math. Phys. 26, 1439-4l.
2Lounesto, P.: 1993, 'Clifford algebras and Hestenes spinors', Found. Phys.to appear in
May 1993 issue, special edition to honor D. Hestenes 60th birthday (P. Lounesto, ed.).
3Hestenes, D.: 1979, Am. J. Phys. 47, 399.
4Fierz, M.: Z. Phys., 1937104, 553.
5Pezzaglia, W.: 1992, Found. Phys. Lett. 5, 57.
6Pezzaglia, W.:1993, 'Dextral and Bilateral Multivector Gauge Field Description of Light-
Unflavored Mesonic Interactions' ,Found. Phys.to appear in May 1993 issue.
7Greider, K.:1984, Found. Phys.14,467; :1980,Phys. Rev. Lett.44, 1718.
BHestenes, D.: 1965, Spacetime Algebra, Gordon and Breach Pub!., 38;
: 1971, Am. J. Phys. 39, 1013-27.
9Hestenes, D. and Gurtler, R.: 1971, Am. J. Phys. 39, 1028.
lOMohapatra, R.N.: 1987, Phys. Rev. Lett 59, 1510.
ELECTRONS, PHOTONS, AND SPINORS IN
THE PAULI ALGEBRA
WILLIAM E. BAYLIS
Department of Physics, University of Windsor
Windsor, Ontario, Canada N9B 3P4

Abstract. The multivectors (" cliffors") of three-dimensional Euclidean space form a com-
plex four-dimensional vector space with the Minkowski metric. In fact all elements of the
real Clifford algebra of Minkowski space (the 'Dirac' or 'spacetime' algebra) can be mapped
(in two mappings) onto the Pauli algebra. The Pauli algebra is used here to provide a co-
variant description of elementary charges and electromagnetic radiation fields in terms of
'spinors' which represent Lorentz transformations describing their motion.

Key words: electrons - photons - spinors - Pauli algebra - Dirac equation

1. Introduction

One of the simplest Clifford algebras with widespread applications to the

physical world is the real geometric algebra of three-dimensional Euclidean
space (Hestenes 1966, Baylis 1992). It is a convenient starting point for math-
ematicians in their abstract study of more general Clifford algebras since,
in spite of its simplicity, it is sufficiently complex to include the complex
field (as its centre) and the quaternions (as its even subalgebra), and these
together with its non abelian product and zero-divisors produce a mathe-
matical structure of considerable richness. Its standard representation in
terms of Pauli spin matrices is also familiar to physicists, especially in quan-
tum theory, although its real power and beauty is better revealed in the
representation-free algebraic form. Here for simplicity, the Pauli algebra P
is used to refer to the geometric algebra of three-dimensional Euclidean
space; no specific matrix representation is implied. Similarly, the Dirac al-
gebra V refers to the real Clifford algebra of Minkowski spacetime, and the
real quaternion algebra is labeled by 'H.
The aims of the present contribution are to summarize applications of
P to the study of 'elementary' particles and radiation fields (with much
more emphasis on the former), to show how naturally the algebra models
the structure of spacetime, and to see how elements of P provide covariant
descriptions of particles and fields and yield a simple interpretation of the
Dirac equation. The following section reviews the structure of the Pauli alge-
bra and its relation to Minkowski space. While the mathematics is familiar

97
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 97-106.
© 1993 Kluwer Academic Publishers.
98 w. E. BAYLIS

to anyone with an interest in Clifford algebras, the physical content is so

beautiful in its simplicity and so magnificent in its power that it deserves
close attention. The following formulation also serves to introduce the id-
iosyncrasies of my notation. Although'it is given for convenience in terms of
flat spacetime, it can be extended to curved manifolds (Rastall 1964).

2. From Euclid to Minkowski

Recall that the Pauli algebra P is generated by an associative product of
real three-dimensional vectors which satisfies the condition a 2 = a . a for
any vector a in P. From three orthonormal, anticommuting basis vectors
{el, e2, e3} of the underlying Euclidean space, the resulting fundamental
relation

ejek + ekej = 20jk (1)

allows eight independent real basis forms of P to be constructed:

(2)
As in V and 1i, the canonical element squares to -1, but in contrast V
and 1i, the canonical element el e2e3 of P, because of the odd number of
dimensions of the ground space, also commutes with all other elements. It
may therefore be identified with the imaginary i:

(3)
The product of i with a vector gives the bivector dual to it. In particular

(4)
The identification of the canonical element of P with the imaginary i
allows one to consider P to be spanned not only by the eight basis forms
(2) over the reals, but also by the four forms {1,el,e2,e3} over the complex
field. That is, a general element pEP is the sum Po + p of a scalar Po and a
three-vector p = pkek' both of which may be complex. Thus, the products
of real three-dimensional vectors generate in P both complex numbers and a
four-dimensional space, and the imaginary i has geometric content: i times a
scalar is a pseudoscalar (trivector) and represents a volume; i times a vector
is a pseudovector (bivector) and represents a plane.
There are two fundamental involutions on P. Reversion p ____ pt plays the
role of complex conjugation and changes the sign of the imaginary part of a
element whereas spatial reversal p ____ j5 changes the sign on the three-vector
part. Thus if

p = Po +p = p/-Le/-L (5)
ELECTRONS, PHOTONS, AND SPINORS IN THE PAULI ALGEBRA 99

where eo = 1, then
p!1*elJ (6)
Po - p. (7)

The two may be combined into the automorphism p --+ pt = pt. If p = pt,
it is real; if p = p, it is a (possibly complex) scalar. The scalar part of a
product pr is indicated as a dot product:

p. r = (pr + pr)/2 = para + p' q. (8)

N ate that p. q = 1 . (pr) = 1 . (rp).

An element p is invertible if there exists another element, say q, whose
product with p is a nonzero scalar: pq = pq = ijp. It is straightforward to
see that q is proportional to p and that the inverse of p is

p-l = pf(pp). (9)

If the scalar norm pp vanishes, the element p is a zero-divisor and is not
invertible. (Such elements add to the structure of the algebra by allowing
proper ideals.) A metric for the space is naturally defined for real elements
by the bilinear scalar norm
- 2 2
pp = Po - P . (10)

It is seen to be exactly the Minkowski metric, with the scalar part of p

playing the role of the 'time' component.
A statement made above can now be strengthened: the products of real
three-dimensional vectors generate in P both complex numbers and a four-
dimensional space with the Minkowski metric.

3. Lorentz Transformations and Covariance

Physical four-vectors form a real subspace M of P. They include (in units
with c = 1) the four- velocity u = 1+ u, the (four- )momentum p = E + p, the
(four- ) current density j = p + j, and the four-potential A = 1> + A. Trans-
formations which leave the norm of any four-vector invariant are called (ho-
mogeneous) Lorentz transformations. It follows that such transformations
also leave invariant scalar products of the form p. A. The restricted (i.e.,
proper orthochronous) subset of these transformations, when operating on
any four-vector p, can be written (Baylis and Jones 1989a)

(11)
where L is any unimodular element ofP: LL = 1. (The proper nonorthochronous
Lorentz transformations take the form p --+ -LpLt whereas the improper
100 W. E. BAYLIS

ones are p -> ±LpLt.) The elements L which give the Lorentz transforma-
tions form the subalgebra sl(2, C) of P and can always be written as the
product
L = sn = exp(w/2)exp( -i8/2) (12)
of a boost by rapidity wand a rotation by () in the plane i8. The rotation
elements nconstitute the Lie algebra su(2) C sl(2, C).
The separation of any four-vector in P into scalar and vector parts corre-
sponds to the physical partition into 'time' and space components. This cor-
responds to an obvious differentiation which every conscious observer makes,
and it is desirable that any algebra modeling the physical world contain an
analogous natural partitioning. However, the separation is obviously frame
dependent since boosts generally scramble time and space components. How
can P provide a covariant description of nature, i.e., one in which the basic
physical equations take the same form in all inertial frames?
The obvious answer is simply to avoid splitting elements of P into scalar
and three-vector parts. Of course it is the essence of Clifford algebras that
they are most useful when elements are not expanded in components or
expressed as the sum of homogenous (scalar, vector, etc.) parts. Just as the
power of complex numbers is largely lost when every such number is written
as the sum of a real number and an imaginary one, so is the efficiency of P
degraded by expressing four-vectors as sums of scalars and three-vectors.
When four-vectors are expressed as single elements of P, relations among
them can be covariant, for example p = mu. Products in P involving four-
vectors also appear frequently in covariant relations. They transform simply
under Lorentz transformations if four-vectors p E M are alternated with
barred four-vectors ij EM:
(13)

If pij is expanded in scalar and vector parts:

pij=P'ij+p/\ij (14)
where the scalar part is as above p' ij = 1· (pij) = (pij + qp)/2 and the vector
part is
P /\ ij := (pij - qp)/2 = pij - p. ij = -q /\ p, (15)
then the scalar part, as seen above, is invariant whereas the vector part
transforms as p /\ ij -> Lp /\ ijL. The vector part of the product pij of p E M
and q E M is called a six-vector because it generally has three real-vector
and three pseudovector (imaginary-vector or bivector) components.
An important six-vector is the electromagnetic field
F = a /\ A = aA - a . A = E + iB (16)
ELECTRONS, PHOTONS, AND SPINORS IN THE PAULI ALGEBRA 101

where 0 is the four-vector operator Ot - \7. The Lorentz-force equation in P

for a particle of momentum p takes the covariant form

dp/dr = eR(Fu) (17)

where r is the proper time, and Maxwell's equations are just the scalar,
pseudoscalar, vector and pseudovector parts of
aF = 411"K] (18)
where J( is a constant depending on units: J( = 1 in Gaussian units, (411" )-1
in Heaviside-Lorentz units, and (41rfO)-1 in SI units. Note that just as the
Pauli product pij = p. ij +p A ij contains both inner and outer parts, so does
differentiation with o. In terms of differential forms, 0 = {j + d.
Thus the Pauli algebra, while allowing a physically intuitive separation
of four-vectors into 'time' and space parts, also generates a Minkowski space
and provides a naturally covariant formalism for problems in relativity. For
problems in physics, it is usually unnecessary and often needlessly inefficient
to use the Dirac algebra 'O. To underscore this point, it is worthwhile to ob-
serve that any element of V can be mapped onto P (Baylis and Jones 1988).
The mapping from even elements of V onto P is a well-known isomorphism.
Odd elements of V are changed into even elements through multiplication by

,0.
a basis vector of the ground space M of V; traditionally one uses the time-
like vector These elements can then also be mapped onto P by the same
isomorphism. Of course the result is a two-to-one mapping, which means
that a given type of element in P can represent two types in 'O. In practice,
however, this causes no problems. Thus a scalar in P can be either a Lorentz
scalar or the 'time' component of a four-vector, and a three-vector in P can
be either part of a four-vector or part of a six-vector (a Dirac bivector),
but no one is likely to confuse the possiblilites. Indeed a covariant algebraic
notation keeps the identities quite distinct.

4. Electrons and Neutrinos

In this section, the Pauli algebra is applied to descriptions of 'elementary'
particles (Baylis 1992). The spin and translational motion of an 'elementary'
particle is described by a characteristic Lorentz transformation A: the trans-
formation of the particle from its rest frame to the observer's frame (Giirsey
1957; Rastall 1964, 1988). Of course a compound particle like a nucleon may
require several Lorentz transformations to fully describe its motion, one, say,
for each quark, but I want to assume that there exist particles which are
'elementary' in the sense that they require only one Lorentz transforma-
tion for their full description. However, it is not necessary to assume that
the 'elementary' particle is a point, only that its motion is described by a
single Lorentz transformation. The particle must be structureless but may
102 W. E. BAYLIS

have a finite extent. In order that a single Lorentz transformation describe

a non-point particle, the time-development of the transformation must be
governed by a linear equation of motion, and this together with the Lorentz-
force equation, can be shown to constrain the g-factor of the particle to be 2
(Baylis 1992). The idea of having a single such transformation for a particle
is essentially classical; the transition to a quantum picture involves replacing
the single Lorentz transformation by a field.
By using A, one can transform any property of the rest frame to the lab
frame of the observer. For example, anyone of the orthonormal unit vectors
eJ.l'J.l = 0,1,2,3 in the rest frame is transformed as a four-vector to the
corresponding Frenet vector

UJ.I = AeJ.l At (19)

in the lab frame. In P the timelike unit vector is simply eo 1 and the
corresponding Frenet four-vector is the four-velocity

U == Uo = AAt. (20)
The six-vectors constructed from the Frenet four-vectors are Minkowski-
space bivectors:

(21 )

The characteristic Lorentz transformation A is usually different for dif-

ferent observers. If the frame of the one observer is transformed to that of
another by a transformation element L, then A is transformed according to

A -+ LA. (22)

The transformation (22) is just what is needed for the Frenet four-vectors
and six-vectors to be covariant: they are the bilinear covariants of the clas-
sical theory. However, the form (22) shows that A itself can not be a four-
vector or any product constructed from four-vectors; its transformation be-
haviour is distinct. Transformation (22) is appropriate for spinors, and A
may be seen to be a vector in the representation space of the group SL(2, C)
of restricted Lorentz transformations. Although the space is reducible, it is
the smallest space to give a faithful representation of S L(2, C). A is called
the eigenspinor of the particle. For restricted transformations, it is unimod-
ular: AA = 1.
The eigenspinor of an accelerating particle is a function of the proper
time 7 of the particle: A = A( 7). Eigenspinors at different times are related
by a Lorentz transformation L( 72, 7J) which serves as the time evolution
operator of the particle:
ELECTRONS, PHOTONS, AND SPINORS IN THE PAULI ALGEBRA 103

The four-momentum p of a classical particle is its mass times its four-

velocity u:
p = mu = mAAt. (24)
Because the eigenspinor A for a restricted transformation is unimodular, the
trivial identification (24) can be put in a form which is linear in the space
containing both A and At:
pAt = mAo (25)
This is the classical Dirac equation. It is put in its traditional quantum form
= m'I/J by defining a four-component column spinor
'YJi.pJi.'I/J

(26)

where the Weyl two-spinors enter as columns of A

(_[t, i]t). Defining the rest-frame two-spinors

(27)

one can write i]t and ~ as transformations from the rest frame:
i]t = Ata t , ~ = Af3. (28)
(Note that in terms of spinors with abstract indices, the bar lowers the index,
and the dagger dots it.)
The correspondence between the classical eigenspinor A and the quantum
four-spinor 'I/J is further strengthened by calculating the bilinear covariants
and the CPT transformations in terms the Weyl-spinor components (Baylis
1992). Comparisons with the quantum forms shows that the quantum am-
plitude 'I/J must, within a normalization constant and an arbitrary initial
rotation of the rest frame, represent the Lorentz transformation of the par-
ticle from its rest frame to the lab frame. An association of'I/J with a Lorentz
transformation of the particle is not new. It was made by Giirsey (1957),
Rastall (1964, 1988), and Hestenes (1975, 1990).
As a sample calculation of A in P, consider the eigenspinor at a given
proper time, say T = O. Like any other Lorentz transformation, it can be
written as the product of a boost and a rotation:
A(O) = B(O)R(O). (29)
From u = AAt = B2 we find B = U 1 / 2. It is readily verified that the solution

t
can be either timelike (Bd or spacelike (B 2 ):

B1 = ± vi2mp E+m
m )' B2 = Bd>· (30)
104 W. E. BAYLIS

The solution Bl is unimodular and therefore a proper boost, whereas B2 is

anti-unimodular and therefore an improper boost. If A = Bl R is a solution
of the Dirac equation (25), it must represent a positive-energy particle:
l'p=mA·At>O (31)
whereas A = B2 R, as an anti-unimodular solution of (25), represents a
negative-energy one:
1 . p = -mA . At = -mB2 . B2 < O. (32)
Unfortunately, this conflicts with Eq. (24) which B2 was supposed to satisfy:
the sign of p is different! Either one changes the sign of p in B2 so that it
represents the momentum of the negative- energy particle or one leaves the
sign and reinterprets p in B2 as the momentum of the lem antiparticle.
Since rotations are unitary as well as unimodular, R = nt. Consequently
the matrix representation can be expressed in terms of a single 2-spinor
x: R = (-X t , X). When this is combined with the standard matrix repre-
sentations of the boost parts Bl and B2 , the usual expressions (within a
constant normalization factor) are obtained for the momentum eigenstates
'lfJp(O) of quantum theory. Now the full momentum eigenstates are plane
waves with the dependence 'lfJp( x) = 'lfJp(O) exp( -ip . rln). Since the Lorentz
scalar p. r = mr, the relation of'lfJ to the eigenspinor A means that
A( r) = A(O) exp( ie3mr In). (33)
In words: the particle spins about its rest-frame direction -e3 at the Zitter-
bewegung (proper) frequency 2mln = 2mc2 In.
The physical interpretation of the quantum phase as a rotation or spin is a
great success of the Clifford-algebra approach to the Dirac theory (Hestenes
1975, 1990). Many other useful insights also surface. For example, at low
velocity, the rotational two-spinor X gives the rotation R of the spin from
-e3 to its direction in the lab frame, and the intrinsic parity of fermions and
antifermions is given immediately. The theory in P has so far been essentially
classical, but the transition to the full quantum theory is surprisingly easy.
The first step is to demand local gauge invariance which now corresponds
to invariance under a rotation about the spin axis. This invariance demands
the introduction of a gauge field and the replacement of the rotation factor
exp(ie3mr) = exp(ie3P' r) by exp[ie3J(P + eA)· drj. However, this factor
is now path-dependent. The second and crucial step is to sum over the A
from all contributing paths. The result is a path integral over contributions
which all satisfy the differential relation
8A(ie3) = (p + eA)A. (34)
The combination of this differential relation with the classical Dirac equa-
tion (25) gives the full Dirac theory in a form ripe with physical significance
and often very convenient for computations.
ELECTRONS, PHOTONS, AND SPINORS IN THE PAULI ALGEBRA 105

The concept of a rest frame becomes meaningless for massless particles

like neutrinos, but the theory developed above for 'elementary' charged par-
ticles works well in the limits e -+ 0 and m -+ O. Of course the wave-
function must be normalized to correspond, say, to pl/2 rather than to
U 1 / 2 , but then it is finite and its 'plane-wave' solutions vary as m 1 / 2 A(r) =
m 1 / 2 A(O) exp( ie3P' r/n). As with electrons, any spin state can be written as
a linear combination of the two helicity states, but in the limit m -+ 0 one
of the helicity states vanishes.

5. Electromagnetic Radiation
The vector potential for circularly polarized electromagnetic plane waves
can be written in the same form as the neutrino eigenspinors:
A(r) = A(r) = A(O)exp(il\:kk. 1') (35)
where for simplicity the transverse ('radiation') gauge has been adopted:
k· A = 4> = 0 and I\: = ± 1 is the helicity. The Lorentz-gauge condition is also
satisfied: &. A = O. The vector potential (35) is a real vector which rotates
about the propagation direction. This is more obvious if (35) is written
A(r) = nA(O)nt, (36)
where n = exp( -il\:kk . r /2) is the rotational eigenspinor of the wave. Of
course a boost can also be applied.
Maxwell's equation (18) for A for source-free space requires the prop-
agation four-vector to be null, that is to be a zero divisor: kk = o. As a
consequence, the associated electromagnetic field is
F = E + iB = & 1\ A = il\:kA(r) = ill;kA(O) exp( -ill;k· r). (37)

The unit vector k could be dropped because kk = k.

6. Conclusions
The three real basis vectors of physical space generate in the Pauli algebra
P a four-dimensional space with the Minkowski metric and complex num-
bers. It provides a covariant framework for problems in relativity without
appealing to higher-order Clifford algebras. Application of the Pauli alge-
bra to radiation fields gives simple real expressions for the fields in terms
of a rotating vector. Application to 'elementary' particles turns the inno-
cent kinematical relation p = mu into the classical Dirac equation, with the
quantum probability amplitude identified with the Lorentz transformation
(the 'eigenspinor') of the particle. The plane waves of quantum theory imply
a rotation about the spin axis at the Zitterbewegung frequency. Local gauge
invariance requires a gauge field A which gives path-dependent eigenspinors,
106 w. E. BAYLIS

and a linear superposition of contributions from different paths leads to path

integrals and the differential operator for the canonical momentum. The full
Dirac theory results, but in a form that clearly displays physical features
which are well hidden in traditional treatments. In the nonrelativistic limit,
the energy eigenvalue equation from the Dirac theory gives the Schrodinger
equation with the Pauli Hamiltonian, and even there all appearances of the
imaginary i in the differential operators and in commutation relations are
seen to originate with the spin of the particle (Hestenes 1971, 1975, 1990;
Baylis et al. 1992).
The above assertion is liable to engender disbelief: "Surely you don't
mean that all quantum particles must have spin?" No, just the 'elementary'
ones. Of course, if a scalar Higgs boson is discovered, there may be more
work to do.
Satisfying progress can be reported in applications of the Pauli algebra to
fundamental problems of physics, including research not reported here on the
electroweak theory, radiation reaction, and general relativity. However, much
work remains to be done. Especially pressing are problems in many- body
interactions and second-quantization, and there are potential applications
in quantum gravity which will probably occupy a generation of physicists.

7. Acknowledgements
This work has been supported by the Natural Sciences and Engineering
Research Council of Canada.

References
Baylis. W.E.: 1992, 'Classical eigenspinors and the Dirac equation', Phys. Rev. A 45,
4293-4302
Baylis, W.E. and Huschilt, J. and Wei, J.: 1992, 'Why i?', Am. 1. Phys. 60, 788-797
Baylis, W.E. and Jones, G.: 1988, 'Special relativity with Clifford algebras and 2 x 2
matrices, and the exact product of two boosts', 1. Math. Phys. 29, 57-62
Baylis, W.E. and Jones, G.: 1989a, 'The Pauli-algebra approach to special relativity', 1.
Phys. A 22, 1-16
Baylis, W.E. and Jones, G.: 1989b, 'Relativistic dynamics of charges in external fields: the
Pauli-algebra approach', 1. Phys. A 22, 17-29
Giirsey, F.: 1957, 'Relativistic kinematics of a classical point particle in spinor form', Nuovo
Cimento 5, 784-809
Hestenes, D.: 1966, Spacetime Algebra, Gordon and Breach: New York
Hestenes, D.: 1975, 'Observables, operators, and complex numbers in the Dirac theory',
1. Math. Phys. 16, 556-572
Hestenes, D.: 1990, 'The Zitterbewegung interpretation of quantum mechanics', Found.
Phys. 20, 1213-1232
Rastall, P.: 1964, 'Quaternions in relativity', Rev. Mod. Phys. 36, 820-832
Rastall, P.: 1988, 'New forms of the Dirac equation', Nuovo Cimento 101 B, 479-494
TWISTORS
TWISTORS AND SUPERSYMMETRY
DMITRIJ VOLKOV
Kharkov Institute of Physics and Technology,
Kharkov 310108, the Ukraine

In modern theoretical physics there are two branches of research which

are closely related to fundamental problems of space-time. These are twistors
and supersymmetry. Recently a growing tendency to intertwining these two
branches has been appearing, and it is now widely believed that twistors
may play an essential part in the foundation of supersymmetric theories.
The main point where twistors and supersymmetry are in touch is the inte-
grability conditions for N-extended and D = 10 super Yang-Mills theories
and supergravity which, in its turn, is connected with the fact that the tra-
jectory of a massless superparticle in superspace is not a world line but a
supersurface with one bosonic and a number of fermionic directions.
The way of reasoning leading from supersymmetry to twistors is rather
complicated and grounded more on guesswork than on a solid constructive
basis. A more direct way which incorporates twistors into supersymmetric
theories as a primary ingredient has been proposed recently (Volkov 1988,
1989, 1990, Sorokin 1989a,b, 1990) and developed in a number of works 1 .
The proposed "twistor like" reformulation of the superparticle and su-
perstring actions contains two kinds of off-shell supersymmetry: the global
supersymmetry of the super Poincare group in the target superspace and a
local su persymmetry on the world line (sheet) of the superparticle (su per-
string). The latter holds off shell and in a particular gauge transforms into
the Siegel symmetry.
In respect to the local supertransformations on the world line the Grass-
mann () coordinates of the target superspace and the components of twistor
connected with the momentum of the superparticle are superpartners. This
fact is of great importance and reflects a fundamental role which twistors
play in the proposed formulation.
Here (after reviewing some aspects of supersymmetry and twistor theory)
we discuss motivation, underlying ideas and an outcome of the proposed
reformulation. The D = 3,4,6 and 10 superparticle action will be considered.

1 The references are given in the following text.

109
Z. Oziewicz et at. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 109-119.
© 1993 Kluwer Academic Publishers.
110 DMITRIJ V. VOLKOV

1. Twistors
Twistors has been introduced by Penrose (1967) as an alternative to the
coordinate description of space- time and simultaneously as a bridge between
space-time and quantum properties of matter. The latter appeared hopeful
because of the complex structure of both the twistor space and the quantum
wave functions. In "the twistor programme" (Penrose 1977) a relation of
twistors to the theory of elementary particles has been argued and unified
twistor description of fundamental interactions had been preconceived. The
unification of twistor theory with supersymmetry may contribute to revival
of some ideas of "the twistor programme" .
In the case of the D = 4 space-time a twistor is defined by four complex
numbers which are usually represented as a pair of two-component complex
spinors Ac» J.l0l. Their relation to the space time coordinates is given by the
basic equation

(1.1 )

which for constant J.l0l and AOI satisfying

1 -c,
S = "2(A OI J.lOl + A TIc,) = 0 (1.2)

describes a light-like line in Minkowski space. (1.1) together with the Cartan
relation for the momentum of a massless relativistic particle

(1.3)

allows one to transform the world line action

(1.4)

to the twistor world line action

(1.5)

or to the third one

S3 = - J drpo<c, (iOlC, - A01 xc, ) (1.6)

which is intermediate between (1.4) and (1.5) as it contains the coordinates

and the twistor components simultaneously. In performing the transforma-
tion it is important to note that p2 = 0 is the constraint equation which
follows from (1.4). The Cartan relation (1.3) solves this constraint explicitly.
As a result the actions (1.5) and (1.6) are invariant under reparametrization
of r without the presence of the one-dimensional metric e( r). This is very
TWISTORS AND SUPERSYMMETRY 111

essential for the subsequent supersymmetric generalization. The action (1.5)

will not be used below. It is written down to stress that the approach under
consideration is tightly connected with the standard twistor theory. It is
evident that the supersymmetrized actions also admit a number of transfor-
mations similar to (1.4-6), which give a way to formulating different variants
of supertwistor theory.2 We will not touch this interesting question here.
The action (1.6) and its generalization to the string provide us with a tool
to incorporate twistors into the superparticle and superstring theory. Now
we proceed to those problems of the superstring and superparticle theory
which require their reformulation.

2. Supersymmetry
While up to now twistor theory, providing a rather powerful technique for
the investigating modern field theories, has not noticeably influenced the
basic concepts of space-time, supersymmetry has lead to their revision and
pretends now to be a physical theory. In the first papers on supersymme-
try (Golfand 1971, Volkov 1972, Wess 1974) the Poincare group has been
generalized to the super-Poincare group. The next years contributed to the
development of supersymmetric field theories, supergravity and superstring
theory.
Here we consider flat superspace with even coordinates xm and odd Grass-
mann spinor coordinates ()Oi their transformation law being defined as
()' () + f
(2.1)

which together with the Poincare transformations is the super Poincare

group.3
In this section we use the notation being conventional for the description
of spaces of dimensions D = 3,4,6 and 10; i.e. spinors are Majorana, rm are
symmetric, the Fiertz identity is fulfilled.
A remarkable peculiarity of the supersymmetry theories is that to a great
extent their content depends on the symmetry properties of free objects
propagating in superspace, which, in turn, can be defined classically. The
proper starting point of any geometrical theory of objects propagating in a
target space should include postulating that the inner space of objects is a
subspace of the target space. Postulates of such kind were widely applied by
2 See, for example, (Berkovits 1991)
3 This formulation of superspace as well as the technique of using the invariant differ-
ential forms (2.2) for constructing supersymmetric actions was first proposed in (Volkov
1972) and in the widely unknown paper (Volkov 1974) where to describe the fermion
Goldstone particles superspace (2.1) was used as a target space and Minkovski space was
used as an inner space.
112 DMITRIJ V. VOLKOV

E. Cartan to his classical geometrical constructions and the method of ex-

ternal differential forms created by him gave nice mathematical formulation
to them.
The objects (superparticles, superstrings, supermembranes) propagating
in superspace sweep some supersurfaces, which locally in the small neigh-
borhood are described by the differential forms

(2.2)
II( d) dO
being invariant under supertransformations (2.1). With the use of the differ-
ential forms (2.2) manifestly super Poincare actions are easily constructed.
The Brink- Schwarz superparticle action (Brink 1981) is

SB-S = ~ J dre- 1(x - iOrO)2 (2.3)

where e is a one-dimensional metric. The Green-Schwarz superstring action

consists of two terms (Green 1981)
Sl + S2 (2.4)
- 2.
211"
J d2 o-v'hh"'i3II",II i3 (2.5)

where II~ = o",x m - iOArmo",oA and A = 1,2. h"'i3 is the two-dimensional

metric and A=1,2, which is the direct supersymmetrization of the bosonic
string action and
S2 = ~ J d2 o- { -i(o:i3o",x m(ol r m0f301 - o2r mOf3(2) +
(2.6)
(''' f3 01r m o,,,olo2 r m0i302}
which is a Wess-Zumino term, corresponding to the differential 3-form
(2.7)
SB-S, Sl and S2 are manifestly invariant under reparametrization. The ac-
tions (2.3) and (2.4) also have a remarkable fermionic symmetry under the
local transformations
(2.8)
where IIm = iormiJ and 8e = 0 for the superparticle case and II~ = o",x m -
iOrmo",O, 8( v'hh"'i3) = 0 and K ----; K", for the superstring case. Spinor as well
as extended symmetry indices are omitted. 4
4 The thorough discussion of the actions (2.3) and (2.4) and of the Siegel symmetry is
given in the monograph (Green 1987) so we only make same general remarks.
TWISTORS AND SUPERSYMMETRY 113

Afterwards Green and Schwarz succeeded in writing down their cele-

brated superstring action (2.4) which is invariant under (2.8) only if the
both terms Sl and S2 are present.
The discovery of the Siegel symmetry solved, in principle, the problem
of unwanted fermionic degrees of freedom appearing due to disbalance in
the number of the vector and spinor coordinates of superspace. Neverthe-
less, because of the presence of singularity in (2.8) no Lorentz covariant
procedure for solving it practically has been proposed up to now, which is
a serious drawback of the GS superstring theory. So, despite all the attrac-
tiveness of the GS superstring action the above mentioned drawback is an
unsurmounted barrier to further development of the theory.
Therefore it may be instructive to go back to another formulation of
superstring which was proposed earlier by Ramond, Neveu and Schwarz
(Ramond 1971, Neveu 1971) and reanalyze its virtues and shortcomings. As
it is well known, this formulation nicely deals with the unphysical fermionic
components since a local superconformal invariance of the superstring world
sheet has been introduced into the theory from the beginning. On the other
hand, the existing procedure for checking the invariance of the RNS super-
string under the super Poincare transformations in the target superspace
is highly artificial and tedious. So, we see that there is striking duality of
virtues and shortcomings of GS and RNS superstrings with respect to their
target space and world sheet supersymmetries.
Just as the GS superstring has a simplified version which is the BS su-
perparticle, and which is a useful training ground for learning the symmetry
properties of the theory, there similarly is a simplified version of the RNS
superstring. This is the so called spinning particle. The latter has attracted
less attention than the BS superparticle. But as we will see later the theory
of the spinning particle may be looked at from a rather unexpected point of
view as a consequence, twistors appear on the scene and a new approach to
the theory of superparticles and superstrings arises.

3. The spinning particle

A covariant action for a spinning particle has been proposed in (Brink,

1976). The authors aimed to demonstrate the presence of supersymmetry
in the Dirac equation and to give a simple one-dimensional model for inter-
acting matter and supergravity. In the massless case the proposed action is
following

(3.1)
114 DMITRIJ V. VOLKOV

(3.1) is invariant under r reparametrizations and the local su persymmetry

transformations

where 1j;m is a fermionic vector superpartner of x m , e is the (one-dimensional)

vielbein field and ~ its fermionic superpartner. (3.1) represents the limit of
the RNS superstring when the string tension goes to infinity. The RNS
superstring action contains the same functions x m , Pm ,and 1j;m depending
on the world-sheet coordinates and two- dimensional supergravity multiplet.
The absence in (3.1) of any spinor variable makes it difficult to formulate
supersymmetry properties in the target space. Varying (8) which respect to
e and ~ gives the constraints

o (3.3)
o (3.4)

which are the conditions for the particle to have zero mass and to obey the
Dirac equation. We begin considering the twistor representation ofthe action
(3.1) with the D = 3 case, as it does not contain complications peculiar to
the D = 4,6 and 10 cases. In section 1 the equation (3.3) has been solved
by taking into account the Cartan relation:

(3.5)

As (3.4) is a superpartner of (3.3) it is naturally to try to solve it by a

similar substitution. We try

(3.6)

where ()DI is a real Grassmann spinor. Under local supersymmetry A and ()

transform as bA = 0, b() = a(r)A in accord with (3.2). In terms of spinor
variables A and () the action of spinning particle (3.1) reads

The equations of motion for X Dl !3 generated by the action (3.7) have the form

(3.8)

Taking into account Eqs. (3.5) and (3.8), we may rewrite the (3.7) in the
form

(3.9)
TWISTORS AND SUPERSYMMETRY 115

It is wonderful that the representation (3.9) is invariant under the super

Poincare transformations
1
8001 = EOI , OA OI = 0, OX Ol {3 = "'2(BOI E{3 + B{3EOI) (3.10)
and coincides with the Brink-Schwarz superparticle action

SB-S = Jdr {POI{3 (:i;0I{3 - iB OI iJ(3) - ~p2} (3.11)

So the action (3.1) and (3.11) are classically equivalent on the mass shell.
Upon quantization in the D = 4 case (3.1) leads to the Dirac equation, and
(3.7) and (3.11) lead to the Dirac equation for a Majorana spinor and the
Klein-Gordon equation for a complex scalar field. So in the both cases there
is a fourfold degeneration of the states, and the quantum systems are also
equivalent. Nonequivalence arises either when an interaction is included or
as a result of second quantization when the difference in statistics comes
into play.
Consider now the Siegel transformations. For the action (3.11) they are
8BOI = POI{3KP,8x Ol {3 = i(B0I8B{3 + B{38Ba),8pOI = 0,8e = 4B 0I 8BOI (3.12)
Due to the Cartan relations (3.12) transforms into
8BOI = a(r)AOI ,8AOI = 0,8x Ol {3 = ia(r)(AOIB{3 + A{3BOI)' (3.13)
where a = AOI""OI. Comparing (3.13) with (3.2) we see that they coincide.
Since the transformations (3.2) are an off-shell symmetry of (3.9), we get
an off-shell formulation of the Siegel symmetry. Note that the relation be-
tween oB and Ox in (3.12) and (3.13) has the opposite sign from that for the
super Poincare group. No explanation to this fact has been proposed. Now,
because of the above relation between the Siegel and local superconformal
transformations, this can be explained as usual difference of the signs of
the left and right Cartan differential forms on a group manifold. The above
consideration can be generalized up to the D = 4,6,10 for the D = 4 case,
for example, eqs. (3.12) and (3.13) are the same, excepting that "", B, and A
are complex and
(3.14)
retaining to be real. It is natural to generalize (3.14) to complex a, so that,
for example, (3.14) gets the form
8x a/3 = ia(r)(Aa O/3 + X/3Ba) (3.15)
One can easily show that (3.9) is also invariant under the complex trans-
formations, but only on shell for an imaginary a( r). To get the off shell
local supersymmetry it is necessary to consider its n = 2 generalization. For
D = 6 and 1O,n = 4 and 8 generalizations are needed. Their consideration
is more convenient in superfield formulation.
116 DMITRIJ V. VOLKOV

4. Twistor like action for superparticle

In the preceding section we have proved the equivalence of actions of the
spinning particle and the superparticle, and we have also got an impor-
tant conclusion that the B-coordinates and the ,X-components of twistor are
superpartners under the local world line superconformal group. Using the
mass shell equations has been an essential drawback of our reasoning. Here
we give a twistor like reformulation of the D = 3,4,6 and 10 superparticle
action with the off-shell n = 1 local superconformal invariance, which will
be our primary principle. To do this we use the superfields

(4.1 )
where Pm and 'ljJm are Grassmann superpartners of Pm and Xm ,respectively.
Now with the use of (4.1) we supersymmetrize the action (1.6) and get

S = -i J drdTJPm(DX m - i0r mD0), (4.2)

where D = :'1} + iTJ t.r.

Note that, due to the structure of the 0-superfield
which contains ,X as a multiplier of TJ , the second term in (4.2) contains
derivatives of 0 . Besides, the sum of the first and the second term being su-
persymmetrized independently is super Poincare invariant. Integrating over
TJ gives the component form of (4.2)

S= J dr{Pm(xm-iOrmO+Xrm,X)+ipm('ljJm+orm,X)}, (4.3)

Excluding some variables from (4.3) one can get different forms of the action,
including that of the BS superparticle and of the spinning superparticle.
Since the action (1.6) is off-shell reparametrization invariant the action
(4.2-3) is off-shell invariant under superconformal transformations of rand
TJ independently from the fact that the derivative D does not contain viel-
bein variables and has the form corresponding to flat superspace. It is also
invariant under D - 2 Siegel transformations one of which coincides with
superconformal one. 5 To get the full equivalence of the Siegel and local su-
perconformal transformations for superspaces with D =
4,6 and 10 it is
necessary to consider the n = 2,4 and 8 extensions of the local supercon-
formal group. This has been done in a number of papers (Sorokin 1989 a,b;
Berkovits 1991; Howe 1991, Ganntlet 1991, Galperin 1992; Pashnev 1992;
Chikalov 1992) which have elaborated a route from D = 3 to D = 10 dimen-
sions. Because of the lack of place for reviewing all of them, here we briefly
5 Formally this is related to the fact that the expression under the integral sign in (4.2)
is the I-differential form on superspace in which even commuting differential d7jis substi-
tuted by the product of two anticommuting differentials drd'l/. The substitution transforms
the I-differential form on superspace into Berezin integral. The detailed discussion of (4.2)
as Chern-Simons action is given in (Howe 1991)
TWISTORS AND SUPERSYMMETRY 117

present only some of the results received recently. In (Galperin 1992) the
general form of D = 3,4,6 and 10 superparticle action has been proposed.
It looks as follows

s= J dTdnT/Pam(DaXm - iD a0·..r0) ( 4.4)

and is the most straight-forward generalization of the action (4.2). In (4.4)

Da = a~a + iT/a t7"'
a = 1,2, ... , nand n = 1,2,4 and 8 depending on the
dimension of superspace. (4.4) is gauge invariant under local superconformal
transformations, which in the superfunction form is 6

(4.5)

where A is an unconstrained world-line superfunction. Secondly, there is a

large abelian gauge invariance
(4.6)

where the spinor parameter (abc is totally symmetric with respect to its
indices.

5. Twistor like actions for strings

To generalize the twistor-like approach to strings it is convenient to find a
twistor realization of the constraints
(5.1)
(which is analogous to constraint for the bosonic particle in the form of the
Cartan relation (1.3). It can be easily shown that such a realization exists
and has the form

(5.2)
Using (5.2), the following form of bosonic string action can he written
(Soroka 1990, Pashnev 1992)

S = J -
dTdadet(e~»\rmpae~A(a/l.Xm
T2_ b
- TArmpbe/l.A). (5.3)

Generalized to two spinors (A=1,2) the analogous to (5.3) form of the action
for the type II GS superstring has been recently received (Chikalov 1992),
which is invariant under diffeomorphism transformations on the world-sheet
superspace, the latter is achieved by using a new ingeniously constructed
6 It has been recently shown that this gauge invariance can be extended to the whole
superdiffeomorphism group (Chikalov 1992) .
118 DMITRIJ V. VOLKOV

representation of the diffeomorphism group, in which all the coordinates

of superspace and spinor superfields representing the left and right moving
modes transform simultaneously and nonlinearly.
Much attention has been paid to twistor like reformulation of the het-
erotic string action (Berkovits 1989, 1991; Tonin 1991, Aoyama 1992; Delduc
1992). Since the heterotic superstring is supersymmetric only for right mov-
ing modes it is similar, to a certain extent, to the superparticle. This con-
siderably simplifies consideration. n = (1,0), (2, 0), (4,0) and (8,0) variants
of world-sheet conformal supersymmetry has been proposed.

References
De Azcarraga, J.A. and Lukierski (1982) Supersymmetric particles with internal symme-
tries and central charges. Phys. Lett. BIl3, 170.
Aoyama, S. Pasti, P. Tonin, M. (1992) The GS and NRS heterotic strings from twistor
string models. Preprint DFPD/92/TN/9.
Aoyama, S. Kowalski-Glikman, J. van Holten, J.W. Lukierski, J. (1987) The spinning su-
perparticle. Phys. Lett. 201B, 487. Kowalski-Glikman, J., Lukierski, J. (1989) Mod.
Phys. Lett. A4, 2437
Berkovits, N. (1989) A covariant action for the heterotic string with manifest space time
supersymmetry and world sheet conformal invariance. Phys. Lett. B 232, 184.
Berkovits, N. (1991a) A supertwistor description of the massless superparticle in ten-
dimensional superspace. Nucl. Phys. B350, 193.
Berkovits, N. (1991b) The heterotic Green-Schwarz superstring on an N = (2,0) super-
world sheet. Preprint ITP- SB-91-63.
Brink, L. and Schwarz, J.H. (1981) Supersymmetric dual string theory. Nucl. Phys. B181,
502.
Chikalov, V. Pashnev, A. (1992) Twistor - like type II superstring and bosonic string.
Preprint JINR-E2/92-337.
Delduc, F., Ivanov, E., Sokatchev, E. (1992) Twistor -like superstring with D = 3,4 and 6
target superspace and N = (1,0), (2, 0), (4, 0) world-sheet supersymmetry. ENSLAPP-
L- 371/92 BONN-HE-92-11.
Galperin, A. and Sokatchev, E. (1992) A twistor - like D = 10 super particle action with
manifest N = 8 world-line supersymmetry. Preprint Jhu-TIPAC-920010 BONN-HE
92-07.
Gauntlet, J.P. (1991) A K, - symmetry calculus for superparticle EFI 91-4l.
Gol'fand, Y.A. and Lichtman, E.P. (1971) Extension of the algebra of Poincare group
generators and violation of P invariance. JETR. 13, 323.
Green, M.B. Schwarz, J.H, Witten E. (1987) Superstring theory, v.l.
Howe, P.S. and Towndsend (1991) The massless superparticle as Chern-Simons mechanics.
Phys. Lett. 259, 285.
Ivanov, E.A. and Kapustnikov, A.A. (1991) Towards a tensor calculus for K, -
supersymmetry. Phys. Lett. 267, 175.
Neveu, A. Schwarz, J .H. (1971) Factorizable dual model of pions. Nucl. Phys. B31, 86.
Pashnev, A. and Sorokin, D. (1992) Note on superfield formulation of D = 2,3,4,6 and
10 superparticle. Preprint JINR E2-92-27.
Penrose, R. (1967) Twistor algebra J. Math. Phys. 8, 345.
Penrose, R. (1977) The twist or programme. Rep. Math. Phys. 12, 65.
Ramond, P. (1971), Dual theory for free fermions. Phys. Rev. D3, 2415.
Siegel, W. (1983) Hidden local supersymmetry in the supersymmetric particle action.
Phys. Lett. 128B, 397.
Sorokin, D.P. (1990) Double supersymmetric particle theories Fortshr. der Phys. 38, 923.
TWISTORS AND SUPERSYMMETRY 119

Sorokin, D.P., Tkach, V.I. and Volkov, D.V. (1989a) Superparticles, twistors and Siegel
symmetry. Mod. Phys. Lett. A4, 901.
Sorokin, D.P., Tkach, V.I., Volkov, D.V. and Zheltukhin, A.A. (1989b). From the super-
particle Siegel symmetry to the spinning particle proper - time supersymmetry. Phys.
Lett. B216, 302.
Tonin, M. (1991) ,. ,. -symmetry as world sheet supersymmetry in D = 10 heterotic
superstring DFPD/91/TN/29.
Volkov, D.V. and Akulov, V.P. (1972) On a possible universal interaction of neutrino.
JETR Lett. 16 621.
Volkov, D.V. and Akulov, V.P. (1974) Goldstone fields of 1/2 spin. Theor. Math. Phys.
18,39.
Volkov, D.V. and Zheltukhin, A.A. (1988) Extension of the Penrose representation and
its use to describe supersymmetric models. JETP Lett. 48 61.
Volkov, D.V. and Zheltukhin, A.A. (1989) On the equivalence of the massless Dirac and
supersymmetric particles Lett. Math. Phys. 17, 141.
Volkov, D.V. and Zheltukhin, A.A. (1990) Lagrangian for massless particles and strings
with local and global supersymmetry. Nucl. Phys. B33S, 723.
Wess, J. and Zumino, B. (1974) Supergauge transformations in four dimensions. Nucl.
Phys. B70, 39.
A TWISTOR-LIKE DESCRIPTION OF D 10
SUPERSTRINGS AND D = 11
SUPERMEMBRANES
IGOR BANDOS and ALEKSANDR ZHELTUKHIN
Kharkov Institute of Physics and Technology
Kharkov, 310108, Ukraine

1. Introduction

The covariant description of the supersymmetric theories of particles and

strings is hampered by the problem of the local ",-symmetry covariant de-
scription [1,2]. Recently the new approach [3-13] for solving this problem
has been suggested, where twist or-like variables have been introduced in
addition to the world coordinates.
The introduction of twistor-like fields gives the simple solution of the
problem of covariant division for the primary Grassmannian spinor con-
straints for superparticles and superstrings. After this covariant division of
the constraints a new way is opened for the solving problem of the covari-
ant BRST-BFV quantization along the line considered in [9a,b,d,e] for null
superstrings and supermembranes in D = 4.
Here we give the general prescription for the constructing of the su-
perstring and super p-brane actions in the extended space of world co-
ordinates and twistor-like (or spinor harmonic) variables. This prescrip-
tion is reduced to constructing the realization of the Cartan moving repere
nm (T, an) == n~) (T, an) in D-dimensional space-time in terms of the gen-
eralized Newman - Penrose "dyades" or Lorentz harmonics v~ (T, an) E
Spin(I,D - 1) [14,7, 9b,e,f, 11,12]. This realization has the form

(1)

For the most interesting case D = 10 superstring we show the mech-

anism of the covariant division for the Grassmannian constraints and find
the covariant irreducible representation for the k-symmetry generators. The
harmonicity (or generalized pure spinors) conditions for D = 11 supermem-
brane are proposed.

121
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 121-127.
© 1993 Kluwer Academic Publishers.
122 LA. BANDOS AND A.A. ZHELTUKHIN

2. Super p-brane in twist or-like harmonic formulation

The suggested formulation for the super p-brane action in D-dimensional
space-time is (see also [15])

SD,N,P = J+dP 1( e [- );;; ej ( nUl WI') + c] + sl:f.~,P , (2)

where e£(f;11 = O, ... ,p) is the world-hypersheet vielbein, w;;' == 0l'xm-

iOI'OI Crml}I (I = 1, ... , N), c is a dimensionless ("cosmological") constant,
Sl:f.~,P denotes the Wess-Zumino term [16]. The symbols nUl denote a set
of (p+ 1) tangent to world hypersheet vectors of the Cartan moving repere
nee) == (nUl((),n(i)(O)

(n(e) . n(k») == n~) n(k)m = q(e)(k) = diag(l, -1, ... , -1) . (3)

The orthogonality conditions (3) are automatically satisfied if the twistor-

like representation (1) for nee) is used and the harmonicity conditions for
the majorana spinor 2[D/2l X 2[D/2l matrix v~( () are taken into account
(a,a = 1, ... ,2[D/2l).

1 Tr
-:::
~M - = S P (r(n)C-
v V ml .•• mk ) -
- 0, k = 2, ... ,[D/2] (4)

Eqs. (4) have been considered for the cases D = 4 in [7], D = 3,6,10
[12,11] and D = 11 in [gc,f] and may be treated as the generalized "pure
spinors" -type defining conditions [17].
After the substitution of Eqs. (1) into (2a) we get the twistor-like repre-
sentation for SD,N,P

SD,N,P =
= f dP+1 ( e [- R;[D/21 ej wI' Sp (v T C r vr Ul C- + c] + Slf.~,P'
1)

(5)
Let prove the classical equivalence of the representation (2) to the stan-
dard Dirac-Nambu one.
The motion equations bSD,N,P / bel' = 0 give
f

ef(O = _P_
I' cH
(WI" nUl) => -1 el'
H f
(WI" nUl) = p + 1 c
p
(6)

The variations of SD,N,P (2) with respect to the additional variables

n(e)(O or v~(O must take into account the restrictions (3) or (4). Then
A TWISTOR-LIKE DESCRIPTION OF D = 1 SUPERSTRINGS 123

the motion Eqs. 8SD,N,P / 8v~ I::::M=o may be presented in the form

(WJL. n(i)) = 0 =? n[J](O = cR eJLf wJL (7)

Using Eqs. (6) and (7) we find that

gJLv(O = e£ efv = CR) 2 (WJL . Wv )

(8)
P )P+l
e(O=dete£== ( eN Y(-l)Pdet(w JL .wv)

and after use of Eqs. (6-8) we get the Dirac-Nambu representation for
SD,N,P (2), i.e.

SD,N,P = - (~) P (Ja,r+l J dP+l(J( -l)p det (wJL. w v) + Sif.'~,P

3. D = 10 Superstring in twistor-like approach

In this case (D,N,P) = (10,2,1) and v~(O is a majorana spinor 16 X 16
matrix. Since the local plane tangent to the superstring world-sheet is two-
dimensional (p = 1) its local fixation is defined by the choice n[o] == n(O) and
n[l] = n(9) as the vector tangent to the world-sheet. This choice reduces the
local Lorentz group 50(1,9) to it's subgroup 50(1,1) X 50(8) and index
a is splitted a = (A+,A-). Here A = 1, ... ,8, A = 1, ... ,8 are the indices
of 8(s) and 8(c) spinor representations of 50(8) group, and +, - belong to
unit weight's spinor indices of 50(1, 1) local tangent group.
Now the harmonicity conditions (4) for v~(O == (V';A' v~A)' are presented
as [11]
(9a)

-=~ =- _1_ (v :-
128 O/A
(j0/{3
n
v :-) (v +
(3A pA
(jnp,\ v
.\A
+) - 1 == 0 (9b)

and the tangent light-cone vectors n[±2] == n(O) ± n(l) together with the 8
vectors n(c) orthogonal to n[±2] are parametrized as [9b,f]

m'" = ~v + (j0/{3 1
n[+2](i) V + n[-2] (i) = -v :- (j0/{3 v :-
8 O/A n {3A' m'" 8 O/A m {3A'
(9)
124 I.A. BANDOS AND A.A. ZHELTUKHIN

The orthonormality conditions (3) are automatically satisfied due to (9)

and the identity a~ai3a'Y)8m = O. Among 1261 harmonicity conditions only
210+1 = 211 ones are independent. However, due to invariance of Eqs.
(9) under the transformations from 50(1,1) x 50(8) gauge group, killing
28+ 1=29 components of v~( 0, only 16=256-211-29 of them are independent
degrees of freedom. Therefore the harmonics v~ may be considered as the
coordinates of the coset space 50(1,9)/50(1,1) X 50(8). Together with
the two non pure gauge degrees of freedom

and p[-2]T = _e_ (e T + eT ) of e~(O

,j(;i 0 1

the 16 independent components of v~(() parametrize the 18 independent

components of the two light-like combinations K 1(1 = 1,2) [9b,f]

T.,-I - P n
I\n = - ---;;-;;r
(_1)1 (!:lVaXn -
2'!:l
l Va
!:l AI)
01 an 0 + Va (lOa)

of the Virasoro reparametrization constraints

(lOb)

Therefore we prove that the introduction of the harmonic variables does

not introduce any additional degrees of freedom.
The action (2) for D = 10 N = I I B Green-Schwarz superstring is
rewritten in the form [9b,f]

(11)
Note that 5 10 ,2,1 (11) may be presented in the Chern-Simons-like form
after the redefinition of the combinations p[+2]l'n[-2] and p[-2]l'n[+2] (con-
taining the world-sheet metric e£ )
into new momentum-like variables. The
A TWISTOR-LIKE DESCRIPTION OF D = 1 SUPERSTRINGS 125

action (11) is invariant under the k-symmetry transformations [9f]

8p[+2]1l =£
ca'
f+fllVV +OOOlI
A OIA v ,

(12)

where f1, fA are the k-symmetry Grassmannian parameters, D i [± 2L the co-

variant derivatives generating the Lorentz boosts and (i[±2] are

The primary Grassmannian spinor constraints i5~ corresponding to the

k-symmetry (12)

i5~ == -7r~ + i [Pm + (_1)1/ a ' (OuXm - iOuOlamOI) 1a":(30(3I-

(13)

are the mixture of the first and second class constraints. The covariant di-
vision of these constraints may be done with the help of vA-CO and v:i+(O
harmonics.
The irreducible first class constraints D!- and D~t are the 16 generators
of the k-symmetry

(14a)

The irreducible second class constraints 1)~- and 1)~t are presented as the
following products

1)2- =
A -
'" 0
v AOl - jj201"" 1)1,+ = ~+
A -VA
jjiOI~'
~ 0 (14b)
Due to the limited volume of the report we have no possibility to repro-
duce all constraints characterizing the twistor-like representation (11).
Since all these constraints [9] are covariant and irreducible, the BRST-
BFV formalism may be employed for the covariant quantization of D = 10
lIB superstring, as it has been done for the null super p-brane (p = 0,1,2)
in D = 4 [9a,d,e],
126 LA. BANDOS AND A.A. ZHELTUKHIN

Concluding the twistor-like description of D = 10 superstring list its

motion equations

v -. e lt (+2]
crA
aIt
Ocrl =0 ,

(15)

which is to be completed by Eqs. (6) and (7).

4. D = 11 supermembrane in twistor-like approach

The local space tangent to the supermembrane «D,N,P) = (11,1,2»
world-volume is 3-dimensional and its basis is built from the repere vec-
tors nlfl == (n(O), n(9), n(IO»). Therefore the local rep ere group SO(I, 10) is
reduced to SO(I, 2) X SO(8) one. Then the majorana spinor 32 X 32 matrix
v~«() is presented as v~«() == (VO~A,Vo,A~)' where a = (1,2) is spinor index
of SO(1, 2,) group. The harmonicity conditions for D = 11 have the form
[9c,f].

==
~-
va
0
C°(3 vb(3 - 0
- cab - ,

ml···ms == Va0 (Cr ml···ms )0(3 Vb(3 (r(n)C- ) ab

::::(n) 1 =0
(16)
and reduce the number of independent variables among V~ to 55=1024-496-
11-462. Due to the invariance of Eqs. (16) under SO(I,2) X SO(8) gauge
group the resulting number of independent components of v~ equals to 24
and coincides with the dimension of the coset SO(1, 10)/ SO(1, 2) X SO(9).
SO(I, 2)x SO(8) invariant representations for cab and the r(m) == (rlfl, r(i»)
in D = 11 are

(17)

ri =( 0
lib - (i)
- f I AB

Then the twistor-like expression (2) for D = 11 N = 1 supermembrane is

A TWISTOR-LIKE DESCRIPTION OF D = 1 SUPERSTRINGS 127

presented in the form [9c,f).

SU,1,1 = J d3 (e [ c - 1
r:J e,I-' wI-'
(Ii
VO'A is f
v.6A bc -
ya'5!
(18)
- vO'Ac vpAb f
Ii c) 1'1i[Jlb (Cr)O'.6] + SW X
11,1,2 ,

where wI-' == {}I-'X - ifJ/JO' (rC- 1 t.6 ().6. The motion equations generated by
Sl1,1,2 (18) are

v . - ",Vlb el-' {} 80'

O'Ab fa ,I-'
= 0,

{}
I-' ,m
n[Jl] -
[e el-' _4_
a' c2 fl-'VP WI-'n (r mn C- 1) {} () = 0P ,

and Eqs. (6) and (7). The presence of v~(() in S11,1,2 (18) provides the
Grassmannian constraints covariant division into irreducible constraints of
the first and second class and carrying out the covariant quantization along
the line [9a,b,d,ej.

References
[1] De Azcarraga, LA. and Lukierski, l., (1982) PllYs. Lett. B113, 170; Siegel, W., ibid.
B128, (1983) 397
[2] Green, M.B., Schwarz, l.H. and Witten, E., Superstring Theory, V.l. Cambridge Univ.
Press 1987
[3] Volkov, D.V. and Zheltukhin, A.A. «1988), Pis'ma Zh. Eksp. Teor. Fiz. 48, 61; Nucl.
Phys. B335 (1990) 723
[4] Sorokin, D.P., Tkach, V.1. and Volkov, D.V., (1989) Mod. Phys. Lett. A4, 901
Sorokin, D.P., Tkach, V.I., Volkov, D.V. and Zheltukhin, A.A., (1989) Phys. Lett.
B216,302
[5] Nissimov, E., Pacheva, S. and Solomon, S., Nucl. Phys. B296 (1988), 469; B297 (1988)
349; B317 (1989) 344
[6] Berkovits, N., Phys. Lett. B232 (1989) 184; B241 (1990) 497
[7] Bandos, LA., Yad. Fiz. 51 ( 1990) 1429
[8] Bengtsson, I. and Cederwall, M., Nucl. Phys. B302 (1988) 104
[9] Bandos, LA. and Zheltukhin, A.A., Pis'ma Zh. Eksp.Teor. FIz. 51 (1990) 547 a ); 54
(1991) 421 b); 55 (1992) 81 c ); Phys. Lett. B261 (1991) 245 d ); Teor. Mat. Fiz. 88 (1991)
358 e ); Yader. Fiz. 55 (1992) 3082f)
[10] Ivanov, E.A. and Kapustnikov, A.A., Phys. Lett. B267 (1991) 175
[11] Galperin, A.S., Howe, P.S. and Stelle, K.S., Nucl. Phys. B308 (1992) 248
[12] Deldue, F., Galperin, A. and Sokatchev, E., Nucl. Phys. B368 (1992) 143
[13] Howe, P.S. and Townsend, P.K., Phys. Lett. B259 (1991) 285
[14] Penrose, R. and Rindler, W., Spinor and Space-time v. 1,2 Cambridge Univ. Press,
1986
[15] Volkov, D.V. and Zheltukhin, A.A., Ukrain. Fiz. Zhurnal, 30 (1985) 809;
Zheltukhin, A.A., Teor. Mat. Fiz. 77 (1988) 377
[16] Bergshoeff, E., Sezgin, E. and Townsend, P.K., Phys. Lett. B189 (1987) 75
[17] Howe, P., Phys. Lett. B258 (1991) 141
BORN'S RECIPROCITY IN THE CONFORMAL
DOMAIN
ARKADIUSZ J ADCZYK
Institute of Theoretical Physics, University of Wroclaw,
pl. Maksa Borna 9, PL-50-204 Wroclaw, Poland

Abstract. Max Born's reciprocity principle is revisited and complex four dimensional
Kahler manifold D4 ~ SU(2, 2)/ S(U(2) x U(2» is proposed as a replacement for space-
time on the micro scale. It is suggested that the geodesic distance in D4 plays a role of a
quark binding super-Hamiltonian.

1. Introduction
Some 55 years ago, in the Scottish city of Edinburgh, Max Born wrote
'A suggestion for unifying quantum theory and relativity'[Born, 1938], the
paper that introduced his 'principle of reciprocity'. He started there with
these words:
'There seems to be a general conviction that the difficulties of our present
theory of ultimate particles and nuclear phenomena (the infinite values of the
self energy, the zero energy and other quantities) are connected with the problem
of merging quantum theory and relativity into a consistent unit. Eddington's
book, "Relativity of the Proton and the Electron", is an expression of this
tendency; but his attempt to link the properties of the smallest particles to
those of the whole universe contradicts strongly my physical intuition. Therefore
I have considered the question whether there may exist (other possibilities of
unifying quantum theory and the principle of general invariance, which seems
to me the essential thing, as gravitation by its order of magnitude is a molar
effect and applies only to masses in bulk, not to the ultimate particles. I present
here an idea which seems to be attractive by its simplicity and may lead to a
satisfactory theory. '
Born then went on to introduce the principle of reciprocity - a primary
symmetry between coordinates and momenta. He explained that
'The word reciprocity is chosen because it is already generally used in the
lattice theory of crystals where the motion of the particle is described in the
p-space with help of the reciprocallattice.'
A year later, in a paper" Reciprocity and the Number 137. Part f', [Born,
1939] he makes an attempt to derive from his new principle the numerical

129
Z. Oziewicz et al, (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 129-140.
© 1993 Kluwer Academic Publishers.
130 ARKADIUSZ JADCZYK

value of the fine structure constant. 1 The most recent and clear exposi-
tion of the principle of reciprocity appears in his paper 'Reciprocity Theory
of Elementary Particles " published in 1949 in honor of 70th birthday of
Albert Einstein [Born, 1949]. The following extensive quotation from the
Introduction to this paper brings us closer to Born's original motivations.
'The theory of elementary particles which I propose in the following pages
is based on the current concepts of quantum mechanics and differs widely from
the ideas which Einstein himself has developed in regard to this problem.( ... )
Relativity postulates that all laws of nature are invariant with respect to such
linear transformations of space time xk = (x, t) for which the quadratic form
R = xkXk = t 2 - x 2 is invariant ( ... ). The underlying physical assumption is
that the 4-dimensional distance r = R! has an absolute significance and can
be measured. This is natural and plausible assumption as long as one has to do
with macroscopic dimensions when measuring rods and clocks can be applied.
But is it still plausible in the domain of atomic phenomena? ( ... ) I think that
the assumptions of the observability of the 4-dimensional distance of two events
inside atomic dimensions is an extrapolation which can only be justified by
its consequences; and I am inclined to interpret the difficulties which quantum
mechanics encounters in describing elementary particles and their interactions
as indicating the failure of this assumption.
The well-known limits of observability set by Heisenberg's uncertainty rules
have little to do with this question; they refer to the measurements and mo-
menta of a particle by an instrument which defines a macroscopic frame of
reference, and they can be intuitively understood by taking into account that
even macroscopic instruments must react according to quantum laws if they
are of any use for measuring atomic phenomena. Bohr has illustrated this by
many instructive examples. The determination of the distance R! of two events
needs two neighboring space-time measurements; how could they be made with
macroscopic instruments if the distance is of atomic size?
If one looks at this question from the standpoint of momenta, one encounters
another paradoxical situation. There is of course a quantity analogous to R,
namely P = = p2 =
Pkp k E2 - p2, where Pk = (p, E) represents momentum
and energy. But this is not a continuous variable as it represents the square of
the rest mass. A determination of P means therefore not a real measurement
but a choice between a number of values corresponding to the particles with
which one has possibly to do. ( ... ) It looks therefore, as if the distance P in
momentum space is capable of an infinite number of discrete values which can be
roughly determined while the distance R in coordinate space is not an observable
quantity at all.
This lack of symmetry seems to me very strange and rather improbable.
There is strong formal evidence for the hypothesis , which I have called the
principle of reciprocity, that the laws of nature are symmetrical with regard to
space-time and momentum-energy, or more precisely, that they are invariant

1 He failed, but many years later Armand Wyler [Wyler, 1968,1969,1971) obtained
a reasonable value by playing, as we shall see, with a similar geometrical idea. Wyler
failed however in another respect: he was unable to formulate all the principles that are
necessary to justify his derivation. His work was criticized (cf. [Robertson, 1971; Gilmore,
1971; Vigier, 1976)), his ideas not understood, his name disappeared from the lists of
publishing scientists.
BORN'S RECIPROCITY IN THE CONFORMAL DOMAIN 131

under the transformation

Xk ---> Pk, (I.l)

The most obvious indications are these. The canonical equations of classical
mechanics
(I.2)
are indeed invariant under the transformation (1.1), if only the first 3 compo-
nents of the 4-vectors xk and Pk are considered. These equations hold also in
the matrix or operator form of quantum mechanics. The commutation rules

(I.3)

and the components of the angular momentum,

(I.4)

show the same invariance, for all 4 components. These examples are, in my
opinion, strongly suggestive, and I have tried for years to reformulate the fun-
damentallaws of physics in such a way that the reciprocity transformation (1.1)
is valid ( ... ). I found very little resonance in this endeavor; apart from my col-
laborators, K. Fuchs and K. Sarginson, the only physicist who took it seriously
and tried to help us was A. Lande ( ... ). But our efforts led to no practical re-
sults; there is no obvious symmetry between coordinate and momentum space,
and one had to wait until new experimental discoveries and their theoretical
interpretation would provide a clue. ( ... ) There must be a general principle to
determine all possible field equations, in particular all possible rest masses.( ... )
I shall show that the principle of reciprocity provides a solution to this new
problem - whether it is the correct solution remains to be seen by working
out all consequences. But the simple results which we have obtained so far are
definitely encouraging ( ... ).'
2 The very problem of a serious contradiction between quantum theory and
relativity was addressed again, in 1957, by E.P. Wigner in a remarkable
paper 'Relativistic Invariance and Quantum Phenomena', [Wigner, 1957].
Wigner starts with the assertion that 'there is hardly any common ground
between the general theory of relativity and quantum mechanics'. He then
goes on to analyze the limits imposed on space-time localization of events
by quantum theory to conclude that:
'The events of the general relativity are coincidences, that is collisions
between particles. The founder of the theory, when he created this con-
cept, had evidently macroscopic bodies in mind. Coincidences, that is,
collisions between such bodies, are immediately observable. This is not
2 It must be said that later on, in his autobiographical book 'My life and my views',
[Born, 1968], Born hardly devoted more than a few lines to the principle of reciprocity.
Apparently he was discouraged by its lack of success in predicting new experimental facts.
132 ARKADIUSZ JADCZYK

the case for elementary particles; a collision between these is something

much more evanescent. In fact, the point of a collision between two el-
ementary particles can be closely localized in space-time only in case of
high-energy collisions.'
3

Wigner analyzes the quantum limitations on the accuracy of clocks, and

he finds that "a clock, with a running time of a day and an accuracy of
10- 8 second, must weigh almost a gram-for reasons stemming solely from
uncertainty principles and similar considerations". 4

2. Reciprocity the Twistor Way

Max Born's original idea of reciprocity was clear but imprecise. We will try
to interpret it using more modern concepts. The interpretation below is ours.
And so are its faults.

2.1. INTERPRETATION

We will interpret the reciprocity symmetry (1.1) as a tangent space symme-

try rather than as a global one. So, we assume that the fundamental arena
D in which relativistic quantum processes take place is an 8-dimensional
manifold with local coordinates (xll-,pll-). The symmetry (Ll) should hold
in each tangent space. Since the square of the operation (1.1) is -1, we in-
terpret (1.1) as the requirement that D should be equipped with a complex
structure, which is respected by the fundamental equations. It is clear from
Born's papers that D should be also endowed with a metric tensor. The sim-
plest complex Riemannian manifolds are those that are Kiihlerian symmet-
ric domains. I choose the Cartan domain D4 ::::! SU(2, 2)/ S(U(2) x U(2)) ::::!
SO( 4,2)/ S( 0(2) x 0(2)) as the candidate. It has many nice properties -
some of them will be discussed later. There are also many possible objec-
tions against such a choice. Let me try to anticipate some of them.
D4 has positive-definite metric - it cannot contain Minkowski space
= True, indeed. On the other hand one can argue that according to
Born's original idea, and according to the analysis by Wigner, Min-
kowski space-time of events is only an approximation. High-energy
or high-mass approximation. Thus it is reassuring that the Shilov
3 I will return to this conclusion when interpreting space-time as the Shilov boundary
of the conformal domain D 4 •
4 In 1986 Karolyhazy et al. in the paper 'On the possible role of gravity in the reduction
of the wave function', [Karoly hazy, 1986], presented another analysis of the imprecision
in space-time structure imposed by the quantum phenomena. They proposed 'to put the
proper amount of haziness into the space-time structure'. Their ideas, as well as the ideas
of a "stochastic space-time" most notably represented by E. Prugovecky (d. [Prugovecki,
1991]) and references therein) all point in a similar direction.
BORN'S RECIPROCITY IN THE CONFORMAL DOMAIN 133

boundary of D4 (the important concept that will be discussed later)

is naturally isomorphic to the (compactified) Minkowski space en-
dowed with its indefinite conformal structure. Let us interpret the
points of D4 as elementary micro event-processes, that is micro-
events accompanied by energy transfers. A coordinate of such an
event is zk = xk + hpk /p2, with p2 = E2 - p2 > 0 (see Sec. ???).
In the limit of large energy transfers h 2/p2 --+ 0 the positive definite
metric blows up. What remains is the Minkowskian conformal metric
for zk = xk + 0 - the finite part of the Shilov boundary. The positive
definiteness of the Riemannian metric on D4 can be thus viewed as
an advantage rather than as a fault.
D4 is not invariant under time inversion.
= Indeed, time inversion is not a symmetry of D4 - it would change the
complex structure into the opposite one. We will see that when real-
izing D4 as a part of the Grassmannian in C 4 one gets automatically
two copies of the domain. Then time inversion can be thought of as
the transposition of these two copies. We consider D4 as useful for the
modeling of physical processes on a micro-scale (say, inside mesons
and hadrons). We know that on this scale time-inversion need not
be a symmetry. On the other hand such a primary arrow-of-time on
a micro-scale may well be connected with the observed macroscopic
irreversibility as dealt with in thermodynamics. Thus breaking of the
time-inversion symmetry can also viewed as an advantage rather than
as a fault.
D4 has constant curvature and it is hard to imagine how models based
on D4 can be constructed that include gravity and/or gauge fields.
= True, one of the original reasons for introducing the principle of reci-
procity was unification of gravity and quantum theory. On the other
hand, let us recall that, according to Born, gravitation 'is a molar
effect and applies only to masses in bulk, not to the ultimate par-
ticles. 'If so, and according to our interpretation above, there is no
place for gravitation (and for other gauge fields as well) inside a me-
son or a hadron. Of, course, one could object that then there is also
no place for space,time,energy and for momentum. It is of course an
extrapolation, perhaps unjustified, that these concepts apply to such
a micro-scale. However, extrapolating Einstein's scheme of general
relativity into this domain would be unjustified even more. Therefore
the idea that the primary arena of elementary event-processes is ho-
mogenous under a sufficient "zoom" may be rather attractive than
appalling. 5
There is nothing new in the idea. Everything has been already said.

5 I heard this idea from Rudolph Haag.

134 ARKADIUSZ JADCZYK

= This objection is a serious one. There are extensive papers dealing

with the domain D 4 , mainly by Roger Penrose and his group (cf. [Pen-
rose and Rindler, 1986] and references there), but also by Odzijewicz
and collaborators (see [Odzijewicz, 1976; Karpio et al., 1986; Odzi-
jewicz, 1988] and references there), and by Unterberger [Unterberger,
1987]. Many of these papers are too difficult for me to understand all
their conclusions. Therefore there is a chance that the ideas presented
here are simplistic and naive, mainly owing to my inadequate knowl-
edge. If so, I will beg your pardon, and I will do my best to (at least)
present those ideas that, I believe, deserve propagation. 6

3. Algebraic description of the conformal domain D4

There are many ways of describing the same domain D 4 . I choose the alge-
braic description because it is simple. On the other hand it so happens that
many years ago I studied its geometry, by algebraic means, without being
fully aware of the full impact of the study [Jadczyk, 1971].
Let V be a complex vector space of complex dimension n = p+q, equipped
with a Hermitean scalar product ( , ) of signature (p, q). The domain D;; is
then defined as the manifold of p dimensional, positive linear subspaces of
V7. In the following we will write Dn to denote D;;. Let L(V) denote the
algebra of linear operators on V. For each subspace W E Dn let Ew denote
the orthogonal projection on W, and let Sw == 2Ew - I. Then Sw = Sw,
S?v = I, and (v, w)sw == (v, Sww) is a positive definite scalar product on
V. The last statement follows from the fact that Sw reverses the sign on
Wl.. Conversely, if S E L(V) satisfies the three conditions above, then the
subspace W == {v : Sv = v} is in Dn and S = Sw. Geometrically, Sw plays
the role of a geodesic reflection symmetry with respect to the point W E Dn.
The parametrization ofthe points of Dn through their symmetries is in many
respects the most convenient one - the fact that is little known! Whenever
we speak about a point of D n , we have in mind one of its representing
objects: subspace W, projection E, or symmetry operator S. We will use the
'*'symbol to denote the Hermitean conjugate with respect to the indefinite
scalar product on V. Given S E D n , the Hermitean conjugate of Y E L(V)
with respect to the positive-definite scalar product (u, v)s will be denoted
by yS. Notice that yS = Sy*S, y* = Sys.
It is evident from the very definition that the unitary group U(V) of
(V,(, )), which is isomorphic to U(p,q), acts transitively on Dn with the

6 A review with a different emphasis can also be found in [Coquereaux and Jadczyk,
1990]
7 The orthocomplements of the subspaces from D~ are q dimensional negative sub-
spaces. They form D;;. For p = q this is the second copy of D~ - as mentioned in the
discussion of time inversion above.
BORN'S RECIPROCITY IN THE CONFORMAL DOMAIN 135

stability group U(p) X U(q). The same is true about SU(p,q), which acts
effectively on D n , so that

Dn ~ SU(p,q)jS(U(p) X U(q))

. By differentiating the defining equations

S = S*, (1)

of Dn we find that the tangent space Ts at S can be identified with the set
of operators X E L(V) such that

X = X*, and X S + SX = o. (2)

Suppose now that p = q, thus n =

2p (the most symmetric case). Call
a basis {ei} in V isotropic if the scalar product of V in this basis reads
(v, w) = vtGw , where G is the block matrix

G = (O.p ilp) . (3)

-zip Op
Fix an isotropic basis, then Dn is isomorphic to the space of all pxp complex
matrices T such that

i(T* - T) > 0, (4)

the isomorphism W ¢:::::> T being given by

(5)

This parametrization defines complex structure on Dn. In terms of the op-

erators X of Eq.(2) the complex structure Js of the tangent space Ts at S
is given by the map Js : X -+ iX S. Notice that (in the chosen isotropic
basis) the orthogonal subspace to WT is

(6)

Dn is naturally equipped with an U(V) - invariant positive definite Rieman-

nian metric:

g(X, Y)s == -Tr(XY), X,Y E Ts. (7)

That g is positive definite follows from X = X* = _X s , thus

g(X, X)s = -Tr(XX) = Tr(Xs X) > 0 for X ~ o. (8)

136 ARKADIUSZ JADCZYK

Dn carries also an U(V) - invariant symplectic structure w: 8

w(X, Y)s == g(X, JsY)s = iTr(SXY). (9)

Dn is a homogeneous Kahlerian manifold. For p = q = 2 its interpretation
as a conformal-relativistic phase space comes from the T-parametrization:
9 with T as in Eq. (5), we write

ql-'
T = tl-'(11-' = (xl-' + -q2 )(11-" (10)

where ql-' = inpl-', and (11-' = {Iz, (1} are the Pauli matrices. The condition
(4) reads now p2 = (pO)2 _p2 > O. Thus topologically, and also with respect
to the action of the Poincare group, D4 is nothing but the future tube of
the Minkowski space, endowed with a nontrivial Riemannian metric. It is
to be stressed that special conformal transformations act on the variables
pI-' not in the way one would normally expect. Thus (xl-', pI-') refer to some
extended process rather than to a point event. Till now no interpretation of
the points of D4 in terms of space-time concepts, i.e. an interpretation that
would explain their transformation properties, has been given.
The second important representation of Dn is as a bounded domain in
C p2 • This representation can be obtained via the Cayley transform from the
T -representation:
T - i 'T _ Z + i
Z=iT+i' z - Z -z.' (11)

Geometrically, Z can be thought of as an orthogonal graph of the subspace

WT with respect to a fixed subspace Wo = W{T=i}' The condition (4) reads
now zzt < I. The topological boundary aDn is (p2 - 1) dimensional. The
Shilov boundary tJDn is defined as consisting of those points of aDn at
which functions analytic on the domain reach their maxima. tJDn is isomor-
phic to the set of p X P unitary matrices; thus, for p = 2, to the compactified
Minkowski space. tJDn carries a unique U(p,p)-invariant conformal struc-
ture of signature (p,p). For p = 2 - the one induced by a flat Minkowski
metric. The Cayley transform maps Minkowski space tl-' = xl-' + 0 onto the
finite (affine) part of tJD 4 • We see from Eq. (10) that Minkowski space can in-
terpreted as the very-high-mass, or very-high-energy-momentum-transfer
limit of D 4 • Elementary micro-processes that are characterized by very high
energy-momentum transfers can be described as pure space-time events. It is
only for such processes that the standard concepts of space, time and causal-
ity are applicable. For generic micro-processes there is no distinction between
8 Although it is clear that w is a non-degenerate, U(V)-invariant two-form, to prove
that it is closed needs a computation.
9 A justification for such a parametrization can be found in [Odzijewicz, 1976), [Co-
quereaux and Jadczyk, 1990)
BORN'S RECIPROCITY IN THE CONFORMAL DOMAIN 137

space and time, no distinction between space-time and energy-momentum.

This would be an extreme manifestation of the Born reciprocity idea! Thus,
we propose to consider D4 as the replacement for space-time on the micro
scale. In an analogy to the harmonic oscillator, the (square of) geodesic dis-
tance in D4 may playa role of the quark binding super-Hamiltonian. One
obtains in this way, again in the spirit of Born's reciprocity, an interesting
and non-trivial version of the relativistic harmonic oscillator. Here we can
only sketch the idea. 10
Given two points 5,5' in D n , the fundamental two-point object is the
unitary operator t( 5',5) == (5' 5) ~. Many of the algebraic properties of
these operators (including the case of n = 00) have been studied in (Jadczyk,
1971]. Notice that t(5',5) is unitary w.r.t the indefinite scalar product of
V, but positive w.r.t both p.d. scalar products (u,v)s, (U,V)SI. In the next
paragraph we will show that the map

X 1--+ t(5'5)Xt(5' 5)*

is the geodesic transport from the tangent space at 5 to the tangent space
at 5'.

3.1. REDUCTIVE DECOMPOSITION OF U(V)

For the Lie algebra of U(V) we have:
Lie(U(V)) = {Y E L(V) : Y = -Y*}, (12)
while L(V) coincides with the complexified Lie( U(V)). The Killing form
B(X, Y) is then given by
B(X, Y) ~ Tr(XY). (13)
Given 5 E D n , the isotropy subalgebra J(s at 5 is
Ks = {X E L(V): X* = -X,[X,5] = a}. (14)
Every X E L(V) can be uniquely decomposed as

X = xt + iX s,
where
[XP,5]'f = O.
The decomposition is given by

xt = ~(5X5 + X),
10 More can be found in the forthcoming Thesis of W. Mulak (cf. also [Mulak, 1992] for
an SU(l,l) version)
138 ARKADIUSZ JADCZYK

_ t
Xs = 2(SXS - X).

We have (X*)~ = (xt)*, and also

VX,Y E L(V).
Therefore the orthogonal complement of J(s w.r.t. the Killing form B(X, Y)
is the subspace Ms C Lie(U(V)) given by

Ms = {X E L(V): X* = -X,XS+ SX = O}.

3.2. t(S', S) AS THE GEODESIC PARALLEL TRANSPORT

We will show that t( S', S) implements parallel transport from the tan-
gent space at S to that at S', and also how it can be used for comput-
ing of the geodesic distance between the two points. First notice that each
geodesic through S is generated by a unique element X E Ms as follows (cf.
[Kobayashi, 1969], p.192):
t I----> Set) = e tX Se- tX = e2tX S, (15)
the last equality follows from X S + SX = O. If Yet) is a parallel vector field
along Set), then (because Dn is a symmetric space; see [Chavel, 1972], p.64)
Yet) = S(tj2)S(0)Y(0)S(0)S(tj2), (16)
which by (15 ) gives
Yet) = etxY(O)e- tx . (17)
On the other hand
t(S(t),S(O)) = (S(t)S)~ = (e2tX)~ = etX , (18)
and so
Yet) = t(S(t), S)Y(O)t(S(t), S)-1, (19)
which proves that t(S(t)S) is the parallel transport operator. To find the
geodesic distance formula, notice that e2tX S is a geodesic through S with
the tangent vector field S = 2Xe 2tX S of length -Tr(S2) = 4Tr(X2). For
Tr(X2) = ~, Set) is parametrized by its length. But, from Eq.(18), we have
that tX = In t(S(t)S), t 2X2 = In2t(S(t)S), thus
dist(S, Set)) = t = 4Tr(ln2t(S(t)S))), (20)
or
dist(S, S') = Tr(ln 2(SS')). (21)
BORN'S RECIPROCITY IN THE CONFORMAL DOMAIN 139

4. Conclusions: quantum conformal oscillator

The relativistic quark model based on the Lorentz-covariant harmonic os-
cillator has been considered by many authors (cf. [Kim and Noz, 1991], and
references there). Extending Max Born's reciprocity principle we propose to
investigate a similar model, but based on the geometry of D 4 •
For simplicity let us consider the spinless two-body problem in D 4 . Quan-
tum states of the two-body system will be described by analytic functions l l
W(S, S') on D4 X D 4, integrable with respect to an appropriate invariant
measure. We take for super-Hamiltonian H of the system the Toeplitz pro-
jection of dist(S, S')2. One can prove that by introducing the 'center of mass
'coordinates, the problem reduces to a one body problem. The spectrum of
H can be computed in terms of the coherent states on D4 (cf. [Mulak,
1992]). Such a model is nonrealistic, as it does not take into account spin.
To consider spinning quarks we have to take for a model Hilbert space the
space of sections of an appropriate vector bundle. The most natural one is
the holomorphic tautological bundle Q+ that associates to each S E D4 the
subspace Ws = {u E V : Su = u}. This bundle is endowed with a natu-
ral Hermitean connection. The operators t(S, S') provide a natural parallel
transport also in this bundle. Using its natural connection a Dirac-like op-
erator can be constructed on Q+. Much work must still have been done in
order to see if models constructed along these lines have anything to do with
reality.

References
Born, M., 'A suggestion for unifying quantum theory and relativity', Proc. Roy.Soc.,
AI65:291-303, 1938
Born, M., 'Reciprocity and the Number 137. Part I', Proc. Roy. Soc. Edinburgh, 59:219-
223, 1939
Born, M., 'Reciprocity Theory of Elementary Particles', Rev. Mod. Phys., 21:463-473,
1949
Born, M., 'My Life and my Views', Charles Scribner's Sons, New York, 1968
Chavel, 1., 'Riemannian Symmetric Spaces of Rank One' Marcel Dekker, Inc. New York
1972
Coquereaux, R. and A. Jadczyk, 'Conformal theories, curved phase spaces, relativistic
wavelets and the geometry of complex domains', Rev. Math. Phys., 2:1-44, 1990
Gilmore, R., 'Scaling of Wyler's Expression for 0", Phys. Rev. Lett., 28:462-464
Jadczyk, A., 'Geometry of indefinite metric spaces', Rep. Math. Phys, 1:263-276, 1971.
Karolyhazy, F., A. Frenkel, and B. Lukacs, 'On the possible role of gravity in the reduction
of the wave function', in Quantum Concepts in Space and Time, Ed. by R. Penrose
and C. J. Isham, Clanderon Press, Oxford 1986
Kim, Y.S. and M.E. Noz, Phase Space Picture of Quantum Mechanics World Scientific,
Singapore 1991
Kobayashi, S., and K. Nomizu, 'Foundations of Differential Geometry, Volume II' Inter-
science Pub!., New York-London-Sydney 1969

11 More precisely: by holomorphic densities

140 ARKADIUSZ JADCZYK

Karpio, A., A. Kryszen and A. Odzijewicz, 'Two-twist or conformal Hamiltonian spaces',

Rep. Math. Phys. ,24:65-80,1986
Mulak, W., 'Quantum SU(l,l) Harmonic Oscillator'. in Proc. XXV Symp. Math. Phys.,
Torun 1992, Preprint 1FT UWr, Wroclaw 1993 - to appear
Odzijewicz, A. , 'A Model of Conformal Kinematics', Int. J. Theor. Phys. ,15: 576-593,
1976
Odzijewicz, A. , 'On reproducing kernels and quantization of states', Commun. math.
Phys. , 114:577-597, 1988.
Penrose, R., and W. Rindler, 'Spinors and Space-Time. Volume 2: Spinor and Twistor
Methods in Space-Time Geometry', Cambridge University Press, 1986
Prugovecki, E., 'Geometro-Stochastic Locality in Quantum Spacetime and Quantum
Diffusions', Found. Phys., 21:93-124
Robertson,B., 'Wyler's Expression for the Fine-Structure Constant 0", Phys. Rev. Lett.,
27:1545-1547
Unterberger,A., 'Analyse Harmonique et Analyse Pseudo-differentielle du Cone de la
Lumiere', Asterisque, 156, 1987
Vigier,J-P., 'On the geometrical quantization of the electric charge in five dimensions'and
its numerical determination as a consequence of asymptotic SO(5,2) group invariance,
in Call. Int. CNRS, 237:411-418, 1976
Wigner,E.P., 'Relativistic Invariance and Quantum Phenomena', Rev. Mod. Phys.,
29:255-269, 1957
Wigner,E.P., 'Geometry of Light Paths between Two Material Bodies', J. Math.
Phys.,2:207-211 , 1961
Wyler,A., 'On the Conformal Groups in the Theory of Relativity and their Unitary
Representations', Arch. Rat. Mech. and Anal.,31:35-50, 1968
Wyler,A., 'L'espace symetrique du groupe des equations de Maxwell' C. R. Acad. Sc.
Paris,269:7 43-7 45
Wyler,A., 'Les groupes des potentiels de Coulomb et de Yukava', C. R. Acad. Sc.
Paris,271:186-188
SELF-DUAL EINSTEIN SUPERMANIFOLDS
AND SUPERTWISTOR THEORY

SERGEY A. MERKULOV·
Department of Mathematics and Computer Science, Odense University, Campusvej 55,
Df(·5230 Odense M, Denmark

Abstract. A supersymmetric generalization of Penrose's non-linear graviton construction

is presented.

Key words: Supertwistor - Supermanifold - Supergravity

1. Self-dual Einstein supermanifolds. Let M be a complex (414)-
dimensional supermanifold equipped with a superconformal structure
(Manin 1984) which means a pair of integrable rank-Ol2 distributions 'TjM
and TrM satisfying the conditions:
the sum of TIM and TrM in T M is direct;
the Frobenius form
ct>: TIM ® TrM --+ ToM = TM/(TIM + TrM)
X ® Y --+ [X, Y] mod (TIM + TrM)
is an isomorphism.
The rank-(210) holomorphic vector bundles, S = IITIM and !i = IITrM,
II denoting the parity change functor, are called spinor bundles on M. The
tangent bundle to the manifold Mred underlying a superconformal superman-
ifold M factors as a tensor product, T Mred = Sred ® !ired, of two rank-(210)
vector bundles, and hence Mred comes equipped with an induced conformal
structure. It is also clear that the second Stiefel-Whitney cohomology class
of Mred vanishes. In fact any such a conformal 4-manifold is a reduction of
some (non-unique) conformal (414)-supermanifold.
PROPOSITION 1. There is a covariant functor from the category of 4-
dimensional conformal manifolds with vanishing second Stiefel- Whitney
class to the category of ( 414 )-dimensional conformal supermanifolds.
Any 4-dimensional conformal manifold has a distinguished family of curves
called null geodesics and defined usually as solutions of some second-order
differential equation. An analogous concept in supergeometry can be intro-
duced without referring to any connection and associated differential equa-
tions .
• On leave from Theoretical Problems Department, Russian Academy of Sciences,
Moscow, Russia

141
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Qunntum Deformations, 141-146.
© 1993 Kluwer Academic Publishers.
142 SERGEY MERKULOV

DEFINITION 2 (Merkulov 1991a). A null supergeodesic L in a conformal

supermanifold M is a (112)-dimensional subsupermanifold ec;uipped with a
pairof(Oll)-dimensional distributions TtL C TtMIL andTrL C TrMIL such
that their sum in T L is direct and the Frobenius form

if! : TtL ® TrL ---+ ToL = T L/(TtL + TrL)

X ®Y ---+ [X, Y] mod (TtL + TrL)

is non-degenerate.

Thus a holomorphic null supergeodesic in M has an induced structure of

N =2 SUSY curve (Cohn 1987). Since distributions TtM and TrM are in-
tegrable, they define a pair of (412)-dimensional supermanifolds, M t and
M r , whose structure sheaves, OMI and OMr , are those subsheaves of OM
which are annihilated by vector fields from TrM and TIM respectively. The
embeddings OM"r C OM define canonical projections

A supergeometry analogue of the notion of metric is the notion of scale.

DEFINITION 3 (Manin 1984). A scale on a conformal supermanifold M is

a choice of particular non-vanishing volume forms on supermanifolds MI
and Mr.

A choice of scale on M induces both a volume form on M (thus giving a

well-defined integration theory of functions) and non-degenerate symplectic
forms on spinor bundles S and S( thus generating a metric on the underly-
ing 4-manifold Mred). A scale on M determines also a unique Levi-Civita
superconnection (Ogievetsky and Sokatchev 1981) with torsion and curva-
ture tensors being expressed in terms of algebraically independent Weyl
superfields W ao /3" and Wa/Fr' Einstein's superfield gaoco and Ricci scalars R
and R (here and throughout the paper undotted and dotted Greek indice~
take values 1,2 and refer to some trivializations of spinor bundles Sand S
respectively). Now we have a series of definitions mimicking the standard
terminology of classical Riemannian geometry:
A conformal supermanifold is called conformally self-dual if Hi = O.
A scaled conformal supermanifold is called Einstein's if g = O.
A scaled conformal supermanifold is called self-dual Einstein's if W =
g = O.
n
A self-dual Einstein's supermanifold is called self-dual Ricci's if = O.
An Einstein supermanifold is a solution of Einstein's N =1 supergravity
dynamical equations with cosmological constant. 2. Integrable cor.ical
CURVED SUPERTWISTOR THEORY 143

structure on a conformally self-dual supermanifold. Let F be a rel-

ative projective line bundle, PM(5), on a conformal supermanifold M. The
exact sequence
0--+II5 + II5--+T M --+5 ® 5--+0

produces a canonical (213)-dimensional conical structure (Manin 1984) on

M, i. e. an embedding F '---+ G M(213; T M) which is given in a local structure
frame (Ogievetsky and Sokatchev 1981, Manin 1984) as a map

F --+ GM(213;TM)
[11"0,] --+ span(1I"o,'\7 0'0" '\70" '\70,)

where 11"0, are homogeneous coordinates in a fibre of the bundle v : F -+ M.

A Levi-Civita superconnection associated with some scale on M induces
a (213)-dimensional conical superconnection on F which actually does not
depend on a choice of scale used in the construction. The integrability con-
dition for this conical superconnection is the equation W = 0 (Merkulov
1991a,1992a). Supposing that M is conform ally self-dual and sufficiently
"small", we obtain a double fibration

with leaves of the integrable (213)-conical connection as fibres of J.l. The

resulting (311)-dimensional superspace Z is called a twistor superspace as-
sociated with a conformally self-dual supermanifold M. In the flat case
M = F(210, 211; C 4 Il), F = F(ll0, 210, 211; C 4 1l) and Z = G(110; C 4 Il). 3.
Deformations of a standardly embedded rational curve. The projec-
tion J.l : F--+Z embeds fibres of v into Z with one and the same normal
bundle N which fits into an exact sequence

Here 0(1) is the hyperplane section bundle on Cpllo. Thus Z comes

equipped with a (412)-dimensional family of standardly embedded ratio-
nal curves. In fact this family encodes full information about the original
(414 )-dimensional conform ally self-dual supermanifold M.

THEOREM 4. Given any embedding of a rational curve P into a (311)-

dimensional complex supermanifold Z with normal bundle N which fits into
an exact sequence,
O--+Nl--+N --+No--+O
with Nl ~ IIO(l) and N2 ~ C 2 ® 0(1), then there is an associated confor-
mally self-dual (414 )-supermanifold M.
144 SERGEY MERKULOV

Proof. Let Oz be the structure sheaf of the supermanifold Z and J C Oz

an ideal of functions vanishing on the subsupermanifold P ~ C plio,

Then,
O__ J/J 2 __ OZ/J 2 __ 0p __ O,

and, by definition of the conormal bundle,

Hence we can define a sheaf OP' on P by an exact sequence

which in turn fits into an exact sequence

Therefore the pair pI = (P, Op') is a complex supermanifold of dimension

111 which contains P as a subsupermanifold. Since Ni = nO( -1), pI is
biholomorphic to the projective superspace C pIll. Thus we conclude that,
for any rational curve C pliO standardly embedded into a complex (311)-
supermanifold Z, there is a projective superspace Cplll which contains this
curve as a subsupermanifold and is embedded into Z with normal bun-
dle C 2 121 0CP1Il(1) (cf. Merkulov 1991a). Now consider the projectivized
cotangent bundle pz(n l Z) of Z. Following ideas of LeBrun (1986) we de-
fine a subsupermanifold P" c pz(n l Z) as consisting of those I-forms on
Z which vanish when restricted on T pl. Since the normal bundle of the
embedding pI '-+ Z is isomorphic to C 2 121 0CP I I1(l), P"--pl is a trivial
Cpllo-bundle over pI and hence over P. One may check (cf. LeBrun 1986,
1991) that the quadric Q = P" Ip is embedded into pz(n l Z) with normal
bundle N ~ nO(l,O) + nO(O, 1) + J l O(l, 1). Relative deformations of such
quadrics have been investigated by McHugh (1991) who proved that the
locally complete (414 )-dimensional parameter family, M, of deformations of
Q in pz(n 1 Z) comes equipped with a conformal structure,

We recognize M/ as a (412)-supermanifold parameterizing relative deforma-

tions of P, while Mr as a (412)-supermanifold parameterizing relative de-
formations of pI in Z. It is also clear that M comes equipped with an
integrable (312)-conical superconnection. Then the theorem follows from the
fact (cf. Merkulov 1991 b) that such a superconnection always admits a lift to
CURVED SUPERTWISTOR THEORY 145

a Levi-Civita superconnection. 4. Twistor transform of self-dual Ein-

stein and Ricci supermanifolds. Let OF( -1) be the tautological sheaf
on F. Consider the composition

V' F/Z : OF 1I~) OF( -1) ® v* (II(BerM1t 1/ 6 ® n1 M) id0~res

id0~res OF( -1) ® v*II(BerMI)-1/6 ® n1F /Z,
where V' is a Levi-Civita superconnection and res denotes restriction of 1-
forms on F on I-l-vertical vector fields, and define an invertible holomorphic
sheaf, L = 1-l*(kerV'F/z), on Z.
THEOREM 5. There is a one-to-one correspondence between scales on a
conformally self-dual supermanifold M satisfying Einstein's equations 9 = 0
and nowhere vanishing sections of nl Z ® (L *)2 on Z.
This theorem (proved in Merkulov (1991a)) is a supersymmetry extension of
the result due to Ward (1980). The family of rational curves in Z which lie
in the kernel of a global section of n1 Z ® (L *)2 have a special meaning (cf.
LeBrun 1982) - they generate a (312)-dimensional conformal supermanifold,
Y, as defined in Merkulov (1992b). There exists an inverse construction (cf.
LeBrun 1982).
THEOREM 6. Let Y be a (312)-dimensional conformal supermanifold.
There is an associated (414)-dimensional self-dual Einstein supermanifold
M.
Proof. If Y is (312)-dimensional conformal supermanifold, there is an as-
sociated (311 )-dimensional "ambitwistor" superspace Z parameterizing null
supergeodesics of Y (Merkulov 1992b). The supermanifold Z has a family
of rational curves embedded with normal bundle N ~ nO(l) + C 2 ® 0(1).
Hence Theorem 4 can be used to generate a conform ally self-dual (414)-
supermanifold M from this structure. Moreover Z comes equipped with
a contact structure which gives a nowhere vanishing global section of
nl Z ® (L*)2. Hence, by Theorem 5, the supermanifold M satisfies Einstein's
equations.
REMARK 7. Since a general (312)-dimensional conformal supermanifold Y
is specified locally by metric and gravitino fields on ¥red defined up to gen-
eral coordinate and SUSY-transformations, we infer from Theorem 6 that a
solution of self-dual Einstein equations for N = 1, D = 4 supergravity can
be constructed for each choice of 3 even and 2 odd analytic functions of 3
variables.
Let T be a section of n1z ® (L*)2 associated with a self-dual Einstein su-
permanifold M and v*(V') a lift of an associated Levi-Civita connection to
F.
146 SERGEY MERKULOV

LEMMA 8. A (414 )-dimensional distribution on F which is annihilated by

the lift Jl*( T) of the section T coincides precisely with the horizontal distri-
bution v*(V').

This lemma implies

THEOREM 9. A self-dual Einstein supermanifold is a self-dual Ricci super-
manifold if and only if the (211 )-dimensional distribution on Z annihilated
by T is integmble.
The latter statement provides a supersymmetry generalization of Penrose's
(1976) non-linear graviton construction.

References
Cohn, 1. D. : 1987, 'N=2 super Riemann surfaces', Nucl. Phys. B284, 349-364.
LeBrun, C. R. : 1982 'H-space with a cosmological constant', Pmc. R. Soc. Lond. Ser.
A380, 171-185.
LeBrun, C. R. : 1986, 'Thickenings and gauge fields', Class. Quantum Grav. 3, 1039-1059.
LeBrun, C. R. : 1991, 'Thickenings and conformal gravity', Commun. Math. Phys. 139,
1-43.
Manin, Yu. 1. : 1984 Gauge Field theory and Complex Geometry. Nauka: Moscow [English
trans!.: Berlin: Springer 1988].
McHugh, A. P. : 1991 'The space for super light rays for complex conformal spacetimes',
Ph. D. Thesis, State University of New York at Stony Brook.
Merkulov, S. A. : 1991a 'Geometry of simple supergravity and supertwist or theory', In
Classical Field Theory and Gravitation Theory, VINITI: Moscow, 3, pp.59-107 [In
Russian].
Merkulov, S. A. : 1991 b 'Twistor transform of Einstein- Weyl superspaces', Class. Quantum
Grav. 8, 2149-2162.
Merkulov, S. A. : 1992a 'Supersymmetric non-linear graviton' , Funct. Anal. and Its Appl.
26, no. 1 .
Merkulov, S. A. : 1992b 'Superconformal geometry in three dimensions', J. Math. Phys.
33, 735-757.
Penrose, R. : 1976 'Non-linear gravitons and curved twistor theory', Gen. ReI. Grav. 7,
31-52.
Ward, R. S. : 1980 Self-dual space-times with cosmological constant, Commun. Math.
Phys. 78, 1-17.
AN APPROACH TO THE CONSTRUCTION OF
COHERENT STATES FOR MASSLESS
PARTICLES
ANDRZEJ KARPIO
Institute of Physics, Warsaw University Branch,
15-441 Bialystok, ul. Lipowa 41, Poland

1. INTRODUCTION

My intention is to present some results related to the construction

of coherent states [Perelomov, 1987] in a Hilbert space whose elements are
cohomology classes. Such spaces appear in twistor theory and they have very
important physical aplications as the quantum spaces of massless particles.
This work is a small part of my doctoral thesis which will be published soon.
It contains details and proofs of all facts I mention below.
I will restrict my considerations to the manifold of positive, projective
twistors PT+ [Wells, 1979]. From a physical point of view, it is the phase
space of massless particles with helicity greater then zero [Hughston, 1979]
[Tod, 1977] [Karpio, 1986]. The quantization procedure leads to the first
cohomology group Hl(PT+,O(-n - 2)) [Penrose, 1977].
In order to introduce the structure of the Hilbert space one can use the
scalar product given in [Eastwood, 1981] [Ginsberg, 1983]. It was formulated
for Czech's cohomologies and I mean just this realisation when speaking
about cohomology groups. It turns out that this Hilbert space is the first
cohomology group of covering PT+ by two open subsets. The orthonormal
basis is formed by "elementary states" [MacCallum, 1972] which are cocycles
with representatives chosen in the following way:

( < A*,Z > )k2 ( < B*,Z > )1-k2 (1)

>/- < A*,A* > >/- < B*,B* >
X
( < A,Z > )kl+1 ( < B,Z > )n+l-k1 +1
>/< A, A> >/< B, B >
147
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 147-150.
© 1993 Kluwer Academic Publishers.
148 ANDRZEJ KARPIO

where

kl = 0, ... , n + I; k2 = 0, ... , I; 1= 0, ... ,00;

< , > is a twistor form ;

A, B, A*, B* form an orthogonal basis for C4 ; A, B are positive twistors
with respect to the twistor form but A *, B* are negative ones.

These sections are defined on the intersection of two open subsets in PT+ :
UA nUB where

UA = {Z E PT+ : < A, Z > of °}, UB = {Z E PT+ : < B, Z > of °}

The above Hilbert space I will denote by H1B(PT+, O( -n- 2)) . It depends
on the choice of twistors A, B but for another choice we obtain a space which
is isomorphic with the previous one. More details relating to the construction
of H1B(PT+, O( -n - 2)) can be found, for instance, in [Penrose, 1979] and
in my work which will appear soon.

2. COHERENT STATES IN H1B(PT+,O(-n - 2))

The basic concept of the further construction is the Reproduction Ker-
nel for Hilbert space in the sense of Bergman. In order to find it for the space
H1B (PT+, O( -n - 2)) we can use our orthonormal basis. By definition we
have to calculate the sum of the series:

(2)

where * is conjugation defined by the natural duality PT+ :::: PT* + and
the sections Bf~k2 are identified with their pullbacks in the double fibering
PT* + x PT+ over PT* + and PT+. The result of the calculations belongs
to the cohomology group H2(PT* + x PT+, O( -n - 2, -n - 2)) and its
representative is given by the formula:

cp(++)( TV Z) = dn + 1 .!. In (x - A)(Y - A) (3)

n , dAn+1 A xy

where
A =< W Z>' x - <W,A><A,Z> . Y _ <W,B><B,Z> .
, , - <A,A> ' - <B,B> '

TV means conjugation with respect to the twistor form.

AN APPROACH TO THE CONSTRUCTION OF COHERENT STATES 149

I will call it the Reproduction Kernel for the Hilbert space H1B( PT+, O( -n-
2)) which describes quantum states of massless particles with positive he-
licity. This terminology is justified by the reproduction property understood
as below:
(<1>~++)(Z,.),f) = feZ) (4)
where f represents the co cycle from H1B (PT+, O( -n - 2)) , the variable
Z is fixed and ( , ) denotes the scalar product for the cohomology classes
mentioned in the introduction.
The details of these considerations as well as calculations will be pub-
lished later. Having the Reproduction Kernel we define coherent states as
the elements of the Hilbert space obtained by the evaluation of the kernel
in the one of its variables [Perelomov, 1987]. This procedure is obvious if
one considers sections of some bundle but in this place we are dealing with
cocycles, and therefore with more complicated objects. My proposition is
to perform the Penrose Transform to <l>h++) in order to obtain the elements
of H1B(PT+, O( -n - 2)) which come from our Reproduction Kernel. It is
the most natural operation we can do in this case. The use of the Penrose
Transform requires the choosing of an element from M++ ( the manifold of
2-dimensionallinear subspaces in C4, positive with respect to twist or form)
which is determined by the pair of positive twistors C,D and some section
of the universal bundle over PT+ given by the constant spinor field r/' .
Calculations are not difficult but need some patience; their result can be
presented in the following statement:
STATEMENT 1. Coherent states in the Hilbert space H1B(PT+, O( -n-2))
of massless particles with non-vanishing helicity s = ~ obtained by evaluating
the Reproduction Kernel ( 3 ) are cocycles with the following representatives:

for n > 0

<1>(++)([(; D])(Z) _ 2 (6)

o , -<C,Z><D,Z>
for n = 0
were [C, D] is an element from M++ spanned by the twistors C, lJ . They re-
alize the embedding of the flag manifold F+,++ into H~B (PT+, O( -n - 2)) .
This statement needs some remarks. First of all I have considered the case
n=O which does not complicate our considerations but it is very natural not
only in this place but also in the physical applications of the above results.
By F+'++ I mean flag manifold which elements are the pairs: I-dimensional
150 ANDRZEJ KARPIO

positive subspace in C4 contained in 2-dimensional positive subspace in C 4.

Where "positive" means positive with respect to the twistor form. All results
presented in this chapter can be extended to the case of the negative twistors
and negative helicities.

3. CONCLUSIONS
The states appearing in the statement were known earlier and were
the subject of the considerations of many works, see for example [Hughston,
1979] [Eastwood, 1979]. They are the linear combinations of the simplest
elements belonging to the Hilbert space H~B(PT+, O( -n - 2)) . Moreover,
they realize the embedding of F+'++ into H~B(PT+,O(-n - 2)) which is
very important from physical point of view and results in the quantization of
classical objects which are the special congruences of null geodesics, the so
called Robinson congruences [Ward, 1979] . On the other hand, the Penrose
Transform of the coherent states gives us the elements obtained from the
reproduction kernel for the Hilbert space of holomorphic spinor fields on the
future tube M++ [Jacobsen, ] . I think that the facts I have just mentioned
justify using the name "coherent" for the states with so great importance
for physics.
More details, physical interpretation and much more will be published
soon.

References
M. G. Eastwood, M. L. Ginsberg - Duality in twistor theory - Duke Math. J. vol. 48, No
1, 1981.
M. G. Eastwood, L.P. Hughston - Massless field based on a line - in Advances in twistor
theory, ed. L. P. Hughston, R.S. Ward, 1979.
M. L. Ginsberg - Scattering theory and the geometry of multi-twistor spaces - Trans. Am.
Math. Soc. vol. 276, No 2, 1983.
L. P. Hughston - The twistor cohomology of local Hertz potential - in Advances in twistor
theory, ed. L. P. Hughston, R.S. Ward, 1979.
H. P. Jacobsen, M. Verne - Wave and Dirac operators and representation of the conformal
group - preprint
A. Karpio, A. Kryszen, A. Odzijewicz - 2-twistor conformal, hamiltonian spaces - Rep.
Math. Phys. vol. 24, 1986.
M. A. H. MacCallum, R. Penrose - Twistor theory: an approach to the quantisation of
fields and space-time - Phys. Rep. sec C, vol. 6, No 4, 1972.
R. Penrose - The twistor programme - Rep. Math. Phys. vol. 12, 1977.
R. Penrose - Twistor function and sheaf cohomology - in Advances in twistor theory, ed.
L. P. Hughston, R.S. Ward, 1979.
A. M. Perelomov - Obobshchenyje kogierentnyje sostojania i ih primenenia. Moskva
"Nauka" 1987.
P. Tod - Rep. Math. Phys. vol. 11, 1977.
R.S. Ward - Massless fields and sheaf cohomology - in Advances in twistor theory, ed. L.
P. Hughston, R.S. Ward, 1979.
R. O. Wells - Complex manifolds and mathematical physics. - Bull. Am. Math. Soc. vol.
1, No 2, 1979.
CLIFFORD ALGEBRAS
WHAT IS A BIVECTOR?
PERTTI LOUNESTO
In$titute of Mathematic$
Hel8inki Univer$ityof Technology
SF-02150 ESPOO, Finland

May 25, 1993

Abstract. Bivectors do not exist in Clifford algebras over arbitrary fields, especially they
do not exist in a canonical way in char 2. However, there is a natural way to introduce
bivectors in all other char i' 2, whilst the polarization formula gives a one to one cor-
respondence between quadratic forms and 6ymmetric bilinear forms. This paper reviews
Chevalley's construction for a quadratic form Q, and arbitrary, not nece$$arily 6ymmetric,
bilinear forms such that B(x, x) = Q(x). The exterior product is obtained from the Clif-
ford product by Riesz's formula x /\ u = &(xu + (-l)kux), where x E V and u E Ak
V.

Key words: Exterior algebra - contraction - bivedors - Clifford algebra

1. Chevalley's Identification of Cl(Q) C End(A V)

Chevalley 1954, pp. 38-42, introduced a linear operator IX E End(A V)

such that (cf. Oziewicz 1986, page 252 line 3 offormula (23))

Ix(u)=xl\u+xJu for XEV, UE/\V.

From the derivation rule x J (u 1\ v) = (x J u) 1\ v + U 1\ (x J v) and

= 0, x J (x J u) = 0 one can conclude the identity bx)2 = Q(x).
x 1\ x 1\ u
Chevalley's inclusion map V ---> End(A V), x -+ IX was a Clifford map and
could be extended to an algebra homomorphism t/J: Cl( Q) -+ End(A V),
whose image evaluated at 1 E A V yielded the map ¢: End(A V) -+ A V.
The composite linear map () = ¢ 0 t/J was the right inverse of the natural
map AV ---> Cl(Q) and

was the identity mapping on A V. The faithful representation t/J sent Cl( Q)
onto an isomorphic subalgebra of End(A V).
Chevalley's identification works fine with a contraction defined by an
arbitrary, not necessarily symmetric, bilinear form B such that B(x, x)
= Q(x). The following properties uniquely determine the contraction also
for an arbitrary, not necessarily non-degenerate, Q:

153
Z. Oziewicz et al, (eds.), Spinors, Twis/ors, Clifford Algebras and Quantum Deformations, 153-158.
© 1993 Kluwer Academic Publishers.
154 PERTTI LOUNESTO

x Jy = B(x,y) for X,y E V (a)

x J (u /\ v) = (x J u) /\ v + U /\ (x J v) (b)
(u /\ v) J w = u J (v J w) for u, v, w E /\ V (c)
(see Helmstetter 1982). The identity (a) fixes the dependence of the con-
traction on the symmetric bilinear form on V. The identity (b) means that
x E V operates like a derivation (cf. Greub 1978 p. 118 and Crumey-
rolle 1990 p. 35). The identity (c) introduces a scalar multiplication on /\ V
making it a left module over /\ V. The identity (b) allows elements of higher
degree on the right hand side (cf. Oziewicz 1986, p. 249 (13)) and the iden-

1986, p. 248 (12)). Evidently, x J a E /\k-l V for a E V and "k

tity (c) allows elements of higher degree on the left hand side (cf. Oziewicz

x J (Xl /\ X2 /\ .•. /\ Xk)

k
= L( _l)i-l B(X,Xi) Xl /\ X2 /\ ... /\ Xi-l /\ XiH/\"'/\ xk.
i=l

The faithful representation 'IjJ sends the Clifford algebra Cl( Q) onto an iso-
morphic subalgebra of End(/\ V) which as a subspace depends on B.
Remark. Chevalley introduced his identification Cl(Q) C End(/\ V) in
order to be able to include the exceptional case of characteristic 2. In char-
acteristic fc 2 the theory of quadratic forms is the same as the theory of
symmetric bilinear forms and Chevalley's identification gives the Clifford
algebra of the sYlllllletric bilinear form <x,y> = ~(B(x,y) + B(y,x))
satisfying xy + yx = 2<x, y>. _

u E "V.
Remark. We could also define the right contraction v L u of v E /\ V by
The right and the left contractions are related by the formulas
v L u = Uo J Vo + Uo J VI - VI J Uo + VI J Ul and u J v = Vo L Uo - Vo LUI + VI L
Uo + VI LUI. The notation a· b may be used for the contraction when it is
clear from the context which factor is contracted and which is the contrac-
tor. This dot product a· b can be used when at least one of the factors is
homogeneous. If both factors are homogeneous, then we agree that the one
with lower (or not higher) degree is the contractor (a E /\ i V, b E /\i V)

a· b = a J b for is; j and a· b = aLb for i ~ j.

When precisely one factor is known to be homogeneous we agree that it is

the contractor (a E /\i V, U E " V)

a .U = a J u and u . a = u L a.

Note that the contraction obeys the rules 1Ju = u, u E /\ V, and xJ 1 = 0,

x E V, but for the dot product 1· u = u . 1 = u. Note also that the dot
WHAT IS A BIVECTOR? 155

product does not act like a scalar multiplication on the left /\ V -module
/\ V, that is, (a 1\ b) . u fa· (b. u). As an exercise the reader may verify that
for a = el, b = el + e2 and u = el + el 1\ e2 in /\ lR 2 all the expressions
aJ(bJu), aJ(bLu), aL(bJu), aL(bLu) and (al\b)Ju, (al\b)Lu
are unequal with the exception of (a 1\ b) J u = a J (b J u) = ?
- The lack of /\ V -linearity renders less useful any extension of U· v for
arbitrary u, v E /\ V. Such an extension was introduced under the name of
'inner product' by Hestenes&Sobczyk 1984 p. 6 who display only formulas
with at least one homogeneous factor. The non-/\ V-linear 'inner product'
is not consistent with the contraction (in the sense that the 'inner prod-
uct' is not a special case of the contraction), because U· v might differ
simultaneously both from u J v and u Lv. Boudet 1992 p. 345 men-
tioned a formalization of the 'inner product' but his rules are not sufficient
to permit the evaluation of (x 1\ y) . u when u E /\ V, u f/- V (though
they do permit a construction of the 'inner product' with an additional rule
(x 1\ y). u = x· (y. u) where u E /\Ie V, k 2: 2). •
The above remark shows how the asymmetric contraction solves a problem
of Hestenes&Sobczyk 1984, who postulate the 'inner product' to be 0 [p.6,
r. 12 formula (1. 21 b )1if one of the factors is a scalar, and run into difficulties
on p.20 rows 8-18 formula (2.9). However, as the following example shows
the problem is deeper than that since the 'inner product' is not equal to the
contraction even if scalars were excluded.
Exalllple. Let ell e2 be an orthonormal basis for JR2 = lR 2,o. Compute

in the sense of Hestenes&Sobczyk. The same elements have the contractions

(el - el 1\ e2) J (e2 + el 1\ e2) = 1 + e2

(el - el 1\ e2) L (e2 + el 1\ e2) = 1 - el'
This shows that neither the left contraction nor the right contraction coin-
cides with the 'inner product' of Hestenes&Sobczyk. •
In other words, the 'inner product' is not dual/adjoint to the exterior prod-
uct. To sUllunarize: the inner product of Hestenes&Sobczyk is not the
same as the contraction or the interior product of Cartan.
In char # 2 we may re-obtain the dot product in terms of the Clifford
product as follows a· b = (ab)li_jl for a E /\i V and bE /\j V, where (u)1e
is the k-vector part of u E /\ V ~ Cf(Q).
For arbitrary Q but char K # 2 there is the natural choice of the
unique synulletric bilinear form B such that B(x, x) = Q(x) giving rise
to the canonical/ pri vileged linear isomorphism Cf( Q) -+ /\ V. The case
char K = 2 is quite different. In general, there are no symmetric bilinear
 156 PERTTI LOUNESTO

forms such that B(x, x) = Q(x) and in case that there is such a syrrunetric
bilinear form, it is not unique since any alternating bilinear form is also sym-
metric and could be added to the syrrunetric bilinear form without changing
Q. [Recall that antisynllietric means B( x, y) = - B(y, x) and alternating
B(x, x) = 0; alternating is always antisYllllletric, though in characteristic
2 antisymmetric is not necessarily alternating.] Thereby the contraction is
not unique, and there is an ambiguity in IU'
In characteristic 2 the theory of quadratic forms is not the same as the
theory of symmetric bilinear forms.
In the next example we need the matrix of v ~ uv, U = Uo + UI el +
U2 e 2 + U12 e l !\e2 with respect to the basis 1, el, ez, el !\e2 for A V, where
dimK V = 2, B(x, y) = aXIYI +bXIY2 + eX2YI + dxzyz and Q(x) = B(x, x):
aUI + eU2 bUl + dU2 -(ad - be)u12 )
Uo + eU12 dUl2 -(bUl + duz)
-aul2 Uo - bUl2 aUl + eU2 .
-U2 UI Uo + (-b + e)ul2
The conullutation relations are el e2 + eZel = b + e and e~ = a, e~ = d,
and we have the following multiplication table
el e2 el !\ e2
el a el !\ e2 + b -bel + ae2
e2 -el !\ e2 + e d -del + ee2
el !\ e2 eel - ae2 del - be2 - ad + be + (- b + e )el !\ ez

In characteristic :f 2 we find ~ (el e2 - e2el) = el !\ e2 + ~ (b - e) and more

generally for ~ (xy - yx) = x!\y + A( x, y) with an alternating scalar valued
form A(x,y) = ~(B(x,y) - B(y,x)) (cf. the last equation in Oziewicz 1986
p. 252). The symmetric bilinear form associated with Q(x) is
1 1
X· Y = 2 (B(x, y) + B(y, x)) = aXIYl + 2 (b + e)(xIY2 + x2yd + dX2Y2
and we have xy + yx = 2x· y for x,y EVe Cl(Q).
It is convenient to regard A V as the subalgebra of End(A V) with the
canonic.al choice of the sYllillletric B = O. We may also regard Cl( Q) as a
subalgebra of End(A V) obtained with some B such that B(x, x) = Q(x)
and choose the symmetric B in char :f 2.
Exalllple. Let K = {O, I}, dimK V = 2 and Q(xlel + X2e2) = XIZ2'
There are only two bilinear forms Bi such that Bi(X, x) = Q(x), namely
Bdx,y) = XIY2 and B 2(x,y) = X2YI, and neither is syrrunetric. The differ-
ence A = BI - B z , A(x,y) = XIY2 - X2Yl (= ZlY2 + z2yt) is alternating
(and thereby synunetric). Therefore, there are only two representations of
Cl(Q) in End(A V)
WHAT IS A BIVECTOR? 157

0 0
(~ ~Ul
Ul
Ul Uo 0 )
for B 1 : U
0
~
U2 Uo - Ul2
U12 -U2 Ul Uo Ul2

C'
U2 0
for B 2 : U ::::
Ul Uo + Ul2 0 o0 )
U2 0 Uo U2
Ul2 -U2 Ul Uo + Ul2
These representations have the following multiplication tables
Bl el e2 el 1\ e2
el 0 1 + el 1\ e2 -el
e2 -el 1\ e2 0 0
el 1\ e2 0 -e2 -el 1\ e2
B2 el e2 el 1\ e2
el 0 el 1\ e2 0
e2 1 - el 1\ e2 0 e2
el 1\ e2 el 0 el 1\ e2
with respect to the basis 1, el, e2, ell\e2 for" V. In this case there are only
two linear isomorphisms "V ---+ Gt( Q) which are identity mappings when
restricted to K + V and which preserve parity. It is easy to verify that the
above tables describe actually the only representations of Gt( Q) in " V. In
this case there is no canonical linear isomorphism" V ---+ Gl(Q), Le.,
neither of the above multiplication tables can be preferred over the other.
In particular, ,, 2 V cannot be canonically embedded in Gt( Q), and there
are no bivectors in characteristic 2. •
In the next section we try to answer the question: Are there bivectors in
characteristics other than 2?

2. Riesz's Introduction of an Exterior Product in Gt(Q)

We start from an object, which we suppose well-known here, the Clifford al-
gebra Gt( Q) over K, char K -I 2, and introduce another product in Gt( Q).
The isometry x ---> -x of V when extended to an automorphism of Gl(Q)
is called the grade involution U ---+ u. Define the exterior product of
x E V and u E Gf(Q) by (see Riesz 1958 p. 61-67)
1 _ 1 _
x 1\ U = 2 (xu + ux), U 1\ x = 2 (ux + xu)
and extend it by linearity to all of Gl( Q) which then becomes isomorphic as
an associative algebra to " V. The exterior products of two vectors x 1\ y =
~ (xy - yx) are simple bivectors and they span ,, 2 V. The exterior product
 158 PERTTI LOUNESTO

of a vector and a bivector x /\ B = ~(xB + Bx) is a 3-vector in A. 3 V. The

subspace of k-vectors is constructed recursively by

1 k-l
1\ V
k
X /\ a = "2 (xa + (_1)k- 1 ax) E for a E 1\ V.
Riesz's construction shows that bivectors do exist in all characteristics
'" 2.
Introduce the contraction of u E Ci(Q) by x E V so that (see Riesz
1958 p. 61-67)
1 _
x J u = "2 (xu - ux)

and show that this contraction is a derivation of Ci( Q) while

x J (uv) = ~ (xuv - Uvx) = ~ (xuv - uvx)

1
= "2 (xuv - uxv + uxv - uvx) = (x J u)v + u(x J v).
Thus one and the same contraction is indeed a derivation for both the exte-
rior product and the Clifford product. Kahler 1962 (p. 435 (4.4) and p. 456
(10.3)) was aware of the equations x J (u /\ v) = (x J u) /\ v + u /\ (x J v)
and x J (uv) = (x J u)v + u( x J v). Provided with the scalar multiplication
(u /\ v) J w = u J (v J w), the exterior algebra A. V and the Clifford algebra
Ci( Q) are linearly isomorphic as left A. V-modules.

References
Boudet Roger: 1992, 'Les algebres de Clifford et les transformations des multivecteurs',
in A. Micali et a1. (eds.): Proceeding& of the Second Worbhop on "Clifford Algebra&
and their Application& in Mathematical Phy&ic&," Montpellier, France, 1989, Kluwer,
Dordrecht, pp. 343-352.
Chevalley Claude: 1954, 'The Algebraic Theory of Spinors', Columbia University Press,
New York.
Crumeyrolle Albert: 1990, 'Orthogonal and Symplectic Clifford Algebras, Spinor Struc-
tures', Kluwer, Dordrecht.
Greub Werner: 1978, 'Multilinear Algebra' 2Dd Ed., Springer, New York.
Helmstetter Jacques: 1982, 'Algebres de Clifford et algebres de Weyl', Cahiers Math. 25,
Montpellier.
Hestenes David, Sobczyk Garret: 1984, 1987, 'Clifford Algebra to Geometric Calculus',
Reidel, Dordrecht.
Kahler Erich: 1962, 'Der innere Differentialkalkiil', Rendiconti di Matematica e delle &ue
Appl,eaz,oni (Roma) 21, 425-523.
Oziewicz Zbigniew: 1986, 'From Grassmann to Clifford', in J.S.R. Chisholm, A.K. Cornmon
(eds.): Proeeedings of the NATO and SERC Worbhop on "Clifford Algebras and Their
AppZ,cations m Mathematical Phy&ic&," Canterbury, England, 1985, Reidel, Dordrecht,
pp. 245-255.
Riesz Marcel: 1958, 'Clifford Numbers and Spinors', Univ. of Maryland.
MONOGENIC FORMS ON MANIFOLDS
VLADIMiR SOUCEK
Mathematical Institute
Charles University
Sokolovska 83
18600 Praha
Czechoslovakia

Abstract. A generalization of holomorphic forms on Riemann surfaces to higher-dimen-

sional manifolds with a spin-structure (so called monogenic forms) is described. Monogenic
forms on R", were defined by Delanghe, Sommen and Soucek in recently published mono-
graph on Clifford analysis. In the paper, the definition is generalized to the curved case
(i.e. to manifolds with a spin-structure). Monogenic forms of any degree are defined in such
a way that monogenic O-forms are harmonic spinors, i.e. solutions of the Dirac equation.
An analogue of the Cauchy theorem is proved for monogenic forms.

Key words: Clifford analysis - monogenic forms - Dirac operator - Cauchy theorem

1. Introduction

Starting from 30's the Dirac equation was established as the most appro-
priate generalization of Cauchy-Riemann equations to higher dimensions
by effort of many people (see references in (Brackx, Delanghe, Sommen
1982), (Delanghe, Sommen, Soucek 1992a)). Complex-valued functions were
replaced in higher dimensions by spinor-valued maps (the spinor space in
dimension 2 being the space of complex numbers).
We shall discuss here the next natural question, namely what is a natural
generalization of holomorphic forms on Riemannian surfaces to higher di-
mensions. Note that this generalization is going to a quite different direction
than holomorphic forms in several complex variables. The forms discussed
here are defined on (real) manifolds with a spin structure, not on complex
manifolds; they are not ordinary forms, but they have values in the cor-
responding spinor bundle and O-forms are solutions of the Dirac equation
instead of being holomorphic functions of several complex variables. Never-
theless, in the special case of the general situation - in the plane, i.e. for
m = 2 - everything is reduced back to the standard case of holomorphic
differential forms.

159
Z. Oziewicz et al. (eds.), Spirwrs, Twistors, Clifford Algebras and Qunntum Deformations, 159-166.
© 1993 Kluwer Academic Publishers.
160 VLADIMIR SOUCEK

Such a generalization of holomorphic forms - so called monogenic forms

- was described in flat case, i.e. on subset of R m , in (Delanghe, Sommen,
Soucek 1992a), (Sommen 1984), (Sommen, Soucek 1985), (Delanghe, Soucek
1992), (Sommen 1992), (Sommen, Soucek 1992). The purpose of the paper
is to show that this definition can be extended from the flat case to any
manifold endowed with a spin structure and to prove a basic property of
monogenic forms - a generalization of the Cauchy theorem for monogenic
functions.

2. A decomposition of spinor-valued forms

The object of our study is the space of all spinor-valued differential forms
on a spin-manifold M. This is a space of sections of the bundle A*(T;) Q9 S,
where T; is the complexified cotangent bundle and S is the spinor bundle.
It is a vector bundle associated to the representation A*( C~) Q9 S, where
S is a basic representation of the group Spin (m). We want to decompose
the space of S-valued forms into smaller pieces while keeping the invariance
properties with respect to the Spin group.
A systematic way how to describe all possible pieces in the decompo-
sition is to consider the representation Ai( C~) Q9 S, to decompose it into
irreducible pieces Ei,j and then to define the subspaces of S-valued forms
by considering the associated bundles to the representations Ei,j. The de-
composition of the tensor product Ai(C~) Q9 S was described in (Delanghe,
Soucek 1992) (and in the language of Clifford algebras in (Sommen, Soucek
1992), (Delanghe, Sommen, Soucek 1992a». Using the standard characteri-
zation of Spin-modules by the highest weights (for more details see (Brocker,
tom Dieck 1985), (Delanghe, Soucek 1992», the result can be summarized
as follows.

THEOREM 1. Let S denote a basic spinor representation.

The product Aj(C~) Q9 S ~ Am-j(c~) Q9 S,j = 1, ... , [m/2] decomposes
into j + 1 irreducible parts

The weights ILL, I = 0, ... ,j are given by:

1. Ifm is even and S ~ (~'''''~)'
th en ILL -- (32' ... , 2'
3 1 1 ( 1)j-I1)
2' ... 2' - 2'
2. Ifm is even and S ~ (~, ... ,~,-~),
then ILL = (~, ... ,~,~, ... ~,(_1)j-I+1~).
3. If m is odd, then ILL = (~, ... ,~,~, ... ,~) .
The component ~ appears I times in all cases.
MONOGENIC FORMS ON MANIFOLDS 161

Let us denote for simplicity the part El-'I in the decomposition of the
product Aj(C~) 0 S, resp. Am-j(c~) 0 S, j = 1, ... , [m/2] by Ej,l, resp.
Em-j,t.

3. Monogenic forms on a spin-manifold

Let (M, g) be an oriented Riemannian manifold of dimension m. Let us
choose a spin-structure on M, i.e. let us suppose that we have chosen a
principal fibre bundle P over M with the group G = Spin (m) together
with the corresponding 2:1 covering map P -+ P onto the bundle P of
oriented orthonormal frames. The Levi-Civita connection on P induces then
a covariant derivative D on the spinor bundle 5 = P XSpin S associated to
the basic spinor representation S.
Let us denote the space of smooth 5-valued differential forms of degree j
on M by [j(5). The covariant derivative D maps [°(5) to [1(5) and can
be extended (see e.g. (Wells 1973)) to the maps D : [k(5) 1-+ [k+1(5) for
all k = 1, ... , m - 1.
To define monogenic differential forms, we are going to define a splitting
of 5-valued k-forms on M into two parts

It imitates the definition of holomorphic forms in dimension 2. Let us

recall the definition of holomorphic forms in the plane. The space of I-forms
can be split into a direct sum of two pieces

a
and the value of the -operator on a function f is defined to be the com-
position of the de Rham operator d and the projection onto the (0,1 )-part
of df. Holomorphic functions and holomorphic I-forms are elements of the
kernels of the maps a.
U sing the splitting [k = [k' ffi [kll , we are going to consider the diagram

The operators d" are defined as the composition of d with the projection
onto [kll. Monogenic forms will be defined as elements of the kernels of the
operators d".
162 VLADIMiR SOUCEK

The requirement of invariance tells us that we have to choose for the

primed and double primed parts sums of pieces in the decomposition de-
scribed above. Let [j,k denote the bundle on M associated to the repre-
sentations Ej,k described in the Theorem 1. Then we have the following
definition.

DEFINITION 1. (i) Let us define for k ~ [m/2]

EB [k,k- j; [k" = EB [k,k-j (1)

05,j5,k 05,j5,k
j even j odd

and for k < [m/2]

[(m-k)' = [m-k,k-j., [(m-k)1I = EB [m-k,k-j (2)
05,j5,k 05,j5,k
j odd j even

- - - I - I I
(ii) Let us define spaces [k',[kll,e(m-k) ,[(m-k) by (1) for k < [m/2] and
by (2) for k ~ [m/2].

Note that the both spaces (with or without tilde) coincide in any odd
dimension and that they are different only in the middle dimension if the
dimension is even.
As an illustration, let us consider the cases of dimension m = 6 and
m = 7. In the odd-dimensional case, we have the following picture (the
numbers indicate the dimension of Ek,j counted in multiples of dim S). The
pieces, belonging to [k' , are indicated by boxes. In the top row the dimension
of the full spaces [k is written. The spaces [k' and k ' coincide. t
1 7 21 35 35 21 7 1

In the even-dimensional case (m = 6), we get two possibilities:

MONOGENIC FORMS ON MANIFOLDS 163

(i) The spaces [II':

1 6 15 20 15 6 1

(ii) The spaces fk' :

1 6 15 20 15 6 1

The monogenic forms are now defined in the following way.

DEFINITION 2. Let us define

Mk = {w E [k'i d"w = O}
and
Mk = {w E fk' I d"w = O}.
The forms w E Mk (resp. w E Mk) will be called monogenic k-forms.

As was shown in (Fegan 1976) and (Bures, Soucek 1986), the space MO
is just the usual space of solutions of the Dirac equation (i.e. the space of
harmonic spinors).
Monogenic differential forms on domain in Rm are defined and studied
in (Delanghe, Sommen, Soucek 1992a), (Sommen, Soucek 1992) and (De-
langhe, Soucek 1992). The main properties proved there are the description
of homology of domains in Rm by homology of the sequence of monogenic
forms and an analogue of the Cauchy theorem. In the next section, we are
going to show how the Cauchy theorem can be proved for monogenic forms
on spin-manifolds.
164 VLADIMiR SOUCEK

4. The Cauchy theorem

Let el, ... ,em be an orthonormal basis in R m , the vectors ej are considered
as elements in the corresI>0nding Cifford algebra Ro,m' Let us define an
(m - I)-form dfj = E( -1)J+ 1 ejdi'j with values in RO,m'
The standard form of the Cauchy theorem for monogenic functions in
Clifford analysis is the statement that the Clifford algebra valued differential
form w = f dfj 9 is closed whenever the functions 9 and f are left, resp. right,
monogenic.
In the fiat case, the space Mm-l of (left) monogenic (m-l )-forms consists
of all forms w = dfj f, where f is a solution of the Dirac equation. Hence the
form w is a product of a monogenic O-form and a monogenic (m - I)-form.
More generally, it is proved in (Delanghe, Sommen, Soucek 1992a) that a
product of a (right) monogenic j-form and a (left) monogenic (m - j - 1)-
form is a closed form. We are going to show that this generalized Cauchy
theorem holds for monogenic forms on manifolds as well.
In this more general situation, the notion of multiplication of Clifford-
valued forms will be substituted by a duality given by an invariant Her-
mitean form on the spinor space S. It is well known that such an invariant
Hermitean form on S exists, it induces then a Hermitean structure on the
associated bundle S. To see the connection with left and right multiplication
used in Clifford analysis on fiat space, it is convenient to use a realization
of the spinor space as a left ideal in the complexified Clifford algebra Cm
described in (Delanghe, Sommen, Soucek 1992a), Sect. 1.4.7. It is showed
there that a dual space to S can be realized conveniently as a right ideal
S in Cm , where the bar map is the composition of a main antiinvolution in
the Clifford algebra Cm and the complex conjugation. The product S . S is
one-dimensional complex space and when we identify it with C, an invari-
ant Hermitean scalar product on S is given by the Clifford multiplication
< s, s' >= s's. In our general situation, we shall use the scalar product in-
stead of the multiplication. As a consequence, we shall not need to introduce
left and right monogenic forms.
So having available the Hermitean scalar product on < .,. >x in each
fiber Sx,x E M, we can define a map < .,. > from £j(S) X ekeS) into £j+k
by

< w ® s,w' ® s' >x= w Aw' < s,s' >x,w E £jx,w' E £kx;s,s' E Sx.

The covariant derivative D induced by the Levi-Civita connection is com-

patible with the Hermitean structure, i.e. we have

d < W,T >=< DW,T > +(-I)j < W,DT >,w E £j(S),T E ekeS)
(for O-forms it is proved e.g. in (Lawson, Michelsohn 1989), it can be checked
that it is true for general forms).
MONOGENIC FORMS ON MANIFOLDS 165

Now we can formulate the Cauchy theorem.

THEOREM 2. For any k = 1, ... , m - 1 and for all W E Mk(S) and T E

Mm-k-l(S)
d < W,T >= O.

The proof of the theorem is based on the multiplicative properties of

elements in A*(R~) ® S.

LEMMA 1. Let Ej,k be the decomposition of Aj (c;,) ® S described in The-

orem 1 and let k = 1, ... , m - 1; i,j < k; i =I j. Then for each W E Ek,iand
T E Em-k,j
< W,T >= O.
Proof
Using the realisation of the spinor space as a left ideal in the Clifford
algebra and using the realization of the Hermitean scalar product described
above, the lemma follows from the Lemma 5.2 in (Delanghe, Sommen,
Soucek 1992) . •
Proof of the Cauchy theorem
-I
Using the defintion of [i' and [m-j , it is easy to check that they never
contain elements at the same row in the decomposition diagrams (see e.g.
diagrams shown above).
Then it is sufficient to observe that

d < W,T >=< DW,T > +(-1)j < W,DT >

and that the monogenicity conditions implies the condition Dw E [(j+l)'
and DT E t(m-j)' . •
The Cauchy theorem for monogenic forms is the key property which made
possible to define a general notion of residue for monogenic forms for higher
dimensional singularities introduced in (Delanghe, Sommen, Soucek 1992a).
So the generalization proved here will make possible to define the notion
of a residue of a monogenic form with a higher dimensional singularity on
spin-manifolds. This topic will be described, however, in another paper.

References
Brackx, F., Delanghe, R., Sommen, F.: 1982, Clifford analysis, Pitman: London
Brocker, T., tom Dieck, T.: 1985, Representations of Compact Lie Groups, Springer-Verlag:
Berlin
Delanghe, R., Soucek, V.: 1992, 'On the structure of spinor-valued differential forms',
Complex Variables 18, 223-236
Delanghe, R., Sommen, F., Soucek, V.: 1992a, Clifford Algebra and Spinor-valued Func-
tions, Kluwer Academic Publishers, 485 pp.
166 VLADIMiR SOUCEK

Delanghe, R., Sommen, F., V.Soucek, V.: 1992b, 'An explicit realization of spinor spaces
and its application to Clifford analysis', ,to appear
Fegan, H.D.: 1976, 'Conformally invariant first order differential operators', Q.Jour.Math.
27,371-378
Gentili, G., Mariconda, C., Tarallo, M.: 1990, 'Quaternionic regular maps and a-type
operators', , SISSA preprint
Lawson, H.B., Michelsohn, M.-L.: 1989, Spin Geometry, Princeton Univ. Press
Sommen, F.: 1984, 'Monogenic differential forms and homology theory', Proc. Royal Irish
Ac. 84A, 87-109
Sommen, F.: 1982, 'Monogenic differential calculus', , to appear
Sommen, F., Soucek, V.: 1985, 'Hypercomplex differential forms applied to the de Rham
and the Dolbeault complex', Sem.Geom. 1984, Univ. Bologna, 177-192
Sommen, F., Soucek, V.: 1982, 'Monogenic differential forms', , to appear
Soucek, V.: 1984, 'H-valued differential forms on H,', Suppl.Rend.Circolo Mat. Palermo,
ser.II 3, 293-299
Wells, R.O., jr.: 1973, Differential analysis on complex manifolds, Prentice.Hall, N.J.
ON INVERTIBILITY OF CLIFFORD
ALGEBRAS ELEMENTS
WITH DISJOINT SUPPORTS

PAVEL SEMENOV
Moscow State Pedagogical University
Kosmodemjanski 9A, kv. 27
125190 Moscow, Russia

Let Cl(n) be the classical associative Clifford algebra over field lffi. with
generators el, e2, ... ,en and relations

eie j + e jei = 0, i ::fi j

e~, = -1.
It's well known ( see [1]) fact that algebras Cl(n) for different n are isomor-
phic to some matrix lffi.-algebra or to direct sum of some matrix lffi.-algebras.
Therefore, from the formal point of view, the question about invertibility
in Cl( n) is equivalent to the question about calculating of determinants of
matrices. But,these matrices have sizes approximately equal to 2[n/2] X 2[n/2]
and really such calculatings are impossible. But for some classes of elements
of algebra Cl( n) a criteria of invertibility may be obtained without above
mentioned matrix realizibility of Clifford algebra Cl(n). The trivial example
of such a class is the set of all vectors

x = Lx;e; E lffi.n C Cl(n).

Indeed we have

and hence vector x is invertible in Cl( n) iff x ::fi O.

The main emphasis of the present lecture is given to questions about
invertibility of bivectors x = L Xijeij and of elements with "disjoint sup-
ports", namely of elements x = L Xiea, with nonempty, mutually disjoint
sets ai.
Case n=3. Let x = ae12 + {3e23 + ,el3. It is easy to see that x 2 =
_(a 2 + (32 + ,2) and hence x is noninvertible iff x = O.

167
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 167-169.
© 1993 Kluwer Academic Publishers.
168 P.V. SEMENOV

Case n=4.

THEOREM 1. Let x =L Xijeij be a bivector in CI( 4). Then x is

noninvertible iff

Case n=5. For any bivector x = L Xijeij in Ct( n) we denote

and for any a = {i < j < k < n} C {1,2,3,4,5}

we denote s",(x) = 2(XijXkn - XikXJn + XinXJk)

THEOREM 2. Bivector x is non-invertible in CI(5) iff

where the sum is taken over all a consisting of four elements.

Case n=6. In Ct(6)J dont know a criteria for invertibility of bivectors in

general case. But for bivectors with disjoint sup ports the following theorem
may be obtained.

THEOREM 3. Let x = ae12+j3e34+,e56 and let

y = a( _a 2+j32+,2)e12 +j3(a 2_j32+,2)et4

Then
xy = (a + j3 + ,)( a + j3 - ,)( a - j3 + ,)( a - j3 - ,)
Hence, the bivector x is non-invertible

This theorem answers Prof. P. Lounesto's hypothesis (Montpellier, 1989,

private communication). The proof of theorem 3 may be easily obtained if
ON INVERTffiILITY OF CLIFFORD ALGEBRAS 169

we remark that x 2 lies in subspace V4 = = span {1,e1234,e1256,e3456} and

that subspace V4 is in fact sub algebra of algebra Cl(6).
Theorem 3 (but not its proof) has the following generalization for ele-
ments of Clifford algebra Cl(n). For any set a C {I, 2, ... , n} we denote I a I
the number of elements of this set.

k i m s
Xl= Laiea"x2= Lbje.B),x3= Lcpe-yp,x 4 = LdqeTq
i=l j=l p=l q=l

sets {al, ... ,ak,.B1, ... ,T s } are nonempty, mutually disjoint and I
ai 1= 1( mod 4), l.B j 1= 2) mod4) lip 1= 3(mod4), I Tq 1= O(mod4).
Then
a) in the case M = L: a;- L: c~ > 0 the element x is noninvertible
iff
II {(bl-fl M- f 2 b 2- ... - flbl)2+(dl-b2d2- ... - bs d s )2}= 0
,),c5 q =±l

b) in the case M = L: ar- L: c~ < 0 the element x is non-invertible

iff
II {(b 1 - f2b 2-". - flbl)2+(d 1 -b 1 V-M-b 2d 2 - ... - bs d s )2}= 0
'J,c5q =±l

c) in the case M = 0 the element x is non-invertible iff

II {(b 1 - f 2 b 2- ... - flbl)2+(dl-b2d2- ... - bs d s )2}= 0

'J,c5q =±l

Open problems

Problem 1. Find an analog of theorem 4 for elements with property of " small
intersection" of supports.
Problem 2. What about an invertibility in algebras Cl(p,q) ?
Problem 3. When may be an element of C l( n) written as a product of some
elements, each of them is an element with disjoint supports?

References
M. Karoubi K-Theory, Springer Verlag. 1978.
CLIFFORD ALGEBRAS AND ALGEBRAIC
STRUCTURE OF FUNDAMENTAL FERMIONS
WOJCIECH KROLIKOWSKI
Institute of Theoretical Physics, Warsaw University
Hoza 69, PL - 00 - 681 Warszawa, Poland

Dedicated to Jan Rzewuski

on the happy occasion of
his 7IJ h birthday

Abstract. We show, how an idea of leptons and quarks composed of algebraic partons
(defined by a sequence of Clifford algebras) can explain the existence of three and only
three families of these fundamental fermions. In this argument, the theory of relativity,
the probability interpretation of quantum mechanics and the Pauli exclusion principle, all
extended to the algebraic partons, playa crucial role. As a consequence, a semiempirical
mass spectral formula for charged leptons is discussed. In terms of experimental me and
m!-" it gives successfully m T = 1783.47 MeV or 1776.80 MeV (two options, the second
fitting excellently to new measurements of mT)'

1. Introduction

The most puzzling feature of today's particle physics is perhaps the phe-
nomenon of three families of leptons:

Ve vp. VT (?) (charge 0)

(1)
e J.L T (charge - 1)
and quarks:

u c t (?) (charge 2/3)

(2)
d s b (charge - 1/3)
differing by nothing but their masses. Among them, the tauonic neutrino
VT and top quark t are not yet observed directly, though indirect evidence
leaves practically no doubt as to their existence. In particular, the recent
CERN measurements of total decay width for ZO gauge boson manifesting
itself as a resonance at ca. 91 GeV of eM energy in the process
e+ e- ---+ ZO ---+ anything, (3)
have shown that the number of different neutrino versions lighter than
~mz ~ 46 GeV is just three. Moreover, this result strongly suggests that

171
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum De/onnations, 171-181.
© 1993 Kluwer Academic Publishers.
172 WOJCIECH KR6LIKOWSKI

the number of all lepton and quark families is equal to three if all neutrinos
are light (or massless).
In this lecture, we are going to show that there are three different, phys-
ically distinguished versions of the Dirac equation

[f . (p - 9A) - M]1/J = 0, (4)

where

{Pi, f"} = 2gl-'lI. (5)

Here, gf· A symbolizes the standard-model coupling, identical for all three
versions, while the mass operator M may depend on the version. So, we shall
be tempted to connect these versions with the three experimental families
of leptons and quarks.
Our argument will express an idea of algebraic compositness of fundamen-
tal fermions that accepts an act of algebraic abstraction from the familiar
notion of spatial compositness (so useful, for instance, in the case of pseu-
doscalar and vector mesons built up of quark-antiquark pairs moving in the
physical space).

2. An example of algebraic compositness

Let us start from the familiar Duffin-Kemmer-Petiau equation describing a
particle with spin 0 Ell 1 (for instance, a pseudoscalar or vector meson). In
the free case, it can be written in the form

1
["2( 11 + 12) . P - M]1/J(X) = 0, (6)

where I i and I~ are two sets of commuting Dirac matrices,

{ IiI-' ,Ii"} -- 29 1-'" ,/1

[I-' ,12"] -- 0 . (7)

Here, ~(ti + I~) are the 16 X 16 Duffin-Kemmer-Petiau matrices.

It can be readily seen that Eq. (6) may be considered as a point-like
limiting form of the following two-body wave equation (Krolikowski 1987,
1988):

[/1 .(~P + p) + 12' (~P - p) - m1 - m2 - S(x)] 1/J(X, x) = 0, (8)

where (for simplicity) masses are assumed equal: m1 = m2 (what, for

instance, is the case for a pair of a quark and an antiquark of the same
sort). The internal interaction S(x) in Eq. (8) can be related to the more
CLIFFORD ALGEBRAS AND ALGEBRAIC STRUCTURE 173

familiar internal interaction I( x) appearing in the Bethe-Salpeter equation

(Bethe 1957) through the formula

S(x) = [71 . (~p + p)I +I I

- m1 + it: 72 . (~p - p) - m2 + it:
I(x).(9)

Then, any of these two four-dimensional integral operators allows to cal-

culate perturbatively, step by step, a three-dimensional integral operator
playing the role of internal interaction Vex) (internal interaction energy) in
the one-time two-body wave equation having the conventional form of the
state equation. This equation, derived many years ago by Jan Rzewuski and
myself (Krolikowski 1955, 1956), reduces to the familiar Salpeter equation
(Salpeter 1952) in the case of an instantaneous internal interaction.
Now, an intriguing question might be asked, what would happen, if in
Eq. (6) the commuting 7i and 7~ were replaced by anticommuting 7i and
7~ (Krolikowski 1986). Then, there would be

(10)

instead of Eq. (7). Note that the Clifford algebra (10) could be represented
by

(11)

with 71l-, 1 and 7 5 = i707 1 7 2 7 3 being the usual Dirac 4 X 4 matrices.

In the case of Eq. (10), the counterpart of the Duffin-Kemmer-Petiau
equation (6) (with the convenient coefficient 1/V2 in place of 1/2),

[~ Cli + 7n· P - M] 1jJ(X) = 0, (12)

might be considered as a point-like limiting form of the two-body wave

equation

(~p + p) +
[V2 71 .
(13)
+V272· (~p - p) - m1 - m2 - Sex)] 1jJ(X,x) = 0,

but the latter, in contrast to Eq. (8), could not be derived from the conven-
tional quantum field theory. This is a consequence of the fact that the parti-
cle kinetic-energy operators in the Fock space 7? (Ii· Pi + m) all commute,
if they are derived from the field kinetic-energy operator J d~ 1jJ+( x)f° (I .
P + m )1jJ(x), so, in such a case, all7f must commute for different i (at least,
when massive particles are considered; if an interaction with an external
174 WOJCIECH KR6LIKOWSKI

scalar field is introduced, also massless particles cannot escape from this
conclusion ).
Thus, while Eq. (12) ( with Eq. (10)) may be investigated for some hy-
pothetical particles, it cannot be considered as a point-like limiting form of
a two- body wave equation following from the conventional field theory. So,
'¢ = ('¢C>lc>J displays an algebraic structure that, now, does not coexist with
any spatial internal structure, at any rate, in the framework of the conven-
tional quantum field theory (Krolikowski 1991). This illustrates, therefore,
the notion of algebraic compositness. In Eq. (12) the Dirac bispinor indices
0:1 and 0:2 describe "algebraic partons", agents of the idea of this composit-
ness.
Let us emphasize that the logical relationship between the notions of
spatial compositness and algebraic compositness reminds the logical rela-
tionship between the notions of orbital angular momentum and spin. In
fact, in these cases we have to do with similar acts of algebraic abstraction
from some notions of spatial character.
It is important to note that due to the Clifford algebra (10) the matrices
1
fit = ../2 (,i + ,t) (14)

appearing in Eq. (12) satisfy the Dirac algebra (5). This implies that Eq.
(12) has the form of the Dirac equation (4) (in the free case). Thus, the
hypothetical particles described by Eq. (12), when coupled to the magnetic
field, should display (magnetically "visible") spin 1/2 though any of them
is a composite of two algebraic partons of spin 1/2. There exists, therefore,
another (magnetically "hidden") spin 1/2. It is related to the matrices
(1/../2) Cli -,n also fulfilling the Dirac algebra (5) and anticommuting
with the matrices fit.
Note further that the matrices (14) may be represented in the convenient
form
fit = ,iJ. @ 1, (15)

if the representation (11) is changed into

,i,2 = ~ (,iJ. @ 1 ± ,5 @ h 5 ,il) . (16)

So, Eq. (12) can be rewritten as

C/C>1f31 . P - bC>1f31 M ) '¢f31C>2(X) = 0, (17)

where the second Dirac bispinor index 0:2 is free. Such an equation is known
as the Dirac form (Banks 1982) of the Kahler equation (Kahler 1962).
CLIFFORD ALGEBRAS AND ALGEBRAIC STRUCTURE 175

3. A sequence of Dirac-type equations

As can be easily seen, the Dirac algebra (5) admits the remarkable sequence
N = 1,2,3, ... of representations

1 N
r lL = -- L If, (18)
VN i =1

where the matrices If, i = 1,2,3, ... , N, satisfy the sequence N = 1,2,3, ...
of Clifford algebras

(19)

With the matrices (18), Eq. (4) gives us a sequence N = 1,2,3, ... of
Dirac-type equations (Krolikowski 1990, 1992). Of course, for N = 1 Eq.
(4) (with the matrices (18) inserted) is the usual Dirac equation, while for
N = 2 it is equivalent to the Dirac form of the Kahler equation already
discussed in Section 2 (in the free case). For N = 3,4,5, ... it provides us
with new Dirac-type equations.
Except for N = 1, the representations (18) are reducible since they may
be realized in the convenient form

r lL = III Q9 1 Q9 ••• Q9 1 (20)

'-..-"
(N-l)times

with III and 1 standing for the usual Dirac 4 X 4 matrices. In fact, for any
N > lone can introduce, beside ri == rl" given in Eq. (18) N - 1 other
Jacobi-type independent combinations r~, ... , r~ ,

rlL _ 1 (I" r lL _ 1 (I"

2- V2 II - 12
IL)
, 3- J6 IL
II +'2 - 2'3
IL)
, ... , (21)

such that

{rl"i, rv} - 2C.

j v'Jg ILV
- (22)

(in consequence of Eq. (19)). In particular, for N =3 one may use the
representation

ri = III Q9 1 Q91 , r~ = 1 5 Q9 i,5,I" Q91 , rj = 'l Q9,5 Q9,IL. (23)

In the representation (20), the Dirac-type equation (4) for any N can be
rewritten as

(24)
176 WOJCIECH KROLIKOWSKI

where M 0i1 {31 = b0i1 {31 M. Here, t/J = (t/J Oi l 0i 2 •.• Oi N) carries N Dirac bispinor
indices ai, i = 1,2, ... , N, of which only the first one is affected by the
Dirac matrices ,11 and so is coupled to the particle's momentum and to the
standard-model gauge fields (among others, to the electromagnetic field).
The rest of them are free. Thus, only a1 is "visible", say, in the magnetic
field, while a2, ... ,aN are "hidden". In consequence, a particle described by
Eq. (4) or (24) can display, say, in the magnetic field only a "visible" spin
1/2, though it possesses also N - 1 "hidden" spins 1/2.
Our first crucial assumption will be that the physical Lorentz group of
the theory ofrelativity, if applied to the particle described by Eq. (4) or (24)
for any N, is generated both by the particle's visible and hidden degrees of
freedom. Then, the form t/J+r~rit/J is no relativistic covariant for N > 1,
though Eq. (4) with r ll == ri implies that always

(25)

In contrast, the form t/J+ r~rg ... r~ ri t/J is a relativistic vector for any N,
but Eq. (4) with r ll == ri shows that

(26)

only for N odd.

Thus, the interplay of the theory of relativity and the probability inter-
pretation of quantum mechanics requires that (i) only the odd terms

N = 1,3,5, ... (27)

should be present in the sequence of the Dirac-type equation (4) (if these
are considered as wave equations), and (ii) the probability current should
have the form

J'11 = TfN 'f/

.'.+ror o rON r ll1 .,.'f/ •
1 2'" (28)

Here, TfN is a phase factor making the matrix of hidden internal parity

lludden = TfN rg ... r~ (29)

Hermitian. Since due to Eq. (26) lludden is a constant of motion, one can
consistently impose on the wave function t/J in the wave equation (4) the
constraint

Phidden t/J = t/J (30)

in order to guarantee the probability density to be positive:
(31)
CLWFORD ALGEBRAS AND ALGEBRAIC STRUCTURE 177

4. Hidden exclusion principle

The Dirac-type equation (4) with fit == fi distinguishes the visible bispinor
index 0:1 from N - 1 hidden bispinor indices 0:2, ••• ,O:N . About the latter
indices, appearing in this scheme on the equal footing, we will make our
second crucial assumption that they represent physically nondistinguishable
degrees of freedom obeying the Fermi statistics along with the Pauli exclu-
sion principle. Then, the wave functions 1/J = (1/JOIl 0I2 ...OIN) in the sequence
(27) of the Dirac-type wave equations (4) or (24) should be completely an-
tisymmetric with respect to the hidden indices 0:2, •.. , O:N • This implies
that the sequence (27) must terminate at N = 5 ,

N=1,3,5, (32)

leaving us with three and only three terms (32) in the sequence of the Dirac-
type wave equations (4) or (24).
In the case of N = 5 our exclusion principle requires that

(33)

Thus, in this case there are 4! = 24 equivalent nonzero components (car-

rying the index o:d, all equal (up to the sign) to one Dirac function 1/J~~) .
This reduces the Dirac-type equation (4) or (24) to the usual Dirac equation.
Here, of course, spin is 1/2 and it is provided by the visible spin, while four
hidden spins sum up to zero.
The case of N = 3 is more complicated since then one should consider
five candidates for relativistic covariants, viz.

(34)

(35)

(36)

resentation, where ,5
Here, C denotes the usual charge conjugation matrix that in the chiral rep-
= diag( 1, 1, -1, -1), may be written as

~ ~)
o i -
C- l . (37)
-i 0
178 WOJCIECH KR6LIKOWSKI

Making use of Eq. (23), one can write the hidden internal parity (29) in the
form

.
Rhidden -- 'roro -
Z 2 3 - 1 to,
'(Y
° °
J ®J , (38)

where in the chiral representation

0010)
0=
( 0001 (39)
J 1 0 0 0
o1 0 0
Then, the constraint (30) implies that

1/JCt133, 'l/JCt 122 1/JCt144 , 'l/JCtl12 1/JCt 134 , 1/JCt121

(40)
1/JCt 131 , 1/JCt124 1/JCt142 , 1/JCt114 1/JCt132, 1/JCt141

Thus, the constraint (30) and our exclusion principle (requiring that 1/JCt l Ct 2 Ct 3 =
-1/JCtlCt3Ct2 ) leads to the conclusion that from all components 1/J Ct l Ci 2 Ct 3 only

./. _ ./. _ ./. _ ./. _ ./.(3)

'f'Ct l12 - -'f'Ct121 - 'f'Ci134 - -'f'Ci143 = 'f'Cil (41 )

and

(42)

may be nonzero. Then, after a simple calculation,

-4 i 1/JCi I12' (43)

{ -4i~Cil14 for f-l o (44)
for f-l 1,2,3 '
tl-'l1
Ctl O. (45)

But, the theory of relativity applied to the vector V~l given in Eq. (44)
requires that V~l = 0 since V~l = 0 for f-l = 1,2,3 . Hence, 1/JCt l14 = 0 .
In this way, we can see that all components 1/JCi lCt2 Ci 3 must vanish except
those in Eq. (41). So, in this case there are 4 equivalent nonzero components
(carrying the index 01 ), all equal (up to the sign) to the Dirac function 'I/J~~) .
This reduces the Dirac-type equation (4) or (24) to the usual Dirac equation.
Here, spin is evidently 1/2 and it is given by the visible spin, two hidden
spins being summed up to zero.
Concluding, in each of the three allowed cases N = 1,3,5 there exists
one and only one Dirac particle (for any given color and up/down weak
CLIFFORD ALGEBRAS AND ALGEBRAIC STRUCTURE 179

flavor described by the standard model). So, it is natural to connect these

three versions of the Dirac particle with the three experimental families of
leptons and quarks. This happy existence of three and only three versions of
the Dirac particle is a consequence of an interplay of the theory of relativity,
the probability interpretation of quantum mechanics and the Pauli exclusion
principle, all extended to the particle's hidden degrees of freedom.
As for the wave functions with N = 1,3,5 the number of equivalent
nonzero components (carrying the visible bispinor index) is 1, 4, 24 ,
respectively, the following overall wave function comprising three sectors
N = 1,3,5 (or three fundamental-fermion families) may be constructed:

.1,(1 ) .1,(1)
'PC>! 'PC>!

1
J4'1j;~~) =p .1,(3) ( 46)
v'29 'PC>!

J24'1j;~~)
Here, the sector-weighting (or family-weighting) matrix

p=- 1 (1oJ40 0) 0 (47)

v'29 0 0 J24
appears.

5. Mass spectral formula for charged leptons

The three-family wave function (46) implies the following form of the mass
matrix for any triple of fundamental fermions ordered in one line in Eqs. (1)
and (2):

if = php. (48)

Here, h denotes a Higgs coupling strength matrix, while P IS given as in

Eq. (47). So, there are four different matrices (48) corresponding to triples
of neutrinos, charged leptons, up quarks and down quarks, respectively.
Among all 12 fundamental-fermion masses, the masses me, mJ1. , m T of
charged leptons e-, {l- , r- are the best known. On the base of some nu-
merical experience, we can propose the following phenomenological ansatz
(in two options) for the matrix h in the case of charged leptons:

(49)
180 WOJCIECH KR6LIKOWSKI

with

(50)

where N = 1,3,5 . Here, Mo > 0 and c: 2 denote two real constants in-
dependent of N. Then, the eigenvalues of the mass matrix (48) take the
form

Mo 2
=f2"9C: ,

~9 Mo
29
(80 2)
=f c: , (51 )

24 Mo (
25 2"9 624 =f c:
2) ,
since the Dirac masses are defined as nonnegative ( a priori, the second option
seems to be more attractive). From the system of three equations (51) we
obtain in terms of experimental me and ml-' the predictions (in two options)
for the mass m".,

6 { 1783.47 MeV
m". = 125 (351mI' ± 136me ) = 1776.80 MeV' (52)

and for the parameters Mo and c: 2 ,

29 { 86.3629 MeV
Mo = 320 (9ml-' ± 4me) = 85.9924 MeV (53)

and

c: 2 = 320me _ { 0.171590 (54)

9ml-' ± 4me - 0.172329·

We can see an excellent agreement between the predictions (52) for m".
and its experimental value

m". = 1784.1~~:~ MeV (55)

cited for several years by Particle Data Group (1992) or

m". = (1776.9 ± 0.4 ± 0.3)MeV , m". = (1776.3 ± 2.4 ± 1.4)MeV (56)

reported recently by Beijing Electron-Positron Collider Group (Qi 1992) and

ARGUS Collaboration (Albrecht 1992), respectively.
CLIFFORD ALGEBRAS AND ALGEBRAIC STRUCTURE 181

This strongly supports the phenomenological ansatz (50) operating with

the number N of "algebraic partons" involved in the families N = 1,3,5
and described by the Dirac bispinor indices as these appear in the Clifford
algebras (19) or, more conveniently, (22). The algebraic partons are agents
of the idea of algebraic compositness. In the picture which emerges from our
argument, any fundamental fermion with N = 1,3,5 is composed of one
"visible" algebraic parton of spin 1/2 and N - 1 = 0,2,4 "hidden" algebraic
partons of spins 1/2, the latter forming relativistic scalars.

References
H. Albrecht et al. (ARGUS Collaboration), DESY report 92-086 (June 1992).
T. Banks, Y. Dothan and D. Horn,Phys. Lett. B 117, 413 (1982).
Cf. e.g. H.A. Bethe and E.E. Salpeter, in Encyclopedia of Physics, Vol. 35, Springer,
Berlin-Gottingen-Heidelberg, 1957.
E. Kahler, Rendiconti di Matematica 21, 425 (1962); d. also D. Ivanenko and L. Lan-
dau, Z. Phys. 48, 341 (1928), published soon after the discovery of Dirac equation
(P.A.M. Dirac,Proc. Roy. Soc. 117, 610 (1928)). I am indebted to Andrzej Trautman
for his calling my attention to this early paper.
W. Krolikowski and J. Rzewuski, Nuovo Cim. 2, 203 (1955); 4, 974 (1956).
W. Krolikowski, Acta Phys. Pol. B 17,813 (1986).
W. Krolikowski, Acta Phys. Pol. B 18, 111 (1987); Nuovo Cim. 100 A, 701 (1988).
W. Krolikowski, Acta Phys. Pol. B 21, 871 (1990);Phys. Rev. D 45, 3222 (1992).
For a formal quantization scheme beyond the conventional quantum field theory d.
W. Krolikowski, Acta Phys. Pol. B 22, 613 (1991); [E.,Acta Phys. Pol. B 23, 83
(1992)].
Particle Data Group, Review of Particle Properties, Phys.Rev. D 45, Part 2 (June 1992).
Nading Qi, Bull. Am. Phys. Soc. 37, 957 (April 1992), caption only; information in: CERN
Courier 32, 13 (July 1992).
E.E. Salpeter, Phys. Rev. 87, 328 (1952).
CLIFFORD ALGEBRA OF TWO-FORMS,
CONFORMAL STRUCTURES, AND FIELD
EQUATIONS
INGEMAR BENGTSSON
Institute of Theoretical Physics
S-41296 Goteborg,
Sweden

Abstract.
I review the equivalence between duality operators on two-forms and conformal struc-
tures in four dimensions, from a Clifford algebra point of view (due to Urban tke and
Harnett). I also review an application, which leads to a set of "neighbo urs" of Einstein's
equations. An attempt to formulate reality conditions for the "neighbours" is discussed.

There is a deep theory for how to solve the self-dual Yang-Mills equations

*FO/{3i = 9 -1/2 90/-y9{38 rT:rt8

i = 1/29 -1/2 90/-y9{38 f -y61'1/FI'I/i = FO/{3i (1)

where the duality operator is defined with respect to some fixed conformal
structure, Le. a metric up to a conformal factor (and some useful notation
- the twiddle - has been introduced as well). Some time ago it occurred to
Urbantke (1984) to pose this problem backwards: Given a field strength,
with respect to which conformal structure is it self-dual? There is an elegant
solution to this curious question, and an elegant proof - due to Urbantke
and Harnett (1991) - based on the Clifford algebra of two-forms in four
dimensional spaces. For the moment, let me state the result and then indicate
how I want to use it. We need a triplet of two-forms, which is non-degenerate
in the sense that it may serve as a basis in the three-dimensional space of
self-dual two-forms. In particular, the index i ranges from one to three. Then

(2)

is U rbantke's formula. It gives the metric with respect to which FO/{3i is

automatically self-dual (the fiJk are the structure constants of SO(3), and
the conformal factor TJ is so far arbitrary).
Some work by Capovilla, Jacobson and Dell (1989) may be regarded as
a more ambitious version of Urbantke's formula. (See also Plebanski 1977,

183
Z. Oziewicz et al. (eds.). Spinors. Twistors. Clifford Algebras and Qtmntum Deformations. 183-188.
© 1993 Kluwer Academic Publishers.
184 INGEMAR BENGTSSON

Capovilla, Dell, Jacobson and Mason 1991.) We may regard Fo:(3i as the
self-dual part of the Riemann tensor, considered - at the outset - as just an
SO(3) field strength, with no connection to the metric. Then the question
arises whether it is possible to formulate a set of differential equations, using
the SO(3) connection (and the Levi-Civita tensor densities) alone, such that
the above metric becomes Ricci flat. The answer turns out to be yes; more
specifically, the answer is the field equations following from the action

(3)

where 'TJ is a Lagrange multiplier and

n·· -
IJ -
(0:(3-y8 F0: (3'1 Fe·
-YOJ (4)

The existence of this action is closely related to Ashtekar's (1987) formu-

lation of the 3+1 version of Einstein's theory - in fact the CDJ action is
a natural Lagrangian formulation of Ashtekar's variables. The action which
leads to Einstein's equations including a cosmological constant is less ele-
gant.
The next question is: What happens if we use the above building blocks
to write an arbitrary action

(5)

where the only restriction on t: is that it has density weight one? (Due to
the characteristic equation for three-by-three matrices, there are only three
independent traces.) The action is certainly generally covariant. Suppose
that we solve the field equations and use Urbantke's formula to define a
metric. Is that reasonable, and relevant for physics? What happens if we
change the structure group from SO(3) to something else?
Now that you know where I am going, we return to prove Urbantke's
formula. For any four-dimensional vector space V, the two-forms give a six-
dimensional vector space W, with a natural metric

(6)

There is a corresponding Clifford map to the space of endomorphisms on

V EB V*:

(7)

We see that the original vector space V now becomes the space of Weyl spin
ors for the Clifford algebra of two-forms.
CLIFFORD ALGEBRA OF TWO-FORMS 185

Now we introduce a metric on V, so that we can define the duality oper-

ator *. W then splits into two orthogonal subspaces W+ (self-dual forms)
and W- (anti-self-dual forms). We choose Euclidean signature, so that **
= 1, and without loss of essential generality we choose the determinant of
the metric to equal one. Then

(8)

where

(9)

Using a well-known property of six-dimensional ,-matrices, and Swedish

indices in W, we can find a totally anti-symmetric tensor zuiio such that

(10)

This determines Z uniquely, and we observe that

(11)
z~ = -~Z (~E w-).

We need a little bit more information about Z.

To prove the result we are after, we will commit the atrocity of choosing
a basis in W. First we choose an ON-basis in V, and then we set

(12)

Clearly, the X's (Y's) form a basis for W- (W+), and

(X;,Xj) = -Dij (Yi,Yj) = Dij; (13)

Looking back on eq. (11), we see that we can set

(14)

- that is to say that Z is the unit volume element of W+. But, since W+
is three-dimensional, this is all we need. In terms of an arbitrary basis ~a,:3i
on W+, eq. (10) now becomes

(15)
186 INGEMAR BENGTSSON

This is Urbantke's formula.

When the metric on V has neutral signature, the metric on W+ bec
omes indefinite, but the discussion is similar, while it becomes slightly more
subtle if the metric on V is Lorentzian.
With this understanding of eq. (2), let us return to the action (5). Our
main result so far (Capovilla 1992, Bengtsson and Peldan 1992, Bengtsson
1991, Peldan 1992) is that this action admits a 3+ 1 decomposition, and that
the resulting formalism is a natural generalization of" Ashtekar's variables"
for gravity. As is well known, the constraint algebra of general relativity
actually singles out the space-time metric by means of its structure functions.
For the 50(3) case, it turns out that - up to some ambiguity concerning
the conformal factor - the "Hamiltonian" metric is precisely the same as
Urbantke's. We refer to the models in this class as "neighbours of Einstein's
equations" , since they all have the same number of degrees of freedom. I will
not discuss the case of arbitrary structure groups here.
There are several holes that have to be filled before we can claim that we
have really been able to generalize Einstein's equations in an unsuspected
way. For the case of Euclidean signatures, we have to show that the field
equations derived from the action (5) ensure that the metric (2) is posi-
tive definite, rather than neutral. This can be done in specific cases. As an
example, consider the action

(16)

As is clear from the preceding discussion, there must be some property of

the field equations that ensure that the matrix n ij has definite signature.
To see this, choose a gauge such that the matrix becomes diagonal. Then it
is a straightforward exercise to show that the constraint that results when
varying the action with respect to 'fl implies that the matrix nij has definite
signature if and only if

0: ~ -1/2. (17)

In particular, 0: = -1/2, which leads to Einstein's equations, is all right. (I

owe this observation to Ted Jacobson.) Although it is not quite clear what
a general statement is, it is clear that, in general, the requirement that the
metric should have Euclidean signature will lead to some restrictions on the
allowed actions.
A similar discussion can be given for neutral signature, provided that the
de finition of the traces in the action is appropriately changed.
Our understanding of the Lorentzian case is in much worse shape. It is
necess ary to show that propagation is causal with respect to the metric
that we have defined. Moreover (since self-dual two-forms are necessarily
CLIFFORD ALGEBRA OF TWO-FORMS 187

complex in this case) the variables in the action are complex valued, and
one must show how to impose restrictions that imply that the metric is real
in any solution. I believe that the latter problem is the crucial one, and that
the former property somehow follows from the latter. It will not come as a
surprise if I state that the conformal structure is real if and only if

(18)

where the bar denotes complex conjugation. However, this condition is not
very helpful in itself. It is not difficult to write down solutions with real
Lorentzi an metrics - a small zoo of real solutions is already known, for
various "neighb ours" (generalizations of Schwarzschild, de Sitter, Kasner,
... ). On the other hand, there will always be some solutions for which the
metric is not real - also in the Einstein case. The correct formulation of the
problem is presumably to require that the space of real solutions should be
"reasonably" big - of the same order as the space of solutions of Einstein's
equations, say. It seems natural to switch to the Hamiltonian form of the
equations, and to address the problem from an initial data point of view.
Unfortunately, as soon as this is done, one discovers that the reality proper-
ties of the metric can be discussed easily (Ashtekar 1987) if and only if we
deal with the Einstein case - for the more general models contained in the
action (5), the calculations tend to b.
Which is where the matter stands at the moment. It is perhaps appro-
priate to add that we have investigated, in a preliminary way, whether the
"neighbours" can be used to explain any property of the real world. The
preliminary answer was not very encouraging, but perhaps the final verdict
is not in yet. Certainly the more difficult case of arbitrary structure groups
(Peldan 1992), which was not discussed here, should be carefully studied in
this regard.
Acknowledgements: I thank Helmuth Urbantke and Ted Jacobson for ex-
plaining things, and the organizers for a nice stay in the castle.

References
A. Ashtekar (1987): New Hamiltonian Formulation of General Relativity, Phys. Rev. D36
1587.
1. Bengtsson (1991): Self-Duality and the Metric in a Family of Neighbours of Einstein's
Equations, J. Math. Phys. 32 3158.
I. Bengtsson and P. Peldan (1992): Another "Cosmological" Constant, Int. J .Mod. Phys.
A71287.
R. Capovilla (1992): Generally Covariant Gauge Theories, Nucl. Phys. B373 233.
R. Capovilla, J. Dell and T. Jacobson (1989): General Relativity without the Metric, Phys.
Rev. Lett. 63 2325.
R. Capovilla, J. Dell, T. Jacobson and L. Mason (1991): Self-dual 2-forms and Gravity,
Class. Quant. Grav. 8 4l.
G. Harnett (1991): The Bivector Clifford Algebra, unpublished manuscript.
188 INGEMAR BENGTSSON

P. Peldan (1992): Ashtekar's Variables for Arbitrary Gauge Group, Phys. Rev. D, to
appear.
J. Plebanski (1977): On the Separation of Einsteinian Substructures, J. Math. Phys. 18
2511.
H. Urbantke (1984): On Integrability Properties of SU(2) Yang-Mills Fields.
I. Infinitesimal Part, J. Math. Phys. 25 2321.
DIRAC FORM OF MAXWELL EQUATION
LZn-GRADED ALGEBRAS

JAIME KELLER
Division de Ciencias Bcisicas, F. Q., and Quantum Theory of Matter Project
Facultad de Estudios Superiores-CuautitIcin, Universidad Nacional Autonoma de Mexico
Apartado Postal 70-528, 04510 Mexico, D. F., J(eller@UNAMVMl

(Received: March 31, 1993)

Abstract. A Dirac form of Maxwell's equation is derived. The interrelationship between

tensors and spinors is discussed in terms of ~2-graded algebras. A generalization to LZn-
graded algebras is given.

Key words: Clifford algebra - twistor - Maxwell equation - LZn-graded algebra

1. Introduction

This paper consists of two separate parts as reflected in the title.

Elie Cartan and then Marcel Riesz (in 1946) defined spinors as ideals in
Clifford algebras l . Twistors as ideals in Clifford algebra have been considered
by Chevalley in 1954, by Atiyah, Bott and Shapiro in 1964, by Crumeyrolle
since 1974 and also by Ablamowicz, Oziewicz and Rzewuski (1982). We use
their results in the derivation of a Dirac form of the Maxwell equation.
The Cartan and Riesz definition of spinors suggests that tensors are the
more fundamental. The tensor product of spinors (or twistors) is mapped
to tensors by the Cartan map. These simple facts are best expressed in
terms of ~2-graded (=:super) algebras. The ~ 2-graded Lie algebras were
introduced by Volkov and Akulov in 1973 and independently by Wess and
Zumino in 1974. Here we generalize this construction to ~n-graded algebras.
We are inspired by the lecture by Kerner on ~3-graded algebras in these
proceedings.
One can consider ~n-graded algebras in which the subalgebra of zero
grade consists of the direct sum of n Clifford algebras generated by the
set of n different pseudo-riemannian spaces. In the case of ~2-grading, the
first space can be chosen as Minkowski space-time and the second space
as an internal space. A similar theory (for groups rather than algebras)
was developed over the last ten years by Professor Jan Rzewuski (see his

1 Spinors and twistors as ideals must be inhomogeneous combinations of skewsymmetric

tensors of the even and odd degrees.

189
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 189-196.
© 1993 Kluwer Academic Publishers.
190 JAIME KELLER

contribution in this volume).

2. Twistors
Let {M,g} be the real pseudo-riemannian space-time with signature (+ -
- - ), and let AC be a module of differential one-forms over an algebra of com-
plex valued functions on M. The Witt decomposition (1937) of the differen-
tial one-forms is AC = Fl EB F2 where the Fi are maximal (two-dimensional)
isotropic <T-spaces, glFi == O. Denote by fi the two-forms fi E 1\2F (fl = 0).
The Clifford <T-algebra of {M,g} is denoted by Cl1,3'
DEFINITION 1 (Twistor). The left principal ideals ofC11,3'

7(fi) == CI1,3!i <J CI~,3'

genemted by the two-forms fi E 1\2 Fi C Cl1,3' are called the twist or spaces.
A twistor space 7 is a four dimensional linear <T-space with a hermitian
correlation (== Dirac conjugation) of signature ( + + - -) (Crumeyrolle 1974
and 1990).
Let Lin 7 denote the algebra of linear endomorphisms of the twistor CI-
module and let, be a Dirac representation (== algebra homomorphism) of
the Clifford algebra Ch,3 in the twistor space,
, : Cl1,3 --- Lin T. (1)
This is preciesly the meaning of the Dirac ,-matrices (Oziewicz 1986). For
the coordinate frames

Let {elL} be an g-orthonormal co-frame in Minkowski space-time.

The de Witt co-frame is the Sachs co-frame e AB == a1 B e IL such that
Fl == span{e li ,e l2 }, F2 == span {e 2i ,e 22 }.
Let {Xa} be a basis in a twistor left Clc-module 7 == clcMR, and let
{X a} be dual basis in the dual twistor right Clc-module 7* ==1R Melc,
b ,b
X Xa = Va'

Following Ablamowicz, Oziewicz and Rzewuski (1982) we choose the

bivector 12 and a basis in the twistor space {Xa} E 7(12) in terms of the
co-frame {elL},
Xl (eO-e 3 )(ie l -e 2)==h, (2)
X2 (1- eOe 3)(ie l e2 - 1),
X3 (1- eOe 3)(ie l - e2),
X4 (eo-e 3 )(ie l e2 -1).
DIRAC FORM OF MAXWELL EQUATION 191

The first two and last two spinors span the two-dimensional ([-spaces of the
Weyl spinors W.
Let
'!9~ == Xa ® l E Lin T,
be the spinorial basis in the matrix algebra Lin T.
A Dirac representation (1), ,IL
== ,(elL) E Lin T, in terms of the spinor
basis is (Keller and Rodriguez 1992)

,( eO) = '!91 + '!9; + '!9r + '!9L ,( eli) = V2 ( '!9r + '!9;), (3)

,(e l ) = i(-'!9~ + '!9~ + '!9~ - '!91), ,( elZ) = V2 ( '!9~ + '!91),
,( e2 ) = '!9~ + '!9~ - '!9~ - '!91 , ,( e 2i) = V2 (-'!91 + '!9~),
,( e 3 ) = '!91 - '!9; - '!9r + '!9L ,( e 2Z ) = V2i( -'!9~ + '!9~).
For every representation, and one-forms a and (3 E Cl we have,

,(a 1\ (3) = 21 b (0'),,((3)]. (4)

From this we calculate

,(eO 1\ e l ) i( -'!9~ + '!9i + '!9~ - '!9~), (5)
, ( eO 1\ e2) '!9~ + '!9i - '!9~ - '!9t
, ( eO 1\ e3 ) '!9t - '!9~ - '!9~ + '!9:,
,( e 1 1\ e 2 ) i( -'!9t + '!9~ - '!9~ + '!9:),
,( e 3 1\ e 1 ) i( -'!9~ - '!9i - '!9~ - '!9~),
,( e2 1\ e3 ) -'!9~ + '!9i - '!9~ + '!9~.

3. Dirac-Kahler operator
Let 9 denote a pseudo-riemannian structure. We have two mutually dual
Clifford algebras: the Clifford algebra of the multi vector fields, and the Clif-
ford algebra of the differential forms Cl == Cl g • Let , be the unique left
adjoint representation 2 (Oziewicz 1986),
, : Cl ---+ Lin Ct. (6)
This means that the Clifford product of two arbitrary differential forms
a, (3 E Cl is not denoted by juxtaposition, but by

,(0',(3) == (fO')(3==,cx(3 E Cl.

If a and (3 are differential I-forms, then

,cx(3 == 0'1\(3 + g(O',(3).
2 Note that (1) is a irreducible summand of (6): CI =T EEl ....
192 JAIME KELLER

For arbitrary differential forms 0 and /3, and multivectors X and Y, we use
"e" for exterior (Grassmann) product and "i" for interior product

(ixo)Y == o(X 1\ Y).

It follows that if 0 is a differential 1-form (covector) then

10/ == eO/ + igO/ E LinCl (7)

(d. Chevalley 1954, pp. 38-42). Let {Xa} be a Cartan frame of vector fields
and {wa} a dual co-frame of the differential one-forms,

We use the notation

When the representation I (6) is restricted to one-vector fields we get, in

these dual frames, the operator valued differential form

Similarly when we restrict I to differential one-forms we get the operator

valued vector field
1= I(w a ) @ Xa = @ Ow ,I-'
This LinCI-valued differential form was introduced by Fock and Ivanenko
in 1929 as "a line element". For differential one-forms 0 and /3 we have

10/ 0 , ,6 + 1,6 0 ' 0/ = 2g(0,,8)· idcl E LinC!.

If we define (!@/)(0@/3) == 10/ o,{3, then on symmetric second degree tensor

fields, I @ 1= 9 . idcl·

DEFINITION 2. The Dirac-Kahler operator on a Clifford algebra of the dif-

ferential forms is defined b'!f

This operator is both frame and coordinate independent because the defi-
nition uses dual frames. The Dirac-Kahler operator depends on the scalar
3 The early versions of the Dirac-Kahler operator was considered by Darwin (1928),
Landau and Ivanenko (1928) and by Marcel Riesz (1958). The most adequate references
are (Kahler 1962 and Hestenes 1966). David Hestenes denotes the Dirac-Kahler operator
by 0 and calls it the gradient.
DIRAC FORM OF MAXWELL EQUATION 193

product 9 and on the connection V. From the definition of the Clifford pro-
duct (7) it is clear that the Dirac-Kahler operator consists of the sum of two
frame-independent parts
v= d'V + b('V,g) ,
where

ew a 0 Vx a , (8)
= igw a 0 VXa'

One can show that

d'V = d + T,
where d is the connection independent Cartan exterior differential, (d 2 == 0),
and T is the torsion of the connection. For the co-differential (== divergence)
b('V,g) in (8), we also separate the connection independent part b == bg ,
(b 2 == 0) (d. Tucker 1986, p.180),

b('V,g) = b + co- Torsion + "V9 - dependent term".

DEFINITION 3. The Dirac operator is the restriction of the Dirac-Kilhler

operator to the ideal (spinor or twistor fields) in the Clifford algebra of the
differential forms (Tucker 1986, Secs 6 and 8).

4. Dirac form of the Maxwell equation

The Maxwell equations dF = 0 and bF =j were presented by Marcel Riesz
(1958) in the Dirac-Kahler form

VF = j E CIg,

where V == d+b is the Dirac-Kahler operator for the torsion-less riemannian

connection, V 9 = 0,

Consider the Dirac representation (1) of this equation in the twistor Cl-
module,

(9)

Here V XaF is a two-form:

194 JAIME KELLER

Therefore equations (4) and (9) gives

(10)

If t E T, then "Ijt E T. Equation (10) evaluated on twistor t is said to be

(t-dependent) Dirac form of a Maxwell equation. In this way every twistor
t E T linearly maps differential forms into subspaces in twistor <T-space T,

"I.t : A 3 Q f----> "Ic,l E T.

Let j == jp,eP, E AI. In basis (2) and using (3) we get

j f----> "IjXI (jo + j3)X3 - (ijl - j2)X4, (11)

j f----> "I j X2 (ijl + h)x3 + (jo - h)X4,
j f----> "IjX3 (jo -13)x1 + (ijl - j2)X2,
j f----> "IjX4 -(ijl + j2)x1 + (jo + h)X2.

This means that every basis twistor Xa E T maps non injectively co-vectors
into a Weyl <T-spinors in W. Moreover

F f----> "I FXI ihl - (F2 + iFI )X2, (12)

F f-----+ "IFX2 (iFI +F2)XI -F3X2'
F f-----+ "IFX3 -F3X3 + (iFl - F2)X4,
F f-----+ "IFX4 -(iFI + F2 )X3 + F3X4.
The Dirac representation of the Maxwell equation (10) was considered
by Oppenheimer (1931), Ohmura (1956), Moses (1958) and by Keller and
Rodriguez-Ramo (1991).

5. ~n-graded algebras
DEFINITION 4. An IR-algebra }( is said to be ~n -graded if

and if the algebra multiplication is a zero grade map

DIRAC FORM OF MAXWELL EqUATION 195

We are abreviating this last property by KiK j c Ki+ J.

Next we define the matrix ~n-graded algebra K. Let {Ab k = 1, ... , n}
be a sequence of associative lR-algebras with units. Let K be a direct sum
of the (Ak - A/)-bimodules
K == EB (kM/) == EB (AkMAI)·
The ~n-grading in K can be introduced in different ways. Every bimodule
kM/ is ~n-homogeneous. The simplest possibility is to put
deg (kMI) = I - k mod n.
The multiplication in K is defined by means of the Cartan map

EBk (iMk ® kM J ) --+ iMj.

The Cartan map is equivalent to matrix multiplication if we display K in
the matrix form

We are assuming that all (Ak-A/)-bimodules hMt} are the tensor product
of left and right modules,

and that the left and right Ai-modules are mutually lR or (l' -dual,
ev: Mi ® iM --+ lR or (l'.

If
dim(Ai) = (dim iM)2,
then Cartan map is a linear (or an algebra) isomorphism and we can identify
kMk == Ak. In this case (Kk)m = Kmk. It follows that if k does not divide
n, the algebra K is generated by every subspace Kb
K = gen {Kd. In the simplest case K = gen{EBi(iMi+d EB (nMd}.
The general ~2-graded algebra is

K == (~l
We can now let: Al be the complexified Dirac-Clifford algebra of Minkowski
space-time (or real Clifford algebra of the de Sitter space) CI'3I = CI 5 and
A2 == lR. This example of ~2-graded algebra is called a geom~tric supemI-
gebm in (Keller and Rodriguez 1992).
196 JAIME KELLER

Acknowledgements
The technical assistance of Mrs. Irma Aragon is greatly appreciated.

References
Ablamowicz Rafal, Zbigniew Oziewicz and Jan Rzewuski: 1982, 'Clifford algebra approach
to twistors', J. Math. Phys. 23, 231-242
Chevalley Claude: 1954, The Algebraic Theory of Spinors, Columbia University Press,
New York, 38-42
Crumeyrolle, A.: 1990, Orthogonal and Symplectic Clifford Algebras. Spinor Structures,
Kluwer A.P.: Dordrecht
Fock V. and D. Ivanenko: 1929, C.R. Acad. Sci. (Paris) 188, 1470
Hestenes David: 1966, Space- Time Algebra, Gordon and Breach, 26 and 64
Keller Jaime: 1991, 'Spinors and mutivectors as a unified tool for spacetime geometry
and for elementary particle physics', International Journal Theoretical Physics 30,
(2) 137-184
Keller Jaime and Rodriguez-Romo Suemi: 1991, International Journal of Theoretical
Physics 30, 185-196
Keller Jaime and Adan Rodriguez: 1992, 'Geometric superalgebra and the Dirac equation',
Journal of Math. Phys. 33 (1), 161
Kerner Richard: 1992, 'Zh-graded algebras', Journal of Math. Phys. 33 (1),403
Kerner Richard: 1993, ';;Z 3-graded structures', These Proceedings, p.
Moses H. E.: 1958, Nuovo Cimento 8 (Supp!. 1), 18
Ohmura T.: 1956, Progress of Theoretical Physics 16, 604
Oppenheimer R.: 1931, Physical Review 38, 725
Oziewicz Zbigniew: 1986, in Chisholm J.S.R. and A.K. Common, ed(s)., Clifford Algebras
and Their Applications in Mathematical Physics, From Grassman to Clifford, Reidel:
Dordrecht, NATO ASI C-183, pp. 245-255
Riesz Marcel: 1946, 'Sur certaines notions fondamentales en theorie quantique relativiste',
Dixieme Congres des Mathematiciens Scandinaves, Copenhaque , 123-143
Riesz Marcel: 1958, Clifford Numbers and Spinors, University of Maryland, The Institute
for Fluid Dynamics, Lecture Series #38
Rzewuski Jan: 1993, 'Structure of matrix manifolds and a particle model', These Proceed-
ings ,
Tucker R.W.: 1986, in Chisholm J.S.R. and A.K. Common, ed(s)., Clifford Algebras and
Their Applications in Mathematical Physics, A Clifford calculus for physical field the-
ories, Reidel: Dordrecht, NATO ASI C-183, pp. 177-199
Volkov D.V. and V.P. Akulov: 1973, Physics Letters 46B, 109
Wess Julius and Bruno Zumino: 1974, Nuclear Physics B70, 39
TRAVELLING WAVES WITHIN THE
CLIFFORD ALGEBRA
BERNARD JANCEWICZ
Institute of Theoretical Physics, University of Wroclaw,
pl. Maksa Borna 9, PL-50-204 Wroclaw, Poland

Abstract. It is shown that travelling electromagnetic waves (radiation fields) cannot

exist in a conducting medium. With the aid of algebra Cl 3 , however, an arbitrary plane
wave can be decomposed into a sum of two electromagnetic fields travelling in opposite
directions. These fields separately do not satisfy the Maxwell equations.

1. Introduction

David Hestenes in his works (1966, 1971, 1974a, 1974b, 1986) has demon-
strated the importance of Clifford algebras in various branches of classical
physics. Among others electrodynamics obtains a beneficial synthesis when
expressed in terms of CC 1 ,3, the Clifford algebra of the Minkowski space
(Jouvet and Schidlof 1932, Mercier 1935, Riesz 1958) or CC 3 , the Clifford
algebra of E3 (Hestenes 1966, Jancewicz 1988).
As was shown in Hestenes 1966, p. 29, when discussing the Maxwell
equations within CC 3 , it is useful to form the Clifford number (we propose
the term cliffor) E + Be123, but this is practical only in empty space. In the
presence of a material medium (and when a system of units is used in which
E and B have different physical dimensions) one has to take into account the
electric permittivity € and the magnetic permeability J.L. The best possibility
is f = vEE + ~Be123 which has the dimension J J 1m3 in SI system of
units, that is square root of the energy density. In this respect it resembles
the wave function of quantum mechanics which has the dimension of square
root of the probability density. We call this combination an electromagnetic
cliffor. Having denoted e = veE,
*b = ~Be123 we we write it explicitly
as a sum of vector and bivector:

With the notation f~ for the reversion of cliffors one obtains

-1ff~ = 1
w+-s
2 u

197
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 197-201.
© 1993 Kluwer Academic Publishers.
198 BERNARD JANCEWICZ

where
w = ~(€E2 + .!.B2) = ~(I e 12 + 1*b 12) = ~ 1f 12
2 Ii 2 2
is the energy density of the electromagnetic field, S = E X H = u( *b) . e
is the Poynting vector and u = ~ is the phase velocity of light in the
medium.
Having introduced V = V' + ~?t, the Fueter operator (Lounesto 1984)
and J = -!h
+ ViIj (here p is the charge density and j - the current density)
one may write the Maxwell equations in the following single formula

Vf= J (1)
ifthe medium is uniform in space and constant in time. We call Eq. (1) the
synthetic Maxwell equation. When J = 0 we call f a free electromagnetic
field; then
Vf =0. (2)
There exist solutions to Eq. (2) in form of the harmonic plane waves:

fer, t) = e±I(kr-nwt) N (3a)

or
fer, t) = e±I[wt-n(k-r)] N (3b)
where 1= e123, k =1 kin is a constant vector, W = u I k I and N is a vector-
plus-bivector such that Nk = -kN. The solutions (3) are plane waves. They
were found in a quaternionic language by Imaeda (1983) and written in the
above form in Jancewicz 1988. They were also considered independently by
Baylis and Jones (1989) in the Pauli algebra, the matrix representation of
ce 3 • • The novelty of expressions (3) lies in the combination of scalar and
vector present in the exponents of the exponential functions. This is the
reason why the phase velocity can not be introduced for them. Solution (3a)
can be called a plane wave with a round polarization, (3b) - a plane wave
with a spiral polarization (Jancewicz 1988, Section 4.4).

2. Travelling plane waves

It has been proved (J ancewicz 1988, Section 2.5) that the inequality

I S I~ uw
holds for an arbitrary electromagnetic field f. If one introduces the velocity
v of energy transport through the relation S = wv, the above inequality
shows that v is bounded from above:

1v I~ u.
TRAVELLING WAVES WITHIN THE CLIFFORD ALGEBRA 199

One can prove that equality occurs in the above relation iff a unit vector
n exists such that nI = -In and the following equivalent conditions are
satisfied:
(i) e = n(*b),
(ii) *b = ne,
(iii) nI = I,
(iv) I = (1 + n)e = (1 + n)(*b).
In this case we say that the electromagnetic field I is travelling in the direc-
tion n. Conditions (i) - (iv) are also equivalent to

(4)

Such a field is also known as a radiation field.

Waves (3) generally are not travelling fields in this sense. This fits the
observation that the phase velocity can not be defined for them. However,
with the aid of two idempotents P±n = ~(1 ± n) satisfying Pn + P- n = 1
and PnP- n = 0, one can decompose (3) into sums

of two fields travelling in opposite directions ±n (Jancewicz 1988, Section

4.2). They separately satisfy the synthetic Maxwell equation Eq. (2).
Solutions (3) are decomposed into the form

I = e±I(k.r-wt) N + + e±I(k.r+wt)N_ (5a)

I = e±I(wt-k-r) N + + e±I(wt+k-r) N_ (5b)
where N± = P±nN. The exponents show now that the phase velocities can
be introduced and they are opposite: ±un in the two terms of both sums.

3. Plane waves in a conducting medium

We assume now p = 0 and j = aE where a is the conductivity of a medium.
Then the synthetic Maxwell equation assumes the form

(6)

and has a particular solution in form of the harmonic plane wave:

(7)

where k = Tzu J ~ + 1, k = kn, 'Y = -;;t/v'1 + K,2 - 1, C is a con-

stant vector orthogonal to n, K, = a/cw (Jancewicz 1988, Section 5.1).
200 BERNARD JANCEWICZ

The exponent in the periodic exponential in (7) shows that a phase ve-
locity can be introduced and is equal to 7fn. The field (7), however, does not
satisfy (4), so it is not the travelling field in our sense. The same is valid for
any other plane wave solutions if a =1= 0, therefore we claim that travelling
waves can not exist in a conducting medium. Some authors (e.g. Jackson
1975, p. 270) consider that the possibility of introducing phase velocitites is
sufficient to call a solution the travelling wave. In our opinion such solution
should be called differently, let it be an advancing-phase wave.
With the aid of the same idempotents P±n, field (7) can be decomposed:

f = Pnf + P-nf = f+ + f-
into two travelling fields:

which, however, separately do not satisfy the Maxwell equation (6). What is
striking, the both fields travelling in opposite directions ±n have the same
phase velocity in the direction +n.
The ratio of energy fluxes of the two travelling fields is independent of
time and position:
R =IS_(r,t)l=k-~ ()
8.
I S+(r,t) I k+ ~
This yields R ~ ~~ for small conductivity a and R ~ 1 for large a.
If there is an interface perpendicular to n at n . r = Xo and a dielectric
medium with the same f and f.l is present for n· r < xo, the continuity of the
electromagnetic field implies that fields f± pass smoothly into free travelling
electromagnetic waves:

with the same ratio (8) of the intensities of the two waves. Here f+ can be
interpreted as the incident wave and f _ as the reflected wave from the con-
ductor. Thus (8) can be viewed as the reflection coefficient of the conducting
medium.
It is, moreover, possible to show (Jancewicz 1991) that travelling electro-
magnetic waves also can not exist in a nonhomogeneous medium.

References
W. E. Baylis and G. Jones: "Relativistic dynamics of charges in external fields: the Pauli
algebra approach", J. Phys. A22,17-29(1989).
D. Hestenes: Space-Time Algebra, Gordon and Breach, New York 1966,1987.
TRAVELLING WAVES WITHIN THE CLIFFORD ALGEBRA 201

D. Hestenes: "Vectors, spinors and complex numbers in classical and quantum physics",
Am. J. Phys.39,1013-1027(1971).
D. Hestenes: "Proper particle mechanics" J. Math. Phys. 15, 1768-1777(1974a).
D. Hestenes: "Proper dynamics of a rigid point particle" J. Math. Phys. 15,1778-
1786(1974b).
D. Hestenes: New Foundations for Classical Mechanics, D. Reidel, Dordrecht 1986.
J.D. Jackson: Classical Electrodynamics, Wiley, New York 1962, 1975.
B. Jancewicz: Multivectors and Clifford Algebra in Electrodynamics, World Scientific, Sin-
gapore 1988.
B. Jancewicz: "Plane electromagnetic wave propagating parallel to the gradient of the
refractive index" 1. Opt. Soc. Am. A 8, 1529-1535(1991).
G. Jouvet and A. Schidlof: "Sur les nombres hypercomplexes de Clifford et leurs appli-
cations a l'analyse vectorielle ordinaire, a l'electromagnetisme de Minkowski et a la
theorie de Dirac", Bull. Soc. Neuchate/ Sci. Nat. 57,127-147(1932).
P. Lounesto: "Spinor function theory" Simon Stevin Quart. 1. Pure and Appl. Math.
58,193-218(1984).
A. Mercier: Expression des equations de l'electromagnetisme au moyen des nombres de
Clifford. These no 935, Universite de Geneeve 1935.
M. Riesz: Clifford Numbers and Spinors, Lecture Series No. 38, University of Maryland,
College Park 1958.
HAMILTONIAN MECHANICS WITH
GEOMETRIC CALCULUS
DAVID HESTENES
Department of Physics and Astronomy
Arizona State University
Tempe, Arizona 85287
USA

Abstract. Hamiltonian mechanics is given an invariant formulation in terms of Geometric

Calculus, a general differential and integral calculus with the structure of Clifford algebra.
Advantages over formulations in terms of differential forms are explained.

INTRODUCTION
In the recent renaissance of Analytic Mechanics, the calculus of differential
forms has become the dominant mathematical language of practitioners.
However, the physics community at large has been slow to adopt the lan-
guage. This reluctance should not be attributed solely to the usual resistance
of communities to innovation, for the calculus of forms has some serious defi-
ciencies. For one thing, it does not articulate smoothly with vector calculus,
and it is inferior to vector calculus for many applications to Newtonian
mechanics. Another drawback is that the calculus of forms has accreted
a veritable orgy of definitions and notations which make the preparation
required to address even the simplest problems in mechanics inordinately
excessive. This is evident, for example, in the pioneering textbook of Abra-
ham and Marsden (1967), which provides nearly 200 pages of preparation
before attacking any significant problem in mechanics. The same high ra-
tio of formalism to results is characteristic of more recent books in the field,
such as Libermann and MarIe (1987). All this goes to show that the calculus
of forms is not quite the right tool for mechanics.
Without denying that valuable insights have been gained with differen-
tial forms, the contention of this paper is that a better mathematical system
is available for application to analytical mechanics; namely, the Geometric
Calculus expounded by Hestenes and Sobczyk (1984, henceforth referred to
as [GC]). In contrast to differential forms, this calculus includes and general-
izes standard vector calculus with no need to change standard notation, and
it has proven advantages in applications throughout Newtonian mechanics,
most notably in rigid-body mechanics (Hestenes, 1985). Geometric Calculus

203
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 203-214.
© 1993 Kluwer Academic Publishers.
204 DAVID HESTENES

also includes and generalizes the calculus of differential forms, as explained

in [GC]. In particular, it embraces the quaternion theory of rotations and
the entire theory of spinors, which are completely outside the purview of
differential forms. This apparatus is crucial to the efficient development of
rigid-body dynamics (Hestenes, 1985).
This paper shows how to employ Geometric Calculus in the formulation of
Hamiltonian mechanics, though space limitations preclude the discussion of
applications or advanced theory. However, the fundamentals are discussed
in sufficient detail with supplementary references to make translation of
standard results in symplectic geometry and Hamiltonian mechanics into
the language of Geometric Calculus fairly straightforward.

1. VECTOR SPACE VERSION

The reader is presumed to be familiar with Clifford algebra and Hamiltonian
mechanics, but familiarity with [GC] will be needed for full comprehension
of the ideas, as well as for their applications. Definitions, notations, and
results from [GC] will be employed without explanation. Though Geomet-
ric Calculus makes a completely coordinate-free approach possible, it also
facilitates computations with coordinates. Coordinates are employed here
primarily to establish a relation to conventional formulations.
For a mechanical system described by coordinates {ql, ... , qn} and cor-
responding momenta {PI, ... ,Pn}, we first define configuration space as an
n-dimensional real vector space nn spanned by an orthonormal basis {ekl
with
(1.1 )
for j, k = 1,2, ... ,n. The state of the system can then be represented by the
pair of vectors
q = L.qkek, P = L.Pkek· (1.2)
k k

The vectors in configuration space generate a real Geometric Algebra, nn =

(J(nn), with geometric product

qp = q . p + q 1\ p. (1.3)

Differentiation with respect to vectors is defined in [GC, Chap.2] along with

the necessary apparatus to perform computations without resorting to co-
ordinates. However, it will suffice here to introduce the vector derivative Oq
by specifying its relation to the coordinates:

(1.4)
HAMILTONIAN MECHANICS WITH GEOMETRIC CALCULUS 205

Equation (1.2) can be solved to express the coordinates as functions of the

vector q instead of as independent variables; thus

(1.5 )

Then the basis vectors ek are given as gradients

(1.6)

The simple linear form (1.5) for the coordinate functions obtains only for
orthogonal coordinates, but the general case is treated in [GC]. It should
be noted, also, that the "inner product" in (1.1) and (1.5) has no physical
significance as a "metric tensor." It is merely an algebraic mechanism for
expressing functional relations. Among other things, it performs the role of
contraction in the calculus of differential forms.
For a Hamiltonian, H = H(q,p), Hamilton's equations of motion can be
expressed in configuration space as the pair of equations
q= &pH, (1.7)
P = -&qH. (1.8)
Since p and q are independent variables, we can reduce this pair of coupled
equations to a single equation in a space of higher dimension. However, to
be useful, the extension to higher dimension must preserve the essential
structure of Hamilton's equations in a way which facilitates computation.
We now show how such a computationally efficient extension can be achieved
with Geometric Calculus.
To that end, we define momentum space as an n-dimensional real vector
space fin spanned by an orthonormal basis {ek} with

(1.9)

so the momentum of our mechanical system can be expressed as a vector

(1.10)

Now we define phase space R2n as the direct sum

(1.11)

This generates the phase space (geometric) algebra R2n = g(R2n), which is
completely defined by supplementing (1.1) and (1.9) with the orthogonality
relations
(1.12)
206 DAVID HESTENES

The symplectic structure of phase space is best described by introducing

a symplectic bivector
J='Lh (1.13)
k
with component 2-blades

(1.14)
The bivector J determines a unique pairing of directions in configuration
space with directions in momentum space, as expressed by
ek = ek·J = ek·Jk = ekJk = -hek, (1.15)
ek = J. ek = Jk . ek = Jkek = -ekJk. ( 1.16)
Each blade h pairs a coordinate qk with its corresponding momentum Pk.
Moreover, since each h satisfies

Jr, = -1, ( 1.17)

it functions as a "unit imaginary" relating qk to Pk. Thus, the bivector
J determines a unique complex structure for phase space. The symplectic
structure on phase space can be described without the reference (1.14) to
basis vectors by defining the symplectic bivector J through a specification of
its general properties. The symplectic bivector determines a skew-symmetric
linear transformation l which maps each phase space vector x into a vector

x = lx = x·J. (1.18)

This, in turn, defines a skew-symmetric bilinear form

X· Y = y·(lx) = x·J·y = J. (y /\ x) = -y. x. (1.19)

This bilinear form is nondegenerate if and only if x is nonzero whenever

x is nonzero, or, equivalently, if and only if < In >2n is a nonvanishing
pseudoscalar. With respect to the basis specified by (1.14),

< r >2n = '---.00-'

J /\ ... /\ J = n!( -1) [n/21EnEn' (1.20)
n times

where En = ele2 ... en = el /\ e2 /\ ... en, En = ele2'" en, and [n/2] is the
greatest integer in n /2. The "complex structure" expressed by (1.17) can be
characterized more generally by
(1.21 )

It follows that
( 1.22)
HAMILTONIAN MECHANICS WITH GEOMETRIC CALCULUS 207

which can be regarded as a "hermitian form" associated with the complex

structure.
The group of linear transformations on phase space which preserves sym-
plectic structure is called the symplectic group. It has recently been shown
that the symplectic group has a natural representation as a "spin group"
(Doran et. at., 1992). This promises to be the ideal vehicle for characterizing
symplectic transformations.
Now, to define Hamiltonian mechanics on phase space, from the "position
and momentum vectors" (1.2) we can describe the state of our physical
system by a single point x in phase space defined by

x = if + p = p + q. J. (1.23)

The derivative with respect to a phase space point is then given by

(1.24)

and we have
(1.25)
The Hamiltonian of the system is a scalar-valued function on phase space

H = H(x) = H(q,p). (1.26)

Accordingly, Hamilton's equation for a phase space trajectory, x = x(t), of

the system assumes the simple form

x=8H. (1.27)

The transcription of the entire theory of Hamiltonian systems into this

invariant formulation is now straightforward. For example, for any scalar-
valued phase space function, G = G( x), the Poisson bracket can be defined
by
{H,G} = (8H).aG = -{G,H}. (1.28)
Its equivalence to the conventional definition in terms of coordinates is pro-
vided by

{H, G} = (8pH - aqH). (ap + 8q)G

= (apH). aqG - (aqH). (apG)

(1.29)

= Lk [(~~) (~~) - (~~) (;~)]

The definition (1.28) does not actually require that G be scalar-valued, so it
can be applied to any multivector-valued function, M =M(x), describing
208 DAVID HESTENES

some "observable" property of the system. It follows that the equation of

motion for the observable is given by

if = x·oM = (aH).oM = {H,M}. (1.30)

For M =x we have
{H,x} = (8H).ox = 8H, (1.31 )
so Hamilton's equation (1.27) can be expressed in the form

x={H,x}. (1.32)

According to (1.30), M is a constant of the motion if {H,M} = o. It

follows that H is a constant of the motion, since

{H, H} = (8H).(oH) = J. (oH 1\ oH) = o. (1.33)

Our next task is to generalize this approach to Hamiltonian mechanics on

manifolds.

2. VECTOR MANIFOLD VERSION

The initial characterization of configuration space in the preceding section
depends on the choice of coordinates. There is a "canonical" choice, though.
For a system of N particles a configuration space of dimension n = 3N is
naturally defined by
n3N = n3 EEl ••• EEl n3 (2.1 )
"------v------ '
N times
where a separate copy ofthe 3-dimensional "physical space" is alloted to each
particle. Whatever the choice of "generalized coordinates," its relation to
physical space must be maintained, so a mapping to the choice (2.1) must be
specified. For many purposes, however, this mapping is not of interest, so we
desire a formulation of mechanics where it can be suppressed or resurrected
as needed.
For a system of particles or rigid bodies with constraints, the space of
allowable states is a manifold of dimension 2n equal to the number of inde-
pendent degrees of freedom. Although this manifold can be mapped locally
into the vector space representation of phase space in the preceding section,
this is awkward if the system has cyclic coordinates. Alternatively, we can
describe here the representation of phase space as a 2n-dimensional vector
manifold M2n. The mathematical apparatus needed for differential and in-
tegral calculus on vector manifolds has already been developed in [GC]. The
phase space manifold M2n can be regarded as embedded in a vector space of
higher dimension (e.g., of dimension 6N for an N particle system), but this
HAMILTONIAN MECHANICS WITH GEOMETRIC CALCULUS 209

is not required except, perhaps, to describe the relation to physical space

expressed by (2.1).
The mathematical apparatus in [GC] enables us to adapt our vector space
version of Hamiltonian mechanics to a vector manifold version with compar-
atively minor alterations. The main difference is that the algebraic relations
of interest will be defined on the tangent spaces of the manifold instead of
on the manifold itself.
Each point x on the phase space manifold M 2m represents an allowable
state of the system. The symplectic bivector J of the preceding section
becomes a non degenerate bivector field J = J( x) on M 2 n with values in the
tangent algebra [GC, ChapA]. For vector fields v = v(x) and u = u(x), in
the tangent space at each point x, J (x) determines a linear transformation

v=lv=v·J (2.2)
and a corresponding nondegenerate bilinear form

u· v = -v.ii, (2.3)
just as in (1.18) and (1.19). However, a direct analogue of (1.21) is not
feasible, because it may conflict with requirements on the derivatives of J.
Instead, however, we can introduce another bivector field K = K(x) with
the property
(2.4)
Thus, K = l-l is the inverse of l. Now the Jacobi identity [GC, p.14]
implies that

K lv = K·(v·J) = (K.v).J +v·(K X J),

where K X J = !(K J - J K) is the commutator product. So if l is to be
the inverse of K, we must have

K XJ = 0, (2.5)
or the equivalent operator equation

K l = lK = 1. (2.6)
To specify the relation of K to J more precisely, we note that, as in (1.13),
they can each be expressed as a sum of n commuting blades.

J = 'LJk, K = "LKk. (2.7)

k k
Moreover, we can select each Kk to be proportional to Jk. Then the condition
K = l-l can be expressed by the more specific condition
F
l~k = J-1
k (2.8)
2\0 DAVID HESTENES

for each k. This generalizes the condition J;l = -h in (2.8). Incidentally,

we note that
J·K = 'LKk·h = n. (2.9)
k
Modern approaches to Hamiltonian mechanics (Abraham and Marsden,
1967; Libermann and MarIe, 1987) begin with symplectic manifolds. A man-
ifold M2n is said to be symplectic if it admits a closed, non degenerate 2-form
w. As shown in [GC], this is equivalent to admitting a closed, non degenerate
bivector field K on the vector manifold. Indeed, the 2-form can be defined
by
w = ](.(dxAdy), (2.10)
where dx and dy are tangent vectors. The 2-form is said to be closed if its
"exterior differential" vanishes, that is, if

dw = (dxAdyAdz),(oAK) = O. (2.11)
This condition is obviously satisfied if K has vanishing curl:

oAK = O. (2.12)

Actually, though, (2.11) implies only the weaker condition of vanishing

cocurl:
'V A K = P(o A K) = 0, (2.13)
where P is the projection into the tangent algebra of M2n (see [GC, p.140]).
The tangent algebra is essentially the same thing as the "Clifford bundle"
which "pastes" Clifford algebras on manifolds, instead of generating them
from a vector manifold as in [GC]. The coderivative 'V as well as the deriva-
tive 0 is an essential concept for calculus on vector manifolds, and its prop-
erties are thoroughly discussed in [GC, Chapt.4], so we can exploit some of
its properties without establishing them here.
Instead of translating the "differential forms approach" into geometric
algebra, it is more enlightnening to ascertain directly what condition on
the bivector field J = J (x) are required to ensure the essential features of
Hamiltonian mechanics on M2n. Hamilton's equation (1.27) can be adopted
without change. The Hamiltonian H( x) determines a vector field h = h( x)
on M2n given by
h = 8H = (oH) .J. (2.14)
Hamilton's equation
x(t) = h(x(t)) (2.15)
determines integral curves of this vector field. This condition that these
curves describe an "incompressible flow" is given by Liouvilles Theorem

'V. h = o· h = o. (2.17)
HAMILTONIAN MECHANICS WITH GEOMETRIC CALCULUS 211

Since
8·h = 8·(-J·h) = -(8·J)·h + J·(8Ah).
and 8 A h = 8 A 8H = 0, the condition

\l.J = P(8·J) = 0 (2.18)

suffices to imply Liouvilles Theorem. We adopt (2.18) instead of the weaker

condition h· (8.J) = (hAS)·J = 0, because it appears to be essential for
the theory of canonical transformations outlined below.
The definition (1.28) for the Poisson bracket can be taken over to M 2 n
without change. However, the role of J in determining its properties must
be examined. Scalar-valued functions F = F(x), G G(x), H = H(x)
determine vector fields

(2.19)

Let us refer to such fields as symplectic vector fields. It follows from (2.19)
that
8· f = -J. (8 A J) = 0, (2.20)
but (2.18) implies the stronger condition

8·1 = -8·f = o. (2.21)

Therefore, all nonvanishing symplectic vector fields generate incompressible

flows on (or automorphisms of) M2n.
The Poisson bracket can be written in a variety of forms, including

{F, G} = - J . (J A g) = 1· 9 = -g. f (2.22)

= 8· (Fg) = -8· (GJ).

Alternatively, using (2.4), one can write

{F, G} = J(. (1 Ag), (2.23)

which, according to (2.10), expresses the bracket as a 2-form evaluated on

symplectic vector fields. This is closer to conventional formulations in terms
of differential forms. However, (2.22) is simpler because K is not involved.
An essential property of the Poisson bracket is the Jacobi identity

{F,{G,H}} + {G,{H,F}} + {H,{F,G}} = o. (2.24)

Using (2.22) to express the left side of (2.24) in terms of vector fields, we
obtain
-8· [(g. h)f + (h.J)g + (f.g)hl
212 DAVID HESTENES

= 8.[J.(jAgAh)]
= (jAgAh).(§AJ)+(JA8).(jAgAh)

=(jAgAh).(8AJ)-!(JAJ).[aA(jAgAh)]. (2.25)
This computation employed the algebraic identities

J. (j Ag A h) = J. (j Ag)h - J. (j A h)g + J. (g A h)f (2.26)

= (g. f)h + (j.h)g + (h·g)f
[GC, eqn.(1-1.40)], and

(JAJ)·a=2JA(J·a)= -2JA8, (2.27)

(JAJ)·[aA(jAgAh)] = [(JAJ).a].(jAgAh) (2.28)
[GC, eqn. (1-1.25b) or (1-4.6)].
The last term in (2.25) vanishes identically when f, 9 and h are gradients.
Therefore, from (2.25) it follows that the Jacobi identity (2.24) obtains if
and only if
VAJ = P(8AJ) = o. (2.29)
This condition is not independent ofthe "incompressibility condition" (2.18),
for from (2.27) we obtain the relation

! V . (J A J) = (V· J) A J - (J . V) A J
(2.30)
= J A (V· J) + V A J.
Thus, (2.24) and (2.29) together imply

V·(J AJ) = O. (2.31)

In analogy with (1.30), a multivector field M = M(x) which is invariant

1
under the flow generated by a symplectic vector field = 8F satisfies

{F,M} = j.VM = O. (2.32)

Note the use of j. V instead of 1· a

when M is not scalar-valued. A flow is
said to be a canonical transformation when it leaves the symplectic bivector
J invariant, that is, when

{F,J} = j.VJ = O. (2.33)

The differentiable vector fields on a manifold compose a Lie algebra under

the Lie bracket defined for vector fields u = u( x) and v = v( x) by

[u,vj = u·av - v·au = V· (uAv) + uV·v - vV·u. (2.34)

HAMILTONIAN MECHANICS WITH GEOMETRIC CALCULUS 213

The properties of the Lie bracket are studied at length in [GC]. For sym-
plectic fields we derive the identity

[j,g] = j.ag - g.aj = {F,8G} - {G,8F} (2.35)

= 8{F,G} + f· (g·aJ) - g. (j.aJ).
According to (2.33), the last two terms in (2.35) vanish for canonical trans-
formations. Therefore, the canonical transformations compose a closed Lie
algebra on M 2 n, and the Poisson bracket of "canonical generators" F and
G is also a canonical generator. This should suffice to show how the general
theory of canonical transformations can be developed on vector manifolds.
As a final point, the crucial role of the symplectic bivector J in canonical
transformations suggests that it should be more intimately linked with the
Hamiltonian H in the theory. One attractive possiblility for linking them is
to introduce a bivector field n given by

n=HJ. (2.36)

Then (2.18) implies

h = (aH).J = \j.(HJ) = \j·n, (2.37)

and Hamilton's equation (2.15) takes the form

x=\j·n. (2.38)

Thus, n is a bivector potential for Hamiltonian flow, and H plays the role
of an integrating factor for this bivector field. This is very suggestive!

3. CONCLUSIONS
Experts will have noted that phase space is identified with its own dual
space in the preceding formulation of Hamiltonian mechanics. Some may
claim that the conventional formulation in terms of differential forms is
preferable because it does not make that identification. On the contrary, it
can be argued that such generality is excessive, contributing little if anything
to deepening analytical mechanics, while introducing unnecessary complica-
tions. Be that as it may, it should be recognized that the identification of
phase space with its dual is a deliberate choice and not an intrinsic limita-
tion of geometric algebra. Indeed, the geometric algebra apparatus needed
to separate phase space from its dual is available in Doran et. ai. (1992) and
ready to be applied to mechanics. Ironically, that apparatus automatically
produces a kind of quantization, something which can only be imposed ar-
tificially in conventional approaches. It remains to be seen if that fact has
significant physical import.
214 DAVID HESTENES

The purpose of this short paper has been to lay the foundation for a
reformulation of analytical mechanics in the language of geometric calcu-
lus. Translation of standard results into this language is not difficult, but it
will not be without surprises and new insights as the treatment above al-
ready suffices to show. Though the emphasis here has been on an invariant
methodology, a powerful apparatus for dealing with coordinates is available
in [GC]. One especially promising possibility is an extention of the invariant
formulation for rigid-body mechanics in Hestenes (1985) to a phase space
formulation for systems of linked rigid bodies. That is likely to have impor-
tant applications to robotics.

References
R. Abraham and J. Marsden (1967), Foundations of Mechanics, W.A.Benjamin, New York.
C. Doran, D. Hestenes, F. Sommen, and N. VanAcker (1992), Lie Groups as Spin Groups,
J.Math.Phys. (submitted).
D. Hestenes (1985), New Foundations for Classical Mechanics, Kluwer, Dordrecht/Boston;
paperback ed. (1987); fourth printing with corrections (1992).
D. Hestenes and G. Sobczyk (1984), Clifford Algebra to Geometric Calculus, Kluwer,
Dordrecht/Boston; paperback ed. (1987); third printing with corrections (1992).
P. Libermann and C-M Marie (1987), Symplectic Geometry and Analytical Mechanics,
Reidel, Dordrecht/Boston.
GRASSMANN MECHANICS, MULTIVECTOR
DERIVATIVES AND GEOMETRIC ALGEBRA
CHRIS DORAN *
MRAD, Cavendish Laboratories, Madingley Road, Cambridge CB30HE, UK
ANTHONY LASENBY
DAMTP, Silver Street, Cambridge, CB3 9EW, UK

and
STEVE GULL
DAMTP, Silver Street, Cambridge, CB3 9EW, UK

Abstract. A method of incorporating the results of Grassmann calculus within the frame-
work of geometric algebra is presented, and shown to lead to a new concept, the multivec-
tor Lagrangian. A general theory for multi vector Lagrangians is outlined, and the crucial
role of the multivector derivative is emphasised. A generalisation of Noether's theorem is
derived, from which conserved quantities can be found conjugate to discrete symmetries.

1. Introduction
Grassmann variables enjoy a key role in many areas of theoretical physics,
second quantization of spin or fields and supersymmetry being two of the
most significant examples. However, not long after introducing his anticom-
muting algebra, Grassmann himself [Grassmann, 1877] introduced an inner
product which he unified with his exterior product to give the familiar Clif-
ford multiplication rule

ab = a·b + aAb. (1)

What is surprising is that this idea has been lost to future generations of
mathematical physicists, none of whom (to our knowledge) have investigated
the possibility of recovering this unification, and thus viewing the results of
Grassmann algebra as being special cases of the far wider mathematics that
can be carried out with geometric (Clifford) algebra [Hestenes & Sobczyk,
1984].
There are a number of benefits to be had from this shift of view. For
example it becomes possible to "geometrize" Grassmann algebra, that is,
give the results a significance in real geometry, often in space or spacetime.
Also by making available the associative Clifford product, the possibility of
• Supported by a SERe studentship.

215
Z. Dziewicz et al, (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 215-226.
© 1993 Kluwer Academic Publishers.
216 CHRIS DORAN ET AL.

generating new mathematics is opened up, by taking Grassmann systems

further than previously possible. It is an example of this second possibility
that we will illustrate in this paper.
A detailed introduction to these ideas is contained in [Lasenby et ai.,
1992b], which is the first of a series of papers [Lasenby et ai., 1992a; Lasenby
et at., 1993; Lasenby et at., 1992c; Doran et at., 1993) in which we aim
to show that many of concepts of modern physics, including 2-spinors,
twistors, Grassmann dimensions, supersymmetry and internal symmetry
groups, can be expressed purely in terms of the real geometric algebras
of space and spacetime. This, coupled with David Hestenes' demonstration
that the Dirac and Pauli equations can also be expressed in the same real
algebras [Hestenes, 1975], has led us to believe that these algebras (with
multiple copies for many particles) are all that are required for fundamental
physics.
This paper starts with a brief survey of the translation between Grass-
mann and geometric algebra, which is used to motivate the concept of a
multi vector Lagrangian. The rest of the paper develops this concept, making
full use of the multi vector derivative [Hestenes & Sobczyk, 1984). The point
to stress is that as a result of the translation we have gained something
new, which can then only be fully developed outside Grassmann algebra,
within the framework of geometric algebra. This is possible because geo-
metric algebra provides a richer algebraic structure than pure Grassmann
algebra.
Throughout we have used most of the conventions of [Hestenes & Sobczyk,
1984), so that vectors are written in lower case, and multivectors in upper
case. The Clifford product of the multivectors A and B is written as AB.
The subject of Clifford algebra suffers from a nearly stifling plethora of con-
ventions and notations, and we have settled on the one that, if it is not
already the most popular, we believe should be. A full introduction to our
conventions is provided in [Lasenby et ai., 1992b].

2. Translating Grassmann Algebra into Geometric Algebra

Given a set of n Grassmann generators {(i}, satisfying
(2)
we can map these into geometric algebra by introducing a set of n indepen-
dent vectors {ei}, and replacing the prod uet of Grassmann variables by the
exterior product,
(i(j ~ ei /\ ej. (3)
In this way any combination of Grassmann variables can be replaced by a
multivector. Note that nothing is said about the interior product of the ei
vectors, so the {ei} frame is completely arbitrary.
GRASSMANN MECHANICS 217

In order for the above scheme to have computational power, we need

a translation for the second ingredient that is crucial to modern uses of
Grassmann algebra, namely Berezin calculus [Berezin, 1966]. Looking at
differentiation first, this is defined by the rules,
a(j
a(i = 0"'J (4)

8 (5)
(j a(i = Oij,

(together with the graded Leibnitz rule). This can be handled entirely within
the algebra generated by the {ei} frame by introducing the reciprocal frame
{ e i }, defined by

(6)

Berezin differentiation is then translated to

a (7)
a(i ( ...... e i .(

so that

a(j ...... ei .ej = oj. (8)

a(i
Note that we are using lower and upper indices to distinguish a frame from
its reciprocal, rather than to simply distinguish metric signature.
Integration is defined to be equivalent to right differentiation, i.e.

(9)

In this expression f(O translates to a multivector F, so the whole expression

becomes

(10)
where En is the pseudoscalar for the reciprocal frame,

(11)
and (FEn) denotes the scalar part of the multivector FEn.
Thus we see that Grassmann calculus amounts to no more than Clifford
contraction, and the results of "Grassmann analysis" [de Witt, 1984; Berezin,
1966] can all be expressed as simple algebraic identities for multivectors. Fur-
thermore these results are now given a firm geometric significance through
218 CHRIS DORAN ET AL.

the identification of Clifford elements with directed line, plane segments etc.
Further details and examples of this are given in [Lasenby et ai., 1992b].
It is our opinion that this translation shows that the introduction of
Grassmann variables to physics is completely unnecessary, and that instead
genuine Clifford entities should be employed. This view results not from
a mathematical prejudice that Clifford algebras are in some sense "more
fundamental" than Grassmann algebras (such statements are meaningless),
but is motivated by the fact that physics clearly does involve Clifford alge-
bras at its most fundamental level (the electron). Furthermore, we believe
that a systematic use of the above translation would be of great benefit to
areas currently utilising Grassmann variables, both in geometrizing known
results, and, more importantly, opening up possibilities for new mathemat-
ics. Indeed, if new results cannot be generated, the above exercise would be
of very limited interest.
It is one of the possibilities for new mathematics that we wish to illustrate
in the rest of this paper. The idea has its origin in pseudoclassical mechanics,
and is illustrated with one of the simplest Grassmann Lagrangians,
L -- l(.(·.
2 "
- lE"kW'('(k
2 'J 'J , (12)

where Wi are a set of three scalar consants. This Lagrangian is supposed

to represent the "pseudoclassical mechanics of spin" [Berezin & Marinov,
1977; Freund, 1986]. Following the above procedure we translate this to

L = tejAej - W, (13)

where
(14)
which gives a bivector - valued Lagrangian. This is typical of Grassmann
Lagrangians, and can be easily extended to supersymmetric Lagrangians,
which become mixed grade multivectors. This raises a number of interesting
questions; what does it mean when a Lagrangian is multivector-valued, and
do all the usual results for scalar Lagrangians still apply? In the next section
we will provide answers to some of these, illustrating the results with the
Lagrangian of (13). In doing so we will have thrown away the origin of the
Lagrangian in Grassmann algebra, and will work entirely within the frame-
work of geometric algebra, where we hope it is evident that the possibilities
are far greater.

3. The Variational Principle for Multivector Lagrangians

Before proceeding to derive the Euler-Lagrange equations for a multi vector
Lagrangian, it is necessary to first recall the definition of the multi vector
GRASSMANN MECHANICS 219

derivative Ox, as introduced in [Hestenes, 1968; Hestenes & Sobczyk, 1984].

Let X be a mixed-grade multivector

(15)
r

and let F(X) be a general multi vector valued function of X. The A derivative
of F is defined by

(16)

where * denotes the scalar product

(17)

We now introduce an arbitrary vector basis {ej}, which is extended to a

basis for the entire algebra {eJ}, where J is a general index. The multivector
derivative is defined by

(18)

Ox thus inherits the multivector properties of its argument X, so that in

particular it contains the same grades. A simple example of a multivector
derivative is when X is just a position vector x, in which case Ox is the usual
vector derivative (sometimes referred to as the Dirac operator). A special
case is provided when the argument is a scalar, a, when we continue to write
00/.
A useful result of general applicability is

ox(XA) = Px(A) (19)

where Px(A) is the projection of A onto the terms containing the same
grades as X. More complicated results can be derived by expanding in a
basis, and repeatedly applying (19).
Now consider an initially scalar-valued function L = L(Xi,Xi) where Xj
are general multivectors, and Xi denotes differentiation with respect to time.
We wish to extremise the action

(20)

Following e.g. [Goldstein, 1950], we write,

Xj(t) = X?(t) + EYj(t) (21 )

220 CHRIS DORAN ET AL.

where Y; is a multivector containing the same grades as Xi, £ is a scalar,

and Xp represents the extremal path. With this we find

(22)

(23)

(summation convention implied), and from the usual argument about sta-
tionary paths, we can read off the Euler-Lagrange equations

(24)
We now wish to extend this argument to a multivector-valued L. In this
case taking the scalar product of L with an arbitrary constant multi vector A
produces a scalar Lagrangian (LA), which generates its own Euler-Lagrange
equations,

Ox. (LA) - Ot( ox. (LA}) = o. (25)

An 'allowed' multi vector Lagrangian is one for which the equations from each
A are mutually consistent. This has the consequence that if L is expanded
in a basis, each component is capable of simultaneous extremisation.
From (25), a necessary condition on the dynamical variables is

(26)
For an allowed multi vector Lagrangian this equation is also sufficient to
ensure that (25) is satisfied for all A. We will take this as part ofthe definition
of a multivector Lagrangian. To see how this can work, consider the bivector-
valued Lagrangian of (13). From this we can construct the scalar Lagrangian
(LB), where B is a bivector, and we can derive the equations of motion
Ge• (LB) - Gt( Ge•(LB}) o (27)
:::} (ei + £ijkWjek)·B O. (28)
For this to be satisfied for all B, we simply require that the bracket vanishes.
If instead we use (26), together with the 3-d result
(29)

we find the equations of motion

(30)

Thus, for the Lagrangian of (13), equation (26) is indeed sufficient to ensure
that (27) is satisfied for all B.
GRASSMANN MECHANICS 221

Recalling (14), equations (30) can be written compactly as [Lasenby et al.,

1992b]
.
ei = ei ·w, (31)
which are a set of three coupled vector equations - nine scalar equations
for nine unknowns. This illustrates how multi vector Lagrangians have the
potential to package up large numbers of equations into a single entity, in a
highly compact manner. Equations (31) are studied and solved in [Lasenby
et at., 1992b].
This example also illustrates a second point, which is that, for a fixed
A, (25) does not always lead to the full equations of motion. It is only by
allowing A to vary that we arrive at (26). Thus it is crucial to the formalism
that L is a multivector, and that (25) holds for all A, as we shall see in the
following section, where we consider symmetries.

4. Noether's Theorem for Multivector Lagrangians

One of the most powerful ways of analysing the equations of motion resulting
from a Lagrangian is via the symmetry properties of the Lagrangian itself.
The general tool for doing this is Noether's theorem, and it is important that
an analogue of this can be found for the case of multivector Lagrangians.
There turn out to be two types of symmetry to be considered, depending
on whether the transformation of variables is governed by a scalar or by a
multi vector parameter. We will look at these separately.
It should be noted that as all our results are expressed in the language
of geometric algebra, we are explicitly working in a coordinate-free way,
and thus all the symmetry transformations considered are active. Passive
transformations have no place in this scheme, as the introduction of an
arbitrary coordinate system is an unn~cessary distraction.

4.1. SCALAR CONTROLLED TRANSFORMATIONS

Given an allowed multivector Lagrangian of the type L = L(Xi,Xi), we

wish to consider variations of the variables Xi controlled by a single scalar
parameter, a. We thus write XI = Xi(Xi, a), and define L' = L(X[, XD, so
that L' has the same functional dependence as L. Making use of the identity
L' = {L'A)OA, we proceed as follows:
Oa L' = .
(OaXD*ox,(L' A)OA .
+ (OaXD*ox,(L' A)OA (32)
= (OaXD*(Ox:(L'A) - OtWk:(L'A))) OA + at ((OaX[)*Ox:L'). (33)
If we now assume that the equations of motion are satisfied for the XI (which
must be checked for any given case), we have
oa L' = at ((OaXi)*Ox:L') , (34)
222 CHRIS DORAN ET AL.

and if L' is independent of a, the corresponding conserved current is (o",Xi) *

0x,L'. Note how important it was in deriving this that (25) be satisfied for
all 'A. Equation (34) is valid whatever the grades of Xi and L, and in (34)
there is no need for a to be infinitesimal. If L' is not independent of a, we
can still derive useful consequences from,

(35)

As a first application of (35), consider time translation,

xI(t,a) Xi(t + a) (36)
=:} o",XI!",=o Xi, (37)
so (35) gives (assuming there is no explicit time-dependence in L)

OtL = Ot(Xi*ox,L). (38)

Hence we can define the conserved Hamiltonian by

H = Xi *Ox. L - L. (39)
Applying this to (13), we find
H = ei*oe,L - L (40)
tei Aei - L (41)
= w, (42)
so the Hamiltonian is, of course, a bivector, and conservation implies that
W = 0, which is easily checked from the equations of motion.
There are two further applications of (35) that are worth detailing here.
First, consider dilations
(43)
so (35) gives

o",L'la=o = Ot(Xi*ox,L). (44)

For the Lagrangian of (13), L' = e 2'" L, and we find that

2L = Ot(teit\ei) (45)
= 0, (46)
so when the equations of motion are satisfied, the Lagrangian vanishes. This
is quite typical of first order Lagrangians.
Second, consider rotations
(47)
GRASSMANN MECHANICS 223

where B is an arbitrary constant bivector specifying the plane(s) in which

the rotation takes place. Equation (35) now gives

(48)

where B X Xi is one half the commutator [B, Xi]. Applying this to (13), we
find
(49)
However, since L = 0 when the equations of motion are satisfied, we see that
ei /I. (B 'ei) (50)
must be constant for all B. In [Lasenby et at., 1992b] it is shown that this
is equivalent to conservation of the metric tensor g, defined by
(51)

4.2. MULTIVECTOR CONTROLLED TRANSFORMATIONS

The most general transformation we can write down for the variables Xi
governed by a single multi vector M is
Xi = f(Xi,M), (52)
where f and M are time-independent functions and multivectors respec-
tively. In general f need not be grade preserving, which opens up a route to
considering analogues of supersymmetric transformations.
In order to write down the equivalent equation to (34), it is useful to
introduce the differential notation of [Hestenes & Sobczyk, 1984],

(53)
We can now proceed in a similar manner to the preceding section, and derive,

A*aML' LA(Xi,M)*ax;L' + LA(Xi,M)*ax,L' . (54)

LA(Xi,M)*(ax:(L'B) - at(ax:(L'B)) aB
+ at (LA (Xi , M)*ax;L') (55)

= at (LA(Xi,M)*ax;L') , (56)

where again we have assumed that the equations of motion are satisfied for
the transformed variables. We can remove the A dependence from this by
differentiating, to yield

(57)
224 CHRIS DORAN ET AL.

and if L' is independent of M, the corresponding conserved quantity is

OALA(X;, M)*[fj('L'
, = 8M I(X;, £1) *OJ['L',
, (58)

where the hat on £1 denotes that this is the M acted on by OM. Which
form of (58) is appropriate to any given problem will depend on the context.
Nothing much is gained by setting M = 0 in (57), as usually multivector
controlled transformations are not simply connected to the identity.
In order to illustrate (57), consider reflection symmetry applied to the
Lagrangian of (13), that is

f(ej, n) (59)
=? L' (60)

Since L = 0 when the equations of motion are satisfied, the left hand side
of (57) vanishes, and we find that

(61 )

is conserved. Now

f (
ej,)
n = -aein -1 + nejn - 1an- 1 , (62)
~

so (61) becomes
1 !:I (
'2ua -e;2 an -1 + nein -1 aein -1) 2 -ej2 n -1 - e;·n -1 nein -1 (63)
-n(e;n-l + ej·n- 1e i)n- 1. (64)

This is basically the same as was found for rotations, and again the conserved
quantity is the metric tensor g. This is no surprise since rotations can be built
out of reflections, so it is natural to expect the same conserved quantities
for both.
Equation (57) is equally valid for scalar Lagrangians, and for the case of
reflections will again lead to conserved quantities which are those that are
usually associated with rotations. For example considering

(65)

it is not hard to show from (57) that the angular momentum x A :i; is con-
served. This shows that many standard treatments of Lagrangian symme-
tries [Goldstein, 1950] are unnecessarily restrictive in only considering in-
finitesimal transformations. The subject is richer than this suggests, but
without the powerful multivector calculus the necessary formulae are sim-
ply not available.
GRASSMANN MECHANICS 225

5. Conclusions
Grassmann calculus finds a natural setting within geometric algebra, where
the additional mathematical structure allows for a number of generalisa-
tions. This is illustrated by Grassmann (pseudoclassical) mechanics, which
opens up a new field - that of the multi vector Lagrangian. In order to carry
out such generalisations, it is necessary to have available the most powerful
techniques of geometric algebra. For Lagrangian mechanics it turns out that
the multivector derivative fulfills this role, allowing for tremendous compact-
ness and clarity. Elsewhere [Lasenby et ai., 1993] the multivector derivative
is developed and presented as the natural tool for the study of Lagrangian
field theory.
It is our opinion that the translation of Berezin calculus into geometric
algebra will be of great benefit in other fields where Grassmann variables
are routinely employed. A start on this has been made in [Lasenby et at.,
1992b; Lasenby et ai., 1992a], but clearly the potential subject matter is
vast, and much work remains.

Acknowledegment
Chris Doran would like to acknowledge his gratitude to Sidney Sussex Col-
lege for financial support in attending this conference.

References
Barut, A.O., & Zanghi, N. 1984. Classical Models of the Dirac Electron. Phys. Rev. Lett.,
52(23), 2009.
Berezin, F.A. 1966. The Method of Second Quantisation. Academic Press.
Berezin, F.A., & Marinov, M.S. 1977. Particle Spin Dynamics as the Grassmann Variant
of Classical Mechanics. Annals of Physics, 104, 336.
de Witt, B. 1984. Supermanifolds. Cambridge University Press.
Doran, C.J.L., Lasenby, A.N., & Gull, S.F. 1993. States and Operators in the Spacetime
Algebra. To appear in: Foundations of Physics.
Freund, P.G.O. 1986. Supersymmetry. Cambridge University Press.
Goldstein, H. 1950. Classical Mechanics. Addison Wesley.
Grassmann, H. 1877. Der Ort der Hamiiton'schen Quaternionen in der Ausdehnungslehre.
Math. Ann., 12, 375.
Gull, S.F. 1990. Charged Particles at Potential Steps. In: Weingartshofer, A., & Hestenes,
D. (eds), The Electron 1990. Kluwer.
Hestenes, D. 1968. Muitivector Calculus. J. Math. Anal. Appl., 24, 313.
Hestenes, D. 1975. Observables, Operators, and Complex Numbers in the Dirac Theory.
J. Math. Phys., 16(3), 556.
Hestenes, D., & Sobczyk, G. 1984. Clifford Algebra to Geometric Calculus. D. Reidel
Publishing.
Lasenby, A.N., Doran, C.J.L., & Gull, S.F. 1992a. 2-Spinors, Twistors and Supersymmetry
in the Spacetime Algebra. These Proceedings.
Lasenby, A.N., Doran, C.J.L., & Gull, S.F. 1992b. Grassmann Calculus, Pseudoclassical
Mechanics and Geometric Algebra. Submitted to: J. Math. Phys.
Lasenby, A.N., Doran, C.J.L., & Gull, S.F. 1992c. Twistors and Supersymmetry in the
Spacetime Algebra. In Preparation.
226 CHRIS DORAN ET AL.

Lasenby, A.N., Doran, C.J.L., & Gull, S.F. 1993. A Multivector Derivative Approach to
Lagrangian Field Theory. To appear in: Foundations of Physics.
INTRINSIC NON-INVARIANT FORMS OF
DIRAC EQUATION

JOSEP MANEL PARRA'

Laboratori de Fiaica Matematica, I.E.C.
Departament de Fisica Fonamental
Universitat de Barcelona
Diagonal 6.47, £-08028 Barcelona SPAIN

Abstract. Several results that seem to arise quite naturally from Hestenes geometric
formulation of Dirac's equation, and that conflict with the standard view on the relativistic
invariance of it, are openly discussed. The result is a better understanding of all quantum
theory. On one hand the mathematics of relativistic quantum mechanics is made fully
compatible with classical physical theories. On the other hand, the geometrical content
of these mathematical operations, involving in an intrinsic manner the observer's frame,
elucidates some of the most fundamental problems and profound mathematical results of
quantum mechanics.

Key words: Clifford - Dirac - Hestenes - equation - spinor - relativistic - in variance -

observer - intrinsic

1. The frame-dependent intrinsic Dirac-Hestenes equation

The problem we want to address here is the study of the relativistic inva-
riance of the Dirac-Hestenes equation: liDX el2 + ~ AX + mcX eo = O. This
is a crucial point in Hestenes' theory in spite of being dismissed in Hestenes
(1990a, p. 1221) saying that:
"Equation (33) is Lorentz invariant, despite the explicit appearance
of the constants /0 and i = /2/1 in it. These constants are arbitrarily
specified by writing (33). They need not be identified with the vectors of
a particular coordinate system, though it is often convenient to do so."
We hold that, as seems almost obvious, the explicit appearance of the
constant multivectors eo == /0 and e12 == /12 factors makes the equation
non-Lorentz invariant and that these factors must be interpreted as
belonging to the particular inertial reference frame in which the equation is
written.
In spite of this disagreement, which motivates the full analysis that
follows, we stress from the very beginning that the final results of our analysis

• This work has received financial support from the D.G.C y T. under contract No.
PB90-0482-C02-01

227
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Defonnations, 227-232.
© 1993 Kluwer Academic Publishers.
228 J.M. PARRA

of the problem give additional support to Hestenes' geometric interpretation

of Dirac's theory.
We start with Dirac-Hestenes' equation written in the inertial frame E
of a Minkowski's space of metric diag( -g, g, g, 9 )withg = ±1 (Parra, 1992a)

Ti( -geoaxo + g( el axl + e2ax2 + e3ax3)X el2 +

~(¢eo + AIel + A 2e2 + A3e3)X + mcXeo = 0 (1)
c
where X = 0: + g(EleOl + E2e02 + E3e03 + Ble23 + B2e31 + B3e12) + Ae0123
is the geometric Dirac field in that frame.
Now we consider another inertial frame E' obtained from E by means of
a constant Lorentz transformation L. The geometric algebra formulation of
the (passive) transformation law of the components of any geometric object
Z is

z = (za ea ) = za(L b'e'b)e'a(LC'e'c)-l =

(L 6' e' b)( zae ,a)( Le' e'c)-l = (Zb' e' b) (2)
where the subindexes a, b, c cover the sixteen linear dimensions of the exterior
algebra and L is the Lorentz transformation operator given by

L = (Lae'a) = f3 + f{l eo1 + f{2 e02 + f{3eo3 +

(3)
with f3(J = K . Rand f32 - K . K + R· R - (J2 = 1, in such a way that only
six constant parameters are needed to specify it. Scalar 1x1, vector 4x4, and
bivector 6x6 matrix representations of the Lorentz group are replaced by
the single operational law (2).
The transformation of equation (1) in the E' frame (in fact, a mere re-
writing, because no geometric object is transformed) gives:

+~(AIL'
c
e' IL)( X a' e' a) + mc( X a' e' a)( Lbe' b)e'O( LCe' C )-1 = 0, (4)

which retains, expressed in E' accordingly to eq. (2), the constant frame
fields e12 and eo of E. This means that it is not a form-invariant equation.
This result is not a defect of Hestenes' theory, but a characteristic of Dirac's
equation itself. This fact was indeed clearly appreciated by Darwin (1928, p.
657) " ... here we have a system invariant in fact but not in form". Hestenes'
equation being a geometric intrinsic formulation of Dirac's equation, its lack
of form-invariance implies its non-invariance in fact. All terms and opera-
tors in it possess well-defined transformation properties that follow from its
geometric nature, and no miraculous remedy is possible.
INTRINSIC NON INVARIANT FORMS OF DIRAC EQUATION 229

What happens, then, with the well-established validity of the original

Dirac equation in all inertial frames? Does Hestenes' formulation deny this
fact? These apparently strong objections are easily resolved in the geometric
formulation. We need only multiply eq. (4) at right by the constant Lorentz
factor (La e,a) to obtain the Dirac equation that in ~' describes the same
physics that is described by (1) in ~. We obtain:

n(-ge'oox1o +g(e'toxll +e'2oxI2 +e'30xI3)(Lae'a)(Xbe'b)e'!2

+ ~(AI-" e' 1')(L ae' a)(Xbe' b) + mc(Lae' a)(Xbe'b)e' 0 = 0 (5)
c
Equation (5) is effectively the Dirac equation in ~' we were looking for. In
this equation the geometric element that plays the role of the Dirac geometric
field is not X

(6)

but a different intrinsic geometric object Y

(7)

This last equation (7) also explains the reasons for our apparently cumber-
some notation. When intrinsic geometric language does not suffice to clarify
the points at issue, as is the case here, then the safest way is to write the
objects in their full operational form including the specific set of basis op-
erators e a used. Writing the essential formula (7) in the "compact" form
y = y' = LX = X'L will fit at the same time the schemes of both Bour-
bakists and physicists infected by the "coordinate virus MV Ie", described
in Hestenes, 1992. But it will be misunderstood by both.
In this process of obtaining Hestenes' equation in ~', the Lorentz transfor-
mation that relates ~ and ~' performs two different kinds of transformation.
From (1) to (4) it does nothing to the geometrical entities but changes their
components (a passive transformation). From (4) to (5) it seems to do hardly
anything on the algebra of the equations, but in fact it "actively" transforms
the differential form in which the Dirac field consists. Said otherwise, while
(1) and (4) are different frame representations of the same intrinsic equa-
tion, (4) and (5) are different intrinsic equations expressed in the same frame:
Equation (5) has an extra geometric factor L at right.
We have then a nice explanation of the apparent paradox that puzzled
Darwin, and also those that approach Hestenes' theory coming from the field
of differential geometry: Dirac's equation in ~ is not Dirac's equation in ~'.
They imply each other, so there is no problem about their experimental
validity or the actual working of the original matrix form in any inertial
frame. But the Dirac field is, as an intrinsic geometrical object, different in
each inertial frame. It is X in ~ and Y in ~'. There is an intrinsic Dirac
230 J.M. PARRA

equation attached to each inertial reference system. Thus, precisely because

it is intrinsically different in each system, it is meaningless to speak about
invariance. Imposing it may produce meaningless results.
We have seen in Dirac's equation the emergence of the inertial reference
system as an intrinsic ingredient ofthe theory. Moreover, put into mathema-
tically compelling terms, this is Bohr's answer to all interpretational prob-
lems of quantum mechanics: until the full experimental arrangement has
been set out and properly taken into account, one cannot ask meaningful
questions in quantum mechanics. Hestenes theory, a transcription of Dirac's
theory in terms of sound, powerful, and well-understood mathematics, shows
precisely of what this experimental arrangement consists. It consists of the
orthogonal tetrad attached to the inertial observer who writes the equa-
tion and performs the experiment. This orthonormal tetrad enters into the
quantum mechanical description as an essential part of the wave function.
This also answers the longstanding and misleading philosophical debate
about the involvement of the observer and his/her consciousness in the quan-
tum mechanical description of nature: all that matters and finds its place
into the equations is the orthonormal tetrad, that is the observer's time and
oriented length standards that will be used in the experiment, and from
which the standards and space-time splitting of all other physical magni-
tudes are derived.
All the preceeding considerations, based on mathematical expressions
that follow naturally from equation (1), using only the geometric algebra,
are not only in full agreement with Hestenes geometric interpretation of the
Dirac field. They provide its mathematical foundation. Full understanding of
the relationship between the two Dirac equations, of eq. (1) for X in ~ and
eq.(5) for Y in ~', is achieved considering how the wave function generates
field observables through the same kind of bilinear transformation formula
(2) that led from (1) to (5):
(8)
where X t means the reversed or trasposed element of X, which differs from
X only in the sign of its bivector component. For the Lorentz transforma-
tion (3) the reversion is equivalent to the inversion, so (2) and (8) are really
the same. Variation of the index a in (8) produces essentially all the six-
teen bilinear covariants of Dirac theory, from which all physical observables
are constructed. The geometric field X, whose components are in one- to-
one correspondence with the eight real components or degrees of freedom of
the complex Dirac spinor, acts then as a field of Lorentz transformations.
A space-time dependent field that connects a space-time dependent tetrad,
attached to the electron in some way to be further elucidated (Hestenes,
1990a,1990b), and the constant tetrad attached to the inertial frame of ob-
servation.
INTRINSIC NON INVARIANT FORMS OF DIRAC EQUATION 231

Leaving aside the problems related to the two remaining degrees of freee-
dom of the Dirac's wave function (Hestenes, 1990a, p. 1220), and limiting
our analysis to the well-established kinematical factor or Lorentz rotation,
we can describe the physical content of Hestenes theory in a more formal
geometrical way.
To each observer there corresponds, for the same electron, a different field
of Lorentz transformations, a different section of the Lorentz bundle over the
Minkowski space-time. This should not be confused with a section of the
frame bundle, that would give the wave function a status independent of
the observer, a fact incompatible with the analysis performed above. Dirac's
wave function is the operator that globally relates the constant (flat) frame
section of the observer to the space-time dependent (twisted) frame section
of the electron. The (passive) change of inertial observers, implied in any
consideration of relativistic invariance, is precisely a change of the flat section
of the frame bundle in which the electron's proper frame is to be expressed
according to (8). The left translation of the Dirac field by a space-time
constant Lorentz transformation relating two "flat" inertial frames, makes
it possible for the second observer to find his own relationship with the same
electron's proper frame from the data used by the first.
The physical content of \Vigner's representation theory is then fully un-
derstood, and can no longer stand as a paradigmatic example of an "un-
reasonable effectiveness of mathematics" . The left action of the full Lorentz
group upon the wave function of an electron described in one inertial frame
gives, as its "orbit", the set of all inertially-equivalent descriptions ofthe elec-
tron. And this set, considered as a whole, obviously constitutes an observer-
free description of an electron state.

2. Spinor theory versus geometric algebra

It is the (active) left-translation expressed in (7) that has been mistakenly

conceived and "understood" as the spin-1/2 (passive) transformation for-
mula. The origin of this confusion was the unjustified belief that Dirac's
equation is an invariant equation. The analysis we have performed shows
that it is "intrinsically" not invariant. Further, there is no need to postulate
such invariance for obtaining full agreement not only with the experimen-
tal evidence of its validity in all inertial frames but also with some of the
most fundamental points of Bohr's interpretation of quantum mechanics.
Plain acceptance of this fact renders the introduction of spinor fields truly
meaningless: they were defined in terms of a (passive) transformation law
that assured the invariant character of Dirac's equation. In the Hestenes
formulation of Dirac theory the spinor fields are simply absent.
It is now possible to express clearly the fact that the even multivectors
called spinors by Hestenes, of which the X in the Dirac equation is an
232 J.M. PARRA

example, have nothing to do with the "mathematical" spinors. Their com-

ponents behave, under any passive transformation, as those belonging to an
aggregate of differential forms of different degree (Graf, 1978), not as any
spin-l/2 representation. Hestenes' spinors are not a space of representation
of constant (passive) Lorentz transformations expressed in the matrix form
of their covering group. Hestenes' spinors are, essentially, space-time depen-
dent Lorentz transformation operators expressed in a direct intrinsic (and
tensorial) geometrical form, and it is obviously misleading to persist in call-
ing them spinors. We end by an attempt to clarify some misunderstandings
that this misuse of the word "spinor" may have produced (see also Parra,
1992b ).
Any claimed physical equivalence between the passive transformation
law (6) and the active transformation law (7) can no longer be supported.
They are not alternative views that can be held in a coherent geometric
interpretation of Dirac theory. Both, applied in succession, were required
to establish the agreement of the theory with the special relativity require-
ments. Geometrical understanding of the theory is greatly improved when
each is conceived as it is. That the physical system of observation is intrin-
sically present in the wave function is not a result that can be dismissed as
a matter of convention. Nor is it a matter of opinion, requiring no closer
examination, that there may be no "spinor matter" and that the physical
application of the extensive mathematical research on spinors and twistors
may be 'flawed' from the very beginning. A sufficient reason for these almost
unbelievable developments may well be that in 1928, when physicists had
to tackle with the problem of the relativistic invariance of Dirac's equation,
geometric Clifford algebra was, still, unbelievably absent!

References
Darwin, C. G.: 1928, 'The wave equations of the electron' Proc. Royal Soc. London 118,
654-680.
Graf, W.: 1978, 'Differential forms as spinors', Ann. Inst. Henri Poincare Sec. A 29,
85-109.
Hestenes, D.: 1990 a, 'The Zitterbewegung Interpretation of Quantum Mechanics' Found.
Phys. 20, 1213-1232.
Hestenes, D.: 1990 b, 'Real Dirac Theory', Tempe, Arizona.(Draft, 89 pp. private commu-
nication).
Hestenes, D.: 1992, in A. Micali, R. Boudet and J. Helmstetter, ed(s)., Proceedings of
the Second International Conference on Clifford Algebras and Their Applications to
Physics, 3-16, Mathematical viruses, I<luwer:Dordrecht
Parra, J. M.: 1992 a, in A. Micali, R. Boudet and J. Helmstetter, ed(s)., Proceedings of
the Second International Conference on Clifford Algebras and Their Applications to
Physics, 463-467, On Dirac and Dirac-Darwin-Hestenes equations, I<luwer:Dordrecht
Parra J. M.: 1992 b, 'The Dirac-Hestenes equation and the algebraic structure of the
Minkowski space-time', XIX ICGTMP, Salamanca (Spain) July-1992, to appear.
2-SPINORS, TWISTORS AND
SUPERSYMMETRY IN THE SPACETIME
ALGEBRAS
ANTHONY LASENBY
MRAO, Cavendish Laboratories, Madingley Road, Cambridge CBS OHE, UK
CHRIS DORAN'
DAMTP, Silver Street, Cambridge, CBS 9EW, UK
and
STEVE GULL
MRAO, Cavendish Laboratories, Madingley Road, Cambridge CBS OHE, UK

Abstract. We present a new treament of 2-spinors and twistors, using the spacetime
algebra. The key role of bilinear covariants is emphasized. As a by-product, an explicit
representation is found, composed entirely of real spacetime vectors, for the Grassmann
entities of supersymmetric field theory.

1. Introduction
The aim of this presentation is to give a new translation of 2-spinors and
twistors into the language of Clifford algebra. This has certainly been con-
sidered before [Ablamowicz et al. , 1982; Ablamowicz & Salingaros , 1985],
but we differ from previous approaches by using the language of a particular
form of Clifford algebra, the spacetime algebra (henceforth STA), in which
the stress is on working in real 4-dimensional spacetime, with no use of a
commutative scalar imaginary i. Moreover, the quantities which are Clifford
multiplied together are always taken to be real geometric entities (vectors,
bivectors, etc.), living in spacetime, rather than complex entities living in
an abstract or internal space. Thus the real space geometry involved in any
equation is always directly evident.
That such a translation can be achieved may seem surprising. It is gen-
erally believed that complex space notions and a unit imaginary i are fun-
damental in areas such as quantum mechanics, complex spin space, and
2-spinor and twistor theory. However using the spacetime algebra, it has
already shown [Hestenes , 1975] how the i appearing in the Dirac, Pauli and
Schrodinger equations has a geometrical explanation in terms of rotations in
• Supported by a SERe studentship.

233
z. Oziewicz et al. (eds.),
Spinors, Twistors, Clifford Algebras and Quantum Deformations, 233-245.
© 1993 Kluwer Academic Publi~hers.
234 ANTHONY LASENBY ET AL.

real spacetime. Here we extend this approach to 2-spinors and twistors, and
thereby achieve a reworking that we believe is mathematically the simplest
yet found, and which lays bare very clearly the real (rather than complex)
geometry involved.
As another motivation for what follows, we should point out that the
scheme we present has great computational power, both for hand work-
ing, and on computers. Every time two entities are written side by side
algebraically a Clifford product is implied, thus all our expressions can be
programmed into a computer in a completely definite and explicit fashion.
There is no need either for an abstract spin space, containing objects which
have to be operated on by operators, or for an abstract index convention.
The requirement for an explicit matrix representation is also avoided, and
all equations are automatically Lorentz invariant since they are written in
terms of geometric objects.
Due to the restriction on space, we will only consider the most basic
levels of 2-spinor and twistor theory. There are many more results in our
translation programme for 2-spinors and twistors that have already been
obtained, in particular for higher valence twistors, the conformal group on
spacetime, twist or geometry and curved space differentiation, and these will
be presented with proper technical details in a forthcoming paper [Lasenby
et al. , 1992c]. However, by spending some time being precise about the
nature of our translation, we hope that even the basic level results presented
here will still be of use and interest. A short introduction is also given of the
equivalent process for field supersymmetry, and we end by discussing some
implications for the role of 2-spinors and twistors in physics.

2. The Spacetime Algebra

The spacetime algebra is the geometric (Clifford) algebra of real 4-dimensio-
nal spacetime. Geometric algebra and the geometric product are described
in detail in [Hestenes & Sobczyk, 1984]. Our own conventions follow those
of this reference, and are also described in [Lasenby et al. , 1992a]. Briefly
we define a multivector as a sum of Clifford objects of arbitrary grade (grade
o = scalar, grade 1 = vector, grade 2 = bivector, etc.). These are equipped
with an associative (geometric) product. We will also need the operation of
reversion which reverses the order of multivectors,
(ABf= BA, (1)
but leaves vectors (and scalars) unchanged, so it simply reverses the order
of the vectors in any product.
The Clifford algebra for 3-dimensional Euclidean space is generatated by
three orthonormal vectors {ad, and is spanned by
1, (2)
2-SPINORS, TWISTORS AND SUPERSYMMETRY IN THE SPACETIME ALGEBRAS 235

where i = 0"10"20"3 is the pseudoscalar (highest grade multi vector ) for the
space. The pseudoscalar i squares to -1, and commutes with all elements
of the algebra in this 3-dimensional case, so is given the same symbol as the
unit imaginary. Note, however, that it has a definite geometrical role as on
oriented volume element, rather than just being an imaginary scalar. For fu-
ture clarity, we will reserve the symbol j for the uninterpreted commutative
imaginary i, as used for example in conventional quantum mechanics and
electrical engineering. The algebra (2) is the Pauli algebra, but in geometric
algebra the three Pauli O"k are no longer viewed as three matrix-valued com-
ponents of a single isospace vector, but as three independent basis vectors
for real space.
A quantum spin state contains a pair of complex numbers, 'l/JI and 'l/J2

(3)

and has a one to one correspondence with an even multivector 'I/J. A general
even element can be written as 'I/J = aD + akiO"k, where aD and the a k are
scalars (summation convention assumed), and the correspondence works via
the basic identification

I'I/J) =( aD + 3
+ j.a 1 ) +-+ 'I/J = aO+ka .ZO"k· (4)
-a 2 Ja
We will call 'I/J a spinor, as one of its key properties is that it has a single-sided
transformation law under rotations (section 3).
To show that this identification works, we also need the translation of the
angular momentum operators on spin space. We will denote these operators
ak, where as usual
A

O"x = (01 1) '

0
A

O"Y = (0j -j)

0 ' O"z
A

= ( 10 -10) . (5)

The translation scheme is then

(k=1,2,3). (6)
Verifying that this works is a matter of computation, e.g.

(-:~ ! ;:: ) +-+ _a 2 + a 3iO"I - a Di0"2 + a 1i0"3 = 0"1 (aD + akiO"k) 0"3, (7)
demonstrates the correspondence for ax. Finally we need the translation for
the action of j upon a state I'I/J). This can be seen to be
(8)
236 ANTHONY LASENBY ET AL.

We note this operation acts solely to the right of 'Ij;. The significance of this
will be discussed later.
An implicit notational convention should be apparent above. Conven-
tional quantum states will always appear as bras or kets, while their STA
equivalents will be written using the same letter but without the brackets.
Operators (e.g. upon spin space) will be denoted by carets. We do not at
this stage need a special notation for operators in STA, because the role of
operators is taken over by right or left multiplication by elements from the
same Clifford algebra as the spinors themselves are taken from. This is the
first example of a conceptual unification afforded by STA - 'spin space' and
'operators upon spin space' become united, with both being just multivec-
tors in real space. Similarly the unit imaginary j is disposed of to become
another element of the same kind, which in the next section we show has a
clear geometrical meaning.
In order to extend these results to 4-dimensional spacetime, we need the
full 16-component STA, which is generated by four vectors 111" This has basis
elements 1 (scalar), II-' (vectors), iO"k and O"k (bivectors), hI-' (pseudovectors)
and i (pseudoscalar) (/-£ = 0, ... ,3; k = 1,2,3). The even elements of this
space, 1, O"k, iO"k and i, coincide with the full Pauli algebra. Thus vectors in
the Pauli algebra become bivectors as viewed from the Dirac algebra. The
precise definitions are
O"k == IklO and i == 10/1/2/3 = 0"10"20"3. (9)
Note that though these algebras share the same pseudoscalar i, this anti-
commutes with the spacetime vectors 1/-1" Note also that reversion in this
algebra (also denoted by a tilde - R), reverses the sign of all bivectors, so
does not coincide with Pauli reversion. In matrix terms this is the differ-
ence between the Hermitian and Dirac adjoints. It should be clear from the
context which is implied.
A 4-component Dirac column spinor 1'Ij;) is put into a one to one corre-
spondence with an even element of the Dirac algebra 'Ij; [Gull, 1990] via

1'Ij;) = ( -:~ !+;:~ )

_b3 jbO +-+ 'Ij; = a
0
+ ak·to"k + t(b
. 0
+ bk·.ZO"k). (10)
_b 1 _ jb 2
The resulting translation for the action of the operators 71' is
7Jl.I'Ij;}+-+II-''Ij;,0 (/-£=0, ... ,3), (11)
which follows if the 7 matrices are defined in the standard Dirac-Pauli rep-
resentation [Bjorken & Drell , 1964]. Verification is again a matter of compu-
tation, and further details will be given in [Doran et al. , 1993]. The action
of j is the same as in the Pauli case,
(12)
2-SPINORS, TWISTORS AND SUPERSYMMETRY IN THE SPACETIME ALGEBRAS 237

3. Rotations and Bilinear Covariants

In STA, the vectors <7k are simply the basis vectors for 3-dimensional space,
which means that the translation (6) for the action of the ak can be recast
in a particularly suggestive form. Let n be a unit vector, then the eigenvalue
equation for the measurement of spin in a direction n is conventionally

(13)

where in this scheme S is a 'vector', with 'components' Sk = (n/2)ak. Now

n'S = ~nkak' so the STA translation for this equation is just
(14)

where n is a (true) vector in ordinary 3-dimensional space. Multiplying on

the right by <73'¢ (,(j; = aO - akiak), yields

(15)

Now 'IjJ,(j; is a scalar in the Pauli case

1'ljJ12 'IjJ,(j; = ,(j;'IjJ (16)

= (aO)2+(a 1 )2+(a 2 ?+(a 3 )2, (17)
so we can write

(18)

This shows that the wavefunction 'IjJ is in fact an instruction on how to rotate
the fixed reference direction 0'3 and align it parallel or anti-parallel with the
desired direction n. The amplitude just gives a change of scale. This idea, of
taking a fixed or 'fiducial' direction, and transforming it to give the particle
spin axis, is a central one for the development of our physical interpretation
of quantum mechanics.
In the relativistic case, 'IjJ,(j; is not necessarily a pure scalar, and we have
'IjJ,(j; = '¢'IjJ = pe if3 . The relativistic wavefunction 'IjJ now specifies a spin axis
s via s = p-l'IjJ'Y3'¢, and a complete set of body axes elJ. via

(19)
eo = v is interpreted as the particle 4-velocity, while pv is the standard Dirac
probabilty current - see [Doran et at. , 1993] for further details. The main
change in viewpoint on going to the STA should now be apparent - instead
of the discrete and discontinous language of operators, eigenstates and eigen-
values we now have the idea of continuous families of transformations. This
238 ANTHONY LASENBY ET AL.

enables us to give a realistic physical description of particle tracks and spin

directions in interaction with external apparatus [Lasenby et al. , 1992b].
One of the great advantages of geometric algebra is the way that rotation
of a general multi vector is achieved in exactly the same fashion as for a
single vector. Thus to discuss Lorentz rotations for example, let us write
1/J = pl/2 ei{3/2 R. Then R is an even multivector satisfying RR = RR = 1
and therefore corresponds to a Lorentz rotation (combination of pure boost
and spatial rotation). To rotate an arbitrary multivector M we just form
the analogue of (19) and write
M' = RMR. (20)
This is a very quick way of obtaining the transformation formulae for electric
and magnetic fields for example. If we use the whole wavefunction, which
incorporates information about the particle density, p, and also the f3 factor,
and use it to rotate a given fixed Clifford entity such as the /0 and /3 con-
sidered above, then we get a physical density for some quantity. For exam-
ple, the spin angular momentum density for a Dirac particle is the bivector
~n1/Jia3-0. (Note the combination 1/J ... -0 preserves grade for objects of grade
1,2 and 3.) Such expressions can generally be written equivalently as bilinear
covariants in conventional Dirac theory notation - for example, pv = 1/J/0-0,
the Dirac current, would be written conventionally as jl-' = (1;&"Ii'I-'I1/J) - but
in the STA version the meaning of the expression is usually much clearer.
We mention this point, since it will transpire that many of the quantities
of importance for 2-spinors and twistors turn out to be bilinear covariants
of the above kind, which could therefore in principle also be translated into
the Dirac notation, but again, look more straightforward in our version.
As a final comment, we should discuss the way in which specific Clifford
elements such as /0 and ia3 enter expressions such as pv = 1/J/0-0, and why
general Lorentz covariance is not compromised by this. What is happening
[Lasenby et al. , 1992c] is that the wavefunction 1/J is an instruction to rotate
from some fixed set of multi vectors to the configuration required (by the
Dirac equation for example) at some given spacetime point. If we desire the
final configurations (at all positions) to be rotated an extra amount R, then
we must use a new wavefunction 1/J' = R1/J. This of course explains the usual
spinor transformation law under a global rotation of space, but also shows
us why we do not want to rotate the elements we started from as well. Thus
general covariance and invariance under global Lorentz rotations is assured
if all quantities appearing to the left of the wavefunction make no mention
of specific axes, directions etc., while those to the right are allowed to do so,
but must remain fixed under such a rotation.
As a complementary exercise, one might decide to rotate the elements
(such as /0, ia3, etc.) we start from, by R say, leaving the final configuration
fixed. In this case we have 1/J' = 1/JR. This is what happens under a change of
2-SPINORS, TWISTORS AND SUPER SYMMETRY IN THE SPACETIME ALGEBRAS 239

'phase' for example, where I'¢) 1-7+ e j9 1'¢). Here the STA equivalent undergoes
'ljJ 1-+ ,¢e 9iq3 , which thus corresponds to a rotation of starting orientation
through 2() radians about the fiducial 0'3 direction. The action of j itself is
thus a rotation through 7r about the 0'3 axis. Note particularly that only
one copy of real sp'tcetime is necessary to represent what is going on in this
process.

4. 2-spinors
Having been explicit about our translation of quantum Dirac and Pauli
spinors, we are now in a position to begin the translation of 2-spinor the-
ory. For the latter we adopt the notation and conventions of the standard
exposition, [Penrose & Rindler , 1984; Penrose & Rindler , 1986].
The basic translation is as follows. In 2-spinor theory, a spinor can be
written either as an abstract index entity ",A, or as a complex spin vector
in spin-space (just like a quantum Pauli spinor) ts,.. We put a 2-spinor ",A in
1-1 correspondence with a Clifford spinor '" via
(21 )
where", is the Clifford Pauli spinor in one to one correspondence with the
column spinor ts,. (via 4). The function of the 'fiducial projector' (1 + 0'3)
(actually half this must be taken to get a projection operator) relates to
what happens under a 'spin transformation' represented by an arbitrary
complex spin matrix R. The new spin vector is RK and has only 4 real
degrees of freedom, whereas an arbitrary Lorentz rotation specified by a
Clifford R applied to a Clifford", gives the quantity R"" which contains 8
degrees of freedom. However, applying R to ",(1 + 0'3) limits the degrees of
freedom back to 4 again, in conformity with what happens in the 2-spinor
formulation.
The complex conjugate spinor K A ' belongs to the opposite ideal under
the action of the projector (1 + 0'3),

K A' ...... - . (1
"'to'2 - 0'3 ) • (22)
This explains why ",A and its complex conjugate have to be treated as be-
longing to different 'modules' in the Penrose and Rindler theory. Note that
in more conventional quantum notation our projectors (1 ± 0'3) would corre-
spond to the chirality operators (1 ±ji5), or in the notation of the appendix
of [Penrose & Rindler , 1986], to (multiples of) II and fl. We do not use
these alternative notations since it is a vital part of what we are doing that
the projection operators should be constructed from ordinary spacetime en-
tities.
The most important quantities associated with a single 2-spinor ",A are
its flagpole f{a = ",AKA', and the flagplane determined by the bivector pab =
240 ANTHONY LASENBY ET AL.

+ fABK;A' K;B'. Here we use the Penrose notation in which a is a

",A ",B fN B'
'lumped index' representing the spinor indices AA' etc. Now in order to get
a precise translation for quantities like ",AKA' , or ",A",B fA' B', it is necessary
to develop 'multiparticle STA' [Lasenby et al. , 1992c]. This still involves
real spacetime, but with a separate copy for each particle. We have carried
this out and thereby found the STA equivalents of 2-spinor outer product
expressions. However, we have also discovered a mapping from the spin-t
space of a single spinor to the spin-l space of general complex world vectors
(as Penrose & Rindler call them), which applied in reverse enables us to
find 'spin-t' (i.e. just one copy of spacetime) equivalents for the lumped
index expressions. It is these equivalents we give now, and proper proofs are
contained in [Lasenby et aZ. , 1992c].
Firstly, if we write 'Ij; = "'(1 + (13)' the flagpole of the 2-spinor ",A is just
(up to a factor 2) the Dirac current associated with the wavefunction 'Ij;,

(23)

We see that the projector (1 + (13) has produced a massless (null) current.
Secondly, the flagplane bivector is a rotated version of the fiducial bivector

(24)

Since /11 anti commutes with i/13, while 10 commutes, P responds at double
rate to phase rotations '" f-+ ",e ia3 () , whilst the flagpole is unaffected. A
convenient spacelike vector L, perpendicular to the flagpole and satisfying
P = LAJ(, is L = (",K,)-1!2"'11K" that is, just the 'body' I-direction.
In 2-spinor theory, a 'spin-frame' is usually written oA, [.A, but for no-
tational reasons, and to draw out the parallel with twistors, we prefer to
write these as wA , 7r A . In our translation, a spin-frame w A , 7r A is packaged
together to form a Clifford Dirac spinor 1> via
(25)

Now
say. (26)

If one now calculates the 2-spinor inner product for the same spin-frame one
finds
(27)

Thus the complex 2-spinor inner product is in fact a disguised version of the
quantity 1>1>. The 'disguise' consists of representing something that is in fact
a pseudoscalar (the i in >.+if,L) as an uninterpreted scalar j. The condition for
2-SPINORS, TWISTORS AND SUPERSYMMETRY IN THE SPACETIME ALGEBRAS 241

a spin frame to be normalized, WA1I"A = 1, is in our approach the condition

for ¢> to be a Lorentz transformation, that is ¢>¢ = 1 (except for a change of
sign which in twistor terms corresponds to negative helicity). We can thus
say "a normalized spin frame is equivalent to a Lorentz transformation".
The orthonormal real tetrad, t a , x a , ya, za, determined by such a spin-
frame [Penrose & Rindler , 1984, p120], is in fact the same (up to signs) as
the frame of 'body axes' eJ1. = ¢>'J1.¢ which we drew attention to in standard
Dirac theory, whilst the null tetrad is just a rotated version of a certain
'fiducial' null tetrad as follows:

la = ~(ta + za) = wAWA' ..... ¢>(/o + 13)¢, (28)

na = ~W - za) = 1I"A7fA' ..... ¢>(/o -/3)¢, (29)

1 .
m a = y'2(x a - 1ya) = w A 7fA' ..... -¢>(/I + h2)¢, (30)

ill" = ~(xa + jya) = 1I" AwA' ..... -¢>(/I - h2)¢. (31)

Note that the x or y axis is inverted with respect to the world vector equiv-
alents, which is a feature that occurs throughout our translation of 2-spinor
theory. Note also that 11 - h2 and 11 + h2 involve trivector components.
This is how complex world vectors in the Penrose & Rindler formalism ap-
pear when translated down to equivalent objects in a single-particle STA
space. We shall find a use for these shortly as supersymmetry generators.

5. Valence-l Twistors

On page 47 of [Penrose & Rindler , 1986] the authors state 'Any temptation
to identify a twistor with a Dimc spinor should be resisted. Though there
is a certain formal resemblance at one point, the vital twistor dependence
on position has no place in the Dimc formalism.' We argue on the contrary
that a twistor is a Dirac spinor, with a particular dependence on position
imposed. Our fundamental translation is
(32)
where 4> is an arbitrary constant relativistic STA spinor, and r = xJ1. , J1. is the
position vector in 4-dimensions. To start making contact with the Penrose
notation, we decompose the Dirac spinor Z, quite generally, as
Z = w !(1 + 0"3) - 11" i0"2 !(1 - 0"3). (33)
Then the pair of Pauli spinors wand 11" are the translations of the 2-spinors
w A and11" A' appearing in the usual Penrose representation

ZO'=(W A ,1I"A'). (34)

242 ANTHONY LASENBY ET AL.

In (34) 1I'A' is constant and w A is meant to have the fundamental twistor

dependence on position

(35)
where w~ is constant. We thus see that the arbitary constant spinor </> in
(32) is

(36)
We note this is identical to the STA representation of a spin-frame.
This ability, in the STA, to package the two parts of a twistor together,
and to represent the position dependence in a straightforward fashion, leads
to some remarkable simplifications in twistor analysis. This applies both
with regard to connecting the twistor formalism with physical properties of
particles (spin, momentum, helicity, etc.), and to the sort of computations
required for establishing the geometry associated with a given twistor.
For present purposes, we confine ourselves to establishing the link with
massless particles, and define a set of quantities to represent various proper-
ties of such particles (most of which are useful in the formulation of twistor
geometry as well). These are basically just the bilinear covariants of Dirac
theory, adapted to the massless case. Firstly, the null momentum associated
with the particle is
(37)
This is constant (independent of spacetime position), since

(38)

p thus points in the flagpole direction of 11'. Secondly, the flagpole of the
twistor itself, defined as the flagpole of its principal part w A , is the null
vector
(39)

Evaluated at the origin, this becomes

Wo = </> (-yo + "Y3) J, = wo (1 + 0'3) Wo "yo· (40)

Thirdly, we define an angular momentum bivector in the usual way for Dirac
theory (see above)

M = Zi0'3Z. (41)

Substituting from (32) for Z yields (in two lines)

M = Mo + rl\p, (42)
2-SPINORS, TWISTORS AND SUPERSYMMETRY IN THE SPACETIME ALGEBRAS 243

where the constant part Mo is given by

Mo = ¢ ia3 4>. (43)

This angular momentum coincides with the real skew tensor field
M ab = iw(A1fB)fA'B' _ iw(A'7r B ')f AB , (44)

on page 68 of [Penrose & Rindler , 1986], who have

M ab = Mob _ xapb + xbpa. (45)
The key calculation showing that (41) is the correct angular momentum,
is to demonstrate that the Pauli-Lubanski vector for this massless case is
proportional to the momentum. In the STA, the Pauli-Lubanski vector (the
non-orbital part of the angular momentum, expressed as a vector) is given
generally by

S= p·(iM). (46)

Now p·(iM) = p.(iMo + irAp) and p·(irAp) = -i(pArAp) = O. Also

piMo = ¢(Jo -/3)4>i¢ ia34>, (47)

so that writing ¢4> = 4>¢ = pe if3 , we have

piMo = _pe- if3 ¢( -/3 + 10) 4> (48)

and therefore

S = -pcosf3p. (49)
The helicity s is thus just minus the scalar part of the product ¢4>.
6. Field Supersymmetry Generators
A common version of the field supersymmetry generators required for the
Poincare super-Lie algebra uses 2-spinors QOI with Grassmann entries:

QOI -_ -t
(& . OlOl , '& )
. &()OI - ta
J1 =<()
J1' (50)

where the ()OI and 7rare Grassmann variables, and J1 is a spatial index
[Freund , 1986; Srivastava, 1986; Muller-Kirsten & Wiedemann , 1987].
A translation of QOI into STA basically amounts to finding real spacetime
representations for the ()Ot variables. Using 2-particle STA we have found such
representations, and they turn out to be two distinct copies of the complex
null tetrad discussed above. The two copies arise in a natural fashion in our
version of 2-spinor theory, but are harder to spot in a conventional approach.
244 ANTHONY LASENBY ET AL.

This has an interesting 'single particle' equivalent, using the 4 quantities

1'0 ± 1'3 and 1'1 ± iT2 as effective Grassmann variables, with the anticommu-
tator {A, B} replaced by the symmetric product {AB}. With

(h = 1'0 + 1'3 01 = 1'0 - 1'3

()2 = 1'1 + iT2 02 = -1'1 + iT2
it is a simple exercise to verify that the Oet satisfy the required supersym-
metry algebra (with {A, B} == (AB})

(51 )

This raises interesting new possibilites, similar to those outlined in [Doran

et at. , 1992], of being able to reduce the arena of 'superspace' to ordinary
spacetime, without in any way diminishing its richness or interest.

7. Conclusions
When 2-spinors and twistors are absorbed into the framework of spacetime
algebra, they become both easier to manipulate and interpret, and many
parallels are revealed with ordinary Dirac theory. In particular the bilinear
covariants of Dirac theory (expressed in STA), turn out to be precisely those
needed to understand the role of higher valence spinors and twistors. As
a byproduct of the translation we have shown that a commutative scalar
imaginary is unnecessary in the formulation of 2-spinor and twistor theory.
Furthermore, had space permitted, we would have presented a discussion of
the mapping we have constructed between lumped vector index expressions,
and spin- ~ equivalents. This would have made it evident that the notion that
2-spinor or twistor space is more fundamental than the space of ordinary
vectors or tensors, is misplaced. In our version the spinor space itself is
imbued with all the metrical properties of spacetime, and the construction
of vectors and tensors using outer products of spinors (as given in Penrose
& Rindler for example) can be shown via our translation to use precisely
the metrical properties already present at the so-called spinor level (which
is in fact just ordinary spacetime).
Normalized spin-frames have been shown to be identical to Lorentz trans-
forms, with spin frames in general identical to constant Dirac spinors (even
multi vectors in the STA approach). Twistors themselves have been shown
to be Dirac spinors, with a particular position dependence imposed, and the
physical quantities constructed from them to be just the standard Dirac bi-
linear covariants. It is therefore clear that some of the claims of the 'strong
twistor' programme, as described in e.g. [Penrose, 1975], must appear in a
new light, though the full implications remain to be worked out.
2-SPINORS, TWISTORS AND SUPERSYMMETRY IN THE SPACETIME ALGEBRAS 245

References
Ablamowicz, R., & Salingaros, N. 1985. On the Relationship Between Twistors and Clifford
Algebras. Lett. in Math. Phys., 9, 149.
Ablamowicz, R., Oziewicz, Z., & Rzewuski, J. 1982. Clifford Algebra Approach to
Twistors. J. Math. Phys., 23(2), 231.
Bjorken, J.D., & Drell, S.D. 1964. Relativistic Quantum Mechanics, vol 1. McGraw-Hill,
New York.
Doran, C.J.L., Lasenby, A.N., & Gull, S.F. 1992. Grassmann Mechanics, Multivector
Derivatives and Geometric Algebra. These Proceedings.
Doran, C.J.L., Lasenby, A.N., & Gull, S.F. 1993. States and Operators in the Spacetime
Algebra. To appear in: Foundations of Physics.
Freund, P.G.O. 1986. Supersymmetry. Cambridge University Press.
Gull, S.F. 1990. Charged Particles at Potential Steps. In: Weingartshofer, A., & Hestenes,
D. (eds), The Electron 1990. Kluwer.
Hestenes, D. 1975. Observables, Operators, and Complex Numbers in the Dirac Theory.
J. Math. Phys., 16(3), 556.
Hestenes, D., & Sobczyk, G. 1984. Clifford Algebra to Geometric Calculus. D. Reidel
Publishing.
Lasenby, A.N., Doran, C.J.L., & Gull, S.F. 1992a. Grassmann Calculus, Pseudoc1assical
Mechanics and Geometric Algebra. Submitted to: J. Math. Phys.
Lasenby, A.N., Gull, S.F., & Doran, C.J.L. 1992b. Particle and Multipartic1e Trajectories,
Tunnelling Times and Spacetime Algebra. In Preparation.
Lasenby, A.N., Doran, C.J.L., & Gull, S.F. 1992c. Twistors and Supersymmetry in the
Spacetime Algebra. In Preparation.
Miiller-Kirsten, H.J.W., & Wiedemann, A. 1987. Supersymmetry. World Scientific.
Penrose, R. 1975. Twistor Theory, its Aims and Achievements. In: Isham, C.J., Pen-
rose, R., & Sciama, D.W. (eds), Quantum Gravity - an Oxford Symposium. Oxford
University Press.
Penrose, R., & Rindler, W. 1984. Spinors and space-time, Volume I: two-spinor calculus
and relativistic fields. Cambridge University Press.
Penrose, R., & Rindler, W. 1986. Spinors and space-time, Volume II: spinor and twistor
methods in space-time geometry. Cambridge University Press.
Srivastava, P. 1986. Supersymmetry, Superfields and Supergravity. Adam Hilger.
QUANTUM DEFORMATIONS
QUANTIZED MINKOWSKI SPACE
JULIUS WESS, BRUNO ZUMIN0 1 ; OLEG OGIEVETSKY" and
W.B. SCHMIDKE
Max-Planck-Institut for Physik und Astrophysik
Werner-Heisenberg-Institut fur Physik
P.O. Box 40 12 12, D - 8000 Munich 40, Germany
1 Department of Physics, University of California, Berkeley

and
Theoretical Physics Group, Physics Division
Lawrence Berkeley Laboratory
1 Cyclotron Road
Berkeley, California 94720

The concept of symmetry groups has a mathematically well defined gen-

eralization in the framework of Hopf algebras. Such generalizations have be-
come known as quantum groups- these are Hopf algebras with an algebraic
structure which depends on one or more parameters q (q E C, q f:. 0), such
that for a particular value of these parameters, say q = 1, the quantum group
coincides with the group. In this sense a quantum group is a deformation of
a group, q being a deformation parameter. With the concept of a group goes
the concept of representations and representation spaces. These represen-
tation spaces find a natural generalization as well, called quantum spaces.
These are algebraic structures that depend on the deformation parameter
q and for q = 1 coincide with the linear space in which the correspond-
ing group is represented. For q f:. 1, the quantum group acts as a linear
morphism of the algebraic structure of the quantum space. The algebraic
structure of the group and the quantum space are closely related.
One of the best known examples ofa quantum group is the group SLq(2, C),
a deformation of the group S L(2, C) which is the covering group of the
Lorentz group. Its representations are known as spinors and tensors. The

• This work was supported in part by the Director, Office of Energy Research, Office of
High Energy and Nuclear Physics, Division of High Energy Physics of the U.S. Department
of Energy under Contract DE-AC03-76SF00098 and in part by the National Science
Foundation under grant PHY-90-21139
•• On leave of absence from P.N .Lebedev Physical Institute, Theoretical Department,
117924 Moscow, Leninsky prospect 53, USSR

249
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 249-256.
© 1993 Kluwer Academic Publishers.
250 JULIUS WESS ET AL.

corresponding quantum spaces are called q-spinors and q-tensors. In this

lecture I am going to to discuss the algebraic structure of a q-deformed four-
vector space.lt serves as a good example of quantizing Minkowski space. To
give a physical interpretation of such a quantized Minkowski space we con-
struct the Hilbert space representation and find that the relevant time and
space operators have a discrete spectrum. Thus the q-deformed Minkowski
space has a lattice structure. Nevertheless this lattice structure is compatible
with the operation of q-deformed Lorentz transformations. The generators of
the q-deformed Lorentz group can be represented as linear operators in the
same Hilbert space.To be more specific and to illustrate the general concepts
just mentioned let us study the 2 X 2 matrix A:

(1)

If the entries a, b, e and d are real numbers, A is an element of GL(2, R), and
if they are complex numbers, A is an element of GL(2, C). A is an element
of G Lq(2) if the entries a, b, e, and d satisfy the following algebraic relations:

ab qba bd qdb ad da + Aeb (2)

ae qea cd qde be eb

where A = q - q-l. These relations seem quite arbitrary, they are, however,
determined by the following properties.
The most important property of (2) is that matrix multiplication pre-
serves these relations. Take a second matrix A', with entries which commute
with the entries of A, and by themselves they satisfy the relations (2) as
well. Take the matrix product AA' = A" and you will find that the entries
of A" satisfy the relations (2) again.
The relations (2) have some further properties:
1) They allow an ordering. The left hand side is alphabetically, the right
hand side antialphabetically ordered.
2) The ordering is invertible. An antialphabetic order can be alphabet-
ically rearranged. It is possible to order a polinomial in any desired order
without changing its degree.
3) The relations are consistent. By this we mean that they do not create
higher order relations. The following example illustrates that this is not
trivial. Take the same relations as in (2) except for bd = q'db, q' # q. Try
to put abd into the order dba by starting to exchange first ab or by first
exchanging bd. Comparing these two calculations you will find the third
order relation b2 e = O. The relations (2) are chosen in such a way that no
such new relations will arise.
4) The relations (2) depend on the deformation parameter q and for q = 1
a, b, e and d commute.
QUANTIZED MINKOWSKI SPACE 251

All algebraic structures that we need in the context of quantum groups

have these four properties. It is property 3 that is most restrictive and leads
to the quantum Yang-Baxter equation. The relations (2) can be written in
the form
(3)
with the 4 x 4 matrix

R=(Hn).
o 0 0 q
(4)

Rows and columns are labelled by (11), (12), (21), and (22). If we define the
8 X 8 matrix
(5)
and similarlyR23 they will satisfy the Yang-Baxter equation
R12R23 R12 = R23R12R23' (6)
This relation arises from the requirement that the product of any three
elements of the matrix A can be rearranged by either first changing the first
two or the last two elements. By reduction it can then be shown that the
consistency condition as formulated in 3 holds for all orders.
Having in mind the importance of the Yang-Baxter equation (6) it is
natural to proceed as follows. First try to find a solution of (6). Note that
in the general case when A is an n X n matrix these are n 6 equations for
n 4 variables. Next impose the 'RTT' relations (in general the matrix A is
called T) to find a deformed group. There will always be a solution to the
RTT equation, a multiple of the unit matrix will always do. We of course
are interested in non-trivial solutions.
The R-matrix also tells us how to define quantum spaces. The matrix R
has eigenvalues and therefore satisfies a characteristic equation. From this
equation we can find the projectors on the eigenvalue. These projectors will
be polynomials in R. In our exam pIe, the matrix (4) has the eigenvalues q
and _q-l. The characteristic equation is
(7)
The projectors are:

(8)
Ps
252 JULIUS WESS ET AL.

They satisfy PAPS = 0, = PA, P'§pl = Ps, and PA + Ps = n. In the limit

q -+ 1 these projectors become:

pij
A 1m
-+ ~(8i8j
2 I m
_ 8mi 8Ij )
(9)

As both projectors are polynomials in R it follows from (3) that the following
equation holds for both projectors:
AAP = PAA. (10)
The matrices are as in (3). If we now define the algebraic relations of a
quantum space by the condition
(11)
it follows from (10) that the linear map
(12)
(where xl and the entries of A commute) preserves the algebraic structure:

(13)
From (9) follows that for q -+ 1 the space defined with PA becomes com-
mutative whereas defined with Ps the coordinates will anticommute. It is
natural to identify the anticommuting space with the differentials.
The R-matrix can be expressed through the projectors:
R
A

= qPS - q
-1
PA. (14)
It therefore follows that a matrix that satisfies (10) for both projectors
will also satisfy the RTT relations (3). In other words, if we find a linear
morphism for the quantum plane and the differentials then we know that
this must be an element of the quantum group defined through (3).
We now aplly another linear morphism
(15)
(A" commutes with x', the entries of A' therefore commute with the entries of
A). It follows that x" will satisfy the same relations as x (for the coordinates
and the differentials). A" = A' A will be an element of the quantum group
again.
This demonstrates the power ofthe R-matrix approach to quantum groups
and quantum planes. Our aim is to construct an R-matrix that allows us to
deform the concept of the Minkowski space. To this end we shall construct
QUANTIZED MINKOWSKI SPACE 253

a four-vector as a bi-spinor and use the relations (11) for spinors to derive
the commutation relations for the four-vector. To be able to define reality
properties we first have to define a conjugation operation on spinors.
We define the conjugate spinor:

(16)

and demand that conjugation is an involution (Xi = xi) with the following
property:

7j = q* (complex conjugation). (17)

From (11) follows for the conjugate quantum plane for real q:

(18)

where we have used the property of the k-matrix k ij kl = kkl ij, the matrix
(4) is symmetric. The x - X commutation relations have to be invented. They
have to be consistent in the sense of property 3. A possible solution is [Wess
1990]

x Xj qR =
i _ A -1 Ii -
kjXIXk. (19)

The relations (11), (18) and (19) define the complex q-spinors, and their
linear morphisms are elements of S Lq(2, C).
Finally, we need a second copy of spinors. By themselves, they are sup-
posed to satisfy (11), (18) and (19). Their commutation relations with x,x
have to be consistent and covariant under SLq(2, C). There are two obvious
choices:
x'yJ kij klykxl
(20)
xifJi k- 1 Ii kjYIXk

or
x'v J k- 1 ij klVkxl
Ii - k (21)
xiv· R kjVIX .
A

The four-vectors of the Minkowski plane are represented as

(22)

and we compute the commutations relation of Xi k with the four-vectors

yjyl and YjVI. This yields two different R-matrices, both 16 x 16 matrices
with the same eigenvalues q2, q-2 and -1. Their characteristic equation is

(23)
254 JULIUS WESS ET AL.

Each of them gives rise to three projectors. In each case, one of the subspaces
can be further decomposed and we obtain four independent projectors alto-
gether. Both n-matrices can be expressed in terms of these projectors:

nI = q2p+ + q- 2p_ - Ps - PT
(24)
nn q2PT+q-2PS-P+-P_.

To identify these projectors we give their values for q = 1:

p,ij
1 ..
T kl -g'J gkl
4

pj kl "21 ( 8;k 8Ij + 8iI 8kj ) - l;j

4"g gkl
(25)
1 ( 8; 8j 8; 8j i ij
P~ kl 4" k I - I k
)
- 4"f kl

p!.j kl 1 ( 8i 8j
4" k I -
8; 8j
I k
) i ij
+ 4"f kl·

They reflect the property that in four dimensions a tensor of second rank
can be decomposed into four irreducable subspaces.
The q-deformed Minkowski space is now naturally defined by

(26)

If we combine this relation with (24) and the fact that P++P-+PT+PS =n
we obtain

(27)

This is the basic relation for our study of the quantized Minkowski space.
For convenience, we introduce the notation:

(28)

and we find from (27)

AB BA - q-l>'DC + q>.D2 BC = CB - q>.DB

AC = CA + q-l>'DA BD = q2DB (29)
AD = q- 2DA CD DC.

These relations are in agreement with our properties 1 to 4. The projector

PT projects on a one dimensional subspace the invariant Minkowski length:

(30)

It turns out to be central, i.e. L2 commutes with A, B, C and D.

QUANTIZED MINKOWSKI SPACE 255

A set of commuting variables is

T
q +qq-l (C + D)
1 _ (31 )
----,.1 (qD - q lC)
q + q-
BA.

T is the 'time' operator, it is central. X 3 is the operator associated with the

3-coordinate and Ri
can be interpreted as the radius in the 1-2 plane. To
specify a point in the Minkowski plane, we need one more operator. It turns
out that the angular momentum operator in the 3-direction commutes with
all the operators of (31), so it can be simultaneously diagonalized.
As a physicist we finally have to relate the derived relations to numbers.
From quantum mechanics we have learned that Hilbert space representations
are a proper tool to derive physical consequences. Thus we study the Hilbert
space representations of (29). We choose C, D, Ri and T3 to be be diagonal
and label the states with the respective eigenvalues: Ic,d,p 2,j). From the
commutations relations it follows that A shifts the eigenvalues:

A Ic,d,p2,j) = 0'.(c,d,p2,j) Ic - q)"d,q 2d,p2 + )"d(q- 1c - qd),q-4j)(32)

where 0'. is a normalization constant. Starting from a particular state with

eigenvalues Co, do, P6 and jo we find that all the eigenvalues of the spectrum
generated by A and B = A are:

(33)

The nomalization can be related to p; because AA = B A = Ri. We find

IO'.nl 2 = p; = P6 + dOq-2((co + do)(q2n - 1) + do(1 - q4n)). (34)
This expression has to be greater than or equal to zero. For q > 1, do > 0
and Co > 0, p;
will tend to minus infinity for n -+ 00. There has to be a
largest n such that for n = N aN = o. This leads to
(35)
where n ~ N.
If we try to find a representation which allows a clock to be at the origin
of the space coordinates we have to demand that p2 and the eigenvalue of
X 3 take the value zero for some n. This leads to the condition
(36)
256 JULIUS WESS ET AL.

where J( is an integer J( < Nand J( S; n S; N. The eigenvalues of X 3 will

be
do (q2n+l + q2n-1 _ q2K -1 _ q2N -1). (37)
q + q-1
which is zero for n = J( = N - 1. The eigenvalues of T for these states are:

TN = doq 2N . (38)
This exhibits the discrete spectrum of the time coordinate. Up to now we
have only used the q-Minkowski space relations (29). For a full discussion
of the homogeneous Minkowski space we have to represent the quantum
derivatives as well as the Lorentz transformations. This would glue together
various representations of the q-deformed algebra (29). This has been done
for the q-Lorentz algebra in [Pillin 1992].

References
J. Wess and B. Zumino: Covariant Differential Calculus on the Quantum Hyperplane.
Nucl. Phys. B (Proc. Suppl.) 18 B, 302 (1990)
J. Wess: Differential Calculus on Quantum Planes and Applications. Talk given at Third
Centenary Celebrations of the Mathematische Gesellschaft, March 1990, based on work
with B. Zumino, preprint KA-THEP-1990-22 (1990)
U. Carow-Watamura, M. Schlieker, M. Scholl and S. Watamura,: Tensor Representa-
tion of the Quantum Group SL q (2, C) and and Quantum Minkowski Space. Z. Phys. C-
Particles and Fields 48, 159 (1990); A Quantum Lorentz Group. Int. Jour. of
Mod. Phys. A 6, 3081 (1991)
W.B. Schmidke, J. Wess and B. Zumino: A q-deformed Lorentz Algebra. Z. Phys. C-
Particles and Fields 52, 471 (1991)
O. Ogievetsky, W.B. Schmidke, J. Wess and B. Zumino: Six Generator q-deformed
Lorentz Algebra. Lett. Math. Phys. 23, 233 (1991)
O. Ogievetsky, W.B. Schmidke, J. Wess and B. Zumino: q-Deformed Poincare Algebra.
pre print MPI-Ph/91-98 (1991), to be published in Commun. Math. Phys.
M. Pillin, W.B. Schmidke and J. Wess: q-Deformed Relativistic One-Particle States.
preprint MPI-Ph/92-75 (1992), submitted to Nucl. Phys. B
D = 4 QUANTUM POINCARE ALGEBRAS
AND FINITE DIFFERENCE TIME
DERIVATIVES
JERZY LUKIERSKI* *-
Institute for Theoretical Physics, University of Wroclaw,
pl. Maxa Borna 9, 50-204 Wroclaw, Poland
ANATOL NOWICKI*-
Institute of Physics, Pedagogical University,
Plac Slowianski 6, 65029 Zielona G6ra, Poland

and
HENRI RUEGG***
Departement de Physique Theorique,
Universite de Geneve 24, quai Ernest-Ansermet,
CH-1211 Geneve 4, Switzerland

Abstract. The contraction of quantum Lie algebras providing real D = 4 quantum

Poincare algebras are briefly reviewed. The case of K-deformation of D = 4 Poincare
algebra with flat nonrelativistic sector is described in some detail. The K-modification of
relativistic dynamics consists in introducing in consistent way the finite difference time
derivatives. The K-Lorentz group has a quasigroup structure introduced by Batalin.

1. Introduction

Firstly we recall that recently several ways of obtaining the quantum defor-
mation of D = 4 of Poincare algebra were proposed:
a) By considering quantum anti-de-Sitter algebra Uq (O(3,2)) and per-
forming the "quantum" de-Sitter contraction [1,2].1

{ R -+ Do} : Rlnq ____--+) - iffl,-l (1)

q-+l R-+oo

where f = 1 for Iql = 1 (see [1]) and f = 0 for q real (see [2]). It appears that
one obtains in such a way the fl,-deformation of Poincare algebra which is
• Presented at II-nd Max Born Symposium by J. L.
Partially supported by KBN grant Nr. 2/0124/91/01
••• Partially supported by Swiss National Science Foundation
1 The contraction (1) for rank one quantum algebras SU q (2) and SU q (l.l) were firstly
introduced by the Firenze group [3,4).

257
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 257-266.
© 1993 Kluwer Academic Publishers.
258 JERZY LUKIERSKI ET AL.

a Hopf algebra, with commuting fourmomenta and the Lorentz generators

not forming a Hopf subalgebra.
b) The q-deformed D = 4 conformal algebra Uq (0(4, 2)) has the following
chain of Hopf algebras [5-9].2

(2)

The q-deformation of Poincare algebra obtained in such a way is embed-

ded in quantum Hopf algebra Uq (P4±:> D) where P4 denotes fourdimensional
Poincare algebra and D the eleventh dilatation generator, and it is char-
acterized by noncommuting fourmomenta (forming quadratic algebra) and
closed quantum Lorentz sub algebra.
c) Following the q-differential calculus on q-deformed Minkowski space
[10,11] another q-deformation of Poincare algebra has been recently ob-
tained [12]. It appears that for this deformation again q-Lorentz algebra
is a Hopf subalgebra, and the eleventh dilatation generator is needed for
defining the coproducts for ten q-Poincare generators.
d) In [6,7] the following contraction limit of Uq (0(4,2)) (q-conformal
algebra) was proposed

[R -+ 00] : R 2 lnq _ _ _--., - i",-2 (Iql = 1) (3)

q-+1 R-+oo

and supplemented by the rescaling of the fourmomentum and dilation gen-

erators. In such a way it was obtained a new ",-deformation of the Poincare
algebra, embedded in 11-dimensional Hopf algebra containing besides Poincare
generators an additional central generator.
In the following we shall present the ",-deformation of D = 4 Poincare
algebra with standard real structure and flat 0(3) sector [2]. It appears that
in our quantum Poincare algebra the fourmomenta are commuting, i.e. one
can introduce four continuous space-time coordinates in conventional way,
as canonically conjugated variables related by a Fourier transform

(4)

The ",-deformation enters if we wish to define free fields by the differential

operators invariant under ",-Poincare transformations. It appears that in
these operators the continuous time derivative is replaced by finite difference
time derivatives.
It should be mentioned that our quantum deformation of relativistic
physics is milder than the one proposed by Wess et all [10-13] where the
q-Minkowski space is described by noncommutative geometry. In such an
2 In [8] the real form of U q (SI(4; C)) providing Uq (O(4,2)) was not given explicitly.
The discussion of all real forms of Uq (SI(4; C)) is given in [9].
D = 4 QUANTUM POINCARE ALGEBRAS 259

approach even classically the measurable values of four-positions and four-

momenta form a discrete lattice, given by the eigenvalue conditions of the
operator-valued q-Minkowski coordinates and momenta.

2. The standard real quantum Poincare algebra.

In order to obtain standard real K-Poincare algebra we proceed as follows:
i) Using the formulae for the commutators and coproducts of antipode-
extended Cartan-Weyl basis of Uq (0(3,2)) (see [2]) we can write the q-
deformation of the 0(3,2) Lie algebra as well as the coproduct relations for
the q-deformed 0(3,2) generators.
ii) We perform further the quantum de-Sitter contraction, obtained by
the conventional rescaling of the 0(3,2) rotation generators
Mil-V unchanged (Mtv = Mil-V)
(5)

and the q --+ 11imit described by (1).

As a result we obtain the following q-deformed Poincare algebra:
a) Three-dimensional 0(3) rotations (M± = Ml + iM2 == M 23 ± iM31 ;
M3 = M 12 )
i) commutation relations:
(6a)
ii) coproducts:
(6b)
iii) antipode:
S (Mi) = -Mi (6c)
b) Boosts sector 0(3,1) (L± = M14 ± iM24 , L3 = M34)
i) commutation relations:
Po
[L+,L_ 1= -2M3 cosh-; 1 1 Po
2
+ 2K2 P3 + 2
K2M3P3 -
.
smh-;

& 1
[L+,L3] = e- I< M+ + -2K (iP3L+ + L 3P_) +
i 1.
- 2K2M3P3P- + 4K 2 (2 - Z)P3 P - (7a)
260 JERZY LUKIERSKI ET AL.

(7b)
[M -, L+l = -2L3 + -M
1
21< - P- + .!.P3M3 -
.
I<
1
-P3
I<

[M+,L31- -L+
_
+ 21<M3P-
1
+ i
21<P-

[M-,L+l = L_ - 211<P+M3 + 2i1<P+

ii) coproducts:
& & 1 &
~L3=L30e21< +e- 2,. 0 L3+ 2h:e-21< (M+0P++M_0 P_)

(7c)
iii) antipode:
i 1
S(L3) = -L3 + 2h: P3 + 2h: (M+P+ + M_P_)
(7d)
1 i
S(L±) = -L± =f -P=f =f -M±P3
h: h:

c) Translations sector (P± = P2 ± iP}, P3, Po )

i) commutation relations (/1, v = 0,1,2,3)
[PI" Pv ] = 0
(8a)
[Mi,Pol = 0

(8b)
D = 4 QUANTUM POINCARE ALGEBRAS 261

ii) coproducts

6Po = Po ® I + I ® Po
(8c)
(i=1,2,3)

iii) antipode
(8d)
Following [14] we have introduced in [2] a nonlinear transformation of the
boost generators

(9)

simplifying the K-Poincare algebra substantially. The new boosts satisfy

the following relations:

[MI,i j ] = ifjjkLk

(10)

It is interesting to observe that the algebra (10) differs from the one
obtained in [10] only by the replacement K ---+ iK. The same holds for the
coproduct formulae.

- )
6 ( Li = Lj !JJ. + e- !JJ. ® Li
- ® e2" 2"
- + 2"fijk
1 ( !JJ.!JJ.
Pj ® Mke2" - e- "Mj 0
2 Pk )
(11)
The coproducts (11) which satisfies the relation (1) permit to define the
tensor product representations in Hilbert space. For completeness we give
also the antipodes:
- ) - 3 i
S ( Li = - Li + 2" ;, Pi (12)
262 JERZY LUKIERSKI ET AL.

One can construct the quantum deformation of quadratic Casimir, de-

scribing quantum relativistic mass square operator. One gets

C1 = pf + pi + pi + 2K2 (1- cosh ~o) = p2 - (2Ksinh :~r (13)

It should be mentioned that recently the D = 4 mass square Casimir

was proposed in [4,15] as the extension of the results obtained for D = 3
Poincare algebra.
The second Casimir can be obtained by introducing the K-deformed
Pauli-Lubanski fourvector
Wo=PM
(14)

where Ii is defined by the formulae (9).

The formula for the second Casimir takes the form:

(15)

3. Finite difference time derivatives from K-Poincare .

Let us consider the simplest realization of the algebra (10) on the scalar
functions 4>( x, t)

(16)
-
L; =i Po)
1 ( XOPi - KXisinh~

generalizing for K < 00 the spinless realization of the Poincare algebra, for
which P . M = 0 and WI' = o. The generators Li act explicitly as follows:

-
Li4>(X, t) = a
-Xo ax; 4>(x, t) - iXi ( t)
D~4>(x, t) (17)

where [2]

J5t.-l.(
",'I' X, t
) = 4>(x, t + ~t) - 4>(x, t - ~t)
2~t
I
t.t=~

and of course
lim D~4>(x, t)
"' ..... 00
= at4>(x, t) (19)
D = 4 QUANTUM POINCARE ALGEBRAS 263

Using the realization (16) one can obtain also the ,..-deformed Klein-Gordon
equation (see [2] )

[~- (2,.. sin g~r] ¢>(x,t) = [~- 2,. 2 (1- cos ~)] ¢>(x,t) = m 2 ¢>(x,t)

(20a)
which can be written as follows

(20b)

One can introduce at least three forms of ,..-deformed Dirac operators, defin-
ing three different ,..-deformed Dirac equations:
i) The Dirac equation obtained by taking square root of the ,..-deformed
Klein-Gordon operator (see [2]). It is the simplest one, but its invariance
properties under ,..-Poincare transformations are quite obscure.
ii) The one derived from the three-dimensional realization of ,..-Poincare
algebra with spin t(see [16]) acting on the functions only depending on
the three-momenta coordinates 3 Such a Dirac operator by construction is
on-shell ,..-Poincare - invariant.
iii) Recently there were found [18,19] the Dirac operators which commute
off-shell with the four-dimensional realization of ,..-Poincare algebra 4.
The linearization of KG operator leads also to the free Hamiltonian Ho
describing scalar particles with relativistic kinematics:

(21)
where
m2 + p2 -
= (Ho + Po)(Ho -
P6 Po) (22)
After ,..-deformation one obtains (C1 = - M2)

M2 + p2 - (2,.. sinh ;~r = (WM + 2,.. sinh ;~) (WM - 2,.. sinh ;:) (23)

The relativistic Schrodinger equation

iOt 1j; = Wm (~V) 1j; (24)

is replaced by
(25)

3 These realizations in the classification of the forms of relativistic dynamics "given by

Dirac [17] are called "instant forms" and describe Hamiltonian dynamics with relativistic
kinematics.
4 The derivation in [18] is using the ,,-Poincare algebra, and in [19] one employs the
finite ,,-Lorentz transformation.
264 JERZY LUKIERSKI ET AL,

The equation (25) can be described equivalently by the following modifica-

tion of the free relativistic Hamiltonian:

(26)

where
'"
wM ' h WM
= 2/'i, arcsm ~ = WM + 6/'i,2
1 wM 3 +0 ( /'1
i,4 ) (27)

In particular for the large values of energies one obtains

WM ___
WM--+OO
--+) 2
InwM + 0(_1_)
WM
(28)

i,e. the high energy behaviour is drastically modified. At present we are

studying the consequences of such a modification e.g. for the description of
confinement.
It should be mentioned here that the modification similar to the one given
by eq. (25) was proposed by Caldirola [20] and studied by his followers. We
would like to point out however two basic differences:
a) In our formalism the finite difference derivative (see (18)) contains ele-
mentary time shift in purely imaginary direction. This property has impor-
tant consequences e.g. the /'i,-deformed kinematics respects the light velocity
as the maximal one (see [21]). In Caldirola formalism there is a maximal
energy at which the velocity achieves infinite value.
b) In the ref. [20] one proposes the replacement of the ordina.ry time
derivative by finite difference time derivative in nonrelativistic Schrodinger
equation, From our considerations it follows that the /'i,- deformation can
be achieved only when the light velocity is finite - i.e. it seems that the
/'i,-Galilei algebra, obtained from the /'i,-Poincare algebra (c finite, /'i, finite)
by the limit c --+ 00 does not exist,5

4. From /'i,-Poincare algebra to /'i,-Poincare group

Let us observe that the /'i,-Poincare algebra is a special example of the
following class of nonlinear algebras:

(29)

5 We would like to mention that in the deformation of Galilei algebra given in [14] one
performs the limit", -+ 0, C -+ 00 ( " 'C fixed) which is not a proper ",-Galilei limit,
D = 4 QUANTUM POINCARE ALGEBRAS 265

If the deformation parameter K, ---+ 00 one recovers the standard Poincare

algebra.
In [19] we show that our choice given by the eq. (10) is a unique possible
solution for large class of the fourmomentum-deper..dent structure constants
in (29).
In order to discuss the finite K,-Poincare transformations one should
introduce the fourmomentum realization dual (in the sense of the Fourier
transform) to the one given by formula (16). Introducing
1 {)
x
J1. = ---
i {)pJ1.
(30)

we see from (16) that the realization of Mi and Li is described by a linear

differential operators in fourmomentum space. The infinitesimal K,-Lorentz
transformations of the fourmomenta are given by the formulae (ai-space
rotations, f3i - boosts):

(31)
opo = Piof3i
The infinitesimal transformations (31) can be integrated and one obtains
the nonlinear formulae, preserving the K,-deformed length of the fourvector,
given by (13).6 Our main observation here is that the general formalism of
finite transformations with the generators corresponding to the infinitesimal
transformations (31) was elaborated by Batalin [24]. It appears that
i) The nonlinear functions on rhs of (31) imply the generalization of the
composition law of two K,-Lorentz transformations. If the integrated form of
(31) looks as follows
(32)
where aA == (ai, f3i) describe the K,-Lorentz parameters, one obtains that

1>J1. (1)(p, a), a') = 1>J1. (p,cp(a,a'jp»

We would like to stress that for the usual Lie groups the function <p describing
the composition law does not depend on the group element (p in our case).
ii) One can relate the nonlinearities of (32) and the momentum-dependent
structure constants.
iii) There is a natural generalization of the Maurer-Cartan equations for
K,-Poincare algebra.
For more details see [19] and further publications.
6 In particular one can obtain the "-generalization of one-parameter Lorentz transfor-
mations, which are described by elliptic functions ([19]).
266 JERZY LUKIERSKI ET AL.

References
[1] J. Lukierski, A. :'{owicki, H. Ruegg and V.N. Tolstoy, Phys. Lett B264, 331 (1991)
[2] J. Lukierski, A. Nowicki and H. Ruegg, Phys. Lett. B 293, 344 (1992)
[3] E. Celeghini, R. Giacchetti, E. Sorace and M. Tarlini, J. Math. Phys. 32 1155 (1991);
ibid. 1159
[4] E. Celeghini, R. Giacchetti, E. Sorace and M. Tarlini, "Contraction of quantum groups",
Proceeding of First ElMl Workshop on Quantum Groups", Leningrad, October -
December 1990, ed. P. Kulish, Springer Verlag 1991
[5] V. Dobrev, "Canonical q-Deformations of Noncompact Lie (Super )-Algebras", Giit-
tingen Univ. preprint, July 1991
[6] J. Lukierski and A. Nowicki. Phys. Lett. B 279.299 (1992)
[7] J. Lukierski, A. Nowicki and H. Ruegg, Boston Univ. preprint BUHEP-91-21 Novem-
ber 1991; to be published in the Proceedings of II Wigner Symposium (Goslar 11.07-
16.07.1991), vol. "Quantum Groups", ed. H.D. Doebner and V. Dobrev, Springer Ver-
lag (1992)
[8] V. Dobrev, Trieste preprint lC/92/13, January 1992; to be published in the Proceedings
oflI Wigner Symposium (Goslar 1l.07-16.07.1991), vol. " Quantum Groups", ed. H.D.
Doebner and V. Dobrev, Springer Verlag (1992)
[9] J. Lukierski, A. :'{owicki and J. Sobczyk, Wroclaw Univ. preprint ITP UWr 1992
807/92/Bochum Univ. preprint \fath. 156/92 (June 1992)
[10] W.B. Schmidke, J. Wess and B. Zumino, Z. Phys. C52. 471 (1991)
[11] O. Ogievetsky, W.B. Schmidke, J. Wess and B. Zumino, Lett. Math. Phys. 23. 233
(1991)
[12] O. Ogievetsky, W.B. Schmidke. J. 'Ness and B. Zumino, "q- Deformed Poincare
algebra", Max Planck and Berkeley preprint MPI-Ph/91-98 LBL - 31703 UCB 92/04,
November 1991
[13] J. Wess, B. Zumino, D. Ogievetsky and W.B. Schmidke, to be published in these
Proceedings
[14] S. Giller, J. Kunz, P. Kosinski, M. Majewski and P. Maslanka, Phys. Lett. B286, 57
(1992)
[15] E. Celeghini, R. Giacchetti, E. Sorace and M. Tarlini, Univ. Firenze preprint DFF
151/11/91
[16] S. Giller, J. Kunz. P. Kosinski, ~I. ~lajewski and P. Maslanka, "On q-covariant wave
functions", Lodz, Univ. preprint, August 1992
[17] P.A. Dirac, Rev. Mod. Phys. 21. 392 (1949)
[18] A. :'{owicki, E. Sorace and M. Tarlini, Firenze Univ. preprint DFF 117 / 12 / 92 ,
December 1992
[19] J. Lukierski, H. Ruegg and W. Riihl, Kaiserslautern Univ. preprint KL - TH - 92 /
22, December 1992.
[20] F. Caldirola, Lett. Nuovo Cim. 16, 151 (1976); 17, 461 (1976)
[21] J. Lukierski, H. Ruegg and A. Nowicki, talk at lnt. Symposium Symmetry VI, Bregenz.
2-7.08.92, to be published by Plenum Press
QUANTUM LORENTZ GROUP AND
q-DEFORMED CLIFFORD ALGEBRA

URSULA CAROW-WATAMURA
Department of Physics, Faculty of Science
Tohoku University, Sendai 980, Japan

Abstract. We explain the construction of the quantum Lorentz Group Fun q (SO(3, 1)),
the quantum Minkowski space and the q-deformed Dirac I matrices.

1. Introduction

When investigating the quantum gravity we have serious problems of con-

ceptual nature, since on one hand the laws of nature have to be covariant
with respect to the group of diffeomorphisms of the spacetime manifold M,
Dif f(M). On the other hand from the point of view of particle physics the
metric itself becomes a dynamical variable and we have to define the dy-
namical variables before the spacetime points have a physical identity. The
algebraic approach to the quantum gravity suggests that the distinction be-
tween spacelike and timelike directions becomes established at scales large
compared to the Planck scale [Fredenhagen 1987]. The standard canonical
quantization approach concludes that the concept of time emerges only at
a classical level [Halliwell 1992], and the idea that the concept of space and
time has to be modified drastically when going beyond the Planck scale,
which is considered as the quantum regime of gravity, is not unfamiliar to
physicists. However we do not have a description of how the classical con-
cepts of space and time may be modified in the transition to the quantum
theory.
The motivation to study the quantum group structure is that it opens
a way to investigate a theory based on a geometry with non-commutative
coordinate function algebra. Therefore, this new class of non-commutative,
non-cocommutative Hopf algebras being available it is interesting to study
the q-deformation of the spacetime symmetry, and as a first step to investi-
gate the quantum Lorentz group.
When talking about quantum groups we always think of the q-deformed
algebra of functions on a certain group G, Funq(G) : G -+ C. So we are
using the approach dual to the one given by [Jimbo 1986].

267
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 267-276.
© 1993 Kluwer Academic Publishers.
268 URSULA CAROW-WATAMURA

2. The q-Plane Approach to the Quantum Lorentz Group

To construct the quantum Lorentz group, instead of starting from the def-
inition of the Funq(SL(2,C)) [see Carow-Watamura 1991a] we report here
about the construction [Carow- Watamura 1990], using the approach of the
q-deformation of the quantum space algebra [Manin 1988]. We introduce the
q-spinor zP = (;), with the q-deformed commutation relation
xy = qyx, (1)
and the anticommuting q-plane e = (~~), with ee + q-1ee 0, and
(e)2 = (e)2 = O.
In the q-plane approach the quantum group commutation relation
R12MIM2 = M2MIR12 (2)
is obtained by the requirement that the coaction LlL(ZP) = Mg ® zU, is an
algebra homomorphism of the comodule algebra generated by zP and e,
where Mg E Funq(SL(2,C)). LlL provides zP with a left Funq(SL(2,C))
comodule structure.
The definition of Funq(SL(2,C)) given in [Reshetikhin 1989] was not
complete as a complex quantum group, see also [Carow- Watamura 1991d],
and thus the Funq(SO(3, 1)) was not constructed. * The construction of the
Funq(SO(3, 1))**, the construction of a possible candidate of the q-deformed
Minkowski space and the q-deformed analog of the Clifford algebra satisfied
by the corresponding 'Dirac' matrices have been given in [Carow-Watamura
1990,1991a] and will be shown here briefly.
To construct the Funq(SO(3, 1)) our strategy is to build a real vector
representation out of the tensor product of two spinor representations. For
this purpose we take a second 'copy' of the algebra of Funq(SL(2,C)) de-
noted by Funq(SL(2, C)), and consider the bigger algebra Funq(SL(2, C))®
Funq(SL(2,C)). Then the reality condition is defined by an inner involu-
tion of this bigger algebra, i.e. we construct a real form of the quantum
double [Faddeev, pivate communication]. One may associate with this the
method to contruct a scalar quantity from two spinors of opposite chirality
in ordinary field theory.
Another property in our construction is that in order to obtain four di-
mensional non-null planes, we need two pairs of q-spinors: one pair trans-
forming under the Funq(SL(2, C)) and a second pair transforming under the
Funq(SL(2,C)). They are denoted by zf and zj, respectively, i,j = 1,2 .
• The authors in [Reshetikhin 1989] themselves have pointed out that the quantum
group Funq(SO(n, m)) for In - ml ;::: 2 is not obtained from their construction .
•• The quantum Lorentz group has also been constructed independently by [Podles
1990]. The algebra of the q-Minkowski space is first constructed in [Carow- Watamura
1990].
QUANTUM LORENTZ GROUP 269

Similarly we denote the q-matrix of Funq(SL(2,C)) by if where the if

satisfy the same commutation relation as eq.(2), and ~L(zn = ifg Q9 zf.
The q-analog of the inner product between two q-spinors can be used to
define the q-analog of a "length" which has to be a central element in our
algebra. Therefore we require the invariant products

to be central where>. and >" are complex numbers. This definition has a
correspondence in twistor theory where it is equivalent to the existence of a
non-zero inner product between two spinors [Wald 1984].
From the requirement of covariance and the consistency with their q-
spinor relations eq.(l) the commutation relation among the Zi (and the Zi)
must be nontrivial. Taking into consideration that the products eq.(3) have
to be central these commutation relations are obtained as
(4)

The R-matrix is the one of the Funq(SU(2)) [Jimbo 1986].

To construct the tensor representation we have to fix the commutation
relations of the q-spinors zP with the i p • It is clear that there are two pos-
sibilities: zP and i P are either commuting or non-commuting.

3. Non-Commuting Case
The quantum group Funq(SO(3, 1)) is obtained by non-trivial commutation
relations between zf and if, i.e. we take the twisted tensor product of the
two algebras. They are given as
z I i2 = k'Ri 2zI , Z2ZI = k'Rz I Z2 ,
1 A 1 A

zlil = k'q RiIZI , Z2 Z2 = k'qRZ2Z2 . (5)

The parameters k' and q are introduced such that the requirement of cen-
trality of the products eq.(3) holds. The choice of the R-matrix in the com-
mutation relations eq.(5) as well as in eq.(4) is a matter of convention. We
could as well have taken R-I. However we will see below that the Ii of the
tensor representation is now determined.
The eq.(5) also fixes the commutation relation between M and if as

(6)
The left coaction on the tensor representation is given by ~L(zpza) =
T(pa)(vl") Q9 ZVzl" = ifpvMal" Q9 ZVzl".
Our matrix T(pa) (VI") satisfies a relation Rl2TI T2 = T2TIR12. However,
due to the two choices of the R-matrix in eq.(2) and in the corresponding
270 URSULA CAROW-WATAMURA

equation for £1, i.e. either R or R- 1 , we have in principle four possibilities

to compose this R 12 . These four choices reduce to two inequivalent choices
since pairwise two combimations are the inverse of each other. So we are left
with two candidates for the Funq(SO(3, 1)). It turns out that the R-matrix
corresponding to the quantum Lorentz group can be specified by investigat-
ing the projector expansion [Carow-Watamura 1991aJ. The solution is
R'(O"P)(WJt) (W"~I1)((1l1pll) -R'pw
- w'p'
R'O"w' W"(J' R'p'Jt p/p" R'-l,,'Jt' J.l"a" (7)
the projector expansion of which coincides with the classical result in the
limit q -+ 1.
It can also be verified easily that this H.-matrix satisfies the Yang-Baxter
equation. The second possible candidate for the R-matrix was excluded from
our present considerations since its projector composition has no classical
analog in the limit q -+ 1. ***

4. Real Representation
In order to construct a real representation one way which is often used is
the operation of the hermitian conjugation and in fact it will also do the job
here. Mathematically we look for a *-conjugation which is an involution of
the algebra and consistent with the Hopf algebra structure of the quantum
group. What we will get is a *-Hopf algebra and our reality condition is
obtained with this definition of the *-conjugation: The hermitian conjugate
of a q-spinor is zp = (x, y). The left comodule structure of the q-spinors
Zp induces a corresponding comodule structure on its hermitian conjugate
as ~L(zp) = Mto" p @ z". The symbol * acts as the complex conjugation
for a usual complex number. Note that S( Me) = fPO" MJt "fJtv. Thus includ-
ing the operation of hermitian conjugation, the quantum space algebra of
zP and of zp are generated by two elements, namely x and y. Including
the spinor metric we find another two: ~L(ZPfp,,) = S(MP,,)@ (ZVfvp) and
~L( f*PO" z,,) = S(Mtp,,) @ (f*"vZv)'
In terms of spinors the reality condition needed to reduce to the real
representations is to identify these algebras. It turns out that it is sufficient
to consider q real. t Then we have the following identification
-p
Zj = f PO"-Zj,O" (8)
This implies that SCM) = Mt. It is straightforward to derive the commuta-
tion relation for the q-spinor and its adjoint as
Z( fZ) = k' R( fZ)Z (9)
For the quantum Lorentz group case see below. For the other case excluded here the
projector expansion has been given in [Carow-Watamura 1991c]'
t For q being a pure phase we cannot get non-null space relations.
QUANTUM LORENTZ GROUP 271

In order to define the commutation relation between M and Mt a heuristic

argument to find this relation has been given in [Carow- Watamura 1991d).
The result is

M tiMi'R' -1 i'j _ R' -1 ii'M j Mti' (10)

i' k j'l - ki' jl I

Note that with the unitarity condition we can consistently restrict our sys-
tem to Funq(SU(2» since it preserves the substructure Funq(SL(2, C» =:l
Funq(SU(2».
The central term in our algebra which can be identified with the length
is obtained by substituting eq.(8) into eq.(3) and taking)..' as the complex
conjugate of A:

(11)

The commutation relations of all q-spinors are given by eq.( 4) and

1 '
Zl(EZt) = klqR(EZdzl ,
1 '
Z2( EZ2) = k'q R( EZ2)Z2 . (12)

and their hermitian conjugates.

Now all relations among the q-spinors are defined. With a tedious but
straightforward calculation one can show that with the following definition
of the 4-dimensional q-space coordinate functions A = Xl Y2 + X2Yl, B =
thX2 + 112xl, C = XIX2 + X2X1 and D = fhY2 + thYl we obtain a closed
algebra:

DC = CD, CA - AC = (1- q2)AD,

BC - CB = (1- q2)DB, DA = q2 AD, DB = q-2 BD,
AB - BA = (q-2 - l)CD + (q2 - 1)D2. (13)

The central terms yield:BA - CD - (1- q2 )D2 = k~q AA *. Note that A = B*,
C and D are hermitian and thus eq.(13) as well as the central term are
invariant under the *-operation.
The structure of the above algebra is preserved under the coaction of
the Funq(SL(2,C» since t!.L(zp,iZu) + zp,)ZU;) = MP'pMuU' @ (Zp"izu ' ) +
-
Zpl,jZ u i).
l

We complete our considerations with showing that the 'rotation group'

Funq(SO(3» appears as a subalgebra of the Funq(SO(3, 1)). For this end
we redefine T = C+D ,X+ = q-l/2A, Z = q-1C-qD, X_ = q1/2B. These
qvlq+q-I viq+q-I
coordinate functions also generate a closed algebra with the commutation
272 URSULA CAROW-WATAMURA

relations given by

[r,x+l = [T,X-l = [T, Zl = 0

-4X+ - q2 X+Z = (l-l)x+T
X_Z - q2ZX_ = (1- q2)TX_
[X+,X-l = (q - q-1)(Z2 - ZT) (14)

The central product is given by: qX+X_ +Z2 +q-1 X_X+ _T2 = qtr1,X,X*.
In this algebra T is a central element and may be put to zero. In thls case
the other quantities form the 3-dimensional comodule of the quantum group
FUnq2 (SO(3)).
Let us define the 'four-vector' U = (X_, Z,X+, T). We can prove by
direct computation that the matrix A of the coaction of Funq(SO(3, 1)) on
U, ~L(U) = A \2) U satisfies the orthogonality relation tACA = C, where
Cis
o
1
(15)
o
o
and A is

( ,"d~J'
qQ
q(ca-db)
Q
aa-q2cc-bb+22dd
Q2
A=
qcb
ab-2 2cd
Q
,1,""+")
r/ aa-24c~+bb-q2 dd
Q2
1
q2 ac+bd
(16)
~ ac-bd
q -Q- ad Q
ba+dc aa±cc-bb-dd ab±cd 22 aa+2 2 cc+bb+dd
qQ Q2 Q Q2

withQ=~.
In this basis the reality condition for the four-vector is ut 1] = tuc
where 1] = diag( 1, 1, 1, -1) and the reality condition for the q-matrix A is
At = 1]A -11]. Restricting to the substructure Funq( SU(2)), i.e. substituting
the condition Mt = SCM) into eq.(16) the matrix A splits into a 3 X 3 and
a 1 X 1 part. The 3 X 3 matrix M3 gives the matrix corepresentation of the
Funq(SO(3)) in terms of the matrix elements of the corepresentation of the
Funq(SU(2)).

aa -Qba -bb)
M3 = ( -Qea (da + qbe) Qdb (17)
-ee Qde dd

satisfying the orthogonality equation t M3C3M3 = C 3.

QUANTUM LORENTZ GROUP 273

We conclude that our above defined 4-dimensional q-space eq.(13) is a

real left comodule of Funq(S L(2, C)) and can be considered as the quantum
Minkowski space. The corresponding q-matrices of the left coaction on this
comodule satisfy the properties of the quantum group Funq(SO(3, 1)) and
are referred to as quantum Lorentz group in the following.
Since the R-matrices of th.: Funq(SL(2,C)) are the building blocks of
the quantum Lorentz group R-matrix we use this fact to derive the pro-
jector decompositi~n of the Funq(SO(3, 1))-R matrix. With the graphical
technique developed in [Carow-Watamura 1991aJ we can derive the charac-
teristic equation of the R matrix as

(18)

From eq.(18) the projectors are found as Ps = Ns(R-q-3)(R+q-l), PA =

NA(R - q-3)(R - q) , PI = Nl(R + q-l)(R - q), where the !,S,NA,Nl
are normalization constants. In terms of these projectors the R-matrix is
gi ven as R = qP s - q-l P A + q-3 Pl' Using the projector expansion we
can write our equations in a very compact form. Identifying (A, B, C, D) =
(Ut, Ui, UI, Un where uij = fki U kj , we can write the algebra eq.(13) as
PA(U®U)=O.
From the explicit form of the PI we know that the length L is proportional
to Pl(U ® U), L = C(;j}(kl)UijU kl = -(q + q-l )(BA - CD - (1- q2)D2).
The projector to the symmetic part gives us the commutation relations
of the q-deformed Clifford algebra.

5. q-Deformed Clifford Algebra

In the non-deformed case the connection between spinor and vector repre-
sentation is given by defining the I matrices. Since one of our requirements
is that in the limit q --+ 1 the ordinary Lorentz group has to be recovered we
expect that for the quantum Lorentz group a corresponding Clifford alge-
bra exists. However due to the non-commutativity properties it is not clear
whether indices can be raised and lowered in the ordinary fashion. Thus
we introduce four sets of Pauli matrices a kl , a~i' a~l, and aZ i , requiring
orthogonality a'.'kl a l/kl = bl/lL , aIL_a
kl l/ = bl/ and completeness a'.'kl a IL = b~1
kl lL klll
k bt'
l'
a~laZ'P = bf' bf·Using these Pauli matrices the four dimensional generators
of the quantum Lorentz group can be represented as

(19)
274 URSULA CAROW-WATAMURA

The the relation between the Pauli matrices (j and 0- can be found as
follows. First we observe that we can build another four vector by using the
0- as

(20)

Since this four vector WI-' belongs to the same representation space as UI-',
the coaction on them must be the same. This leads us to the equations

and (21)

There is a choice of an overall factor in eq.(21) which we have set to 1. Using

the orthogonality we can rewrite this as (jklo-I-'_ = il- 1kl - and o-Ik(j,!
I-' I'k' I'k' I" k'l'
'Ik
R k'l"
These q-deformed Pauli matrices can be used as a basis for representing
tensors.
With the properties of the q-Pauli matrices we can write down the anal-
ogous equations of the ordinary S L(2, C) spinor calculus for the case of the
quantum Lorentz group , for examle Tr{(jl-'o-"} = -
Ekk ' (ji! Ell' (j" _ = q-1CJJ"
kl I'k' •
Furthermore one can prove that the condition for the anticommutation re-
lation is
(22)

The validity of eq.(22) is most easily understood by using the diagrammatics

[Carow-Watamura 1991a]. ~ ~
With the projector decomposition of the R-matrix as qR + 1 = qQPs +
q-lQP1 and using eq.(22) we derive

(23)

(24)

The eq.(23)) and eq.(24) suggest to define the Dirac matrix, as follows

,I" == In
VQ
(0
-qE(j1-'
Eo-Ol-') (25)

Then eqs.(23) and (24) can be combined as

,1-'," + qRI-''' "'1-',,"',1-" = q-lQCI-''' (26)

The eq.(26) is equivalent to the condition ps(r ® r) = 0, where r sym-
bolizes the basis of the Clifford algebra. For the proof see (Carow-Watamura
1991a].
QUANTUM LORENTZ GROUP 275

Finally we present an explicit form of the q-deformed Pauli matrices.

With at == /ck' ak'l we obtain

a
1 (1 0)
ok _ _
I - qVQ 0 1
+k _ ~
,a I -.;q
(0 0

0)
o
3k 1
,a I=VQ
(q-l0 (27)

which in the limit q --+ 1 coincide with the conventional Pauli matrices.
With this basis we can represent our 'Minkowski four vector' as

(28)

where the index J-l = 0, +,3, - and the components of VIL are related to the
quantum four plane by the identification (VO, V+, V 3, V-) = (T,X, Z, Y).
U sing the spinors of the Fun q ( S L(2, C)) we can represent a Dirac spinor
WDirac as

WDirac = Cz:~)k ) (29)

and the conjugate spinor is q, = \[It ,0. The element q, 1 \[12 = \[I h°\[l2 is
central with respect to the algebra of coordinate functions.
The relations among the q- Dirac matrices also fixes the reality condition
of the Lorentz vectors. From ,ILt,Ot = ,0,PCpvr(IL with the metric C ILV in
this basis given as
o o
o o (30)
o
_q-l -1
o
and 1]ILV = diag( 1, -1, -1, -1) we obtain the following condition for V :
vpt = VILC ILv 1]VP. For the transformation matrix we get At = 1]A -11].
In order to compare eq.(30) with the metric of the quantum Minkowski
space we simply have to change the overall sign in the definition of the metric
of the q-Minkowski space.

6. Conclusion
As already pointed out in the introduction the quantum group gives us
a possibility to investigate a theory, the algebra of coordinate functions
of which is non-commuting. In order to reach such a stage one way is to
study the q-deformed generalizations of the known theory. With this aim
we have also investigated the differential calculus on the q-Euclidian space
276 URSULA CAROW-WATAMURA

[Carow- Watamura 1991bj which gave some encouraging results involving the
q-deformed polynomials_ It is our hope that this new approach will give us a
better insight into the problems of formulating a quantum theory of gravity
based on the non-commutative geometry.

References
Carow-Watamura, U., Schlieker, M., Scholl, M. and Watamura, S.: 1990, Z. Phys. C -
Particles and Fields 48, 159
Carow-Watamura, U., Schlieker, M., Scholl, M. and Watamura, S.: 1991a, Int. Jour. of
Mod. Phys. A6, 3081
Carow-Watamura, U., Schlieker M. and Watamura, S.: 1991b, Z. Phys. C - Particles and
Fields 49, 439
Carow-Watamura, U., Schlieker, M., Watamura, S. and Weich,W.,: 1991c, Commun.
Math. Phys. 142, 605
Carow-Watamura, U. and Watamura, S.: 1991d 'Complex quantum group, dual algebra
and bicovariant differential calculus', preprint-TU-382(1991), to be published in Com-
mun. Math. Phys.
Fredenhagen, K. and Haag, R.,: 1987, Commun. Math. Physics 108, 91
Halliwell, J.: 1992, 'Quantum cosmology and time asymmetry', Proceedings of the NATO
Workshop, 'Physical Origins of Time Asymmetry', eds. J. Halliwell, J. Perez-Mercader
and W. Zurek (Cambridge University Press, Cambridge, 1992).
Jimbo, M.: 1986, Lett. Math. Phys. 10, 63
Manin, Yu. I.:1988, 'Quantum groups and non-commutative geometry', Centre de
Recherches Mathematiques, Universite de Montreal.
Podles, P. and Woronowicz, S.L.: 1990, Commun. Math. Phys. 130, 381
Reshetikhin, N.Yu., Takhtadzhyan L.A. and Faddeev, L.D.: 1989 Algebra and Analysis 1
178; translation Leningrad Math. J. 1, 193-225 (1990).
ISOTROPIC q-LORENTZ GROUP

JAKUB REMBIELINSKI
Department of Theoretical Physics
University of L6di
ul. Pomorska 149/153, 90-236 L6di, Poland

Abstract. A new q-deformation of the Lorentz group is proposed and investigated. In

this Hopf algebra the rotation group 80(3) is an authomorphism group.

Quantum deformations of the Lorentz group were considered by number of

authors [1, 1, 1]. In particular the q- Lorentz group considered by Wess et
al. [?, 1] corresponds to a q-Minkowski space-time with a non-commutative
(non-isotropic) space sector. On the other hand the deformation proposed by
Lukierski et al. [1], although isotropic, acts in a fully commutative Minkowski
space-time.
In this paper we propose another deformation of the Lorentz symmetry
acting in an isotropic but non-commutative space-time. Isotropy means that
the standard rotation group 50(3) is an authomorphism group both the q-
Minkowski space and time and the q-Lorentz group. Now, under the isotropy
condition, the hermitean coordinate generators xJ1- satisfy

(1)
for i,j = 1,2,3 and

xOx = qxxO (2)

with Iql = 1. The co-module action of q-Lorentz group reads

6(x)=A0x (3)
where the isotropy condition implies the following form for A

(4)
wher R is a rotation

(5)

277
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 277-279.
© 1993 Kluwer Academic Publishers.
278 J. REMBIELINSKI

i.e. RT R = I and Ri j belong to the center of algebra generated by AIL v.

The q-matrix Lw corresponds to a quantum boosts. It can be parametrized
as follows

(6)

with
(metric condition)
(7)
(co-linearity condition)

(8)

The co-unity is defined by

feR) = I, f(W±) = 0, f(WO) = f(a) = 1 (9)
The co-product reads
Ll(wO)=wo0 wo+w+0w_ }
Ll(w_) = w_ 0 WO + a 0 w_ + ~w_(w+ 0 w_) (10)
Ll(a) =a 0 a
The Ll(w+) can be calculated by means of the form of product

Lw 0 Lw == L.1(w)Rw (11)
where the Thomas precession Rw is given explicitly by

Rw (a-10a-l){w_0wt+

+ (a + w
°q+a w_ X wt) 0 (a + w °q+a w_ X wt) +
q
Ll(wo) + Ll(a)Ll(w-) X
X [W00wt+wt0(a+wo~aw-xw!)]} (12)

From the Eq. (11)

Ll(w+) = Rw [wo 0 w+ + w+ 0 (a + WO ~ a w_ X w+)] (13)

ISOTROPIC Q-LORENTZ GROUP 279

Finally, the antipode has the form

L~l = a- ~ I q~I ) (-:-:-t-a-+-W--:o~:-:-W"':"!-_-X-W----""!'---)

2
(
(14)

The above introduced q-Lorentz group can be extended to q-Poincare

group. Its form, properties and representations will be given in the forth-
coming papers.

Acknowledgements
I am grateful to Prof. W. Tybor, Dr. P. Kosinski and Mr. K.A. Smolinski
for interesting discussions.
This work is supported by University of L6di grant No. 505/771.

References
[1] Schmidke, W.B., Wess, J., Zumino, B.: 1991, preprint MPI-Ph/91-15; Ogievetsky,
0., Schmidke, W.B., Wess, J., Zumino, B.: 1 991, preprint MPI-Ph/91-51.
[2] Carow-Watamura, U., Schlieker, M., Scholl, M., Watamura, S.: 1990, Z. Phys. C48,
150; Carow-Watamura, U., Schlieker, M., Scholl, M., Watamura, S.: 1991, Int.
J. Mod. Phys. A6, 308!.
[3] Lukierski, J., Nowicki, A., Ruegg, H., Tolstoy, V.N.: 1991, Phys. Lett. B264, 331;
Lukierski, J., Nowicki, A., Ruegg, H.: 1992, preprint UGVA-DTP 1992/07-776.
LORENTZ ALGEBRA AND TWISTS
RALF ENGELDINGER, MICHAEL SCHLIEKER* and WOLFGANG WEICH
Sektion Physik der Universitiit Miinchen
Lehrstuhl Professor Wess, Theresienstr. 97, 8000 Miinchen 2
Federal Republic of Germany

The aim of the following paper is to classify the relation between the
deformation of SO (4) and the q-deformation of the Lorentzgroup.
The starting point of our investigation is the universal enveloping alge-
bra Uq equiped with comultiplication .6.(a) = al 0 a2 (a E Uq ), antipode S,
and counit c. This Hopfalgebra should be of standard-type described in ref.
[Drinfel'd 1986, Faddeev et al1987]. It is coassociative but not cocommuta-
tive.
It holds
a 0,6,
(1)
On the tensorproduct over C Uq 0 Uq it is possible to define a natural
Hopfalgebrastructure by the following definitions:
D a 0 id) ,6, 0 ,6,
( id 0
S S0S
e .- c 0 c
This Hopfalgebra is called A in the following.
The Hopfalgebra A has the property that the diagonal embedding
,6,: Uq -+ Uq 0 Uq , ,6,(a) = al 0 a2 (2)
is not a subbialgebra in A, because
(3)
The mapping ,6, 0 ,6, would trivially make this diagonal embedding a
subbialgebra but ,6, 0,6, is coassociative outside ,6,(Uq ). Starting from that
observation it is possible to introduce a new comultiplication which respects
the diagonal embedding by coassociative continuation of ,6, 0 ,6, in the fol-
lowing way: on ,6,( Uq ) one has the following relation between D and ,6, 0 ,6,:
D(a) = R23,6,0,6,R231, a E ,6,(Uq ) (4)
* Talk presented by M. Schlieker

281
Z. Oziewicz et at. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 281-284.
© 1993 Kluwer Academic Publishers.
282 RALF ENGELDINGER ET AL.

The proof of (4) is based on the coassociativity of 6.. Therefore one defines
the following new Hopfalgebra A by

It is easy to proove that A is indeed a Hopfalgebra.

This new Hopfalgebra A now contains 6.(Uq ) as subhopfalgebra because:

1., D(a) 6. 0 6.(a)

if a E 6.(Uq )
2., using a a 6. = R6.R- 1
and a a 6. R:;l 6. R21
it holds
S(6.(a)) 6.(s(a))
if a E 6.(Uq )
Because of the definition of a a 6. there exists a second twist by replacing
R by R 21 . These two twisted Hopfalgebras only coincide on 6.(Uq ).
Using the definition of A it is now easy to calculate the universal R-
matrix for A. in order to do that we define the following exchange operation
ao(a0b0c0d) = (c0 d 0a0 b) (6)
Then it holds:

aa 0 a( Rzl DR 23
R:;l R13 R 24R23 DR zl Rzl R131R41

Therefore the universal R-matrix of A is given by

If R is the R- matrix of S Lq (2), then R is a reparametrization of the Lorentz-

group R-matrix given in ref. [Carow-Watamura et al1990 and 1991].
If one introduces a pairing one obtains a modification of the quantum
group Hopfalgebra A'. The reason for that is the twist of the comultiplica-
tion:
i., the comultiplication on A' is unchanged:

ii., the tensor multiplication is modified in the following form:

LORENTZ ALGEBRA AND TWISTS 283

< 1,(s@t)(g@k) >

=< D(J),s@t@g@k >
= < 1, s . a( R * (t @ g) * R -1) . k >

=> (S@t).(g@k)=s.a(R*(t@g)*R- 1 ).k (7)

with the '*'-action of the universal enveloping algebra on the dual defined
by: 1 E Uq

SI < 1,S2 > (8)

< 1, SI > S2 (9)

Modified tensor multiplication laws have been considered in great detail

by Majid in refs.: [Majid 1990, Majid 1992].
Now it possible to introduce a more canonical system of generators in A.
by the following definition:

.- (id @ id @ T ac @ Tb d)R (10)

(T a c @ Tb d @ id @ id)R- 1 (11)

Using the definition of the generators of Uq : 1± in terms of the R-matrix of

Uq one obtains the following expression:

L + ab cd L bs 1+ a t @ 1+ s d
1+ t c (12)
L_ ab cd L bs La t @ LSd 1+ \ (13)

In the case Uq = Uq (su q (2)) one can explicit ely show, that the above formu-
las for L+, L_ are invertible.
The comultiplication of the L± is given by:

(14)

This has to be expected from their definition, but can also be calculated
by using identities of the form:

(15)
The complex conjugation on these twisted algebras has not yet been
investigated.
The easiest example of the above developed procedure is given by the rela-
tion between the universal enveloping algebra of the SOC 4) and the Lorentz-
group. To see that we start from

Uq = Uq (su q (2))
the untwisted Hopfalgebra A equals Uq ( SOq( 4)).
284 RALF ENGELDINGER ET AL.

The twisted Hopfalgebra A corresponds to the deformation of the en-

veloping algebra of the Lorentzgroup which can be seen from example by
defining a pairing through he generators L± and comparing this pairing with
the standard FRT-pairing using the R-matrix of [Carow-Watamura 1990 and
1991] (and a reparametrization).
The relation between this algebra and the one considered in [Drabant et
al 1992] is now under investigation.

References
Carow-Watamura, U., Schlieker, M., Scholl, M., Watamura, S.: 1990, 'Tensor representa-
tion of the quantum group SLq(2,C) and quantum Minkowski space', Z.Phys.C. 48,
159-165
Carow-Watamura, U., Schlieker, M., Scholl, M., Watamura, S.: 1991, "A quantum Lorentz
group', Int.J.Mod.Phys.A. 6, 17, 3081-3108
Drabant, B., Schlieker, M., Weich, W., Zumino, B.: 1992, Commun. Math.Phys. 174,
625-633
Drinfel'd, V.G.: 1986, 'Quantum Groups' In: Proceedings of the International Congress of
Mathematicians, Berkeley 1986, 798-820
Faddeev, L.D., Reshetikhin, N.Yu., Takhtajan, L.A.: 1987, 'Quantization of Lie groups
and Lie algebras' Algebra Analysis 1, 178
Majid, S.: 1990, J.Algebra 130, 17-64
Majid, S.: 1992, DAMTPI 92/12
ON A NONCOMMUTATIVE EXTENSION OF
ELECTRODYNAMICS

JOHN MADORE
Laboratoire de Physique Theorique et Hautes Energies'
Universite de Paris-Sud, Bat. 211, F-91405 ORSA Y
France

Abstract. The Maxwell vector potential and the Dirac spin or used to describe the classical
theory of electrodynamics both have components which are considered to be ordinary
smooth functions on space-time. We reformulate electrodynamics by adding an additional
structure to the algebra of these functions in the form of the algebra Mn of n x n complex
matrices. This involves a generalization of the notions of geometry to include the geometry
of matrices. Some rather general constraints on the reformulation are imposed which can
be motivated by considering matrix geometry in the limit of very large n. A few of the
properties of the resulting models are given for the values n = 2,3. One of the more
interesting is the existence of several distinct stable phases or vacua.

Key words: noncommutative geometry

1. Introd uction
In the usual formulation of electrodynamics the Maxwell potential and the
Dirac spinor are constructed with components which lie in the algebra C of
smooth functions on space-time. We wish to extend the construction to the
algebra A = C ® M n , where Mn is the algebra of n X n complex matrices.
The Maxwell potential is a I-form on space-time. We must therefore be
able to define differential forms on the geometric structure defined by A.
This involves generalizing the notions of geometry to include the geometry
of matrices. We give a brief review of matrix geometry in Section 2. In
Section 3 a noncommutative generalization of the Maxwell-Dirac action is
given. There are several possible generalizations, depending principally on
the structure of the spinors. At the end of Section 3 we shall make some
assumptions which reduce the possibilities to a set of models parametrized
uniquely by the integer n, a mass scale m and the analog g of the electric
charge. These can be partially motivated by considering matrix geometry in
the limit of very large n, which in a sense which can be made explicit tends
to the geometry of the ordinary 2-sphere. In Section 4 the properties of the
models are sketched for n = 2,3 .
• Laboratoire associe au CNRS.

285
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 285-298.
© 1993 Kluwer Academic Publishers.
286 J. MADORE

2. Matrix Geometry
We recall briefly here some of the details of matrix geometry (Dubois-
Violette et ai. 1989b, 1990a). For an introduction to noncommutative ge-
ometry in general we refer to the work of Connes (1986, 1990). An essential
element in differential geometry is the notion of a vector field or derivation.
It is an elementary fact of algebra that all derivations of Mn are interior.
A derivation X is therefore necessarily of the form X = ad J for some J in
Mn. The vector space Dn of all derivations of Mn is of dimension n 2 - 1.
Let Aa , for 1 :s: a :s: n 2 -1, be an antihermitian basis of the Lie algebra of
the special unitary group in n dimensions chosen with units of a mass scale
m. The product AaAb can be written in the form

A A 1 CC A 1 c 1 2
a b = 2 ab c+ 2D abAc-;;m gab· (1)

The structure constants CC ab are real and have also units of mass. The
Killing metric is given by kab = _Cc adCdbc' It is related to gab by
kab = 2nm2gab.
The tensor kadCdbc is completely antisymmetric. We shall raise and lower
indices with gab. Then C abc is also completely antisymmetric. We shall nor-
malize the Aa such that gab is the ordinary euclidean metric in n 2 - 1 di-
mensions.
The set Aa is a set of generators of Mn. It is not a minimal set but it is
convenient because of the fact that the derivations

form a basis over the complex numbers for Dno Any element X of Dn can
be written as a linear combination of the e a: X = xa ea , where the xa
are complex numbers. The vector space Dn has a Lie-algebra structure. In
particular the derivations e a satisfy the commutation relations

We define differential forms on Mn just as one does in the commutative

case (Dubois- Violet.te 1988). For each matrix J we define the differential of
J by the formula
(2)
This means in particular that
(3)
We define the set of I-forms n1 (Mn) to be the set of all elements ofthe form
Jdg or the set of all elements of the form (dg)J, with J and g in Mn. The
ON A NONCOMMUTATIVE EXTENSION 287

two definitions coincide because of the relation dUg) = f(dg) + (df)g. The
p-forms are defined exactly as in commutative case (Dubois-Violette et al.
I990a) with the product given as usual. The set of all differential forms is a
differential algebra.
There is a basis ()a of the I-forms dual to the derivations ea :
(4)
We have here suppressed the unit matrix which should appear as a factor
of bi: on the right-hand side. The ()a are related to the d>.a by the equations
d>.a = cabc >.b()c, (5)
and their inverse
(6)
They satisfy the same structure equations as the components of the Maurer-
Cartan form on the special unitary group SUn:

d()a = -~cabC ()b()c. (7)

2
The product on the right-hand side of this formula is the product in the
algebra of forms. Using the ()a the exterior derivative can be written as
df = ea()a. We shall consider the ()a as the analog of a moving frame. They
constitute a set of n 2 -1 elements each of which is an (n 2 - 1) X n 2 matrix.
Each ()a takes in fact D n , of dimension n 2 - 1, into M n , of dimension n 2 •
The interior product and the Lie derivative are defined as usual.
From the generators ()a we can construct a I-form () in nl(M) which will
play an important role in the study of gauge fields. We set
() = ->'a()a.
From equation (5) we see that it can be written in the forms

() = -~>'ad>.a
nm
= ~d>'a>.a.
nm
Using () we can rewrite equation (6) as
()a = m-4Cabc>.bd>'c _ nm-2 >.a(). (8)
From equations (5) and (7) one sees that () satisfies the zero-curvature con-
dition:
d() + ()2 = O. (9)
It satisfies with respect to the algebraic exterior derivative the same condi-
tion which the Maurer-Cartan form satisfies with respect to ordinary exterior
derivation on the group SUn.
288 J. MADORE

We shall introduce a metric on Dn by the requirement that the frame ea

be orthonormal. For X = Xa ea and Y = ya ea we define
g(X, Y) = gabxayb.
To within a rescaling g(X, Y) is the unique metric on Dn with respect to
which all the derivations e a are Killing derivations. The first structure equa-
tions for the frame ea and a linear connection Wab can be now written down:
doa + Wab ob = ea.

We see then that if we require the torsion form e a to vanish then the internal
structure is like a curved space with a linear connection given by

(10)

The second structure equation defines the curvature form nab, which satisfies
the Bianchi identities as before.
The complete set of all derivations of Mn is the natural analog of the
space of all smooth vector fields D(V) on a manifold V. If V is parallizable
then D(V) is a free module over the algebra of smooth functions with a
set of generators eO' which is closed under the Lie bracket and which has
the property that if eO'! = 0 for all eO' then! is a constant function. The
matrix algebra Mn has in general several Lie algebras of derivations D with
this property. The smallest such one, D 2 , is obtained by considering three
matrices Aa which form the irreducible n-dimensional representation of SU2 •
These matrices generate the algebra Mn. The most general element of Mn
is a polynomial in the Aa. The equations

imply that ! is proportional to the unit element. The set D2 could also
be considered as the natural counterpart of a moving frame on a manifold
(Madore 1991).
With a restricted set of derivations, one can define the exterior differ-
ential exactly as before using equation (2). However now the set of e a is a
basis of D ~ Dn. The derivations are taken, so to speak, only along the
preferred directions. Equation (3) remains valid, the only change being that
the structure constants are those of the algebra of derivations. A difference
lies in the fact that the forms are of course multilinear maps on the pre-
ferred derivations and are not defined on all elements of Dn. The formula
(4) which defines the dual forms is as before but the meaning of the expres-
sion ea changes. If we choose for example D2 as the derivations then oa is
a 3 X n 2 matrix. It takes the vector space D2 into Mn and it is not defined
on the n 2 - 4 remaining generators of Dn. Equation (5) remains unchanged
but equations (6) and (8) will have to be modified.
ON A NONCOMMUTATIVE EXTENSION 289

Consider the case D = D2 ~ Dn. Let e r for 1 ::; r ::; n 2 - 1 be a basis

of Dn and let e a for 1 ::; a ::; 3 be a basis of D 2. We can choose the e a
to be the first 3 elements of the er . Then the SU2 structure constants cabc
are the restriction of the SUn structure constants C r st. Equation (1) will be
therefore written

(11)

If a basis J a of D2 satisfies the commutation relations [Ja, Jbl = 2ifabcJc

then JaJa = n 2 - 1. On the other hand, from equation (11) we see that
AaAa = -3m 2 In. If we write then
l
Aa = -2rJa,
we find that r is related to m by the equation 12m 2r2 = n( n 2 - 1) and that
the SU2 structure constants are given by

Let (r be the dual basis of er and let oa be the dual basis of ea. We have
then 2 possible expressions for oa. We have
o'a = m- 4 C arsAr dA S _ nm -2 Aao',

with 0' constructed using Ar and we have

(12)

with 0 constructed using Aa • Both definitions satisfy the equation (4). That
is, they coincide as 3 X n 2 matrices. In equation (10) each of the Ar can
be expanded as a polynomialin terms of the 3 elements ).a and using the
Leibnitz rule this yields a long complicated expression for O'a in terms of
the d).a. In equation (12) the expression d).a is a 3 X n 2 matrix but it has
a natural extension to an (n 2 - 1) X n 2 matrix in which case it coincides
with the definition of d).r for the first three values of the index r. The two
expressions for oa can be compared therefore as forms on the complete set
of derivations. Whereas by construction O'a (e r ) = 0 for r 2: 4, in general the
corresponding equation for oa would not be satisfied.
Using the basis of D2 and its dual we can write the differential of a matrix
! as
(13)
The complete differential is given by dl = erlO r • If dl( ea) = 0 then ea! = o.
This means that! is proportional to the unit element and therefore that
d! = O. However if a is a general I-form then the condition a(e a ) = 0 does
290 J. MADORE

not imply that a = O. For example any basis element (r for r 2 4 satisfies the
equation BT(e a ) = O. When we consider the restricted set D2 of derivations
we shall choose the algebra of forms to be the differential algebra generated
by the forms (12). In this case if a is a 1-form which satisfies the condition
a(e a ) = 0 then a = O.
Using the 1-form B we can write the differential of a matrix f as
df = -[B,f]·
If we consider the algebra of all forms as a Z2-graded algebra then we can
define another d acting on any form a by the formula (Connes 1986, 1990)
da = -[1], a], (14)
where 1] is some 1-form and the bracket is Z2-graded. See also Quillen (1985)
and Dubois-Violette et aZ. (1991). If 1]2 = -1 we have d2 = O. Equation (9)
becomes
d1] + 1]2 = l.
The definition (14) is interesting in that it does not use derivations and thus
can be used when considering the case of more abstract algebras which have
none.
We shall now consider an extension of matrix geometry by considering
the algebra of matrix-valued functions on space-time (Dubois- Violette et aZ.
1989a, 1989b, 1990b). Let x/.L be coordinates of space-time. Then the set
(x/.L, Aa) is a set of generators of the algebra A which is the tensor product
(15)
of C the algebra of smooth real-valued functions on space-time and Mn. The
tensor product is over the complex numbers. Let eo: = e~a/.L be a moving
frame on space-time and ea with 1 ::; a ::; 3 a basis of D 2 • Let i = (a,a).
Then 1 ::; i ::; 7. We shall refer to the set ei = (eo:, ea ) as a moving frame on
the algebra A-
For f E A we define df by equation (2) but with the index a replaced by
i. Choose a basis BO: = B~dx>' of the 1-forms on space-time dual to the eo:
and introduce Bi = (BO:,Ba) as generators of the 1-forms !V(A) as a left or
right A-module. Then if we define

we can write Ul(A) as a direct sum:

U1 (A) = Uk EB U~.
The differential df of a matrix function is given by
ON A NONCOMMUTATIVE EXTENSION 291

We have written it as the sum of two terms, the horizontal and vertical parts,
using notation from Kaluza-Klein theory. The horizontal component is the
usual exterior derivative dH I = eaIBa. The vertical component dv, given
by equation (13), is purely algebraic and it is what replaces the derivative
in the hidden compactified dimensions. The algebra W(A) of all differential
forms is defined as usual. It is again a differential algebra.

3. The Maxwell-Dirac action

We shall now write down the analog of the Maxwell-Dirac action in the
geometry defined by the algebra (15). We shall identify a connection with
an anti-hermitian element w of nl(A). We saw above that it can be split as
the sum of two parts which we called horizontal and vertical. We write then

w=A+wv, (16)

where A is an element of n}, and Wv is an element of n~.

In Section 2 we introduced a 1-form B in nl(M) ~ n~. We shall use this
1-form as a preferred origin for the elements of n~. We write accordingly

Wv = B + 1>. (17)

The field 1> is the Higgs field.

We have noted previously that B resembles a Maurer-Cartan form. For-
mula (3.1) with 1> = 0 is therefore formally similar to the connection form
on a trivial principal U1-bundle. We have in fact a bundle over a space which
itself resembles a bundle. This double-bundle structure, which is what gives
rise to a quartic Higgs potential as we shall see below, has been investigated
previously, by Manton (1979), Hamad et at. (1980), Chapline and Manton
(1980), and, more recently, by Kerner et al. (1987) and by Coquereaux and
Jadczyk (1988). The A-modules which we shall consider are the natural
generalization of the space of sections of a trivial Urbundle since Mn has
replaced C in our models. So the Un gauge symmetry comes not from the
number of generators of the module, which we shall always choose to be
equal to 1, but rather from the factor Mn in our algebra A.
Let Un be the unitary elements of the matrix algebra Mn and let Un be
the group of unitary elements of A, considered as the algebra offunctions on
space-time with values in Mn. We shall choose Un to be the group of local
gauge transformations. A gauge transformation defines a mapping of nl(A)
into itself of the form

We define
292 J. MADORE

0' = g-10g + g-ldVg,

A' = g-1 Ag + g-ldH9,
and so ¢ transforms under the adjoint action of Un:

It can be readily seen that in fact () is invariant under the action of Un:

0' = O.
Therefore the transformed potential w' is again of the form (17).
The fact that 0 is invariant under a gauge transformation means in par-
ticular that it cannot be made to vanish by a choice of gauge. We have then
a connection with vanishing curvature but which is not gauge-equivalent
to zero. If Mn were an algebra of functions over a compact manifold, the
existence of such a I-form would be due to the non-trivial topology of the
manifold.
We define the curvature 2-form n and the field strength F as usual:

n=dw+w 2, F=dHA+A2.
In terms of components, with ¢ = ¢aoa and A = A"O" and with

we find

n",8 = F",8, n"a = D,,¢a, nab = [¢a, ¢bj- CC ab ¢c. (18)

The analog of the Maxwell action is given by

SB = J LB, (19)

where

LB = ~Tr(F",8F",8)
4g
+ ~Tr(D,,¢aD,,¢a) -
2g
V(¢). (20)

The Higgs potential V( ¢) is given by

V(¢) = -~Tr(nabnab).
4g
(21 )

It is a quartic polynomial in ¢ which is fixed and has no free parameters

apart from the mass scale m. The trace is the equivalent of integration on the
ON A NONCOMMUTATIVE EXTENSION 293

matrix factor in the algebra. The constant 9 is the gauge coupling constant.
We see then that the analog of the Maxwell action describes the dynamics
of a Un gauge fields unified with a set of Higgs fields which take their values
in the adjoint representation of the gauge group.
The lagrangian (20) is the standard lagrangian chosen for all gauge the-
ories which use the Higgs mechanism. Given a gauge group the theories
differ according to the representation in which the Higgs particles lie and
the form of the Higgs potential. The particular expression to which we have
been lead has been also found by slightly different, group theoretical, con-
siderations in the context of dimensional reduction by Hamad et at. (1980)
and by Chapline and Manton (1980). What our formalism shows is that the
Higgs potential is itself the action of a gauge potential on a purely algebraic
structure. The nab are in fact the components of the curvature nv of the
connection (17):

nv = dw v + w~ = ~nab(ia A (ib.
The connection determines a covariant derivative on an associated A-
module (Connes 1986, 1990). See also Dubois-Violette et af. (1991). Let H be
a Mn-module. It inherits therefore a Un-module structure. Define 1i = C0H.
Then 1i is an A-module as well as a Un-module. The form of the covariant
derivative depends on the module structure of H. The covariant derivative
of '¢ E 1i is of the form

D'¢ = d'¢ +w'¢.

The action of won '¢ is determined by the action of Un on H. We have only
then to define the vertical derivatives ea ,¢ of '¢. Since 1i is a A-module, for
any J in A we must have the relation

ei(f,¢) = (e;j)'¢ + Jei'¢. (22)

Suppose that H is a left module. We shall consider only the case H = en.
From equation (22) we see that we must set

The action of Un can only be left multiplication. We find then that

Suppose that H is a bimodule. We shall consider only the case H = Mn.

From equation (22) we see that we must set
294 J. MADORE

There are now two possibilities for the action of Un. We can choose H to be
a bimodule with the adjoint action or a left module with left multiplication.
We find then in the first case

This is invariant under the adjoint action of Un. In the second case we find

(23)

This is invariant only under the left action of Un.

With the frame (Ji which was introduced above the geometry of the al-
gebra A resembles in some aspects ordinary commutative geometry in di-
mension 7. As n -+ 00 it resembles more and more ordinary commutative
geometry in dimension 6 and the frame ()i becomes a redundant one in the
limit. Let gkl be the Minkowski metric in dimension 7 and "/ the associated
Dirac matrices which we shall take to be given by

,../ = (1 ® '"(", (fa ® ,.l).

The space of spinors must be a left module with respect to the Clifford
algebra. It is therefore a space of functions with values in a vector space P
of the form

The Dirac operator is a linear first-order operator of the form

where Dk is the appropriate covariant derivative, which we must now define.

The space-time components are the usual ones:

The wOP'"Y are the coefficients of a linear connection defined over space-time:

W a f3-
- W a f3(J'"Y
'"Y •

By analogy we have to add to the covariant derivative given above a term

which reflects the fact that the algebraic structure resembles a curved space
with a linear connection given by equation (10). We make then the replace-
ment (Madore 1989)

(24)
ON A NONCOMMUTATIVE EXTENSION 295

The analog of the Dirac action is given by

(25)

where

We have therefore defined a set of theories which are generalizations of

electrodynamics to the algebra A. In order to restrict the generality we
shall make three assumptions. First, we shall suppose there is no explicit
mass term in the classical Dirac action. We have already supposed that the
derivations to be used are the algebra D 2 • Last, we shall suppose also that
H has the module structure which leads to the covariant derivative defined
by equation (23) to which we add the curvature term as in (24). The last
two assumptions can be motivated by showing that in the limit for large n
the covariant derivative tends in a sense which can be made explicit to that
used in the Schwinger model (Grosse & Madore 1992). With the restrictions
we have a set of classical models which for each integer n depend only on the
coupling constant 9 and the mass scale m and given by the classical action

(26)

where SB is defined by equation (19) and SF is defined by equation (25).

Different restrictions result in different models (Dubois-Violette et al.
1989a, 1989b, 1990, 1991, Madore 1989, 1991, Balakrishna et al. 1991a,
1991b). If one uses the exterior derivative (14) one obtains yet different
models (Connes & Lott 1989, Coquereaux et al. 1991) but which are similar
at least in the bosonic sector. The main difference lies in the form of the
Higgs potential which is in fact closer in form to that used in the standard
electroweak model.

4. Models
We shall now consider the action (26) in the case n = 2 (Dubois-Violette
et al. 1989b, 1990b) and examine the resulting physical spectrum. The la-
grangian (20) is a generalization of the Yang-Mills-Higgs-Kibble lagrangian,
with a more elaborate Higgs sector. The most original part is the poten-
tial term V( 1» which comes from the curvature of the vertical part of the
connection. It is not the most general gauge-invariant polynomial in the
Higgs field which would be allowed and there is no reason to suppose that
its form remains invariant under renormalization effects. The fermions are
Dirac fermions which take their values in the space M2 ® (2 and the gauge
group is U2 • There are therefore four U2 doublets.
296 J. MADORE

From equation (21) and the definition (18) of !lab we see that the vacuum
configurations are given by the values J.la of <Pa which satisfy the equation

(27)

The number of solutions to this equation is given by the partition function,

the number of ways one can partition the integer n into a set of decreasing
positive integers. Two obvious solutions are J.la = 0 which corresponds to
the partition (1, ... ,1) and J.la = Aa which corresponds to the partition (n).
If n = 2 there are no others. Matter can exist then in two phases. In the
symmetric phase all the gauge bosons are massless and three of them are
gluon-like. The fermions are quark-like. In units of (1/2V2)m there are two
doublets of mass 3 and two of mass 5. We call this phase the hadronic phase.
We shall suppress in an ad hoc way the U1 component of the gauge group and
reduce it to SU2. There is no photon then and the fermions are all neutral.
In the broken phase, the gauge bosons are all massive if we suppress the U1
component. The fermions are again neutral but of different masses. There
are now two doublets of.mass 5, a doublet of mass 7 and a doublet of split
mass 5 and 7 units. We call this phase the third phase, for reasons to be
made clear below.
In the case n = 3 the fermions are Dirac fermions which take their values
in the space M3 @ (2 and the gauge group is U3 . There are therefore six U3
triplets. Matter can exist now in 3 phases corresponding to the 3 partitions
of 3. In the symmetric phase all the gauge bosons are mas8less and eight of
them are gluon-like. The fermions are quark-like. In the units given above
they have all masses of the order of one. This is the hadronic phase. Since we
do not wish to interpret the U1 component of the gauge group as the photon,
the fermions are neutral. In the broken phase which corresponds to the n = 2
case the gauge bosons are all massive if we suppress the U1 component. The
fermions are then again neutral and again of different masses. But in the
units given above they still have masses of the order of one. This is the third
phase.
The extra phase for n = 3 we call the leptonic phase. It is given by the
solution to the equation (27) of the form

As we shall see below, in this phase there are two massless gauge modes. We
must identify one of the corresponding fields with the photon and again in
an ad hoc way suppress the other mode. Define the matrices 1\;4 and 1\;5 by
o
o
o
ON A NONCOMMUTATIVE EXTENSION 297

We write the gauge potential in the form

A = A 4 ~4 + A 5 ~5 + (0
w+ w-)
z .
Here A4 and A 5 are ordinary I-forms, W+ is a I-form with values in (2,
W- = -(W+)* and Z is a I-form with values in the Lie algebra of SU2 • In
this phase there are therefore 2 charged gauge bosons and 3 neutral ones.
Their masses are given by

mz=
2 4 m.
2

There are two massless bosons. We shall set

and choose A4 to represent the photon then the unit of charge is given by
e = g. All of the 6 triplets of fermions have again masses of the order of m
and these masses are again different from the corresponding masses in the
hadronic and the third phases. Two triplets have charge 1 and the other 4
are neutral. We write then the spinor field in the form

Here, e, f.L and T are charged doublets; each v and I is a neutral doublet.
The coupling of the Higgs field to the fermions is not constrained by
gauge invariance and so there is no reason why the corresponding coefficient
in equation (23) for example should be equal to 1. We could have for any
real number x:

Da1/J = x4>a1/J -1/JAa - ~Cbcalb/c1/J.

The same argument applies to the curvature terms. Under an arbitrary
change of frame,
(Y" 1-+ ()'C1I = Ap(}{3 ,
the spinor 1/J transforms to 1/J' = S-l(A)1/J and the Dirac matrices transform
as

The space-time components D(JI1/J of the covariant derivative of 1/J have been
constructed so that they transform correctly. The same behaviour must be
required of the algebraic components Da 1/J. The covariant derivative we have
298 J. MADORE

used transforms as it should. But in fact each term transforms correctly and
we could have more generally for any real numbers x, y

Da'IjJ = X<pa'IjJ - 'ljJAa - ~yCbcalb/C'IjJ. (28)

There is no way then to fix the renormalized values of the masses of the
fermions. They will depend on the mass scale m and the two parameters x
and y.

5. Conclusions
We have presented some of the details of the simplest noncommutative ex-
tension of electrodynamics, which we have completed in an ad hoc way by
suppressing the abelian component of Un gauge potentials. Even with this
modification none of the models we have presented has the correct phe-
nomenology. There are in general unobserved particles and the mass spec-
trum is too rigid. There is only one mass parameter m and all of the particles
either have mass zero or a mass or order m. Renormalization effects could
however within the context of the model introduce a modification of the
mass spectrum according to equation (28).

Acknowledgements
Much ofthe content of this article was obtained in collaboration with Michel
Dubois-Violette and Richard Kerner.

References
Balakrishna B. S., Giirsey F., Wali K. c.: 1991a, Phys. Lett. B254 430.
Balakrishna B. S., Giirsey F., Wali K. c.: 1991b, Phys. Rev. D44 3313.
Chapline G., Manton N. S.: 1980, Nucl. Phys. B184 391.
Connes A.: 1986, Publications of the I.H.E.S. 62 257.
Connes A.: 1990, Geometrie noncommutative', InterEditions: Paris
Connes A., Lott J.: 1989, Nucl. Phys Proc. Suppl. B18 29.
Coquereaux R., Esposito-Farese G., Vaillant G.: 1991, Nucl. Phys. B353 689.
Coquereaux R., Jadczyk A.: 1988, World Scientific Lecture Notes in Physics 16.
Dubois-Violette M.: 1988, C. R. Acad. Sci. Paris 307 Serie I 403.
Dubois-Violette M., Kerner R., Madore J.: 1989a, Phys. Lett. B217 485.
Dubois-Violette M., Kerner R., Madore J.: 1989b, Class. Quant. Grav. 6 1709.
Dubois-Violette M., Kerner R., Madore J.: 1990a, J. Math. Phys. 31 316.
Dubois-Violette M., Kerner R., Madore J.: 1990b, J. Math. Phys. 31 323.
Dubois-Violette M., Kerner R., Madore J.: 1991, Class. Quant. Grav. 81077.
Grosse H., Madore J.: 1992, Phys. Lett. B283 218.
Harnad J., Shnider S., Tafel J.: 1980, Lett. in Math. Phys. 4 107.
Kerner R., Nikolova L., Rizov V.: 1987, Lett. in Math. Phys. 14333.
Madore J.: 1989, Mod. Phys Lett. A 4 2617.
Madore J.: 1991, Int. Jour. of Mod. Phys. A 6 1287.
Manton N.S.: 1979, Nucl. Phys. B158 141.
Quillen D.: 1985, Topology 24 89.
BICOVARIANT DIFFERENTIAL CALCULUS
AND
q-DEFORMATION OF GAUGE THEORY
SATOSHI WATAMURA
Department of Physics
College of General Education
Tohoku University
Kawauchi, Aobaku, Sendai 980, JAPAN

Abstract. The q-deformation of the BRST algebra, the algebra of the ghost, matter
and gauge field on one spacetime point is constructed using the result of the bicovariant
differential calculus. We define the covariant commutation relation among the fields and
their derivatives consistently with the two nilpotent operation the spacetime derivative
and the BRST operation.

1. Introd uction
It is an interesting question whether one can construct a q-analogue of the
gauge theory by taking the quantum group [Drinfeld 1986, Reshetikhin 1986,
Jimbo 1986, Woronowicz 1987] as a symmetry. One of the interesting possi-
bilities of such a q-deformed theory is that the deformation parameter q may
play the role of a regularization parameter. Furthermore, since the quantum
group is provided by a noncommutative algebra, in such a theory the non-
commutative geometry plays a basic role like the differential geometry in
the usual gauge theory.
There are some proposals to this problem [Aref'eva 1991, Bernard 1990,
Hirayama 1992, Isaev 1992, Wu 1992]. However, it seems that there are still
conceptual problems concerning the definition of the gauge transformation
when we take the quantum group as an algebraic object of the gauge sym-
metry. Since the quantum group is formulated in the language of the Hopf
algebra, it forces us to formulate the whole theory in an appropriate algebraic
language [Brzezinski 1992]. Therefore, the gauge transformation has to be
represented in this abstract language and the notion of the transformation
parameter becomes obscure. Even when we consider only the infinitesimal
transformation, we still have to clarify the definition of the infinitesimal
parameters.
One of the alternative formulations of the gauge theory is given by the
BRST formalism [Becchi 1976, Tyuitin 1975]. There, the gauge transfor-

299
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 299-308.
© 1993 Kluwer Academic Publishers.
300 SATOSHI WATAMURA

mation parameter is replaced by the ghost field and becomes an object of

equal level with the matter and the gauge fields. Therefore, when we con-
sider the q-deformation of the gauge theory, it is very natural to consider
the q-deformed field algebra starting with the BRST formalism.
Here we report recent results about q-deformation of the BRST algebra
which is the algebra of the gauge fields, the ghost fields and appropriate
matter fields on one spacetime point [Watamura 1992]. The gauge trans-
formation of the theory is replaced by the BRST transformation which is
represented by a nilpotent "differential operator" 8B.
For the notation concerning the Hopf algebra [Abe 1980], we take: the
coproduct ~, the antipode K, and the counit e. Through this talk the upper
case roman index I, J, J(, L runs 0, -,3, + and the lower case index like
a, b, c, d runs over the label of the adjoint representation, -,3, +, otherwise
we specify explicitly.

2. Bicovariant Differential Calculus

Before we start to construct the BRST algebra, let us briefly recall some re-
sults of the bicovariant differential calculus [Woronowicz 1989, JUfCO 1991,
Carow-Watamura 1991a]. The one forms are defined by the right invari-
ant bases OJ (i,j = 1,2) where oj* = of. Using the spinor metric fkl =

( q2°1 _q-2),
0
we define Oij = 0ifkj then they have the commutation rela-
tion
aO ij = Okl(a * L~) (1)
for Va E Funq(SU(2)). L is the functional Funq(SU(2)) -+ C defined by
(2)

where the functionals 1 L± are defined using R-matrix as L~j(Mn = R±i7

for generators M] of Funq(SU(2)), and the convolution product is (f*g) =
(f 0g)~.
The right invariant basis Oij can be split into two parts according to its
transformation property, the adjoint representation oa (a = -,3, +) and
the singlet 00 , by using the q-Pauli matrices and 0'&
where 0'1
= 8) 0'&0'1
0 ---'L +
(1=0,-,3,+ ) andO'kl=.,;Qfk['O'kl= (00-1
0) 'O'kl=
- (1000) 'O'kl=
3

.Jb (q~~ q;) with Q = q+ q-l and fkl = _fkl. The projectors can be

1 The functionals f± appearing in ref.[Carow-Watamura 1991a] are equivalent to the

L± in ref.[Reshetikhin 1989] which we use here. Thus, L;:I is equivalent to the functional
f;:d kl 0 Ii in ref.[Carow-Watamura 1991a].
Q-DEFORMATION OF GAUGE THEORY 301

.
WrItten Akl = O"OO"kl'
as pij ij 0 pij ij a h
Ski = O"a O"kl were S
p (PA ).IS th e proJec
. t or lor .j.'

the q-(anti)symmetric product and R = q"iPs - qTPA.

- 1 -3

The q-deformed exterior derivative d is defined as a map from Funq(G) to

the bimodule defined with the basis () requiring that the nilpotency and the
Leibniz rule hold in the standard way [Woronowicz 1989]. Such an operation
can be defined simply as a commutator with the singlet component ()o as
[Carow-Watamura 1991a]
zg 0
da = -[()
w
,a]_ (3)

for any element a E Funq(SU(2)), where w = q - q-l, i = A and g is a

non-zero real constant. 2 Since da is an element of the bicovariant bimodule,
we can expand it with the basis as
(4)
where the right invariant vector field Xij is given by

(5)
One of the suggesting relation given by the bicovariant differential calcu-
lus is the q-analogue of the Maurer-Cartan equation. We gave the expression
in a more familiar form in ref.[Carow-Watamura 1991a]:

(6)

where /\ is the q-deformed exterior product. The fbc is the q-analogue of the
structure constants. Using the general formula for the structure constants
in ref.[Carow-Watamura 1991a] (See also ref.[Carow-Watamura 1992]), we
obtain them for the Funq(SU(2)) as
rt3 = f 3- = q, ft+ = f~3 = _q-l
f~- = - f~+ = 1, fl3 = q - q-l (7)
The commutation relation of the right invariant basis are [Xo, Xa] = 0
and

(8)

where the functional p is p = igE - WXo and PAa! cd = q2;':-2 f~~f:,b, where
fg c == - fbc is the projector onto the antisymmetric product of two adjoint
representations. (see also section 5 of ref. [Carow-Watamura 1991b].)
2 The relation of the constant 9 with the constant No in ref.(Carow-Watamura 1991aJ
.
IS 9 =~
qNo .
302 SATOSHI WATAMURA

3. Gauge Transformation and BRST Formalism

As we explained in the introduction, we need to represent the gauge theory
using an appropriate algebra language which fits to the Hopf algebra struc-
ture. Thus, let us first reconsider the gauge and the BRST transformation
in the non-deformed theory.
When we consider the usual non-deformed gauge theory with a symmetry
group SU(2), the matter like a lepton is represented by the field which is
the section of the associated fiber bundle of the structure group SU(2) with
the spacetime as a base manifold B. Thus the algebra of the matter fields is
the algebra of all possible sections.
Giving the SU(2) valued function g(x) E SU(2) on the base manifold
B 3 x, when the matter is of the fundamental representation the gauge
transformation of the matter ijii( x) can be written as

(9)

where (i,j = 1,2). We wrote the gauge transformation matrix as Mj(g(x))

to clarify the algebraic structure. The matrix element MJ maps the g( x)
to the complex valued function on the base manifold and thus pointwise
Mj is an element of the Fun(SU(2)). Therefore, the gauge transformation
property of the matter field can be translated into the algebraic language
such that the algebra of matter fields is the (left )comodule algebra, and
there is a pointwise (left)coaction b..L of Fun(SU(2)) on the field iji:

(10)
s

where Ts E Fun(SU(2)) are matrix elements of the representation corre-

sponding to the matter iji. For the fundamental representation eq.(10) is
b..L(ijii) = Mj ® ijij and with the corresponding argument we get eq.(9).

The infinitesimal transformation corresponding to the transformation (9)

can be written as
(11)

where a = -,3, + is the label of the adjoint representation of SU(2) and e is

the gauge parameter which is the real function of the spacetime and Xa(Mj)
is a 2 X 2 matrix. In the non-deformed case we can identify Xa with the right
invariant vector fields and the evaluation Xa(Mj) gives the Pauli matrix for
the SU(2) case; thus eq.(l1) is the familiar infinitesimal transformation. In
general, the infinitesimal transformation be of the matter field iji can be
represented by the vector fields Xa and the infinitesimal parameter ~a as

(12)
Q-DEFORMATION OF GAUGE THEORY 303

where (. * .) denotes the convolution product of a comodule element with a

functional. The infinitesimal transformation of the quantum group coaction
is also investigated in section 5 of ref. [Carow-Watamura 1991b].
U sing the above algebraic formulation, we may consider the q-analogue
of the finite and the infinitesimal gauge transformation which we will discuss
elsewhere. 3
Here we want to concentrate on the q-deformation of the BRST algebra
which seems the most appropriate algebra to consider the q-deformation of
the gauge theory.
The BRST transformation of the matter field is defined by replacing the
gauge parameter e e
by the ghost field a [Faddeev 1967]. Thus the BRST
transformation can be written as

(13)
To define the q-deformed BRST algebra we extract appropriate properties
from the non-deformed BRST formalism and impose them as the condition.
We also require that in general under the limit q -7 1 the algebra always
reduces to the non-deformed one.
The BRST algebra is the algebra which contains the matter fields q; and
e
the gauge fields AI and the ghosts I which are the standard field contents of
the BRST formalism. The suffix I corresponds to the adjoint representation
in non-deformed case. However in the q-deformed case we only require that
it contains the adjoint representation and allow to add a singlet component
like the right invariant basis (}~ in the bicovariant differential calculus.
We also have to consider the spacetime derivative. In the BRST algebra
we introduce the spacetime derivative d as the formal mapping:
d
--->
d
---> o. (14)
The fields dq;, dA I and del must be treated as independent generators from
the original fields.
Definition 1: The BRST algebra AB is a comodule algebra over Fun q ( G)
which is generated by the following set of fields:

(15)
where eI represents the ghost, q; the matter and AI the gauge fields. I is
a set of the covariant commutation relations among these comodules, which
we shall determine in the next section.
In the non-deformed BRST formalism of the gauge theory, the exterior
derivative d and the BRST transformation {j B are nilpotent operators. Thus
3 The global transformations are considered in ref. [Brzeziiiski 1992].
304 SATOSHI WATAMURA

°
we require the nilpotency and also the Leibniz rule for each operator. Fur-
thermore we require the following properties: db B + bBd = and under the
* conjugation bB 0 * = * 0 bB' and do * = * 0 d.
In order to define the properties of the ghost in the BRST algebra we
identify them with the right invariant one-form () as a comodule. In the q-
deformed case, the result of the bicovariant bimodule calculus says that the
number of independent bases of the invariant one-forms is 4 for the calculus
on Funq(SU(2)). They include both the adjoint and singlet representation.
Therefore, in the q-deformed BRST algebra, we introduce the four ghosts
C I where the suffix I runs 0, -,3, +.
Definition 2: In the q-deformed BRST algebra based on the bicovariant
differential calculus on Funq(SU(2)), we define the ghost field as a comodule
represented by a 2 x 2 matrix C i j. The left-coaction on it is

(16)

and under the *-conjugation it transforms as a hermitian field: (Cij)* = C j i.

From the properties of the ghosts in the non-deformed case we also require
that they are q-anticommuting and that the BRST transformation bB of the
ghosts has the same form as the Maurer-Cartan equation obtained by the
bicovariant differential calculus.

(17)

where 9 is an arbitrary non-zero real constant. Note that we also decompose

the ghost fields into singlet and adjoint representation as C I = (J'~ i2 C il i2
where I = 0, -,3, +.
Then we define the q-deformed BRST transformation of the matter anal-
ogously to eq.(13) as
bBw = CI(W * XI) = Ca(w * Xa) + CO(W * Xo) (18)
where XI E Uq(SU(2)) is the one given in eq.(5). Note that the last term
does not have a corresponding term in the non-deformed case, and it goes to
zero in the limit q -+ 1. The singlet component of the ghost is not desirable
from the physical point of view. On the other hand, as we shall discuss it
seems it is necessary to include it in order to put the algebra in a simple
form.
Finally we require the existence of the covariant derivative which is rep-
resented by the derivative d and the gauge field A I. The coupling of the
gauge field to the matter field is determined naturally by the structure of
the BRST transformation of the matter field given in eq.(18). Therefore, our
requirement concerning the covariant derivative is:
Q-DEFORMATION OF GAUGE THEORY 305

There exists a covariant derivative '\7 which acts on the matter as

(19)

where (A~)* = Ai. The covariant derivative transforms with the same rule
as the corresponding matter

(20)

and '\7 0 * =* 0 '\7

Requiring the above conditions and the covariance, we define the co-
module algebra AB. The main part of the construction is to define the
commutation relations I.

4. The BRST Algebra

Here we give the commutation among the elements and the BRST trans-
formation of the gauge fields and other relations without proof. The com-
mutation relation of each field among itself can be defined by taking the q-
antisymmetric (q-symmetric) product to vanish if it is a bosonic(fermionic)
field in the limit q -+ 1.
The ghosts are q-anticommuting fields by definition. The gauge fields are
also q-anticommuting since they are I-forms in the limit of q -+ l. We define
the q-anticommutation relation of these field using the same formula used
to define the 1\ product in ref.[Carow-Watamura 1991a]:

(21)
(22)

The pair ~f projectors (P, Q) with P~1 and Q~l is given by (P, Q)~lik:l!?2 =
R- i 2 jl pilj; Qi~j2 Rl~ k~
j;i; kll~ k;12 k2 1l '
The other relations including the derivative of the fields have to be also
defined. Since the operation d relates some of the relations, they are not all
independent, i.e. some of them can be obtained from others by the oper-
ation d. The independent commutation relations are the ones between the
following pairs: ({CI},{dC I }), ({AI}, {dAI}), ({CI},{AI}), ({C I },{1l1}),
({AI}, {1l1}), ({1l1},{1l1}), and ({1l1},{d1l1}). When we require the consis-
tency with other structures, we can fix all these relations. For the derivation,
we refer to the paper [Watamura 1992]. The resulting relations except the
({1l1}, {1l1}) and the ({1l1}, {d1l1}) relations are given by
Proposition 1: Define the ordering of the fields as

(23)
306 SATOSHI WATAMURA

then if X > Y l, the commutation relation is given by

Xyl = ±yJ(X*L)) (24)
where the sign is taken as +(-) if they are commuting(anticommuting) in
the limit q --+ 1 and L) is the functional defined in eq.(2). Note that we take
the I-forms and the ghosts anticommuting with each other.
The relations of dAl and dC l are
(Ps, PA)J/LdAKdA L = 0, and (PA, Ps)J/LdAKdA L = °
°
(25)
(Ps, PA)J/LdCKdCL = 0, and (PA, PS)J/LdCKdCL = (26)
which simply mean that dA I and dC I are q-commuting fields.
The algebra of the matter fields can be defined like a quantum plane,
since the quantum plane algebra is the algebra generated by the comodule
imposing an appropriate commutation relation [Manin 1988]. The algebra
depends on the representation of the matter fields in the model. In our
construction, we do not need to specify the representations of the matter.
The algebra of the ghost and gauge fields which is defined here is applicable
for any representation of the matter. This property provides the flexibility
to consider the model with various matter fields. One can find an example
of the commutation relation of the matter fields in ref.[Watamura 1992].
With the above relations we can find the BRST transformation of the
gauge fields by using the standard logic to define it in the field theory: The
derivative of the field is not covariant under the BRST transformation. Its
transformation is
(27)
On the one hand the covariance under the BRST transformation (20) can
be rewritten as
bB \71J1 = C I (\71J1 * XI) = C I (d1J1 * XI) + C l AJ(1J1 * Xl * XJ) (28)
On the other hand taking the BRST transformation of the r.h.s. of eq.(19)
we get
bB \71J1 = -(dC I )1J1 * Xl + C I (d1J1 * XI) + (8 B AI)(1J1 * xI)
-A I C J (1J1*XI*XJ) (29)
The BRST transformation of the gauge field can be defined by requiring
the equivalence of the eqs.(28) and (29). Thus we get
(8 B AI)(1J1 * XI) = (dC I )(1J1 * Xl) + (AIC J + C l AJ)(1J1 * Xl * XJ) .(30)
Using the commutation relation of AI and C l we see that the (AIC J +
C l AJ) term is proportional to the projector PAd. Then, applying eq.(8)
and comparing the coefficient of the functionals XI we get
Q-DEFORMATION OF GAUGE THEORY 307

Proposition 2: The BRST transformation of the gauge field is given by

bBAo = dCo, and bBA a = dC a - ig(wCo A a + fbcCb A C ) (31)
and it is nilpotent.
We can also define the field strength using the above algebra as:
Proposition 3: The field strength is given by

FO = dAD, and Fa = dAa _

q2
tg
+ q-2
.r,a Ab A C
bc
(32)

The field strength is covariant under the BRST transformation: {jBFf =

C J (Ff * XJ), and satisfies the Bianchi identity:

and (33)

In order to obtain the q-deformation of the BRST formulation of the

gauge field theory, we have to take the structure of the base manifold into
consideration. Using the result here, one may take the base manifold as a
usual spacetime but a more interesting possibility is the one when the base
manifold is also described by the non-commutative function algebra. In both
cases, we have to reconsider the meaning of the usual quantization so that
it fit to the pure algebraic formulation.

References
Abe, E.: 1980, 'Hopf Algebras', Cambridge Tracts in Math., vol. 74, (Cambridge Vniv.
Press, 1980).
Aref'eva, LYa. and Volovich, LV.: 1991, Mod. Phys. Lett. A6, 893
Bernard, D.: 1990, Prog. Theor. Phys. Suppl. 102, 49
Brzezinski, T. and Majid, S.: 1992 'Quantum group gauge theory on quantum spaces.'
Pre print DAMTP /92-27, ; 'Quantum Group Gauge Theory on Classical Spaces'
Pre print DAMTP /92-5l.
C. Becchi, A. Rouet and R. Stora,: 1976, Ann. Phys. , 287
Tyuitin, LV.: 1975 Lebedev preprint FIAN 39 (1975), unpublished.
Carow-Watamura, V., Schlieker, M., Watamura, S. and Weich, W.: 1991a, Commun. Math.
Physics 142 , 605
Carow-Watamura, V. and Watamura, S.: 1991b "Complex Quantum Group, Dual Alge-
bra and Bicovariant Differential Calculus," pre print TU-382(1991) to be published in
Commun. Math. Physics.
Carow-Watamura, V.: 1992, Ph.D. Thesis, 'The quantum group symmetry in the models
of elementary particle physics.'
Drinfeld, V.G.: 1986, 'Quantum Groups', Proceedings of the International Congress Math-
ematicians, Vol.1, 798.
Faddeev, L.D. and Popov, V.: 1967, Phys. Lett. 25B, 29
Hirayama, M.: 1992, 'Gauge Field Theory of the Quantum Group SUq (2)', preprint
TOYAMA-74(1992).
Isaev, A.P. and Popowicz, Z.: 1992, Phys. Lett. 281B, 271
Jimbo, M.: 1986, Lett. Math. Phys. 10, 63
308 SATOSHI WATAMURA

Jurco, B.: 1991, Lett. Math. Phys. 22, 177

Manin, Yu.I.: 1988, 'Quantum groups and non-commutative geometry', Centre de
Recherches Mathematiques, Universite de Montreal.
Reshetikhin, N.Yu., Takhtajan, L.A. and Faddeev, L.D.:1989, 'Quantization of Lie groups
and Lie algebras', Algebra i Analiz, 1 178.
Watamura, S.: 1992, 'Quantum Deformation of BRST Algebra', pre print HD- THEP-92-39,
TU-411(1992).
Woronowicz, S.L.: 1987, Commun. Math. Phys. 111, 613
Woronowicz, S.L.: 1989, Commun. Math. Phys. 122, 122
Wu, K. and Zhang, R.: 1992, Commun. Theor. Phys. 17, 175
CYCLIC PARAGRASSMANN
REPRESENTATIONS FOR COVARIANT
QUANTUM ALGEBRAS
ALEXEYISAEV
Laboratory of Theoretical Physics, JINR, Dubna
SU-l0l 000 Moscow, Russia

Abstract. This report is devoted to the consideration from the algebraic point of view
the paragrassmann algebras with one and many paragrassmann generators 0., 0;+1 =
O. We construct the paragrassmann versions of the Heisenberg algebra. For the special
case, this algebra is nothing but the algebra for coordinates and derivatives considered in
the context of covariant differential calculus on quantum hyperplane. The parameter of
deformation q in our case is (p+1)-root of unity. Our construction is nondegenerate only
for even p. Taking bilinear combinations of paragrassmann derivatives and coordinates
we realize generators for the covariant quantum algebras as tensor products of (p + 1) x
(p + 1) matrices. There is now the extensive literature about finite dimensional cyclic
representations for quantum algebras with q being a root of unity (see e.g. [2],[24]). It
is rather interesting to relate our paragrassmann representations with representations
explored in [2],[24]. At the end of our talk we discuss the paragrassmann extensions of the
Virasoro algebra. This report is largely based on the papers [25-27].

Paragrassmann algebras (PGA) are interesting for several reasons. They

are relevant to conformal field theories [1,2] and to unusual statistics [3], in
particular, to the Green-Volkov parastatistics which was earlier discussed
mainly in the context of the standard field theory [4]. There are also some
hints (e.g., Ref.[5]) that PGA have a connection to quantum groups. Finally,
it looks aesthetically appealing to find a generalization of the Grassmann
analysis [6] that proved to be so successful in describing supersymmetry.
Recently, some applications of PGA have been discussed in literature.
In Ref.[7], a parasupersymmetric generalization of quantum mechanics has
been proposed. Refs.[8]'[21],[22] have attempted at a more systematic con-
sideration of the algebraic aspects of PGA. Using the Green ansatz [4] and
fractional supersymmetry approach a sort of paragrassmann generalizations
of the conformal algebra have been introduced in [8],[28]. Applications to
the relativistic theory of the first-quantized spinning particles have been
discussed in Ref.[9]. Further references can be found in [2],[5],[7],[8],[28].
We start by defining the PGA r p( 1) (or simply r), generated by one
nilpotent variable () (()p+l = 0, p is some positive integer). Any element of

309
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and QUllntum Defonnations, 309-316.
© 1993 Kluwer Academic Publishers.
310 A.P.ISAEV

the algebra, a E r, is a polynomial in 0 of the degree p,

(1)
where ai are real or complex numbers or, more generally, elements of some
commutative ring (say, a ring of complex functions) [10]. It is useful to have
a matrix realization of this algebra. One may regard ai as coordinates of the
vector a in the basis (1, 0, ... , OP). Defining the operator of multiplication
by 0, O( a) = aoO + ... + ap-l OP, we see that it can be represented by the
triangular (p + 1) X (p + 1)-matrix acting on the coordinates of the vector a:

(2)

m, n = 0,1, ... ,p. We may now treat elements of the algebra as matrices.
The next step is the definition of the derivative with respect to O. We
expect a differentiation a == a / aO to act as
a(l) = 0, a(O) = 1, a(on) IX on-I, n> 1, (3)
It is easy to see that the condition a( 0) = 1 together with the standard
Leibniz rule, a( ab) = a( a) . b + a . a(b), completely define the action of a
on any a E r, but this immediately leads to a contradiction 0 == a( Op+l) =
(via Leibniz rule) = (p + 1 )OP. This is a manifestation of the general fact
about nilpotent algebras known even for the Grassmann case: once the nor-
malization conditions of the type (3) are established, the Leibniz rule is to
be deformed.
To introduce a useful definition of a we suggest a generalized Leibniz rule
(g- Leibniz rule)
p-l
a(ab) = a(a)· b + g(a)· a(b) ,g(O) = L Imom+1 , (4)
m=O

where 1m are some numbers and g is an automorphism of the algebra

rp , g(ab) = g(a)g(b). For the Grassmann case (p = 1) we have g(a) =
(-1 )(a)a where (a) is the Grassmann parity of the element a . The automor-
phism g and, hence, the derivative a are completely fixed by the normaliza-
tion conditions a(O) = 1 and a(02) IX O. These, by (4), give 1m = 0 for m > D,
a(l) = 0, a(on) = (1 + 10 + ... + I~-l)On-l and from a(OP+1) == 0 we get
1+ ,0 +., .+,~ = D, so that 10 is fixed to be a root of unity. For the moment,
we choose 10 to be the prime root i.e.: 10 = q == e 21ri /(p+1) = (_1)2/(p+l). By
introducing the notation (n)q == 1 + q + ... + qn-l = \-::.qqn , the action of a
can be performed as a(on) = (n)qon-t, and so the matrix elements of a in
the basis {om},m = D, ... ,p are
(5)
CYCLIC PARAGRASSMANN REPRESENTATIONS 311

a a
Since (p + l)q = 0, the operator is nilpotent, p+1 = o. It is not difficult
a
to see that and 0 satisfy the q-deformed commutation relation
[a, O]q == 8B - qoa = 1 . (6)
The Grassmann case for p = 1 and the classical one in the limit p -- 00 are
evidently reproduced. The last equation is suggestive of a relation between
PGA and much discussed q-deformed oscillators and quantum groups (see,
e.g. Refs.[12] - [14],[18],[23],[29],[30]) with the deformation parameter q
being a root of unity.
Consider now the algebra IIp(l) (or, simply II) generated by both 0 and
a. One can show (see [25]) that the algebra II has the basis {om an}, m, n =
0, ... , p and is isomorphic to the algebra M at(p + 1) with natural "along-
diagonal" grading deg(oma n ) = m - n .
The automorphism 9 from Eq.( 4) is expressed in the operator form as
(7)
Its matrix elements are (g)mn = qmb mn . In the mathematical literature (see,
e.g. Ref. [11 ]), our generalized differentiation (4) is called g-differentiation.
Mathematicians also consider a further generalization, called (g, g)-differentiation
that satisfies the rule
a(ab) = a(a) . g(b) + g(a) . a(b) . (8)
Although we think that Eq.(4) looks more natural than Eq.(8), the latter
can be used to define "real" differentiation, i.e., the one with real matrix
elements. In fact, choosing for 9 and 9 the automorphisms defined by g( 0) =
ql/20 , g( 0) = q-l/20, we find that
qn/2 _ q-n/2
aeon) = [n]..;qon-l, [n]..;q == ql/2 _ q-l/2 . (9)

This is obviously a real representation of a. The operators 9 and 9 have the

matrix elements (g)mn = qm/2b mn , (g)mn = q-m/2b mn , and the following
operator expressions in terms of 0 and a 9 = aO-q-l/2oa, 9 = aO-ql/2oa.
One can easily recognize in these formulas the definition of the quantum
oscillator in the MacFarlane-Biedenharn form (see [12],[18]).
In addition to the g-differentiation, one can also define the integration
over 0 I dOa( 0) = ap that generalizes the Grassmann integration to the
paragrassmann one (see e.g. [15])
In some applications (e.g., in constructing parasupersymmetries) one has
to deal with an (Eq.(l)) taken from the ring of the differentiable functions
of a real or complex variable z i.e., an = an(z). For such an algebra, it is
possible to define a sort of "covariant derivative"

(10)
312 A.P.ISAEV

a
where 00 == and the standard notation is used (p)q! = (p)q(p -1)q ... (l)q.
This derivative obviously satisfies the g-Leibniz rule (4) and may be consid-
ered as a root of a z since DP+1a(z; ()) = aza(z; (}).
Our discussion of the PG A r p( 1) and IIp( 1) was completely general and
did not rely on special matrix representations for (} and a. In fact, different
representations could be classified if we relaxed our assumption for q to be
the prime root of unity, qp = exp(27ri/(p + 1)). Then, one would find that
the structure of the extended algebras r peN) and IIp( N) depend on the
arithmetic properties of (p + 1) (see [25]).
We present here just the explicit inductive construction of r peN). Start-
ing with N = 2, define
(}1 =90 () , (}2 = () ® 1 , (11)
where () and 9 have been defined in (2), (5) and (7). It is easy to see that
= q(}2(}1, (}r+ 1 = O. The crucial fact is that the definition (11) allows for
(}1(}2
nilpotency of any linear combination of (}1 and (}2
(12)
as long as q is a primitive root of unity (see for details [25]).
Suppose now that we have constructed the algebra rp(N) satisfying the
relations
(}i(}j = q(}j(}i, i < j, i,j = 1. .. N , (13)
N
(I: ai(}i)pH = 0 . (14)
i=1

Then, N + 1 matrices {)i satisfying (13) and (14) can be constructed by

analogy with (11)
{)i=g®(}i, i = l ... N,{)NH=(}@1. (15)
The proof of the identity (14) is performed in full analogy with the N = 2
case.
It is rather amusing that the consideration of PGA naturally leads to the
objects introduced in the context of quantum groups. In fact, the generators
of the algebra rp(N), determined by the relations of type (13) and (14),
might be considered as coordinates of a certain nilpotent Manin's quantum
hyperplane similar to those of Refs.[13]' [14]. Such an object and, especially,
its a-extensions (defined by its automorphisms) look rather interesting both
from algebraic [17] and quantum-geometric [16] points of view.
Let us consider an algebra r p( N) with the commutation relations Bi(}j =
qPIJ(}j(}i, i,j = 1 .. . N, where q denotes the prime root of unity. The require-
ment for qPIJ to be a primitive root is equivalent to the requirement for Pij to
CYCLIC PARAGRASSMANN REPRESENTATIONS 313

be invertible elements of the ring Zp+1. Then, let us define differentiations

Oi satisfying the normalization conditions oi(lh) 6ikl and the g-Leibniz
rule
OJ ( ab) = OJ ( a) . b + gi ( a) . Oi ( b) (16)
where the action of the automorphisms gi on (h is gi((h) = qV'k(h. These
conditions determine the commutation relations in the operator form oi(h =
6ik + qV,k(}kOi, OiOj = qP'JOjOi, and, for i "1 k, we have Vik = Pki = -Pik,
while the diagonal Vii remains nonspecified.
I t is possible to construct another interesting extension of r p( N) (where p
is an even number) with the generators (}i and Oi if we even further relax the
g-Leibniz rule (16) to the form familiar from the theory of quantum groups
[16] Oi( ab) = Oi( a) . b + g{ (a) . OJ( b) . This makes it possible to construct the
operators OJ by the inductive procedure similar to (15)

&j = 9 @ Oi, i = 1 ... N, &N +1 = 0 @ 1 , (17)

where we have also slightly modified the definition of 0 and 9
o(} - q2(}O = 1, o(} - (}o = g2 . (18)
i,From these equations and from definitions of (}j and OJ (i = 1, ... , N) we
obtain the following algebra
(}j(}j = q(}j(}j i <j , OiOj = q-10jOj i < j ,
OJ(}j = q(}jOj i "1 j , OJ(}j - q2(}jOi = 1 + (q2 - 1) L (}kOk . (19)
k>i
These are the well known formulas for GLq(N)-covariant differential calculus
on the quantum hyperplane [16]. These formulas may also be interpreted as
the definition of the covariant q-oscillators [18],[23] or, else, as the central
extension of the quantum symplectic space relations for the quantum group
Spq(2N) (see [13]). Note that nilpotency of the linear combinations ai(}j and
biOi as well as non degeneracy of 0 (18) are guaranteed since both q and q2
are primitive roots of unity (for p even integer only). Here we would like
to stress that the representations (15), (17) can be extracted from the first
paper of [18], where the analogous representations have been considered in
the context of q-oscillators. Using the matrix representations (2), (5) for ()
and 0 one can realize 2N variables Oi and Oi as (p+ I)N x (p+ I)N matrices.
This example demonstrates a deep relation between PGA and quantum
groups.
Indeed, relations (19) are invariant under the coaction of the GLq(N)
group
314 A.P.ISAEV

where {Tij} are generators of GLq(N) and S(.) is an antipode. Now, we

would like to consider the composite operators

Eij = ()iOj (21 )

with the coadjoint transformation rule (see (20»

Eij f-+ E: j = T;kEkITI-/ == Ekl ® TikTljl . (22)

For q = 1 these operators realize the Jordan-Schwinger construction for the
generators of the usual gl(N) algebra. It is natural to expect that the oper-
ators Eij (for q :/;1) generate the deformed algebra Uq(gl(N». However, it
is known [29], that q-deformed commutation relations for Eij defined in (21)
are not represented in a unique form. In this situation, there are two ways to
write down the commutation relations for Eij uniquely. First, the covariant
algebra {()i, OJ} can be realized in terms of the MacFarlein-Biedenharn oscil-
lators (see [18]). Then, these osclillators can be used (via Jordan-Schwinger
construction) to construct the quantum algebra in the Drinfeld-Jimbo form.
But it is a rather long way. Second, one can use the covariancy of the algebra
under the coadjoint transformations (22). It is possible to prove [27] that
q-deformed GLq(N)-covariant commutation relations for Eij are unique (up
to some inessential rescaling factors).
To present these relations, let us rewrite the algebra (19) in the R-matrix
form [16],[30]

R12(h()2 = q()2(h, 0102R12 = q0201, ol(h = K8 12 + q(hR1201, (23)

where K is an arbitrary parameter (in Eqs.(19) we have put K = 1), R12 is the
R-matrix satisfying the Hecke relation (P12 R 12 )2 - (q- q-l )(P12 R 12 ) -1 = 0
and P12 is a permutation matrix (for the notations see [13]). Using the
algebra (23) one can obtain the complete set of covariant relations for Eij
(21) (see [27])

R12EIR21E2 - E2R12EIR21 = -(P12EIR21

K

q
- R 12E 1P 12 ) , (24)

(R12 - qP12)(qEIR21E2 + KE1P12 )(R12 - qP12 ) = 0 (25)

The equations (24) give us for q = 1 usual gl(N)-commutators and it is

natural to consider Eqs.(24) as structure relations for the covariant quantum
deformation of the gl(N)-algebra. The Casimir operators Ck for this algebra
are expressed via q-deformed trace: Ck = Trq(Ek) = Tr(DEk) , D =
diag(1, q2, q4, . .. ) . After shifting Y = E - i~-l' one can rewrite eqs.(24) in
the concise form
(26)
CYCLIC PARAGRASSMANN REPRESENTATIONS 315

This is nothing but the well known commutation relations for the operator
Y = (L-)-l L+ (here L± are the Borel sub algebras of Uq(gl(N)) [13)) inter-
preted also as a differential operator in the bicovariant differential calculus
on the quantum group GLq(N) [31]. These also are the structure relations for
the braided algebras [30],[20]. As it was shown in [27], the second set of the
covariant relations (25) in the limit K, = 0 gives us the part of q-deformed an-
ticommutation relations for 1-forms d(T)T-l defined for the GLq(N)-group.
The other part of such relations can be obtained by considering the covariant
differential calculus on the fermionic quantum hyperplane [27]. The realiza-
tion of Ejj = (JjOj in terms of the (p + 1)N X (p + I)N matrices (Jj, OJ (15),
(17) leads us to the matrix paragrassmann representations for the covariant
quantum algebras with commutation relations (24), (26) (qP+1 = 1).
At the end of this report we would like to present a paragrassmann ex-
tension V irp of the Virasoro algebra. This extension has been discussed in
[26]. We define this algebra denoted by Virp as the algebra of generators
for the parasuperconformal transformations z ...... z(z, 0) ,0 ...... 8(z, 0), con-
serving the form of the covariant derivative (10). It means that we have
Da(z, 8) = D(O)iJa(z, 8) for an arbitrary parasuperfunction a(z, 0). As
it was shown in [26], the algebra Virp has generators Tn and G r with the
following commutation relations (we also present here the possible central
extensions)

[Tn' Tm] = (n - m)Tn+m + P~l Ej Cj(n 3 - n)8 n+m,o,

[Tn' G r ] = (P~l - r)Gn+r ,

{Gro ,···, Grp}c = (p + 1)TE r - E Cj[Ei riri+j + p!1]8 E r,O ,

where { ... }c is the cyclic sum of the p + 1 linear monomials {Go, ... , G p}c =
Go' .. Gp + GpG o ... Gp- 1 + ... G 1 ... GpG o , and the number of the central
charges Cj is equal to [E.:}!]. The algebra Virp(p > 1) has the multilinear
commutation relations and, in fact, is not a Lie algebra. Note, that the
special case of this algebra has been considered in [28].
As a final remark, we would like to mention a possible relation of PGA to
the finite-dimensional quantum models introduced by H.Weyl in his famous
book and further studied by J.Schwinger (Refs.[19)). They considered quan-
tum variables described by unitary finite matrices Ui satisfying the relations:
UjUj = qUjUi and (Ui)p+l = 1. (Obviously, q must be a root of unity). They
realized that the p = 1 case is relevant for describing the spin variables
and treated the infinite-dimensional limit p ---+ 00 as a limit in which usual
commu tative geometry is restored.

References
[1] A.B. Zarnolodchikov and V.A. Fateev, Sov.Phys. JETP 62(1985)215.
316 A.P.ISAEV

[2] V. Pasquer and H. Saleur, Nucl.Phys. B330(1990)523.

[3] C. Aneziris, A.P. Balachandran and D. Sen, Statistics in one dimension, Prep. SU-
4228-453, Syracuse, 1990.
[4] A.B. Govorkov, Sov. J. Part. Nucl. 14(1983)520; Y. Ohnuki and S. Kamefuchi, Quan-
tum Field Theory and Parastatistics, Univ. of Tokyo Press, 1982.
[5] V.P. Spiridonov, in: Proc. of the Intern. Seminar Quarks-90, Telavi, USSR, May 1990,
eds. V.A. Matveev et aI., World Sci., Singapore, 1990.
[6] F. Berezin, Intr. into Algebra and Analysis with Anticommuting Variables, Moscow
State Univ. Press, 1983; R. Casalbuoni, Nuovo Cim., A33(1976)389.
[7] V.A. Rubakov and V.P. Spiridonov, Mod. Phys. Lett. A3(1988)1337.
[8] S. Durand, R. Floreanini, M. Mayrand, L. Vinet, Phys. Lett. 233B (1989)158.
[9] V. Gershun and V. Tkach, Problems of Nucl. Physics and Cosmic Rays (Kharkov Univ.
Press), 23(1985)42; G. Korchemsky, Phys.Lett. 267B (1991)497.
[10] N. Jacobson, Structure of Rings, Am. Math. Soc. Pub!., Providence, 1956.
[11] N. Jacobson, Lie Algebras, Interscience Publish, N.-Y.-London, 1962.
[12] A. MacFarlane, J.Phys. A22(1989)4581; 1. Biedenharn, ibid. L873.
[13] L. Faddeev, N. Reshetikhin and M. Takhtajan, Alg. i Anal. 1(1989)178.
[14] Yu. Manin, Quantum Groups and Noncommutative Geometry, Prep. Montreal Univ.,
CRM-1561 (1988); E. Corrigan, D. Fairlie, P. Fletcher, J.Math.Phys. 31(1990)776.
[15] M. Rausch de Traubenberg and N. Fleury, in: Leite Lopes Festschrift, World Sci.,
Singapore, 1988.
[16] J. Wess and B. Zumino, Nucl. Phys. (Proc. Suppl.) 18B (1990)302.
[17] A. Borowiec, W. Marcinek and Z. Oziewicz, On multigraded differential calculus, in:
Proc. of 1st Max Born Sympos., Wroclaw 1991, (World Scient., 1992).
[18] M. Chaichian, P. Kulish and J. Lukierski, Phys.Lett. 262B(1991)43; P. Kulish
and E. Damaskinsky, J.Phys. A23(1990)L415; D. Fairlie and C. Zachos, Phys.Lett.
256B(1991)43.
[19] H. Weyl, Theory of Groups and Quantum Mechanics, Dover. Pub!., N.Y., 1931; J.
Schwinger, Proc. Natl. Acad. of Sci., 46(1960)570.
[20] Sh. Madjid, J.Math.Phys. 32(1991)3246.
[21] C. Ahn, D. Bernard and A. LeClair, Nucl. Phys. B346 (1990)409.
[22] R. Kerner, J. Math. Phys., 33 (1992)403.
[23] W. Pusz and S.L. Woronowicz, Reports Math. Phys. 27 (1989)231.
[24] C. deConcun and V. Kac, Progr. in Math. 92(1990)471; D. Arnaudon and A.
Chakrabarti, Comm.Math.Phys.139(1991)461,605; E. Date, M. Jimbo, K. Miki and T.
Miwa, ibid. 137(1991)133; V. Bazhanov, R. Kashaev, V. Mangazeev and Yu. Stroganov
ibid. 138(1991)393.
[25] A. Filippov, A. Isaev and A. Kurdikov, Mod.Phys.Lett. A7(1992)2129.
[26] A. Filippov, A. Isaev and A. Kurdikov, Paragrassmann extensions of Virasoro algebra,
Prep. JINR E2-92-392,393 (1992).
[27] A. Isaev and P. Pyatov, GLq(N)-covariant algebras and covariant differential calculus,
Prep. JINR E-2-92 (1992).
[28] S. Durand, R. Floreanini, M. Mayrand, V. Spiridonov and 1. Vinet, Parasupersymme-
tries and fractional supersymetries, in: Proc. of 18th Int. Colloquium on Group Theor.
Meth. in Physics, Moscow (1991); S. Durand, McGill Univ. Prep. 92-30(1992).
[29] S.P. Vokos, J.Math.Phys. 32(1991)2979.
[30] P. Kulish, Quantum groups and quantum algebras as symmetries of dynamical systems,
Prep. YITP/K-959(1991).
[31] B. Zumino, Introduction to the differential geometry of quantum groups, Prep. LBL-
31432(1991 ).
HEeKE SYMMETRIES AND BRAIDED LIE
ALGEBRAS

DMITRI GUREVICH
Max-Planck-Institute fur Mathematik, 26 Gottfried-Claren Strasse, 5300 Bonn 3

Abstract. We consider Hecke symmetries of minimal type, i.e., solutions of the QYBE
with two eigenvalues and such that the Poicare series of the corresponding exterior alge-
bras are polynomials of degree 2. We construct the corresponding quantum cogroups and
introduce notion of braided Lie algebra. The examples of Hecke symmetries of minimal
type and of braided Lie algebras are given.

Key words: Quantum Yang-Baxter Equation, Hecke symmetry, bi-rank, quantum cogroup,
braided Lie algebra
Generalized Lie algebras connected with involutive (S2 = 1) solution of
the quantum Yang-Baxter equation (QYBE) have been introduced in our
paper [3]. In [5] (see also references therein for our previous papers) we have
constructed some explicit examples of generalized Lie algebras (or in other
words S-Lie algebras) of gl and sl types, connected with involutive non-
quasi classical (or non-deformation) solutions of the QYBE. The problem of
a proper generalization of this notion to the non-involutive case was open
though a lot of papers were devoted to the problem.
This paper is devoted to two questions. On the one hande we continue to
study some non-quasiclassical non-involutive solutions S of the QYBE (so
called Hecke symmetries). On the other hand we propose the definition of
S-Lie algebras (called here braided Lie algebras to stress non-involutivity of
the operator S) connected with Hecke symmetries.
The paper consists of three Sections. In Section 1 we recall some usuful
facts about Hecke symmetries. We put emphasis on Hecke symmetries of
minimal type, i.e. such that the Poincare series of corresponding exterior
algebras are polynomials of degree 2 with leading coefficient 1. Some of such
type solutions of the QYBE have been independently constructed in [1].
In Section 2 we introduce quantum cogroups connected with Hecke sym-
metries of minimal type and compare these objects with Hopf algebras aris-
ing from non-degenerated bilinear forms defined in [1]. In Section 3 we in-
troduce a notion of braided groups and give their examples connected with
Hecke symmetries of minimal type.

317
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 317-326.
© 1993 Kluwer Academic Publishers.
318 DMITRI GUREVICH

1. HECKE SYMMETRIES: STRUCTURE, EXAMPLES

Let V be a finite-dimensional vector space over a field k of characteristic 0
and S : V02 ---+ V02 a solution of the QYBE
(S @ id)(id @ S)(S @ id) = (id @ S)(S @ id)(id @ S).
Among all the solutions of the QYBE, the most interesting are the so called
closed ones. Fix a base {ej,l :::; i :::; n = dimV} in the space V and put
S( ej @ ej) = Stl ek @ el. Consider the operator T which in the base {ed is
defined by S11 T£:::
= {jj{j~. We call the solution S of the QYBE closed if T
exists.
It is not difficult to show that a closed solution of the QYBE can be ex-
tended up to a braiding operator in a rigid quasi tensor category 2l containing
the space V. According to generally accepted terminology, a quasitensor cat-
egory is called rigid if it satisfies the condition U E Ob2l ---+ U* E Ob2l. The
braiding operator S (or in other words, "commutativity morphism") is a
mophism in the category 2l but it is not involutive in general.
In this paper, we deal only with solutions of the QYBE wich have two
eigenvalues. We call them Hecke symmeties. More precisely we call a solution
S of the QYBE a Heeke symmetry if S satisfies the equation

(qid - S)(id + S) = D.
We assume that q i 0 and qn i 1, n = 2; 3; ...
The Hecke symmetries have a great advantage: it is possible to define for
them an analogue of the symmetric and exterior algebras. Namely we put
!\±(V) = T(V)/{I:d
where T(V) = EElV0 k is the tensor algebra of V and {I+} (resp., {L}) is
the ideal in T(V) generated by the image 1+ (resp.,L) of S + id (resp.,
q id - S). Denote !\~(V) the homogeneous component of degree k of these
algebras and consider the Poincare series P±(t) of the algebras !\±(V):

P±(t) = L dim!\~(V)tk.
We call a Hecke symmetry S (and the corresponding space V) even if it
is closed and the Poincare serie P _ (t) is a polynomial (as it was shown in
[5] this condition is equivalent to following one: P_(t) is a polynomial with
leading coefficient 1). If this polynomial is of degree k we say that V (or S)
has bi-rank kiD and denote it bi-rk V. 1
1 Note that bi-rank is well-defined for odd objects of Hecke type (it is left to the reader
to give a definition of odd spaces). For them we say that bi-rank is equal to 011 and for
some objects V composed in some sens from even and odd spaces it is natural to put bi-
rank V = kll. We dont want to examine this problem in more detail but stress only that
it is not clear yet, whether all involutive closed solutions of the QYBE have a bi-rank.
HEeKE SYMMETRIES AND BRAIDED LIE ALGEBRAS 319

Now we introduce two important operators B = B(S) : V -+ V and C =

C(S) : V -+ V as follows

where {ei} is the fixed base in V.

It is easy to see that this definition does not depend on the choice of the
base. These operators can be defined for any object in any rigid quasi tensor
category but we need them only for an inicial space V equipped with a Heeke
symmetry S.
The following statements are proved in, or can be easily deduced from,
[5].

PROPOSITION 1. 1. For any Hecke symmetry S the relation

holds.
2. If S is even then the polynomial P _ (t) is reciprocal.
3. Moreover if bi-rk V = klO then the operators Band C satisfy the rela-
tion
tr B = tr C = q-k kq

(we denote here and below kq = 1 + q + ... + qk-l).

4. If bi-rk V = 210 then BC = C B = q- 3 id and the operators b = Bq2
and c = Cq2 satisfy the following condition
J) if Jordan form of b or c contains a cell with eigenvalue x it contains
another cell with eigenvalue qx- l (with the same multiplicity).
5. If an operator c : V -+ V satisfies the conditions J) and tr c = 1 + q
then there exists an even closed Hecke symmetry S : V02 -+ V0 2 of bi-rank
210 such that C = C( S) = q- 2 c. There exists the one-to-one correspondance
between the family of all such Hecke symmetries and matrices v satisfying
the condition
(C*)-lq = V-lcv, V = (v ij )

(c* denotes the matrix conjugated to c). If such v is fixed then the corre-
sponding Hecke symmetry is of the form

where U = (Uij) can be deduced from the equality

c = (1 + q)vu* i.e. c{ = (1 + q) vjkuik.

320 DMITRI GUREVICH

Remark that the quantity tr B = tr C, which can be defined for any

element of a rigid category, is usually called its rank (see for example [6]).
So the statement 3 of Proposition 1 establishes the relation for even Hecke
symmetries between rank in this sens and bi-rank in our sense. Here and
further on, we say that a Hecke symmetry is of minimal type if it is even
and has bi-rank 2iO.
Stress also that bi-rank does not change under deformation and therefore,
a quasiclassical Hecke symmetry (i.e.,a deformation of the usual tranposi-
tion) must have bi-rank niO, n =dim V.
Let us give two examples of minimal Hecke symmetries.

EXAMPLE 1. Let dim V =2 and q :j:. 1. Then any pair (c, v) satysfying the
conditions above has in some base form

Then

oq 00 qm-
0 1 0)
0
S = ( 0 m q_ 1 0 where m = -a/b.
o 0 0 q

Stress that the operator N = uv is scalar iff m 2 = q (the role of this operator
will be explained in Proposition 2).

EXAMPLE 2. Let dimV=3. We put c = diag(x,t,q/x) where t is one of

roots ±vfii and x satisfies the equation x + t + q/x = 2. Then assuming v to
be as follows we obtain S

q 0 0 0 0 0 0 o0
oq 0 0 0 0 0 o0
o0 q-x 0 -bx/a 0 -cx/a o 0
o0
v= CO a)
ObO
cOO
,S=
o0
o0
o0
0

0
q

0
0
-at/b 0 q-t
0
0

q
0
0 -tc/b o 0
0 o0
o0 -qa/cx o -qb/cx 0 q-q/x 00
o0 0 0 0 0 0 q 0
o0 0 0 0 0 0 oq
For this example the operator N = uv is scalar if a/c = x/to
Stress that the last example can be easily generalized to arbitrary dimension
n=dim V.
HEeKE SYMMETRIES AND BRAIDED LIE ALGEBRAS 321

2. HECKE SYMMETRIES ARISING FROM BILINEAR FORM

AND QUANTUM COGROUPS
In [1] a method have been introduced to construct a solution of the QYBE
by means of a non-degenerated bilinear form. In this Section we show that
the family of such solutions coincides with subset of Hecke symmetries of
minimal type. We introduce also the quantum groups connected with Hecke
symmetries of minimal type and compare them with Hopf algebras defined
in [1].
Consider a linear space L = V @ V* with base {el = ei @ ei } equipped
with the operator

SQ : L0 2 ---+ L0 2 , SQ( el @ e~) = Sit (S-l ){;,n e;;' @ eb'

Stress here that V* differs from the dual space (right or left one) in the rigid
category mentioned above and moreover the space L does not belong to this
category (wich will appear below as the category of comoduls of a quantum
cogroup ).
It is obvious that this operator SQ satisfies the QYBE and has eigenvalue
1.
Consider the algebra A(S) = T(L)j{I} where {I} is the ideal generated
by the image I of the operator id - SQ. Suppose now that the initial operator
S is Hecke symmetry of minimal type and introduce the so called quantum
determinant det = ui/vike{ @ e~ (in [5] it was defined for any even Hecke
symmetry).
One can see that

for some operator M : L ---+ L. Introduce the formal inverse element der 1
and put
SQ(deC 1 @ e{) = (M- 1 )lke7 @ deC 1
(so the element det der 1 is central) and define the algebra k[GL(S)] as the
quotient of A(S) with the additional generator der 1 by the ideal generated
by elements
deC 1 @ e{ - SQ(deC 1 @ el).
It is natural to do this because

S~(det @ e{) = det @ e{

(see [5]).
If det is a central element of A(S), we introduce also the following algebra
k[SL(S)] = A(S)j{Idetl where {Idet } is the ideal in A(S) generated by
322 DMITRI GUREVICH

det - 1. The algebras k[GL(S)) and k[SL(S)), being equipped with the
usual comultiplication (~e1 = ef@e{) the usual counit (te{ = 151) and some
antipod, are Hopf algebras. We call them quantum cogroups because, like in
deformation case, it is more natural to use the terme quantum groups for
dual objects (although we do not have their description similar quasiclassical
quatum groups Uq(g)).
These quantum cogroups have been introduced in [4) and [5).
PROPOSITION 2. (see f4J,[5j) If S is Hecke symmetries of minimal type
then the element det E A(S) is central iff the operator N = uv(N/ = Uikvkj)
is scalar.

Represent now the construction of [1) in a form convenient for our aims.

PROPOSITION 3. Let B = (Bij) be a non-degenerated bilinear form. Then

the operator SDL
(SDL)fj = t5ft5~ + aB i j(B- 1 )kl

where Bik(B- 1)kl = 15; is a solution of the QYBE iff a+a- 1+Bij(B- 1 )ij = O.

To establish the relation between the construction from [1) and ours, consider
the operator

S = qSDL = qid + qaB @ B- 1

and put Uij = Bij,V k1 = -qa(1 + q)-1(B-1)kl. It is easy to see that the
operator S satisfies the conditions of Proposition 1 iff for B, a and q the
relation above and relation qa 2 = 1 hold.
Hence S is a Hecke symmetry of minimal type with eigenvalues -1 and
a- 2 and the operator SDL has eigenvalues _a 2 and 1. The operator N = uv
is scalar in case under consideration. Therefore the map

{SDd ---+ {Hecke symmetries of minimal type with central det}

is constructed. It is invertible because assuming det to be central we have

v = bu- 1 with some b E k.
PROPOSITION 4. In the algebra k[SL(S») the relations

hold.

The first relation arrises from the follow chain of equalities

HEeKE SYMMETRIES AND BRAIDED LIE ALGEBRAS 323

(1 + q)-1(qukle7 0 e} + UklSiler:: 0 eb(S-1)~n) =

(1 + q)-1(qukle7 0 e} - UmnSijber:: 0 eb) = uijdet = Uij·
The second relation can be proved in the similar way. Here we use the
following lemma.
LEMMA 1. The relations

S( uk/e k 0 el ) = uklS!tea 0 eb = -Uklek 0 el ,

hold
Vice versa any ofthe relations from Proposition 4 yields the equality det = 1.
In [1) some Hopf algebras have been introduced as quotients of T( L) by
the relations from Proposition 4. Due to this Proposition we can conclude
that these algebras coincid with quantum cogroups k[SL(S)) defined above.

3. BRAIDED LIE ALGEBRAS

Let us recall first the definition of S-Lie algebras in the case when the op-
erator S is an involutive solution of the QYBE. We say that the space V is
equipped with a structure of S-Lie algebra if there exists an operator (S-Lie
bracket) [,) : V®2 -+ V satisfying the axioms
1. [,)S = -[,)j
2. [,)[,]12(id + S 12 S23 + S 23 S12) = 0;
3. S[,]12 = [,]23S 12 S23 with usual notation S12 = S0id and so on.
To introduce a braded counterpart of this notion we consider first a notion
of quadratic algebras. Let the space V be fixed. Consider a subspace I C V®2
and so called quadratic algebra corresponding to I : A+(V) = T(V)/{I}
where {I} is the ideal in T(V) generated by I.
Recall now that a quadratic algebra A+(V) is called Koszul algebra if the
complex

is exact. 2 Here Ai(V) is as usually the k-homogenous component of A+(V)j

A~(V) are defined as follows A~(V) = V, A~(V) = I, A::(V) = 10VnV01
and so on, and d is a natural differential (see [7) for details).
Let a map [,1 : I -+ V be given. Define a quadratic-linear algebra (an
analogue of envelop ping algebra) in the natural way U(g) = T(V)/ {J} where
2 In some papers another complex connected with quadratic algebra is considered and
the algebra is called Koszul algebra if the last complex is exact (see [5] where the both
complexes are considered).
324 DMITRI GUREVICH

{J} C T(V) is ideal generated by elements I - [, ]1. Since in this algebra

there exists a natural filtration, it is possible to consider the graded algebra
GrU(g).

PROPOSITION 5. Let us assume that the algebra !\+(V) is Koszul algebra

and that the following conditions

and
[,]([,P2 _ [,j23)(I09 V n V 09 1) = 0
hold. Then GrU(g) is isomorphic to !\+(V).

This Proposition is proved in [7] where the first condition is called correctness
and the second one is called Jacoby identity.
Suppose now that we have an algebra A = k[GL(S)] or A = k[SL(S)] as
above. Consider the category 2l of left comodules of A, i.e., for any V E 2l
there exists a coaction ~ : V -> A 09 V with usual properties.
Let V E 2l. Suppose that there exists a map [,] : V0 2 -> V.

DEFINITION 1. The agregate (V, I E& 1* = V0 2, [,]) will be called a braided

Lie algebra if the following axioms hold
o. the algebra !\+(V) = T(V)/{I} is Koszul algebra;
1. [,]1* = 0;
2. the relations from Proposition 5 are satisfied;
3. I, 1* E Ob 2l and the map [,] is a morphism in 2l.
Let us explain that the last condition means that

~[,] = (p, 09 [,])(~ O9~)

where ~ 09 ~ : V02 -> A02 09 V02 and p, : A0 2 -> A is the multiplication in
the algebra A.
Stress that a S-Lie algebra for involutive S is a particular case of a braided
Lie algebra. If we put I = Land 1* = h where I± E V02 is as in Section
1 (assuming q = 1), all axioms of braided Lie algebras are satisfied for any
S-Lie algebra. The verification of this fact is left to the reader. We note only
that "koszulity" of the algebras !\+(V) have been proved (in more general
context) in [5].
Note also that it is natural to introduce the axiom 0 if we want to ob-
tain a "good" envelopping algebra (see Proposition 5). In the forthcoming
publications we hope to elucidate the important role of this axiom in the
quantization procedure.
Consider now an example of a braided Lie algebra constructed by means
of a Hecke symmetry of minimal type.
HEeKE SYMMETRIES AND BRAIDED LIE ALGEBRAS 325

Let S : V®2 _ V®2 be a Hecke symmetry of minimal type such that det
is central and put A =
k[SL(S)]. Fix the base {ej, 1::; i ::; n dim V}. =
Consider one-dimensional A-comodule Vo = keo (6eo = 10 eo) and denote
V' = V ffi Vo. We put I = L ffi 10 (resp., 1* = 1+ ffi 10 ) where 1± C V02
are the same spaces as in Section 1 and 10,10 C V00 2 ffi Vo 0 V ffi V 0 Vo are
generated by elements {eo 0 ej - ei 0 eo} (resp., {eo 0 eo, eo 0 ej + ej 0 eo}).
In [5] we have proved that "+(V) is Koszul algebra. Using this result
it is not difficult to show that the algebra ,,+(V') = T(V')/ {I} is Koszul
algebra as well. We introduce in V' an A-modul structure putting 6ei =
ef 0 ep, 6eo = 1 0 eo and extend this structure on T(V') in a natural way.
It is obvious that 1,1* are A-comodules and I ffi 1* = V'0 2 • Introduce a
bracket:

where U = (Uij) is as in Proposition 1 and g, Ci E k.

Verify now that this bracket is a morphism in the category m of A-
comodules. First we will check compatibility ofthe bracket [ei, ej] with coac-
tion of A. Indeed by virtue of Proposition 4

gJ.l(ef 0 e])upq 0 eo = 9UijVmnefr, 0 e~upq 0 eo = 10 gUijeO = 6[ei,ej].

It is obvious that the bracket lei, eo] is compatible with coaction of A iff
Ci = C for any i. The verification of the axiom 1 is left to the reader. Verify
now the axiom 2. Since S is a Hecke symmetry of minimal type one has
L 0 V n V 0 L = {o}. Hence the space 10 V' n V' 0 I is generated be the
elements
{vij(ei 0 ej 0 eo - ei 0 eo 0 ej + eo (9 ei 0 ej)}.
Applying the operator [,]12 - [,]23 to an element from this family we have

([,] 12 -[,] 23"

)v'J(ei(gej(geo-ei0 eo(gej+e00 ei(gej)=

Vij(9UijeO 0 eo - cei (9 ej - cei (9 ej - cei (9 ej - cei (9 ej - gUijeO (9 eo) =

-4v ij cei 0 ej E I.
Axiom 2 is satisfied if cg = O. Therefore under this condition all axioms of
braided Lie algebra are satisfied.
Consider the particular case n = 2. In terms of the "envelop ping algebra"
the relations between the generators eo, e1, e2 are of the form
326 DMITRI GUREVICH

where we assume that a, b from Example 1 satisfy the condition m 2 =

(alb? = q and eigher c = 0 or 9 = O. As result we obtaine a braided
deformation of usual Lie algebras, namely of Heisenberg algebra when c = 0
and of the algebra [el,e2] = 0, [el,eo] = 2cel, [e2,eO] = 2ce2 when 9 = 0
(in fact only the first relation is deformed).
Stress that these relations differ from ones arising from representation
of quantum group Uq (sI2) of spine 1 (see [2]). The last example will be
considered elsewhere from the point of view of our definition of braided Lie
algebras.
Acknowledgements. The athor wishes to thank Max-Planck-Institut
fur Mathematik for hospitality during the preparation of the paper. I am
grateful to Organizing Committee of Second Max Born Symposium for in-
vitation to give a talk. I am also very grateful to Professor J.Donin for
stimulating discussions.

References
[1] M.Dubois-Violette and G.Launer, The quantum group of a non-degenerated bilinear
form, Phisics Letters B, 245 (1990), no.2, pp.175-177
[2] I.Egusquiza, Quantum mechanics on the quantum sphere, Preprint, Cambridge, 1991
[3] D.Gurevich, Generalized translation operators on Lie groups, Soviet J. Contempory
Math. Anal.,18 (1983)
[4] D.Gurevich, Hecke symmetries and quantum determinants, Soviet Math. Dokl., 38
(1989), no.3, pp.555-559
[5] D.Gurevich, Algebraic aspects of the quantum Yang-Baxter equation, Leningrad
Math.J., 2 (1991), noA, pp.801-828
[6] S.Majid and J.Sobeilman, Rank of quantized universal algebras and modular func-
tions, Comm. Math. Phys. 137 (1991), no.2, pp.249-262
[7] A.Polistchuk and L.Posicelsky On quadratic algebras, Preprint, Moscow, 1991
ANYONIC QUANTUM GROUPS
SHAHN MAJID"
Department of Applied Mathematics £3 Theoretical Physics
University of Cambridge, Cambridge CB3 9EW, U.K.

Abstract. We introduce non-standard quantum group structures on the finite groups

IZ. n • These determine non-trivial braidings 1f1 in the category of IZ.n-graded vector spaces.
2'11'llvllwl
The braiding is an anyonic one, 1f1 (v ® w) = e - - n - W ® v for homogeneous elements
of degree lvi, Iwl. This category of anyonic vector spaces generalizes that of super vector
spaces, which are recovered as n = 2. We give examples of anyonic quantum groups. These
are like super quantum groups with ±1 statistics generalised to anyonic ones. They include
examples obtained by transmutation of u q (sI2) at a root of unity.

Key words: supersymmetry - anyonic symmetry - quantum groups - braided category

1. Introduction

One of the fascinating aspects of Clifford algebras is their close connection

with fermionic statistics and supersymmetry. One might ask for correspond-
ing algebraic structures that playa similar role when super statistics are
replaced by more general anyonic statistics. In order to do this, one has to
first understand the algebraic structure underlying anyonic symmetry itself,
which is what we do here using the theory of quantum groups and braided
categories. We then give several examples of 'anyonic' algebraic structures
such as anyonic groups and anyonic quantum groups. The construction of
anyonic matrices, anyonic harmonic oscillators etc. remain for further work.
Our study of anyonic symmetry is an application of the general theory
of algebraic structures living in braided categories introduced in [6][7][8],
recalled briefly in the Preliminaries. The role of ±1 statistics in the super
case is now played by the braiding or quasisymmetry 111. For example, our
notion of braided groups and quantum braided groups are modelled on super
groups and super quantum groups respectively. An interesting feature now,
however, is that since the role of transposition is played by a braiding, many
algebraic computations inevitably reduce to braid and tangle diagrams [7][8].
We begin with our formalization of anyonic vector spaces, which means
for us nothing other than a vector space in which the finite group of order n,
IZ. n , acts. The braiding 111 in the category of anyonic vector spaces is the one
familiar to physicists in the context of anyonic statistics, such as [17]. Hence
• SERC Fellow and Drapers Fellow of Pembroke College, Cambridge. This paper is tn
final form and no version of it will be submitted for publication elsewhere.

327
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 327-336.
© 1993 Kluwer Academic Publishers.
328 SHAHN MAJID

the name. The main result of Section 2 is to identify this as the category of
representations of a certain quantum group C::z~, which we imroduce. We
also give formulae for the anyonic dimension and the anyonic trace.
We then proceed in Section 3 to construct our examples of groups and
quantum groups living in such anyonic categories, i.e. anyonic groups and
anyonic quantum groups. Our first example is an anyonic version of the
symmetry group of an equilateral triangle. Our second is an anyonic version
ofthe quantum group U q ( 8[2) at a root of unity and leads to a simple formula
for its universal R-matrix.
In Section 4 we give the general construction that was used to obtain the
anyonic ones. We note that other generalizations of ::Zn-graded spaces have
been considered in the literature, for example to G-graded spaces (where G
may be non-Abelian), such as [1]. By contrast, our generalization by means
of self-dual Hopf algebras appears to be in a new direction.
I thank A.J. Macfarlane and S. Shnider for comments. This is the final
version of a May 1991 preprint of the same title and much the same content.

Preliminaries
A general introduction to quasitensor or braided categories[4] in the context
of the representations of quantum groups is in [5, Sec. 7]. Briefly, a quasiten-
sor category is (C,®,l,CP, w) where C is a category (a collection of objects
X, Y,"', and morphisms or 'maps' between them) and ® is a tensor product
with unit object 1. cpx,Y,Z : X ®(Y ® Z) -+ (X ® Y) ® Z are associativity
isomorphisms for any three objects and WX,Y : X ® Y -+ Y ® X, the braid-
ing or 'quasisymmetry' between any two. Their appearance in physics in
the statistics of quantum fields in low dimensions was recognized in [3]. The
connection with quantum groups leads to link and 3-manifold invariants[14].
We suppress CP, as well as isomorphisms associated with the unit. Then W
obeys
WX,Y@Z = wx,z 0 WX,y, WX@Y,Z = Wx,z 0 wY,Z, wX,l = wl,X = id. (1)
We work over a commutative field k. Our examples are over C A quantum
group over k in the usual sense means for us a quasitriangular Hopf algebra
(H, b.., E, S, R) where H is an algebra over k, b.. : H -+ H ® H the coproduct
homomorphism, E : H -+ k the counit, S : H -+ H the antipode and R the
quasi triangular structure or 'universal R-matrix' obeying [2]
(id®b..)(R) = R13R12' (b..0P®id)(R) = R 23 R 13 , b.. 0P = R(b..( ))R- 1 (2)

where R12 = R ® 1 etc, and b.. 0P is the opposite coproduct. We have written
the middle axiOlll in a slightly unconventional form but one that generalizes
immpdiately to quantum groups in quasi tensor categories. For an introduc-
tion to quantum groups see [5].
ANYONIC QUANTUM GROUPS 329

The axioms of a quasitriangular Hopf algebra (H, f}., f}. op, f,~, n) in a
quasi tensor category C (a quantum braided group) are just the same except
that f}. and n are defined with respect to the braided tensor product algebra
structure[6]. In the concrete cases below, this is

(a0b)(c0d) = a\f!(b0c)d, a,b,c,d E H. (3)

By this we mean to first apply \f! Hi! to b 0 c and then multiply the result
on the left by a and on the right by d. The definition of f}. op in the braided
case when \f!2 :I id is rather more sUbtle[7] and not given simply by \f! 0 f}. or
\f!-l 0 f}.. If f}. op = f}. then we say that we have a braided group. Algebraic
structures in symmetric (not braided) monoidal categories have been studied
by many authors, such as [15]. The novel aspect of our work is to go further
to the truly braided case.

2. Anyonic Vector Spaces

In this section we study quasi tensor or 'braided' categories associated to
Zn, the finite group of order n. Let 9 be the generator of Zn with gn = 1.
As a category of objects and morphisms we take the category Rep (Zn) of
finite-dimensional representations of Zn. Given an object V of Rep (Zn) we
can decompose it under the action of Zn as
27t'Ja
a=O,I,···,n-l; gl>v=e 1 l v,

Here a runs over the set of irreducible representations Pa of Zn and Va is

the subspace of V where 9 acts as copies of Pa. The action is simply denoted
1>. If v E Va, we say that v is homogeneous of degree Ivl = a.
On this category Rep (Zn) we can now define the non-standard braiding
271'"tl1lllwi
\f!v,w(v0w)=e n w0v (4)
on homogeneous elements of degree lvi, Iwl. This is well known to physicists
in the context of anyons[17]. The phase factors in (4) can be called fractional
or anyonic statistics. We denote by Cn the category Rep (Zn) equipped with
this anyonic braiding. The associativity <P is the usual vector space one.
To my knowledge the structure of this quasi tensor category Cn has not
been systematically studied before. Our main result of this section is to
identify it as the category of representations of a quantum group. Although it
is well known that quantum groups (in the strict sense, with quasitriangular
structures) have quasitensor or braided categories of representations, given
such a category it may not come from a quantum group. Our result is,

PROPOSITION 2.1. Let CZn denote the group algebra of Zn. Thi::; is just
the algebra over C generated by 1, 9 and the relation gn = 1. It is a H opf
330 SHAHN MAJID

algebm with jj.g g 0 g, Eg = = 1, Sg = g-l. Then CZ n has a non-trivial

quasitriangular structure
1
2:
n-l
R = - e_271"~ab ga@l. (5)
n a,b=O

We denote the Hopf algebm CZ n equipped with this non-standard quasitri-

angular structure by CZ~. Moreover, en
= Rep (CZ~). If n ~ 3 then CZ~ is
strictly quasitriangular and en
is strictly bmided.

Proof It is easy to verify that the R shown obeys all the axioms (2) for
a quantum group. The prime in CZ~ is to distinguish it from the usual
group algebra CZ n with R = 11211. We compute the braiding in the cate-
gory of representations of CZ~. Recall, e.g.[5, Sec. 7], that for any quan-
tum group H, the category Rep (H) of representations becomes a qua-
sitensor category as follows: The tensor product of representations V, W
is h~( v 121 w) = E h(l)~V 121 h(2)~W for v 121 w E V 121 W and the braiding is

WV,w(v@w) = 2:R(2)~w@R(1)~v, R == 2:R(1)@R(2)

where ~ is the relevant action. For us, the 121 is as for Zn-representations,
1 21nblwl 21nalvl 2'11'"tab 211'&blwl
andwv,w(v@w)=nEa,benw@enve--n-=Ebenw@volvl,b
21nlvllwl
=e n 121 v on homogeneous elements. We use here and below the or-
W

= Ob,c D
1 1 271".a(b-c)
thogonality of Zn representations in the form n E~~o e n
Now in any quasitensor category with duals (as there are here) there
is an intrinsic notion of rank (V) for any object V, and of Tr f for any
endomorphism f. The Tr f is defined as the morphism

1-- V@V*f~dV@V*Ili~'V*@V __ 1
and rank (V) = Tridv. Here 1 is the identity object in the category (in
our case, the trivial representation c). The definition of trace Tr extends
Ilivv'
further as a morphism Hom(V, V) =
V 121 V* ~ V* 121 V -- 1, where Hom
is the internal hom in the category. For the quasitensor categories Rep (H)
where H is a quantum group, the rank was studied in [5][9). For H = Uq (sI2)
it comes out as a variant of the familiar q-dimension. In general it comes out
as[5), rank (V) = Tr PV(1£) where 1£ E H is 1£ = E(SR(2))nV) and PV(1£) is
the matrix of 1£ acting on V. Likewise if f : V -- V is an endomorphism or
indeed any linear map (viewed as for vector spaces in Hom(V, V) = V 121 V*),
TLf = Tr pV(1£)f. (6)
Because of Proposition 2.1 we can apply this general theory to the quasiten-
sor categories en.
It is also evident from this formula that TLf 0 g = Tr go f
ANYONIC QUANTUM GROUPS 331

for endomorphisms f,g. Note that this is not necessarily true if f,g are not
intertwiners for H but merely linear maps. For example, one can show that
mv(h) 0 pv(g) = mv(S-2g) 0 pv(h) for h,g E H. Because of Proposi-
tion 2.1 we can apply this general theory to en. We have,

PROPOSITION 2.2. The intrinsic category-theoretic rank or 'anyonic' di-

mension dim of an anyonic vector space V in en, and the 'anyonic' trace of
a map f : V ~ V are

L =L
n-l 21!'ta2 n-l 21rta2
dim(V) == rank (V) = e--n- dim Va, Tr f e--n-Tr flva (7)
a=O a=O

where Va is the subspace of homogeneous degree a and flva : Va ~ Va is

the restriction of f to degree a. If f is not degree-preserving we project it to
each Va' If n = 2 we recover the usual super-dimension and super-trace.

Proof We compute
~ 27r.(a+b)b
(}n(a) = L...J e n . (8)
b

To compute ILl, let {ea,"Ya} be a basis of V and {ja,"Ya} a dual basis, where
the ea,"Ya are homogeneous of degree a and 'Ya = 1"" dim Va. By cyclicity of
the ordinary trace, we can apply:!! first. So Tr (J) = ~ L:a (}n( a) L:b L:"Yb fb m (
f(gar>ebm)) = ~ L:a,b {}n( a)e 27r~ab (L:"Yb fb,"Yb(J( eb,"Yb))) giving the result 0

3. Anyonic Quantum Groups

An elementary example of a quantum group living in the category of anyonic
vectors spaces is what we call the anyonic enveloping algebra of the line,
denoted Un(k). This by definition has one generator ~ with

(9)
extended to products of the generators as a homomorphism working in en,
i.e., remembering the anyonic statistics of ~. For example, using'll to take
one ~ past another ~, we have ~e = (~® 1 + 1 ®~? = ®1+1® + e e
e e
~ ® ~ + w(~ ®~) = 01 + 10 + (1 + /~.)~ ®~.
In the remainder of this section we show how to obtain further quantum
groups in the category of anyonic vector spaces by means of the general
transmutation theorem in [7]. This theory applies to quasitensor categories
which are generated as the representations of some quantum group H l .
Proposition 2.1 says that en are of this type with generating quantum group
Hl = C;;Z~. The general transmutation theory says that if H is any ordinary
332 SHAHN MAJID

Hopf algebra into which HI maps by a Hopf algebra map f : HI -+ H, then

H acquires the additional structure of a Hopf algebra H in the quasi tensor
category Rep(H). It consists of the same vector space and algebra as H,
but with a modified coproduct. The vector space of H becomes an object
H in Rep (H) by means of the adjoint action via f. H also has a certain
opposite coproduct and if H is a quantum group (with R) then H has a
quasi triangular structure R in Rep (H). In our case we obtain,

PROPOSITION 3.1. If (H, R) is an ordinary quantum group containing a

group-like element 9 of order n, then it has the additional structure of an
anyonic quantum group H in the category en.
The product coincides with
that of H. The anyonic quantum group structure of H is

where ~b = L b(l) 0 b(2) is the usual coproduct. The action of 9 on H

is in the adjoint representation gc.b = gbg- 1 for b E H and defines the
degree of homogeneous elements by gc.b = e n b. The quantity Rg is the
2"'lbl

quasitriangular structure on C;z.~ as given in Proposition 2.1.

Proof These formulae follow directly from the general formulae in [7]. In
the notation there we are computing H = B( C;z.~, H) where C;z.~ is the Hopf
subalgebra generated by g, equipped with the non-standard quasitriangular
structure given in Section 2. In the result shown it is assumed that all
tensor product decompositions are into homogeneous elements. The second
coproduct ~op specified in [7] is not simply ",-1 0 ~ but has something of
the character of this. It comes out as
I
~ oPb = 2:>- 2"oi b(I)llb(2)1
n b(2)g
-21b
(2l 0 bi!l (10)

where ~b = L b(l) 0 b(2)' This then computes to the form stated. Note also
that 9 itself appears in H with degree 0 0
The transmutation formulae in [7] hold slightly more generally in the sit-
uation where there is a Hopf algebra map C;z.~ -+ H that need not be an
inclusion. We limit ourselves here to giving two examples of the transmuta-
tion procedure. In both of these the map is an inclusion.
Our first example is with H the group algebra of a finite non-Abelian
group containing an element g of order n. To be concrete we take for our
example the group 53, the permutation group on three elements, regarded as
the symmetries of an equilateral triangle with fixed vertices 0,1,2, numbered
clockwise. Let 9 denote a clockwise rotation of the triangle by and let2;
ANYONIC QUANTUM GROUPS 333

Ra denote reflections about the bisector through the fixed vertex a. Let CS3
denote the group Hopf algebra of S3. It has basis {1,g,g2,R o, Rl,R2}' Of
course, there are many ways to work with S3: we present it in a way that
makes the generalization to higher n quite straightforward.

EXAMPLE 3.2. Let H = CS3 where S3 is the symmetry group of an equi-

lateral triangle as described. Its transmutation ~ by Proposition 3.1 is the
following anyonic group in C3 . Firstly, some homogeneous elements are

1 b=2
= "3 L
2".,ab
ra e--3 -Rb,
b=O

Together with 1, g, g2 of degree zero they form a basis of ~ as an anyonic

vector space. Its anyonic dimension is dimS 3 = 2e - ¥ . Its algebra and counit
are those of CS3 but now

Proof The reflections have the property that gRag- 1 = Ra+1 (mod 3).
Hence their inverse Fourier transforms r a as shown are homogeneous of de-
gree as stated. The Hopf algebra structure on 9 (of degree zero) is unmodi-
fied. The usual coproduct in the remainder of CS3 is /::iRa = Ra is) R a , hence
/::ira = Lc rc is) r a- c' This then becomes modified as /::ira = Lc rcg c- a is) ra- c.
Now note that in S3, Rag = gRag- 1 = g<;>Ra for all a. Hence rag = gt>ra =
e 2';'" r a giving the result shown. Likewise, the original antipode on the Ra is
SRa = R;;l = Ra. Hence Sra = ra also. From this and g-l Ra = gt>Ra for
s..
all a (so that g-lr a = e 2 ,;,a ra ) we obtain as shown. The computation for
/::i op is similar to that for /::i and comes out the same. The unmodified n of
CS3 is n = 1 is) 1, so that n = n;l 0
For the second example we consider the quantum groups H = uq (sI2)
defined at q a root of unity as in [14]. Here we refer to the finite-dimensional
versions. They are generated by K, X, Y with relations

K X IC 1 = qt X, KY IC 1 = q-ty, [X, Y] = K; - K~2

q"2 - q-"2

and xr = yr = 0, K4r = 1 with q = e 2;' . There is a coproduct /::iX =

X is) K + K-l is) X etc, and a quasitriangular structure[14]. We work with
this quantum group in an equivalent form with new generators and the
[X, Y] relation taking the form
g4 _ 1
.g = K, F = YK- 1 , qEF-FE= - - .
q-1
334 SHAHN MAJID

8",
EXAMPLE 3.3. Let n = 4r, so q = en and H = U q (Sl2) as described. Its
transmutation U q (Sl2) by Proposition 3.1 is the following anyonic quantum
group in en.
As an anyonic algebra it has generators g, E, F with Igl = 0,
lEI = 2, IFI =-2. The algebra and counit are those of u q (sI2)' but now
Em 0 Fm( 1)2m
L (qm - ) q ('
r-l
!:iE =E 0 g4 + 10 E, !:iF = F 01 + 10 F, n = m=O 1 ... q - 1)

!:i0PE=E01+10E, !:i0PF=F01+g 4 0F, £E=-Eg-\ £F=-F.

Proof This follows from direct computation using the form of the gener-
ators shown. The degree of E, F is from gEg- 1 = e n E and similarly for
2"'IEI

IFI. Since 9 has degree 0 its structure is of course unchanged. The formula
for n was in fact obtained by direct computation from the axioms for an
anyonic quasi triangular structure in U q (Sl2). Proposition 3.1 can then be
pushed backwards to obtain a new expression for n in uq(sh), namely

Emg-2m0Fm(q_1)2m (KXr0(I(-ly)m(1_q-l)2m
n g L
r-l

m=O (qm-1) .. ·(q-1)

= L -'-------:------'----,----'-;----'---==,-,---'--
r-l
n K m=O (1-q-m) .. ·(1-q-l)

Its matrices in the standard representations coincide with those in [14].

ng = nK comes from Proposition 2.1 0
Note in this example the general phenomenon of transmutation: it can
trade a non-cocommutative object uq( sh) in an ordinary bosonic category
into a more cocommutative one (see !:iF, !:i°P E) in a more non-commutative
(in this case anyonic) category. There are plenty of other examples along the
lines of the two examples above.

4. General Construction Based on Self-Dual Hopf Algebras

In this section we briefly describe a general construction from which the
results of Section 2 were obtained. For these purposes we work over an
arbitrary field k of characteristic not 2. Let H be an arbitrary self-dual Hopf
algebra. This means a Hopf algebra equipped with a bilinear form < , >:
H0H ---- k such that < !:ig,a0b >=< g,ab >, < h0g,!:ia >=< hg,a >
etc hold, see [5, Sec. 1]. Now for any finite-dimensional Hopf algebra H
there is a quasitriangular one, D(H) (the quantum double of H) introduced
in [2] and built on H 0 H* with certain relations. In [10] we showed how to
generalize this construction to the situation of dually paired Hopf algebras,
and use this now in the self-dual case. Then D(H) is built now on H 0 H
with the product[10]

(h0 a)(g0 b) = L < Sh(1),b(l) > (h(2)g 0 b(2)a) < h(3),b(3) > (11)
ANYONIC QUANTUM GROUPS 335

and the tensor product coalgebra structure. The antipode is 5( h 0 a) =

(5h0 1)(105- l a).

PROPOSITION 4.1. Let H be an involutory self-dual Hopf algebra and D(H)

its quantum double as described on H 0 H. Then

H'- D(H)
-(h01-10h: hEH)'
n = L:r0ea E H I
0H'
a

is a commutative Hopf algebra, which is quasitriangular at least in the

finite-dimensional case. Here {e a } is a basis of Hand {r} another, dual,
basis of H and we use for n their projections to H'. If 5 does not act as the
identity in H' then n is strictly quasitriangular (so that Rep (H) is strictly
braided).

Proof On D(H) we have ~(h 0 a) = L: hell 0 a(l) 0 h(2) 0 a(2)' hence

1
~(h01-10h) = 2L:(hU)01-10h(l))0(10h(2)+h(2)01)
+(h(l) 0 1 + 10 hell) 0(h(2) 01 - 10 h(2))
where we use the same decomposition ~h = L: hell 0 h(2) for the two h's.
Hence the ideal generated by the relations (h 0 1) = (10 h) for all h E H is
a biideal in the sense of [16, p. 87]. It is also respected by 5 if 52 = id (the
condition that H is involutory). Hence the quotient is a Hopf algebra. The
n shown is just the projection of the one on D(H) found by [2]. We used
the conventions of [11]. That H' is commutative follows from the formula
for the product (in our conventions, D(H) includes H on the right with the
opposite product). Note that in any quasi triangular Hopf algebra one has
(S0id)(n) = n- l which leads to n- l = L:5r0ea' Note that because
H' is commutative, its antipode has square 1 0
We used this construction applied to H = CZ n to obtain the structure
of CZ~ described above. To do this we take for H a basis e a = ga for
a = 0, 1, ... ,n-1. The dual basis can be written in terms of Z:n (the character
group of Zn), which we identify with Zn itself to obtain the self-pairing.
Of course, the input Hopf algebra H need not itself be commutative
or cocommutative. For example, let Tn be the group of upper triangular
matrices in Mn (k) with 1 on the diagonal. Then in [12] we constructed
a self-dual Hopf algebra kTnf3 r:><la(kTn )* by means a certain action Q and
coaction {3. The left Tn factor here plays the role of momentum group, the
other of position space. The Hopf algebra itself is then the quantum algebra
of observables in an algebraic approach to quantum gravitY[12]. Physically,
in this setting Hopf algebra duality corresponds to a reversal of the roles of
observables and states in the quantum system, and in this class of models
336 SHAHN MAJID

the dual Hopf algebra is of the same type with the roles of position and
momentum interchanged.
Finally we mention a variant of Proposition 4.1 which avoids some of
the restrictions there. It applies also to H infinite-dimensional provided the
antipode is invertible and that R makes sense.
PROPOSITION 4.2. Let H be a finite-dimensional antisel! dual Hop! alge-
bra. Then H' = D(H)j(h 01 - 10 h: h E H) is a quasitriangular Hop!
algebra (not necessarily commutative) with R as in Proposition 4.1.
Proof This variant differs in that we now suppose that there is a pairing
< , >: H0H --+ k that obeys < f}.h,a0b >=< h,ba > and < Sh,a >=<
h,S-1 a > for all h,a,b E H (the rest as before). In the finite-dimensional
case this says H~H*op where the latter is H* with the opposite product.
The formulae for D(H) are now similar but with ab(2) rather than b(2)a in
(11) and S(h0a) = (Sh01)(10Sa). This means that both H factors in
H 0 Hare sub-Hopf algebras. In this case H' is always a Hopf algebra and
need not be commutative 0
Few antiself-dual Hopf algebras are known so far. One example of H that
is found to be antiself-dual (as well as self-dual) is U+ in [7, Prop. 2.9].
In this example, H' coincides with H and the last proposition recovers its
known quasitriangular structure as in [13].

References
[1] M. Cohen. Contemp. Math. 43 (1985) 49-61.
[2] V.G. Drinfeld. Quantum groups. In Proc. ICM, pages 798-820, AMS (1987).
[3] K. Fredenhagen, K.H. Rehren, and B. Schroer, Comm. Math. Phys. 125 (1989)
201-226; R. Longo, Comm. Math. Phys. 126 (1989) 217.
[4] A. Joyal and R. Street. Mathematics Reports 86008, Macquarie University (1986).
[5] S. Majid. Int. J. Modern Physics A 5 (1990) 1-91.
[6] S. Majid. In Proc. XIXDGM, Rapallo (1990), Spring. Lect. Notes. Phys 375, pages
131-142; Int. J. Mod. Phys. 6 (1991) 4359-4374; Lett. Math. Phys. 22 (1991)
167-176; J. Math. Phys. 32 (1991) 3246-3253.
[7] S. Majid. Transmutation theory and rank for quantum braided groups. Math. Proc.
Camb. Phil. Soc, in press.
[8] S. Majid. Cross products by braided groups and bosonization. J. Algebra, in press.
[9] S. Majid. Comm. Algebra 18 (1990) 3705-3712.
[10] S. Majid. Isr. J. Math 72 (1990) 133-148.
[ll) S. Majid. Comm. Algebra 19 (1991) 3061-3073.
[12) S. Majid. Phd Thesis, Harvard, (1988); J. Algebra 130 (1990) 17-64; J. Classical
and Quantum Gravity 5 (1988) 1587-1606; J. Funct. Analysis 95 (1991) 291-319.
[13) D. Radford. J. Algebra 141 (1991) 354-358.
[14) N.Yu Reshetikhin and V.G. Turaev. Invent. Math. 103 (1991) 547-597.
(15) M. Scheunert, J. Math. Phys. 20 (1979) 712-720; D.l. Gurevich, Leningrad Math.
1. 2 (1991) 801-828.
(16) M.E. Sweedler. Hopf Algebras. Benjamin (1969).
(17) F. Wilczek, ed. Fractional Statistics and Anyon Superconductivity, World. Sci.
(1990); J. Frohlich and P.-A. Marchetti, Comm. Math. Phys. 121 (1989) 177.
ON S-LIE-CARTAN PAIRS
WLADYSLAW MARCINEK
Institute of Theoretical Physics,
University of Wroclaw
Pl. Maxa Barna 9, 50-204 Wroclaw, Poland

Abstract. The noncommutative differential geometry corresponding to an arbitrary tri-

angular Yang-Baxter operator S is described in a purely algebraical way. The concept of
Lie-Cartan pairs of Kastler and Stora is generalized to the noncom mutative case. All vec-
tor spaces, algebras and modules considered in the paper are in th e symmetric monoidal
category C(S) corresponding to S.

1. Introduction

It is known that in the noncommutative geometry the role of the algebra

of smooth functions COO(M) on a smooth manifold M is played by a non-
commutative abstract associative algebras F [1,2,3). The noncommutative
generalizations of operators d, ix, Lx are defined exactly like in the com-
mutative case [2,3). The essential difference with commutative case is that
the algebra der F of derivations of F is not an F-module in general. It is
very interesting that for some special noncommutative algebras, namely for
S-symmetric algebras Fs, where S is a Yang-Baxter operator, the algebra
of generalized derivations (S-derivations) der Fs is an Fs-module. Hence
for such algebras we can construct the noncommutative geometry in the
complete analogical way like in the commutative case. In order to do it we
use the concept of Lie-Cartan pairs. The concept of Lie-Cartan pairs has
been considered by Kastler and Stora [4) as a purely algebraical frame for
describing operators of classical differential geometry. The generalizations
to the supersymmetric case has been presented by Jadczyk and Kastler [5),
Coquereaux and Jadczyk [6] and generalized further to the colour symmetry
by the authors [7,8,9]. Similar approach has been considered by Matthews
[10). In this paper we generalize the concept of Lie-Cartan pairs to the case
of arbitrary triangular Yang-Baxter operator S. We describe the noncom-
mutative geometry correspond to S in the purely algebraical way like in the
commutative case. All vector spaces, algebras and modules considered here
are in the symmetric monoidal category determined by S [11,12).

337
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Defonnations, 337-342.
© 1993 Kluwer Academic Publishers.
338 WLADYSLAW MARCINEK

2. Symmetric monoidal categories

Let E be a vector space over the field ee of complex numbers. A linear

operator S : E @ E --+ E @ E such that

(1)

and
(2)

is said to be a triangular Yang-Baxter or symmetry on E, here S(l) = S@id,

S(2) = id @ S [12,13). An S-vector space or vector space with symmetry is
a given vector space equipped with symmetry defined above. Let E be an
S-vector space. Then there exist a series of representations

(3)

of the symmetric group Sn defined by

k
p( 7r ) = II S(i) , (4)
i=l

where S(i) = id@ ... @S@ ... @id, (S on the i-th place), 7r = T1, .•. , Tk, Ti are
transpositions and p( Ti) = S(i). Moreover, there exist the rigid symmetric
monoidal category C(S) determined by S. The construction of C( S) has been
described by Lyubashenko in Ref. [12), see also References [14,15). The fact
that C(S) is a rigid symmetric monoidal category means that for every pair
U, W of objects of C( S) we have a family of natural isomorphisms S = Su,w,
U @ W --+ W @ U such that

SU0V,W = (Su,w @ idv) 0 (idv @ Sv,w) (5)

(Sv,W @ idu) 0 SU,v0W (idw @ Su,v) 0 SU0V,W (6)

and
Su,W 0 Sw,u = id w0 u . (7)

The category C(S) contains: the underground field ee, the given vector space
E, the left and right duals of E, tensor products of such spaces, some algebras
such as the S-symmetric algebra F, the algebra of S-derivations, S-Lie
algebras, ... , and some F-modules.
ON S-LIE-CARTAN PAIRS 339

3. S-Lie algebra
An algebra L in C(S) equipped with a bracket [ , ls: ( 0 ( ------> ( such
that
[ , ls = -[ , ls 0 S (8)

[ , ls 0 [ , l~) = [ , ls 0 [ , l~) + [ , ls 0 [ , l~) 0 S(1) , (9)

So [ , l~) = [, l~) 0 S(1) 0 S(2) , (10)

is said to be an S-Lie algebra in C(S) [ , 1, here we have the following
notation [ , 1~1) = [ ,
ls 0 id, [ , l~) id 0 [ , ls, see =
[13,151. An (-module is an S-vector space E E C(S) equipped with an
action (X 0 1) E ( 0 E ------> X fEE such that

[X 0 Yl s f = (0 - oS)(X 0 Y)(id 01) (11)

for X, Y E (, fEE, 0 is a composition map. Let F be an S-symmetric

algebra in C(S) equipped with multiplication m : F 0 F --+ F. The S-
symmetry of F means that m = moS. An S-derivation of F is a linear
mapping X : F ------> F such that

X 0 m = m 0 (X 0 id) 0 (id + S) . (12)

see References [12,161. Obviously the space of all S-derivations der F of F
is an S-Lie algebra

[X 0 Yl s = (0 - oS)(X 0 Y) . (13)
We also have

f(gX) = (fg)X, f(Xg) = (fX)g, (14)

for g, f E F, X E der F. This means that der F is an F-module. Moreover

we have

[X 0 fYls = [, l~) 0 ev 0 S(1) (X 0 f 0 Y) + (X 1)Y

for X,Y E derF, f E F and ev(f 0 X) = fX. Let der F be an S-Lie
algebra of S-derivations of an S-symmetric algebra F. A F-linear mapping
w : (der F)0P ------> F such that

w = (sgn 7r)f 0 p(7r) (15)

for every7r E Sp is said to be skew-S-symmetric of degree p, here p : Sp ------>
end( der F)0p is the representation of the symmetric group Sp in the space
340 WLADYSLAW MARCINEK

(der F)0P. The space of all F-linear and skew-S-symmetric mappings of de-
gree p is denoted by AP(derF,F). Ifw E AP(derF,F) and "7 E Aq(der F,F),
then the S-exterior product w A "7 E Ap+q(der F, F) is defined by

(w A "7) (Xl®"'" ®Xp+q)

(16)
= L: sgmr(w®"7) 0 p(7r)(Xl®, ... ,®Xp+q)
trESp,q

for Xl,'" ,Xp+q E der F, where Sp,q = {7r E Sp+q : 7r(1) < ... < 7r(p) and
7r(p + 1) < ... < 7r(p + q)}. The exterior derivative d : AP (der F, F) ----+
AP+l(der F,F) is defined by

dw(X1®, ... , ®Xp+1)

L:( -l)k+1(XkW) 0 P(7rk)(X1®"" ,Xp+l)

k
(17)

where 7rk(l, ... ,p + 1) = (k,l, ... ,k - 1,k + 1, ... ,p + 1) and

7rkl(l, ... ,k, ... ,I, ... ,p+ 1) = (k,I,1, ... ,k-1,k+1, ... ,1-1,1+1, ... ,p+
1).

4. S-Lie-Cartan pairs
Let B be an S-Lie algebra and let F be an S-symmetric algebra, both
algebras Band F are in the category C(S). The pair (B,F) is said to be an
S-Lie-Cartan pair if the following linear mappings are given

ev : (X ® J) E B ® F ----+ X f E F, (18)

. : (f ® X) E F ® B ----+ f· X E B, (19)
and (i) the mapping (18) defines an S-morphism 8 of B into der F, 8 :
X E B ----+ 8x E der F such that

8x 0 m = m 0 (8x ® id) 0 (id + S), (20)

where m : F ® F ----+ F is the S-symmetric multiplication in F, and

8[X0 Y ]s = 0 (8x ® {)y) - oS (8x ® {)y) S-1, (21 )

ON S-LIE-CARTAN PAIRS 341

where o (ax ® 8y) = ax 0 8y, (ii) the mapping (19) makes B an unital
F-module

f 0 (g. X) = m(f ® g). X, IF' X = X (22)

for f,g E F, X E B, IF is the unit in F, (iii) in addition we have

(f. X)g = f· (Xg) for f,g E F, X E B, (23)

and
[X®gY]s = evo[,]~)oS(1)(X®g®Y) + (Xg)·Y (24)
for 9 E F; X, Y E B. Let us take some examples.
Example 1. The pair (der F, F), where F is an arbitrary S-symmetric
algebra, is an S-Lie-Cartan pair.
Example 2. If S is colour symmetry [13], then the S-Lie-Cartan pair
becomes graded Lie-Cartan pair of Refers [7,8,9].
Example 3. If S == T (the transposition), then we obtain the ordinary
Lie-Cartan pair of Kastler and Stora [4].
Let (B, F) be an S-Lie-Cartan pair and let V be an F-module. An F-linear
mapping V' X : V ----+ V such that

V'xfx = evo(id®V'oid)oS(l)(X®f®x) + (Xf)x (25)

for X E B, f E F, x E V; ev(f ® V' X ® x) = fV' X x, and

V'f·xx = f·V'xx (26)

for f E F, X E B, x E V, is said to be a covariant S-derivative. The mapping

V' : X E B----+ V' X E end(V) such that V' X is a covariant S-derivative for
every X E B, is said to be an S-connection on V. An F-linear mapping
R: (X ® Y) E B ® B ----+ RX0Y E end(V) defined by

RX0Y = o(V'x®V'y) - oS(V'x®V'y)S-l - V'[X0Y]s (27)

is said to be an S -curvature of V'. Next one can define the covariant exterior
S-derivation, the generalized inner derivation or covariant S-Lie derivative
in a similar way.

References
[1] A. Connes, Non-commutative differential geometry, Publ. I.H.E.S. 62, 257 (1985).
[2] M. Dubois-Violette, Derivations et calculi differential noncommutative, C.R. Acad.
Sci. Paris Ser I, 307, 403 (1988)
[3] M. Dubois-Violette, R. Kerner and J. Madore, Non-commutative differential geometry
of matrix algebras, J. Math. Phys. 31, 316 (1990).
[4] D. Kastler and R. Stora, Lie-Cartan pairs, J. Geom. and Phys. 2, 1 (1985).
342 WLADYSLAW MARCINEK

[5] A. Jadczyk and D. Kastler, Graded Lie-Cartan pairs II, Annals of Phys. 179, 169
(1987).
[6] R. Coquereaux and A. Jadczyk, Differential and integral geometry of Grassmann alge-
bras, Rev. Math. Phys. 3, 63 (1991).
[7] W. Marcinek, Generalized Lie algebras and related topics, Preprint ITP UWr No 691
and 692 (1987).
[8] W. Marcinek, Generalized Lie-Cartan pairs, Rep. Math. Phys. 27, 385 (1989).
[9] W. Marcinek, Graded algebras and geometry based on Yang-Baxter operators, J. Math.
Phys. 33, 1631 (1992).
[10] R. Matthes, A covariant differential calculus on the "quantum group" Qj~, Proceedings
of the Wigner Symposium, Goslar.
[11] S. Mac Lane, Categories for Working Mathematician, Graduate Text in Mathematics
5, Springer-Verlag 1971.
[12] V.V. Lyubashenko, Vectorsymmetries, in Seminar on Supermanifolds, No 19, 1 (1987).
[13] D. Gurevich, A. Radul and V. Rubstov, Noncommutative differential geometry and
Yang-Baxter equation, I.H.E.S./M/91/88.
[14] S. Majid, Braided groups, DAMTP/90-42.
[15] D. Gurevich, Quantum Yang-Baxter equation and a generalization of the formal Lie
theory in Seminar on Supermanifolds, Stokholm University, Report No 19, 33 (1987).
[16] A. Borowiec, W. Marcinek and Z. Oziewicz, On multigraded differential calculus,
Proceedings of the First Max Born Symposium, ed. by R. Gielerak et all. Kluwer
Acad. Pub. 1992.
NEW REAL FORNIS OF Uq(Q)

JAN SOBCZYK
Institute of Theoretical Physics
Wroclaw University
Pl. Moxa Borna 9, 50-205 Wroc/aw, Poland

Abstract. We consider different co algebra structures in U q (9) induced by its (as algebra)
automorphisms. We prove that for each obtained Hopf algebra a complete classification
of real forms can be done.

1. Introd uction
Recently many authors [1,4,.5,6,7,11] addressed a question of defining real
forms of complex Hopf algebras. If the Hopf algebra in consideration is a
deformation of the universal enveloping algebra of a simple complex Lie alge-
bra its real forms can be viewed as deformations of real Lie algebras. Taking
into account a fundamental role played in physics by e.g. Poincare algebra or
sue n) such study may be relevant in future applications of quantum groups.
Most often (for other approach see [4]) by a real form one understoods a
morphism cJ> with the following properties:

<1>2 = 1 (1)
<1>( aX + j3Y) = a*<I>(X) + 0*<1>(1") (2)
<I>(XY) = <1>(1") <I>(X) (3)
(<I> ® il»i:l(X) = i:l(<I>(X)) (4)
Then it can be shown [9] that <.I> has to satisfy also

<1>0.5'0<1>0.5'=1 (5)

c: (<l>(X) = (c:(X))* (6)

In the case of quantum deformations of enveloping algebra of a complex
simple Lie algebra g a complete classification of such transformations has
been done by Twietmeyer [11] The classification has been done for a definite
coproduct in Uq(g) namely (e±a aud.h" are elements of Cartan-Chevalley
basis)
i:l(e a ) = (" ® 1 + qha @ e a (7)
Ll( e_ a ) = L-" 8) q-h a + 1 ® e_ a (8)

343
Z. Oziewicz et aI. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 343-346.
© 1993 Kluwer Academic Publishers.
344 JAN SOBCZYK

(9)
It is however well known that there are many coproducts in Uq (9) which with
the same algebra structure make it a Hopf algebra [2,3,8]). Thus a natural
question arises if Twietmeyer classification can also be done for coproducts
different from that given in eq8.(7-9).

2. Main theorem
Suppose we have algebra authomorphisms nand 0- 1 of Uq (9) (here viewed
only as an algebra). It can be easily seen that any such n defines a new
coproduct in Uq (9) from an "initial" one 6.. We put namely

(10)
Since

(1 ® 6.0) 0 6.0 (n @ 0 ® n) 0 (1 ® 6.)' 0 6. 0 n- 1 (lla)

(6.0 ® 1) 0 6.0 (n ® n ® n) 0 (6. (» 1) 0 6. 0 n- 1 (llb)

6. ° is coassociative whenever 6. is so.
We can now state a following

THEOREM
<po := n 0 <p 0 n- 1 defines a real form for Uqun with the coproduct 6.°.
Proof:
We calculate
(<po ® <po) 0 6.0 = (n (3:) n) 0 (<I> (2) <1» 0 (0- 1 (0 n- 1 ) 0 no
= (n (0 n) 0 (<I> GI <1» 0 6. n- 1 =
= (n 'v 12) 0 6. 0 1> 0 n- 1 =
= 6.1/ 0 0 0 <I> 0 0- 1 = 6.1/ 0 <1>0
(12)
Remaining requirements are t ri vially satisfied.

Let us ask now what is a J'l~lation between real forms defined by <p and
<po. In the undeformed case they give rise to two real Lie algebras with
generators

{A: 1>(A) = -A} (13a)

{IJ: (1lI/eB) = -B} (13b)
NEW REAL FORMS OF U Q (9) 345

Observe now that A E Q<l> if and only if il(A) E Q<l>n:

cpO(il(A)) = il 0 cp 0 il- l (il(A)) = il( cp(A)) (14)

We conclude that in the limit q ~ 1 il becomes an isomorphism of two real

Lie algebras. Therefore cp and cpo correspond to deformations of the same
real form of Q.

3. Example
As an illustration ofthe above scheme let us present an example of Uq (sl(3)).
We start with (e±1 and e±2 a.re simple roots)

-1)2

(15a)
qho _ q-h u
rea, Cb] = Oab
q - q- I
(15b)

el+2 = [CI, e2]q == Cle2 - QC2 C l (15e)

[el, el+2]q-l = [(1+2. C2]'1-' = 0 (15d)
The coproduct is given in eqs. (7-9). We can take as il

il(qd = C±(1+2) (16a)

il(e2) = e-lq -hI , il(C-2) qh1el (16b)
il(h}) = hI + h2' n(h 2) = -hI (16e)
It then follows that

A
u
O ()
e_(1+2) = e_(1+2) 6;\
u q
-(hI +h 2 )
+ 1 ® C(1+2) (17b)
~O(e_d = e-l @ qhl + 1 @ C-1 (17e)
~O(el) = el @ 1 + q-h, @ el (17d)
If we look at il it should be clear that the new coproduct acts on el+2 and
e_lq-h 1 as if they were simple roots. III ['act n as q --+ 1 becomes an
element of the Weyl group of .si( 3). It should also be added that ~o can be
obtained [10] from ~ by a nonlri\'ial twisting [2], [8J: ~O(X) = F ~(X) F- 1
(nontrivial in a sense that. F is constructed not only from the elements of
the Cartan subalgebra).
346 JAN SOBCZYK

Acknowledgments
I am grateful to V.Tolstoy and J.Lukierski for many useful conversations.

References
[1] V. Dobrev, to appear in "Quillltum Groups", eds. H.D. Doebner and V. Dobrev,
Springer Verlag 1992 (series Let lIIH' Notes in Physics).
[2] V.G. Drinfeld, Qualltum Groups. Proceedings of the Iliternational Congress of Math-
ematicians, Berkeley 1986.
[3] S.M. Khoroshkin, V.N. Tolstoy. Cal t an- We.l'l basis [or (juan t.ized Kae-Moody (super)
algebras and universal R-matl'ix, to appear in "QuantuJJl GroufJ.s", eds. H.D. Doebner,
V. Dobrev, Springer Verlag 19~2 (series Leetule Notes iJl Pll,Ysies).
[4] V. Lyubashenko, Real and imaginary [arms of quantulll groups, Kiev preprint KPI-
2606.
[5] J. Lukierski, A. Nowicki, H. Ruegg. Phys. Lett. D 271.321 (1991).
[6] M. Mozrzymas, Reality condi{ions for quantum algebras, University Bordeaux 1
preprint, LPTB 92-2.
[7] T. Masuda, K. Mimachi, Y. Nakagami. M. Noumi, Y. Saburi, K.Ueno,
Lett. Math. Phys. 19, 187 (1990).
[8] N. Reshetikhin, Lett. Math. Phys. 20 331 (1990).
[9] M. Scheunert, The antipode of <llld star operations in lIopf algebra, Bonn University
preprint HE-92-13.
[10] V.N. Tolstoy, talk given at Max Born Symposiulll, Wroclaw - Sobotka, September
1992.
[11] E. Twietmeyer, Lett. Math. Phys. 24. 49 (19~2).
RELATED TOPICS
Z3-GRADED STRUCTURES*
RICHARD KERNER
Laboratoire de Physique Theorique
GCR - Universite Pierre et Marie Curie,
CNRS - URA 769
Tour 22, Boite 142
4, Place lussieu
75005 Paris
France

Abstract. We investigate some consequences of imposing the Z3-grading on algebraic

structures. It leads to abandoning of algebras defined by binary relations, introducing the
ternary relations instead. Possible analogs of Lie algebras and Grassmann algebras are
discussed, as well as an example of a simple gauge theory on 3 x 3 matrices. We show also
how a cubic root of super-translations can be defined.

1. Why is Z3-grading beautiful?

Among the swarming multitude of new structures which are under inves-
tigation since a few years, such as non-commutative geometries, quantum
groups, braid groups, and the like, all of which generalize the well known
classical algebraic structures such as Lie algebras and Lie groups by trans-
gressing one of the axioms, we would like to point out one of the possibilities
which consists in replacing the Z2-grading by Z3-grading.
By Z3 we mean the cyclic group of three elements, which can be rep-
resented on the complex plane <C as multiplication by j = e27ri / 3 , j2 and
j3 = 1. This simple group is a subgroup of the group of permutations of
three elements, S3, which contains six elements. It can be also represented
faithfully on complex plane <C if we add the involution, which is complex
conjugation. Then the two other involutions are generated by composition
with cyclic elements. Here is the full representation of 53 in the complex
plane:

• Dedicated to Prof. Jan Rzewuski on his 75-th birthday

349
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 349-356.
© 1993 Kluwer Academic Publishers.
350 RICHARD KERNER

Pennutation
ABC (ABC)
(ABC) BCA (ABC)
CAB (ABC)
CBA (ABC)
BAC (ABC)
ACB
Complex j p 1\
*
representation
complex reflexion reflexion
conjugation inp in j

There is another representation obtained by replacing j by j2 and 1\ by *.

S3 is the last group of permutations that possesses faithful representations
on the complex plane. S4 has an unfaithful representation, whereas S5 and
higher do not have representations in <C (besides the trivial one, reducing
everything to S2)'
In the case of Z2 we could not distinguish between Z2 and S2; here Z3
is abelian, whereas S3 is not. Nevertheless, the possibility of representing
it faithfully in the complex numbers suggests that some important ternary
symmetries of wave functions depending on three variables should be inves-
tigated. Perhaps these symmetries will be adequate to describe quarks?

2. Z3-graded derivations, Z3-commutators

Consider a free associative algebra with unit element, over complex numbers,
generated by a finite number of elements a, b, c, .... It can be naturally Z3-
graded if we define <C and the unit element as grade 0, the generators as grade
1 elements, their binary products as grade 2, and afterwards, any product
of p elements as being of grade [p mod 3].
A Z3-graded derivation of grade k of such an algebra is a linear mapping
from this algebra into itself satisfying the generalized Leibniz rule:

il (ab) = (il a) b + l grade(a} a (il b) . (2.1)

with
(k) )
grade ( D a = [grade(a) - k] mod 3 (2.2)

k = 0,1,2.
It is easy to prove that for k = 1 or 2 the third power of any such
derivation vanishes.
Now, contrary to the Z2-case, no binary relations can be imposed on the
products in our algebra. Indeed, suppose that among the generators aO/ some
binary relation exists, given by
(2.3)
Z3-GRADED STRUCTURES 351

Then, deriving it with respect to aa, one gets

(2.4)

which implies
(2.5)

possible only if U a {3 == O.
On the contrary, ternary relations may be imposed, compatible with our
derivation.
They should satisfy

L U a{3'Y aa a{3 a'Y o (2.6)

a,{3,'Y

(2.7)
because the binary products aaa{3 are linearly independent by virtue of (2.5).
One of the simplest solutions to (2.7) is of course U a{3'Y == 1, because 1 + j +
j2 = O.
The binary relations that define the Grassmann algebra in the Z2-graded
case, i.e. aaa{3 + a{3aa = 0 should be replaced by a ternary relation. We have
the choice, depending on the interpretation of the antisymmetry as a Z2 or an
S2 group average, between the following two possibilities of generalization:

(2.8)

i.e. a Z3 invariance property, or

o (2.9)
which is an S3-average without any weight.
The formula (2.8) gives rise to a finite algebra whose dimension depends
on the number of generators N as (N+1)3;(N+1) + 1; e.g.

one generator : 1l,a,a 2 ; a3 = 0 (2.10)

two generators: ll, a, b, a 2 , ab, ba, b2 , ab 2 , a 2 b (2.11)

etc.
The formula (2.9) defines an infinite algebra. The finite algebras defined
by (2.8) are incompatible with the graded derivations except for the sim-
plest case (2.10) given by one generator; the structures defined by (2.9) are
compatible with the graded Leibniz derivation.
352 RICHARD KERNER

The simplest algebra consisting of 11, a, a 2 admits three Z3-graded deriva-

tions defined as follows:
o
o (2.12)

o a

The Z3-grades of these derivations are 1,2 and 0 respectively. They do

not form an algebra, i.e. no binary combination yields another Z3-derivation;
instead, they close under the following ternary rule:

(2.13)

This is an example of a ternary algebra, that we shall encounter in other

realizations, too.
Consider now a Z3-graded commutator defined as follows:

(2.14)
where A, B are elements of a Z3-graded associative algebra, a, b their re-
spective grades.
Let 1] be of grade 1. Then we can define a Z3-grade 1 differential as

dA = [1], Ab = 1]A - ja A1]. (2.15)

Of course, d 2 -::I 0, but d3 = 0 if 1]3 commutes with all element of our algebra.
There is no analog of Jacobi identity for the Z3-graded commutator; nor
the derivation d is a derivation of the commutator algebra. Instead, one has
the following identity:

[{A,B,C}D]Z3 + [{B,C,D},A]Z3 +
(2.16)
+ [{C,D,A},B]Z3 + [{D,A,B},C]Z3 0

where {A,B,C} = [[A,B]Z3,C]Z3 + [[B,Cb ,A]Z3 + [[C,A]Z3,B]Z3'

and only if all the items have the same grade.
The simplest representation of such an algebra is given by 3 X 3 complex
matrices separated into three linear subspaces with grades 0, 1 and 2:

(2.17)
Z3-GRADED STRUCTURES 353

There are three linearly independent grade 1 derivations,generated by a

Z3-commutator with the matrices

0 0
"11= ( 100,"12=
o 1
1)
0
(0
j
0
0
00
j2 0
1) ,"l3=
(0 0 1)
POO.
0 j 0
(2.18)

If we put alA = "11A - ja A "11 etc., we can easily prove that

(2.19)

and, of course, Eff == 0, a? == 0, a~ == o.

Ternary composition rules displaying nice representation properties with
respect to Z3 or 53 permutations can be easily defined on free associative
algebras over C. For example, one can define

{a, b, c} =
dj
abc + j bca + 2
j cab. (2.20)

Obviously, {a, a, a} == 0, and

{a,b,c} = j2 {b,c,a} = j {c,a,b}. (2.21 )

We don't know if an analog of Ado's theorem for the Lie groups can be
proved for the Zrcase, namely, whether any ternary rule satisfying (2.21)
and perhaps some 4-linear analog of Jacobi identity may be realized by
embedding in some associative algebra like in the formula (2.20).

3. Z3-graded gauge theory

Consider the simple model of a Zrgraded algebra provided by the 3 X 3
complex matrices (2.17). Let us choose first one of our grade 1 differentials,
e.g. the Z3-graded commutator with "11. One can easily prove that in our
algebra
Im(d) = K er(d 2 ), Im(d 2) = Ker(d). (3.1 )
A covariant differential can be introduced on a free left module over our
algebra; in such a case, as any element of the module can be represented by
an action of some algebra element on a fixed element of the module, it is
enough to define our covariant differential on the algebra itself.
Let
DB = dB + AB (3.2)
where A E A 1 •
We have
D2B = (d + A)(dB + AB) =
(3.3)
d2 B + AdB + (dA) B + j AdB + A2B
354 RICHARD KERNER

which can not be reduced to the left action of some element on B. However,

(3.4)

because of d3 B == 0, and 1 +j +P = O. (This is to be composed with dA +A 2

in Z3-graded case).
We can define the Z3-curvature of the connection A as

(3.5)

o is an element of Ao.
The natural question to ask is what are the connections that have no
curvature, i.e. the flat ones? The answer is easy to compute: if

0 0
A = ( f3 0
o I

J.
then

!l = [(a + I)(fi + 1)(1 + I) - II (' I (3.6)

Consider now a gauge transformation defined by

B ---+ U- 1 B (3.7)

and
A ---+ U-1AU + U-1dU = A' (3.8)
with U any non-singular 3 x 3 matrix of definite Z3-grade. The covari-
ant differential undergoes a usual transformation only if U E Ao; if not, it
transforms as
(3.9)
where u = grade of U.
Nevertheless the curvature 0 transforms covariantly whatever the grade
of U:
0 ' = U-10U. (3.10)
The action is very poor: if U E A Q , 0' = 0; if not, the diagonal matrix n
will undergo a cyclic permutation of its three entries.
One could imagine the generalization of action if a hermitian product
could be introduced:

(OJO) > 0 if n ~ O.
Z3-GRADED STRUCTURES 355

This can be done if we introduce the notion of hermiticity for our matrices.
This new 3-linear curvature may serve for defining a cubic root of a Zr
graded derivation: suppose that the entries a, (3" in the connection matrix
A are replaced by some differential operators, and suppose that we want
to keep only the linear part of (3.5). Now, as a,(3" do not commute, our
formula for n becomes

n = (
(a+I)(f3+~I)h+I)-I 0
(13 + I)h + I)(a + I) - I o ) .
o h + 1)( a + 1)(13 + I) - I

(3.6)
Keeping the linear part a + (3 +, on the main diagonal means that the
following identities must hold:

a(3 + (3, + a, 0, a(3, 0,

(3, + ,a + (3a 0, fha 0, (3.7)

,a + a(3 + ,(3 0, ,a(3 0.

It is not difficult to realize these relations if we put

a = (3 =, = (AVI + j.lV z) , (3.8)

VI, V z being the two Z2-graded nilpotent supersymmetric derivations, A

and j.l arbitrary numbers. Because of

vi = 0, V~ = 0, V1V z + VZV I = ° (3.9)

we shall have

n (.IV, + ~V,) ( 1 1

This means that we have found the cubic root of the supersymmetry
J (3.10)

translations.
Note that we can generalize our scheme by considering three indepen-
dent exterior derivations induced by 'T/I, 'T/z, and 1]3, and the corresponding
covariant derivations DI, Dz, D3 .
There is enough space then to accommodate other Zz-graded derivations
Vi, Vi, and to find their cubic roots, too.
Our scheme can be now resumed as follows: D I , D z , D3: Z3-graded "cubic
roots" of the Zz-graded supersymmetric translations Va, V/3 (a,/3 = 1,2).
Va, V/3: "square roots" of ordinary translations contained in the Dirac op-
erator:
(3.11)
356 RICHARD KERNER

and finally, the Dirac operator being the "square root" of the Klein-Gordon
operator.
It is tempting to think that the equations

etc. (3.12)

are the analogs of Dirac equation for the entities that could be identified as
quarks.
JORDAN FORM IN ASSOCIATIVE ALGEBRAS

GARRET SOBCZYK
Universidad Nacional Aut6noma de Mexico, FES-C
Apartado Postal #25, Cuautitlan !zcalli 54700, Estado de Mexico
[email protected]

April 30, 1993

Abstract. The Jordan form of an element in an associative algebra over the real number
field is uniquely determined by special generators of the factor algebra of its minimal
polynomial.

Key words: Jordan form - Minimal polynomial - Factor Algebra

1. Introduction

Augustin Cauchy observed in 1847 that the real factor ring lR[Al/< A2 + 1 >
of the principal ideal < A2 + 1 > is isomorphic to the algebra of complex
numbers. In (Sobczyk 1993), we defined a <T-algebra <T{ m1,' .. , m r } which
is isomorphic to the complex factor ring <T[A]/< 'I/J > for a given polynomial
'I/J, and used this result to find the Jordan form of an element in a <T-algebra
AD In this paper, we define an lR-algebra which is isomorphic to the factor
ring lR [A]/< 'I/J >, and find the related Jordan forms.
As examples, we find relevant Jordan forms for elements having minimal
polynomials of degree four or less. These canonical forms make it possible
to extend the domain of any function J to a domain D A C A, where J :
DA --+ A, if the roots of the minimal polynomials of each x E DA are in D,
(Sobczyk 1993).

2. The Algebra lR[A]/<'I/J>

Let {m1"'" m r } and {Q1, ... , Qs} be two sets of non-decreasing positive
integers with only the first h and I of them == 1. We allow the possibility that
one of these sets may be empty. For the set of r distinct real numbers {Ai E
lR; i = 1, ... ,r} and the set of s distinct pairs of real numbers {(O'J,,8j) E
lR 2 ; j = r+ 1, ... , r+s, and,8J > OJ, the most general polynomial 'I/J E lR[A]
can be written in the irreducible form

r+s
II(A - A;)m. II
r
'I/J == [(A - O'J)2 + J3}lQ). (1)
i=l J=r+1

357
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 357-364.
© 1993 Kluwer Academic Publishers.
358 GARRET SOBCZYK

We now extend IR to an associative and commutative IR-algebra IR{ ml,

... , m r ; qr+l,.··, qr+s} which we will show is isomorphic to the factor alge-
bra IR [A)/< tP> .

DEFINITION 1. The set of elements {uj,vj,nt,vknD, for the range of in-

dicies

{I ::; i ::; r + s, r + 1 ::; j ::; r + s, h + 1 ::; k ::; r an d 1 ::; t < mk ,

or r + l + 1 ::; k ::; r + sand 1 ::; t < qd,

make up a basis of an associative and commutative deg( tP )-dimensional IR-

algebra

where the operations of addition and multiplication of the basis elements are
specified by

{Ul + ... + ur+s = 1, UjUj = b,juj, v% = -uk, vJu J = v J'

n;;'k- 1 f- °
but n;;'k = 0, or ni/- 1 f- °but nkk = 0, nkuk = nd.
The elements Uj make up a partition of unity and are mutually annihilat-
ing idempotents. The elements nk are nilpotents with the respective indexes
mk or qk. The nilpotents nk are projectively related to the corresponding
idempotents Uk. The pseudo imaginaries Vj are projectively related to the
corresponding idempotents Uj. We adopt the convention that nj == 0, for
i = 1, ... , h, and for i = r + 1, ... , r + I.
We have dimR IR{ m}, ... , m r; qr+l,··· ,qr+s} = L mj+2 L qj = deg( tP)·
We have IR{I;} = IR, and IR{; I} = (t. Other examples are IR{l, I}, which
are called the real unipodal numbers in (DGM-Sobczyk 1992, p.397), and
IR{I, ... , 1; 1, ... ,1} = IR r X (ts. It is interesting to note that IR{mt, ... ,mr ;
qr+}, ... , qr+s} can be algebraically extended to a Clifford algebra, (Sobczyk
1992, p.57).

3. The Isomorphism Theorem

The investigation of the structure of a linear operator has been considered by

many authors,e.g., (Gantmacher 1960, p.175-214), and (Greub 1967, p.368-
427). The following Theorem and its Corollary provide new tools in this
investigation.

THEO REM 1. The algebra IR {ml, ... , m r ; qr+ 1, ... , qr+s} is isomorphic to
the factor ring IR [AJ/< tP > for the polynomial tP given in (1).
JORDAN FORM 359

r r+s
W == l)AiUi + ni) + L (ajuj + {3jVj + nj) (2)
i=1 j=r+l

Let {i, j, ... , j - I+deg,p} denote the standard basis of IR [A] I < 'lj; >. The
algebra isomorphism

is defined by

-k <I> k
A f-+ W

for k = 0,1, ... ,-1 + deg'lj;.

The proof is completed by observing that the determinant of the mapping
<p between the basis elements of IR [A l/ < 'lj; > and IR{ ml, ... , m r ; qr+I, ... ,
qr+s} is non-zero, which is a consequence of the fact that this determinant
involves the first mk - 1 linearly independent derivatives of the functions
Amk - 1 , together with the first qk - 1 linearly independent derivatives of the
functions AQk - 1 , evaluated at the distinct roots of the polynomial 'lj;.
Q.E.D.
The importance of the isomorphism between these two algebras is that
we can find elements in IR [A]I < 'lj; > which have the same multiplication
rules as the basis elements Ui, Vj, nk E IR{ ml,'" , m r ; qr+b"" qr+s}' We
state this in the following

COROLLARY 1. Their exist polynomials

which have the same multiplication rules as do the elements Ui, vJ ' nk E
IR { m 1 , ... , m r ; qr+ 1 , ... , qr+ s } .

Proof: Define

Q.E.D.
We classify the algebras IR{ ml,"" m r ; qr+l, .. . , qr+s} into Jordan types
according to the various possi bili ties for the sets {ml, ... , m r } and {qr+I, ... ,
qr+s}' We say that IR{mt, ... ,m r ;qr+l, ... ,qr+s} is
360 GARRET SOBCZYK

TYPE a) if {qrH, ... ,qr+s} = {},

TYPE b) if {mI, ... ,m r }={},

and of mixed

Each ofthese types is further broken into Jordan subtypes I, II, III, ...
according to

Algebras of Jordan Type a) can be considered as a special case of the

algebras <L'{ ml, ... , m r } studied in (Sobczyk 1993), and the formulas de-
rived there by a different method apply without modification. We shall use
Theorem 1 and its Corollary to study representatives of the 17 algebras
m{ ml, ... , m r; qr+I, ... , qr+s} for which deg( '¢) == 2: mi + 22: qj ~ 4, and
their relationship to the corresponding factor rings m
[A 11 < '¢ >.

4. The algebras m{mI, ... , m r ; qr+I, ... ,qr+s} for which deg '¢ ~ 4.

The following is a complete list of the various possible 17 Jordan types:

I a): {ml, ... ,mr } = {I}, {I,I}, {I,I,I}, {I,I,I,I},

II a): {ml' ... ' m r } = {l, 1, 2}, {I,2}, {2,2}, {2},

IlIa): {mI, ... ,mr }= {I,3}, {3},

IVa): {ml, ... , m r } = {4}.

I b): {qr+ I, ... , qr+s} = {I, I}, { 1},

lIe): {2;I}.

TYPE I a): As representative of this class we choose {mI, ... , m r }

{I, 1, I}. By theorem 1,
JORDAN FORM 361

for the algebra isomorphism

AL W == AlUl + A2U2 + A3U3 E 1R{1, 1, I}.

Taking powers of wE 1R{1, 1, I} gives

{Ul + U2 + U3 = 1, AlUl + A2U2 + A3U3 = W,

Aiul + A~U2 + A5u3 = w 2 ,}
which can be considered as a system of linear equations in Ui. Solving for
Ui == Ui( w) gives

(w - A2)(W - A3)
Ul( w) = (AI - A2)( Al - A3)'

(w - A1)(W - A2)
U3(W) = (A3 - A1)(A3 - A2)"
(3)

By corollary 1, the elements

have the same multiplication rules as the corresponding mutually anihiliat-

ing idempotents Ul,u2,u3 E 1R{1, 1, I}. A similar construction can be found
in (Turnbull and Aitken 1955, p.163), and in (Herstein 1969, p.48).

TYPE III a): As representative of this class we choose {m1, ... ,mr } =
{1,3}. By theorem 1,

for the algebra isomorphism

Taking powers w E 1R{1,3} gives

{Ul + U2 = 1, Al u1 + A2U2 + n2 = w,

Aiu1 + A~U2 + 2A2n2 + n~ = w 2, A~U1 + A~U2 + 3A~n2 + 3A2n~ = w 3},

which can be considered as a system of linear equations in UI, U2, n2, n~.
Solving this system gives
362 GARRET SOBCZYK

(4)
By corollary 1, the polynomials

have the same multiplication rules as the corresponding elements Ut, U2, n2 E
1R{1,3}.
Alternative formulas for Jordan Type a) have been given in (Sobczyk
1993). We apply the more general techniques of this paper to an example of
Jordan Types II c).

TYPE II c): The single member of this class is {Qr+1,' .. , qr+s} = {2j 1}.
By theorem 1,
1R{2; I} ~ 1R[All < (A - Ad 2[(A - 02)2 + f3~l >,
for the algebra isomorphism

j t... w = A1U1 + nI + 02U2 + f32v2 == w E 1R{2j 1}.

Taking powers of w leads to the system of linear equations

{U1 + U2 = 1, A1U1 + n1 + 02U2 + f32v2 = W,

AiuI + 2A1nI + (o~ - f3Du2 + 202f32v2 = w 2, and
Aiu1 + 3Ain1 + (o? - 302f3~)U2 + (30~f32 - f3i)V2 = w3 }
in U1, VI, U2, V2. Solving this system for U1 == U1 (w), nI == n1 (w), U2 = U2( w),
and V2 == V2( w) gives

[(w - 02)2+ f3i][(2w + 02 - 3A1)(02 - AI) + f3il

U1 = [(AI - 02)2 + f3iF

(w - Ad[(w - 02)2 + f3i][(Al - 02)2 + f3il

n1 = [(AI - 02)2 + f3iF

(w - At)2[(Al - 02)(2w + Al - 30 2) - f3il

U2 = [(>.1 - 02)2 + f3~F
(w - Ad 2{(w - 02)[(A1 - 02)2 - f3~l- 2(A1 - 02)f3n
V2 = f32[(Al - 02)2 + f3iF
By corollary 1, the polynomials

have the same multiplication rules as the elements Ut, U2, n2 E 1R{l,3}.
JORDAN FORM 363

5. A Matrix Example
We shall find the transition matrix and Jordan normal form of the matrix
969 -148 -752
1150 )
40 -5 -32 46
a= (
937 -143 -729 1116
-195 30 150 -229
which has the minimal polynomial 1jJ = (>. - 1)2[(>. - 2)2 + 3 2] of Jordan
Type II c). Letting w = a in the formulas given for this type in the previous
section, we find the Jordan form

a = Ul + nl + 2U2 + 3V2
for the matrices

39
( 54
-8
-5
-42
-30 6446 ) (llS
472
-18
-72
-92
-368
140)
560
Ul = -209 32 163 -248 ,nl = 59 -9 -46 70
-177 27 138 -210 0 0 0 0

C -64) COl
8 42 -46 -234
-39 6 30 -46 -131 20 102 358 )
-156
Uz = 2093 -32 -162 248 ,V2 = 223 -34 -174 266
177 -27 -138 211 -124 19 96 -147

Denote the column vector Cl [1,0,0,0]. From this column vector we

construct the generalized eigenvectors of the matrix a,

The transition matrix c is constructed from these eigenvectors;

Carrying out these calculations, we get

118 54 301
( 472 39 -131 -53)
-39
c = 59 -209 223 209 .
o -177 -124 177

0~
U sing the transition matrix c we calculate the Jordan normal form

,-'a, =
oo 0)
0
2 3
-3 2
of the matrix a.
364 GARRET SOBCZYK

Acknowledgements
I greatly benefited from discussions with Professor Jaime Keller and want
to thank him for inviting me to Universidad Nacional Aut6noma de Mexico
where this work was written. Calculations were done with the help of Stephen
Wolfram's MATHEMATICA. The author wishes to thank the Organizers for
an excellent Conference.

References
Gantmacher, F.R.: 1960, Matrix Theory, Vol. 1, Chelsea Publishing Company, New Vork.
Greub, W.H.: 1967, Linear Algebra, Third Edition, Springer-Verlag, New Vork.
Herstein, LN.: 1976 Rings with Involution, Chicago Lecture Series in Mathematics, Uni-
versity of Chicago Press, Chicago.
Sobczyk Garret: 1993, 'Jordan Form in Clifford Algebras', in Clifford Algebras and their
Applications in Mathematical Physics, Proceedings of the Third International Clifford
Algebras Workshop, Edited by Fred Brackx, Kluwer, Dordrect.
Sobczyk Garret: 1992, in Catto, S. and A. Rocha, ed(s)., Differential Geometry Methods
in Theoretical Physics, , World Scientific Publ. Co., Singapore, pp.397-407.
Sobczyk Garret: 1992, 'Unipotents, Idempotents, and a Spinor Basis for Matrices',
Advances in Applied Clifford Algebras [2], No. l.
Turnbull H.W. and A.C. Aitken: 1969, An Introduction to the Theory of Canonical
Matrices, Dover Publications, Inc ..
UNIFIED THEORY OF SPIN AND ANGULAR
MOMENTUM
LEOPOLD HALPERN
Department of Physics
Florida State University
Tallahassee, Florida 32306-3016
USA

The requirement that a physical theory be expressible in a mathematical

form that is not contradictory within its system of axioms and the fact
that we have no theory of truly everything, together, lend support to the
assumption that we are bound to discover in any good (and therefore clear)
theory "so great an absurdity that no reasonable person can believe in it."
This should result from the limitations on mathematical axioms and physical
assumptions we have to impose.
We consider Einstein's general theory of relativity as probably the best
macroscopic physical theory and we assume here tentatively that an absur-
dity of the prescribed kind is found in its prediction, that a spherical cloud
of dust of uniform density and sufficient extension will have to contract ulti-
mately to a point, irrespective of what the physical short range interaction
between the dust particles may be. The adopted point of view that situations
should occur where either the physical assumptions or the mathematical ax-
ioms or both no longer apply, is in general not shared by most physicists
- notably in the case of general relativity, where the Einstein-Hilbert field
equations successfully predicted a large domain of physical reality.
Schrodinger's discovery of the "alarming phenomenon" of pair creation
in the non-static external gravitational field of the expanding universe [E.
Schrodinger, 1939] and the analogue of the effects of virtual elementary par-
ticle pairs which give rise to the Lamb shift [R. Utiyama, 1962]'[R. Utiyama
and B. De Witt, 1962] indicated a possible way to avoid the absurdity by
supplementing the Lagrangian of general relativity: "f§ R by a linear super-
position with terms nonlinear in the curvature [L. Halpern, 1967]:

v'9 (aR + bR 2 + CRiklm Riklm) (1)

The coefficients of this expression remain unknown - even if the removal of

divergent terms had been solved unambiguously, one is not able to estimate
the effects of all the elementary particles involved.

365
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 365-370.
© 1993 Kluwer Academic Publishers.
366 LEOPOLD HALPERN

The field equations of the above modified Lagrangian admit all the vac-
uum solutions of general relativity. No other solutions are known that could
remedy the discussed problem.
C.N. Yang suggested a gauge theory of gravitation with GL(4,IR) as the
gauge group [C.N. Yang, 1974]. He considered field equations for the vacuum
of the form:

(i,j,k: 1,2 ... r) (2)

These equations admit obviously all the vacuum solutions of the Einstein-
Hilbert equations; they were however shown to admit other, unphysical,
solutions. Yang's equations, due to Bianchi identities, are also expressible in
the form:

(3)

Expressed in the curvature two-form of Riemannian geometry with the

Christoffel connection, equation (3) may be called the Riemannian ana-
logue of Maxwell's electromagnetic vacuum equations. Yang has somehow
excluded torsion from his theory, which should however appear naturally in
a gauge theory of G L( 4, IR).
The present author suggested a gauge theory of gravitation (or more gen-
erally of space-time geometry) for which in its simplest form the gauge group
is the subgroup SO(3,1) ofGL(4,IR), which results in the same curvature
[L. Halpern, 1980],[L. Halpern, 1984].
The principal fibre bundle of the manifold of the anti-De Sitter group
G = SO(3, 2) and its subgroup H = SO(3, 1) (proper Lorentz group) is:

P(G,H,G/H, rr) (4)

with G / H the space of left cosets and the natural projection rr : G -+ G / H.

With the Cartan-Killing metric / on G the metric on the base manifold
B = G / H is obtained by the projection 9 = rr'/; with this metric B becomes
the Anti-De Sitter universe.
The metric / for any n-dimensional semi-simple group fulfills Einstein's
equations with a cosmological member:

1 (n - 2)
Ruv - "2/uv R + /uv 8 =0 (5)

and in our case so does 9 (with a different value of the cosmological mem-
ber). This suggests considering the group manifold with / as the vacuum
solution of a special kind of Kaluza-Klein theory which has the Anti-De Sit-
ter universe as the base manifold with the metric 9 of space-time. There are
UNIFIED THEORY OF SPIN AND ANGULAR MOMENTUM 367

other vacuum solutions with non-vanishing torsion. Cartan has made the
interesting suggestion to relate torsion to spin; the present formalism indi-
cates a bold generalization of this idea, associating all properties of matter
to torsion. The restrictions of this presentation do not, however, allow dis-
cussion of the lengthy general features of equation (5) in our theory with
n = 10. We discuss here only the theory with vanishing torsion but with a
right hand member of equation (5) for the matter source.
The Riemannian curvature is then the gauge field, yet expressible in terms
of the metric g of the base manifold. The analogue of the charge in this
theory is related to elementary particle spin which is convertible into orbital
angular momentum. The connection as well as g is determined by the metric
,. Horizontal vectors are perpendicular to vertical vectors.
We express now the left hand side of equation (5) in terms of the curvature
tensor on the base which we denote also by B. We work in an orthonormal
frame in which horizontal vectors are labeled by capital indices A . .. J( and
vertical vectors by L ... Q. The Einstein summation convention is applied to
each of these separately and also to indices R ... Z which extend over all the
ten components. This convention will henceforth be used without further
warning.
We obtain for the vertical M - N component of the expression (5) the
term:

CJjB' c)5E are structure constants of the group G = SO(3,2).

The vertical component ofthe metric in the original version ofthe Kaluza-
Klein theory was kept constant and thus the analogue of the above expression
would not result from a variational principle and would not be considered for
the field equations. Without sources it would also yield unphysical results.
Requiring the present term (6) to vanish without source would simplify the
remaining equations by elimination of many terms quadratic in the curva-
ture. The gravitational collapse of a cloud of dust seems however not to be
avoidable in this case.
The mixed vertical-horizontal components of equation (5) are expressible
as

1 M 1 [
2REMCAB == 2BABE ;[ (7)

this is nothing other than Yang's term (3) which thus forms part of the
present equation (5). The equations (7) have in our case a source in the
presence of elementary particle spin. This source is the analogue of the
charge and (7) is the analogue of Maxwell's equations. A particle with spin
368 LEOPOLD HALPERN

has in the test particle approximation a vertical component of the world

line's velocity vector.
The horizontal component of equation (5) is

This expression contains besides the Einstein tensor of space-time with

cosmological member, also nonlinear terms which are the analogue of the
energy-momentum tensor of the electromagnetic gauge field. Likewise, this
term is covariantly conserved only together with a term containing the in-
teraction of the spin current with the curvature.
The appearance of our nonlinear term in (8) is nevertheless surprising
because the Einstein term itself has been shown to decompose into the lin-
ear Fierz-Pauli spin two-wave operator with the energy-momentum complex
of the self-interacting spin two-field as source [A. Papapetrou, 1954]. The
present additional term bilinear in the curvature should be related to the
interaction with the curvature of the gravitational field's own spin current.
The spherical symmetric vacuum solution of general relativity is only a
solution of the horizontal and the mixed terms (7,8) but not of the vertical
term (6). A similar feature occurs, as mentioned, in the five-dimensional
theory. The mixed equations (7) and the nonlinear term in the horizontal
equations (8) modify the solutions in the presence of matter, even if matter
is spinless. Modified solutions of this kind have not yet been obtained.
Any representation of a group G is equivalent to a functional realization
in terms of the parameters of the group manifold [F. Bopp and R. Haag,
1959]. Representations of integer as well as half integer spin can thus be
expressed this way on the manifold of our pseudo orthogonal groups G and
H.
Dirac has constructed De Sitter covariant wave equations on the manifold
of the De Sitter universe, using only the generators of the group as differen-
tial operators and for half-integer spin, besides this matrix representations
of the Clifford algebra in four-dimension [P.A.M. Dirac, 1935].
The scalar wave equation can in our formulation be expressed with the
generators on the group manifold as:
RS - -
I AR As't/J = 0 (9)

't/J is a trivial realization of H. Dirac showed that in the limit of increasing

radius of the universe (decreasing magnitude of the cosmological member),
only the terms with the generators of the De Sitter translations (in our case
the AE ) contribute significantly.
Our equation (9) is formulated on G, not on G / H; it is tempting to replace
in this case also the generators of G for which Dirac used elements of the
UNIFIED THEORY OF SPIN AND ANGULAR MOMENTUM 369

Clifford algebra, by generators on the group manifold. The left invariant

generators AM appear indeed suited for this purpose, as they act only on
the fibres and not on the base and commute with all AR • The horizontal
AE have not the former property - but they are the only, remaining of
importance after the limit. In order to have 10 generators of the required kind
one would have to consider a higher dimensional group; G = SOC 4,2) and
H = (SO(3,2) with a five-dimensional base G/H on which electromagnetic
theory can be formulated as a sub Kaluza-Klein theory as indicated in [L.
Halpern, 1992]. In the present case of G = SO(3,2) one can however use
the six AM to form the analogue of the Weyl spinor equation. The present
complete formalism makes this equation appear in its covariant form as the
[4,5] component of a set of equations:
-
A[4,5) = c, bd e A[b,d) A[e,S)
-
(b, d, e : 1,2,3) (10)
The field equations of the general theory of relativity determine even the
geodesic motion of test particles [A. Papaetrou, 1951]. The same is still true
for the original Kaluza-Klein theory. The conclusion for a higher dimensional
generalization of such a theory is not unambiguous [L. Halpern, 1992]. In
the present model we can identify P with the bundle of orthonormal frames
of SO(3, 1) over the Anti De Sitter manifold. This allows even the follow-
ing of the angular position of a spinning body in its rest frame along the
world line; the vertical components of the tangent vector indicate angular
velocity and linear acceleration of the rest frame. Elementary particle spin
can in fact not be ascribed to such a simple model of angular momentum of
a rotating rigid body [E. Schrodinger, 1930]. This justifies our assumption
about the interaction of the spin charge with the curvature. Spin is how-
ever convertible into angular momentum, as demonstrated by the Einstein-
DeHaas effect. We may expect it thus to have the same kind of interaction
mechanism and strength with the curvature as a macroscopic spinning test
body in general relativity [A. Papaetrou, 1951] and to assume here that it
moves along a geodesic in ten-dimensional space. This assumption seemed
to be justified by the projection on space-time of the orbits of geodesics
with vertical components; it is however not correct for the spin precession.
The correct non-geodesic orbit is given in [1. Halpern, 1992]. This appar-
ent contradiction is one of the most interesting features of the model. The
description seems to require modifications of the mathematical structure to
be assumed.
The theory has been extended to the principal fibre bundle of the univer-
sal covering group of SO(3, 2) and its corresponding 6-parameter subgroup.
The topology of the structure admits then a remarkable possibility of de-
scribing multiparticle systems which imply even the spin-statistics relations
[1. Halpern, 1992]. This structure is somewhat of a generalization of features
occurring already in the five-dimensional theory.
370 LEOPOLD HALPERN

Acknowledgments
The author thanks the president of Florida State University and the
University Foundation for their continuing support.

References
F. Bopp and R. Haag, Z. Naturj., 5a, p. 644-635, 1959.
P.A.M. Dirac, Annals of Mathematics, 36, p. 657, 1935.
1. Halpern, Arkiv F. Fysik, 34, p. 539, 1967.
L. Halpern, Physics and Contemporary Needs. 5, Riazudding and Asghar Qadir editors,
p.556, Phenurn Press, 1980.
1. Halpern, Internat. J. Theor. Phys., 23, p. 848, 1984.
1. Halpern, J. Korean Phys. Soc., 25, p.S224-229, 1992.
A. Papaetrou, Proc. R. Soc., A209, p. 248, 1951 and M. Mathisson, Acta Physica Poland,
6, p. 167, 1937 and J. Lubanski, ibid, p. 356.
A. Papapetrou, Proc. R. Irish Acad., 52, p. 1683, 1954.
R. Pavell, Phys. Rev. Lett., 17, p. 1114, 1975.
E. Schriidinger, Sitzber. Pruss. Acad. Science, 24, 1930.
E. Schriidinger, Physica, 6, p. 899-912, 1939 and Proc. R. Irish Acad., A46, p. 25-47, 1940.
R. Utiyarna, Phys. Rev., 125, p.1741, 1962.
R. Utiyarna and B. De Witt, J. Math. Phys. 3, p. 608, 1962.
C.N. Yang, Phys. Rev. Lett., 33, p. 445, 1974.
NOETHERIAN SYMMETRIES IN PARTICLE
MECHANICS AND CLASSICAL FIELD
THEORY
DAN RADU GRIGORE
Department of Theoretical Physics,
Institute of Atomic Physics
Bucharest-Magurele, P.O. Box MG6,
Romania

Abstract. A geometric generalization of the first-order Lagrangian formalism is proposed,

following the original ideas of Poincare and Cartan and the extension to field theory due
to Krupka, Betounes and Rund.
The method is particulary suited for the study of Noetherian symmetries. This point
is proved by explicit study of systems with gauge groups of symmetries.

1. Introduction
There are many advantages in using a geometric framework for the La-
grangian formalism. Most of the papers are based on the Poincan~-Cartan 1-
form, but it was also realized that the formalism became more natural work-
ing with a 2-form having as associated system exactly the Euler-Lagrange
equations (see e. g. [1]). This 2-form is defined on the projective tangent
bundle over the space-time manifold of the system, called by Souriau [2] the
evolution space of the system and can be used for an alternative definition
of the phase space. This formulation also allows a very elegant treatment of
the Noetherian symmetries and of the connection with the symplectic action
of groups appearing in the Hamiltonian formalism.
In this paper we will present a generalization of these ideas to classi-
cal field theory closely related to the point of view of Krupka, Betounes
and Rund [3-6]. The most important property of this generalization is the
possibility of expressing in a geometric way the usual notion of Noethe-
rian symmetry. This definition is very suitable for practical computations.
Namely, we can solve, in principle, the classification problem of Lagrangian
systems with Noetherian groups of symmetry for many important groups
appearing in theoretical physics.
The general theory will be presented in Section 2 and in Section 3 we will
illustrate the method on the case of Abelian gauge theories. The details of
computation will appear elsewere [8].

371
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 371-377.
© 1993 Kluwer Academic Publishers.
372 DAN R. GRIGORE

2. A Geometric Formulation of the Lagrangian Formalism

2.1 Let S be a differentiable manifold of dimension n + N. The first order
Lagrangian formalism is based on an auxiliary object, namely the bundle of
I-jets of n-dimensional submanifolds of S, denoted by J~(S). This differen-
tiable manifold is, by definition:

where J~(S)p is the manifold of n-dimensional linear subs paces of the tan-
gent space Tp(S) at S in the point pES. This manifold is naturally fibered
over S and we denote by 11" the canonical projection. Let us construct charts
on J~(S) adapted to this fibered structure. We first choose a local coordinate
system (xJ.','lj;A) on the open set U ~ S; here JL = 1, ... ,n and A = 1, ... ,N.
Then on the open set V ~ 1I"-1(U), we shall choose the local coordinate
system (xl', 'lj;A, X A1'), defined as follows: if (xl', 'lj;A) are the coordinates of
p E U, then the n-dimensional plane in Tp(S) corresponding to (xl', 'lj;A, XA1')
is spanned by the tangent vectors:

o_a A a
oxJ.' = axJ.' +x I' a'lj;A . (2.1 )

We will systematically use the summation convention over the dummy

indices.
By an evolution space we mean any (open) sub bundle E of J~(S).
2.2 Let us define for a given evolution space E:

Next, one defines the local operator K on ALS by:

Ker:::::i_6_i fJ (o'lj;Al\er), (2.3)

6xl' fJXAI'

where
o'lj;A ::::: d'lj;A - XAJ.'dxJ.' ,
and proves that K is in fact globally defined [7).
We say that er E ALS is a Lagmnge-Souriau form on E if it verifies

K er = o. (2.4)

and is also closed:

der = o. (2.5)
A Lagmngian system over Sis a couple (E,er) where E ~ J~(S) is some
evolution space over Sand er is a Lagrange-Souriau form on E.
NOETHERIAN SYMMETRIES IN PARTICLE MECHANICS 373

It is natural to call the Lagrangian systems (El' O't} and (E2' 0'2) over the
same manifold 5 equivalent if there exists Q E Diff(5) such that a(El ) =
E2 and:
(2.6)
Here a E Diff(J~(5)) is the natural lift of Q.
2.3 The purpose of the Lagrangian formalism is to describe evolutions i.e.
immersions i[1 : M -+ 5, where M is some n-dimensional manifold, usually
interpreted as the space-time manifold of the system.
Let us note that frequently, one supposes that 5 is fibered over M, but we
do not need this additional restriction in developing the general formalism.
Let us denote by ~ : M -+ J~ (5) the natural lift of i[1. If (E, 0') is a
Lagrangian system over 5, we say that i[1 : M -+ 5 verifies the Euler-
Lagrange equations if:
~*izO' = o. (2.7)
for any vector field Z on E.
2.4 By a symmetry of the Euler-Lagrange equations we understand a map
1 E Dif f(5) such that if i[1 : M -+ 5 is a solution of these equations, then
10 i[1 is a solution of these equations also.
It is easy to see that if 1 E Dif f(5) is such that ¢ leaves E invariant
and:
¢*O' = 0'. (2.8)
then it is a symmetry of the Euler-Lagrange equations (2.7). We call the
symmetries of this type Noetherian symmetries for (E, 0').
If a group G act on 5: G 3 g f-> 19 E Dif f(5) then we say that G
is a group of Noetherian symmetries for (E,O') if for any 9 E G, 19 is a
Noetherian symmetry. In particular we have:

(2.9)

It is considered of physical interest to solve the following classification

problem: given the manifold 5 with an action of some group G on 5, find
all Lagrangian systems (E,O') where E ~ J~(5) is on open subset and Gis
a group of Noetherian symmetries for (E, 0'). This goal will be achieved by
solving simultaneously (2.4), (2.5) and (2.9) in local coordinates and then
investigating the possibility of globalizing the result.
2.5 Now we make the connection with the usual Lagrangian formalism.
We can consider that the open set V ~ 11"-1 (U) is simply connected by
choosing it small enough.
The first task is to exhibit somehow a Lagrangian. This can be done as
follows [7]. Form (2.5) one has that

0' = dB. (2.10)

374 DAN R. GRIGORE

Then one can show that by eventually redefining (), one can exhibit it in
the form:

where the smoth functions L~~':"".'1k are completely antisymmetric in the

upper indices and also in the lower indices.
Finally, using the structure equation (2.4) one shows that there exists a
smooth function L : V ---- R such that:

(2.12)

(Pk is the permutation group ofthe numbers 1, ... , k) and lal is the signature
of a). L is called a local Lagmngian. The formulae (2.11)-(2.12) are exactly
those of [3]-[6]. If a is of the form (2.10)-(2.12) then we denote it by aL.
Now one can easily show the following facts. If a = aL, then
1) the local form ofthe Euler-Lagrange equations (2.7) coincides with the
usual one.
2) the Euler-Lagrange equations (2.7) are trivial iff aL = O.
3) let us suppose now for the moment that a is exact i.e. verifies (2.10)
on the whole E. Then one can define the action functional (see e.g. [8]) and
establish that the definition (2.8) is equivalent to the usual definition for the
Noetherian symmetries.

3. Abelian Gauge Theories

3.1 We consider only the case without matter fields. If M the n-dimensional
Minkowski space, then in the general framework of Section 2, we take S =
M X M with coordinates (xJL, AV); 11, v = 1, ... , n. AI' are the components of
the electromagnetic potential.
In the global coordinates (xJL, AI', Xl' JL) on E == J~ (S), the expression of
any a E ALS is:

n 1
(3.1)
c:
JLl, .. ·,JLk
"""
~ (k + 1)! CkTJL1, ... ,JLkbAVQ 1\ ... 1\ bAvk 1\ dXJLk 1\ ... 1\ dxJLn.
n VQ,,,,,Vk
k=O
with a::: and T::: having apropriate antisymmetry properties and:

(3.2)
NOETHERIAN SYMMETRIES IN PARTICLE MECHANICS 375

The structure relations (2.4) is in this case:

k
' " (_l)i+ j aPJ,lll""'!'J, ... ,llk = O. (3.3)
~ lIt,lIl,···,II"··',Vk
i,j=l

for k = 1, ... , n and the closedness condition (2.5) gives:

(3.4)

orIlO, ... ,llk

VO""Vk+l (3.6)
OXllo

for k = 0, ... , n.
Here:
o _ () v ()
(3.7)
bx ll = ()x ll +X Il{)Av'
3.2 We now impose the gauge invariance of the theory. If ~ : M -+ R is an
infinitesimal gauge transformation let us define the following transformation
on S:
<h(xll,A V) = (xll,AV + ({)VO(x)). (3.8)
We say that the system is gauge invariant iff:
(¢e)*a = a. (3.9)

One can prove that in the particular case when a is exact and we have an
action functional, this definition coincides with the usual definition of gauge
invariance.
We also impose Poincare invariance; the action of PIon Sis:

<PA,a(X, A) = (Ax + a, AA). (3.10)

and we require that <PA,a are Noetherian symmetries for any (A,a) E pI:
(3.11)

3.3 One can easily translate (3.9) and (3.11) into conditions on the coeffi-
cients a::: and r::: appearing in (3.1); namely these functions are dependent
only of the field strength variable:

(3.12)
376 DAN R. GRIGORE

and they are Lorentz invariant tensor functions.

From (3.3)-(3.6) it is clear that the functions a::: are constrained only by
(3.3) and (3.4). Using induction we can prove that there exist a F-dependent
function Ly M such that:

a I1O , ..• ,l1k = 8( Ly M )111 ,···,l1k

V1,···,lIk _ (Ly M )110 '·",l1k.
(3.13)
VO ,···,lIk 8 FIIO 110 VO,···, lI k

where:
(LYM )111 ,···,l1k
1 YM
8k L
1I1"",Vk
__ '"'
- k' ~ 8F lI1 8F lIk
(3.14)
• UEPk 11,,(1)'" 11,,(k)

The Lorentz invariance of the tensor function a'" can be used to show
that, without modifying a, one can redefine LYM such that it is a Lorentz
invariant function:

(3.15 )
This fact is of cohomological nature.
3.4 Let us now turn to the functions T· ... We have from the structure
equations (3.3)-(3.6) only:
8 r J-Ll ,···,I-l.k ,Vo,···,Llk
_~ ____ = T J11 ,.. ·,l1k+1,1I0 ,... ,lIk + 1 • (3.16)
8FlIk+1I1k+1

(for k = 0, ... , n). The tensor Till ,· .. ,l1k+l ,II O, .. ·, IIk+l can be shown to be com-
pletely antisymmetric in all indices. Let us use the notation m == [~]. Now
one easily integrates (3.16) and gets that:
1
L II
m P
T I11 , ... ,l1k,1I0 , ... ,Vk = (p _ k)!2P_kCI11, ... ,I1P,1I0, ... ,IIP . F II ,I1,' (3.17)
p=k l=k+1

where C'" are some constants which are completely anti symmetric in all
indices.
The Lorentz invariance of the tensor function T'" is equivalent to the
Lorentz invariance of the tensors C"', so we get two distinct cases: (a) if
n = 2m the tensors T'" are zero for any k, so we have a aLYM; (b) if
n = 2m + 1 then:
CJ-LlI···,/-Lm,vo, ... ,vm = KfJ-LlI ... ,/-lm,VQ, ... ,vm. (3.18)
for some K E R and all the others tensors C'" are zero.
If we define:

II F
m
L cs ( X, A AVO
,x) = (m + (3.19)
K C' II ,I1.
1)!2m cI11, .. ·,l1k, 1I0,···. lI k i=l .

then in this case we have: a = aLYM+L cs '

Here Lcs is the usual expression of the Chern-Simons Lagrangian.
NOETHERIAN SYMMETRIES IN PARTICLE MECHANICS 377

4. Conclusions
The method of analysing Lagrangian systems with group of Noetherian
symmetries illustrated above can be succesfully used for other interesting
physical situations: non-Abelian gauge theories [8], Galilean invariant many-
particles systems [9], string theory [10], gravitation theory [11], etc.

References
[1] J. Klein, Ann. lnst. Fourier (Grenoble) 12 (1962) 1-124
[2] J. M. Souriau, "Structure des Systemes Dynamiques", Dunod, Paris, 1970
[3] D. Krupka, Czech. Math. Journ. 27 (1977) 114-118
[4] D. Betounes, Phys. Rev. D 29 (1984) 599-606
[5] D. Betounes, J. Math. Phys. 28 (1987) 2347-2353
[6] H. Rund, Lect. Notes in Pure and Appl. Math. 100 (1985) 455-469
[7] D. R. Grigore and O. T. Popp, " On the Souriau-Lagrange Form in Classical Field
Theory", submitted for publication
[8] D. R. Grigore, "A Generalized Lagrangian Formalism in Particle Mechanics and Clas-
sical Field Theory" , to appear in Fortschr. der Physik
[9] D.R.Grigore, "Generalized Lagrangian Dynamics and Noetherian Symmetries", to ap-
pear in International J. Mod. Phys. A
[10] D. R. Grigore, J. Phys. A 25 (1992) 3797-3811
[11] D. R. Grigore, Class. Quant. Gravity 9 (1992) 1555-1571
LIENARD-WIECHERT YANG-MILLS FIELDS
K. PAULTOD
Mathematical Institute and St John's College
Oxford

Abstract. We consider the problem of defining Lienard-Wiechert fields in Yang-Mills the-

ory. Trautman (1981a,b) following Arodi (1978) defined them by the form of the potential,
but we choose to define them in terms of the principal spinors of the curvature. This leads
to essentially the same results as those of Trautman for gauge group SU(n), but some
different solutions arise for gauge group SL(n, R).

1. Introduction
The Lienard-Wiechert solution of Maxwell's equations is the retarded solu-
tion corresponding to an electric monopole moving on an arbitrary world-
line in Minkowski space. The charge is conserved, but the solution radiates
energy at a rate proportional to the acceleration of the world-line. It is nat-
ural to ask if there is a solution of the (non-linear) Yang-Mills equations
which corresponds to the Lienard- Wiechert solution in a suitable sense. One
might then seek to see if it is possible to define conserved charges or whether
charges can be radiated, and also whether the non-linearity of the Yang-Mills
equations causes the world-line to be restricted in any way.
In such an investigation, one needs to decide which characterization of
the Lienard-Wiechert solution to choose for generalization to the Yang-Mills
theory. In his study of this problem, Trautman (1981a,b; see also Tafel and
Trautman 1983) chose to characterize the field by its potential, and, fol-
lowing Arodi (1978), to take a form of Yang-Mills potential which gener-
alized that. He concluded that, if the gauge group is any compact, semi-
simple group (and so in particular if it is SU(n)), then there are conserved
"colour" charges and the Yang-Mills Lienard-Wiechert field is a product of
Maxwellian Lienard- Wiechert fields. However, for other gauge groups, he
found solutions which do radiate colour charge.
It is possible to characterize the Maxwellian Lienard- Wiechert fields in
spinorial terms by their principal null directions (Lind and Newman 1974):
briefly, one principal spinor ofthe Maxwell spinor must be tangent to a twist-
free, shear-free congruence of null geodesics (see e.g. Penrose and Rindler
1984 for the definitions of these terms). Such a congruence is necessarily
generated by the future (or past) null cones springing from an arbitrary
world-line in Minkowski space, and the Lienard- Wiechert field is based on

379
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Defonnations, 379-389.
© 1993 Kluwer Academic Publishers.
380 KP. TOO

this world-line. In this paper, my aim is to explore the consequences of

using this other characterization of the Maxwellian Lienard- Wiechert fields
as the characterization of Lienard-Wiechert solutions in Yang-Mills theory
also. This study leads to a more general class of solutions, depending on
the gauge group chosen, but the conclusions for S U (n) are the same as
Trautman's.
In the Section 2, I review the coordinate and tetrad system of Held et al
(1970) and Lind and Newman (1974) which is adapted to a twist-free, shear-
free congruence in Minkowski space. I solve the radial parts of the Yang- Mills
equations in the Newman-Penrose formalism and obtain a reduced system
of equations. In Section 3, I solve the reduced system. The spin-coefficients
from Section 2 and the Yang-Mills equations in the NP formalism are given
in an Appendix.

2. The Coordinate System and Tetrad

We begin by describing the coordinate and tetrad system of Held et al (1970)
which is adapted to an arbitrary time-like world-line r in Minkowski space,
M. Suppose r is given parametrically by

(2.1 )

where T is the proper-time along r, so that

'a
'f/abz Z
·b
= 1 (2.2)

where dot denotes differentiation with respect to To Next we coordinatize

the null cone by (), <p according to

La = (1, sin () cos <p, sin () sin <p, cos B). (2.3)

Define

V = 'f/ab Z'aLb (2.4)

then the coordinate system (T, r, B, <p) adapted to r is defined implicitly by

(2.5)

Here T labels the future null-cones springing from r, () and <p label the null
geodesic generators of these null-cones, and r is an affine paramter along
each generator. The generators taken together constitute the null geodesic
congruence defined by r.
LIENARD-WIECHERT YANG-MILLS FIELDS 381

In these coordinates, the metric is

(2.6)

A suitable null tetrad is defined by

C = dT; n = (1- 2r V )dT + dr; m = -~(dO + isinOd¢) (2.7)

V Vv2
or in covariant form:

o
C = or; n
0 1 (
= OT - 2"
V) or;0 m = rV2
1 - 2r V
V (000 + sini 0 o¢0) .(2.8)
As usual, we write (oA, [A) for the spinor dyad corresponding to this null
tetrad. It is a straightforward matter to calculate the NP spin-coefficients
for this dyad, and they are given in the Appendix.
Given a choice of gauge group G, a Yang-Mills field is a connection on a
principal G-bundle, B, over M. For simplicity, I will assume that B is trivial,
which corresponds to the assumption that there is no magnetic charge. Then
the connection may be represented by a globally-defined, Lie-algebra-valued
1-form Aa. The gauge freedom in the potential is given by

Aa -+
,
Aa = (gAa + Vag)g -1 for g: M -+ G. (2.9)

We exploit this freedom to set CaAa equal to zero. This requires the solution
of the equation

(2.10)

and then we may suppose that

caAa == roo' = o. (2.11)

The curvature of the Yang-Mills connection is represented by a Lie-algebra-

valued 2-form Fab, which can be decomposed into spinors as

(2.12)

The relation between the spinor fields XAB and XA' B' depends on the choice
of gauge group G. Given the gauge condition (2.11), the potential can be
expanded in the tetrad (2.7,8) as

(2.13)
382 K.P. TOO

Again, there will be relations among the components of Aa depending on

the choice of G.
The property of Lienard-Wiechert fields which we are regarding as char-
acteristic is that the spinor oA is a principal spinor of XAB (and similarly
for oA' and XA' B'), In components:
ABO - -A' -B'
XABO 0 == Xo = ; XA'B'O 0 == -
Xo =
0 (2.14)

while
A B - -A'-B' -
XABo l == Xl; XA'B'O l == Xl
A B - -A'-B' -
XABl l == X2; XA'B'l l == X2· (2.15)

The Yang-Mills equations in the NP formalism are given in the Appendix

with the specializations (2.11) and (2.14). We proceed to solve these.
Equation (A.3a) implies

Xl = r12x (r,0,4»; X1 = r12 :\:(r, 0, 4» (2.16)

for some X, X. Next (A.2a) implies

/01' = Vr /(r,O,4»; /10' = Vr 7(r, 0, 4» (2.17)

for some /, 7. From (A.3c) we find

(2.18)

where

150
r
= Vb = V2
1(880 + sini °84>8) (2.19)

From (A.2b) we find

/11' = - ~(X
r
+ X) (2.20)

and

(2.21)

where 00 is the "eth" of Newman and Penrose (1966; see also Penrose and
Rindler 1984), here defined on a spin-weight s quantity by

V2 cot °
s
00 = 150 - (2.22)
LIENARD-WIECHERT YANG-MILLS FIELDS 383

From (A.2c) we find

V· V-
X2 = --i
r - -(box
r2 + [x,i])

- = - -
X2 V., - -V(>vox-
r r2
+ [_X,, ]) (2.23)

so that (2.16,17,20,23) give the radial dependence of all quantities in terms

of " i, X and X, which in turn are functions on S2 X R coordinatized by
((), ¢» and T, respectively. Equation (A.3b) is now identically satisfied, while
(A.3d) reduces to

001 + [1, ,] + (~2) .= 0

0 + [i',i] + (~2) .
5 i' = o. (2.24)

The reduced Yang-Mills equations are therefore (2.18) and (2.21), which
may be viewed as constraints, and (2.24) which is the evolution (and which
preserves (2.21)). There is residual gauge freedom, namely:

(2.25)

where 9 = g((),¢» is a function from SX2 to the gauge-group G.

For the gauge group SU(n), " i, x and X are n X n complex matrices
with

(2.26)

while for gauge group SL(n,R) " i, X and X are n X n complex matrices
with

X = x; i = 1; trx = 0; tTl = O. (2.27)

In the next Section, we set about solving the remaining equations.

3. Solving the Reduced System

We begin by considering the equations

(3.1)
384 K.P. TOD

where h, h, , and i are n X n complex matrices. We may interpret each

of the operators 00 +, and 50 +i as defining a a-operator on the Yang-Mills
bundle B, and then (3.1) is the condition for hand h respectively to be n
holomorphic sections. We can therefore solve these equations at each fixed
T for non-singular matrices hand h, provided the Yang-Mills bundle B is
trivial as a holomorphic bundle at each T. We will make this assumption,
thereby extending the assumption that B is trivial as a smooth bundle,
which was made in Section 2 to eliminate magnetic charges.
Having solved (3.1), we find that (2.18) is solved by

(3.2)

where

ooA = 0; 50A = 0 (3.3)

If X and Xare to be globally regular on S2 then (3.3) implies that A and A are
functions only of T. They have the character of matrices of charges, which,
at this stage, can apparently change with time. Note that the invariants of
X and X are the same as those of A and A respectively, so that these are
also functions only of T.
Now we introduce w = hh- 1 , to find that (2.21) becomes

5000 w - 50ww- 1 00 W + ~2 (Aw - wA) = o. (3.4)

As we shall see below, this equation has something of the character of

an eigenvalue equation on S2, with the matrices A, A of charges as the
eigenvalues.
The evolution equations (2.24) are solved with the aid of

p=- h'h- 1 ;p=-

- h-h-- 1 (3.5)

as

50(w- 1 oopw) + (:2)' + :2]

[p, = 0

00(w50pw-1) + (:2)' + [p, :2] = O. (3.6)

At this point we may summarize the freedom available. Gauge transfor-

mations (2.25) have the effect

(3.7)
LIENARD-WIECHERT YANG-MILLS FIELDS 385

so that A, A,w,p and p are all gauge-invariant. Thus (3.4,6) is a gauge-

invariant formulation of the problem. However, there is also freedom in the
choice of hand h satisfying (3.1), namely

(3.8)

where band b are functions only of T. Call this a b-transformation, then

under a b-transformation we find

A -> bAb-1; A -> bAb- 1

(3.9)

We can exploit this freedom to make A and A independent of T. To do this,

choose band b so that

{ 1 { 1_
lS2 V 2P = 0; lS2 V2 P = O. (3.10)

then integrating (3.6) over the sphere and using

(3.11)

we find at once that A and A vanish, so that the matrices A and A, which
we have identified intuitively as matrices of charges, are constant in time.
To make this intuitive identification tighter, we may recall that a defini-
tion of quasi-local charges for Yang-Mills fields was proposed in (Tod 1983).
Briefly, given a GL(n,C)-Yang-Mills field, the definition associates a pair of
n X n complex matrices up to similarity transformations with any topolog-
ically spherical, space-like 2-surface in Minkowski space. The construction
mirrors Penrose's quasi-local mass construction (Penrose and Rindler 1984),
and the eigenvalues of the given matrices can be regarded as quasi-local
charges for the Yang-Mills field. In the present case, the matrices obtained
at any 2-surface of constant T and r are actually the matrices A and A.
This strengthens the identification of the eigenvalues of these matrices with
charges. Further, by the result above, in this case these quasi-local charges
are constant. We may use a b-transformation with constant band b to put
A and A into canonical form.
How we proceed now depends on the choice of gauge group. For SUe n),
following (2.26), we find

h- I = ht; A = -At; P = _pt; w = wt; trA = 0 (3.12)

386 K.P. TOD

while for SL(n,R)

(3.13)

Consider first the case of SUe n). Then W is positive-definite and Hermitian,
and so it has a positive-definite square-root. Call this n, and integrate (3.4)
over S2 to find

(3.14)

where

E = n- 1 oow.
Suppose A has been diagonalized, say A = diag( AI, ... , An), then a diag-
onal entry on the right-hand-side of (3.14) takes the form

(Ai + Xi )Wii
while the corresponding entry on the left-hand-side is non-negative. Since
W is positive-definite, Wii is positive and therefore Ai + Xi is non-negative for
each i. Since tr A is zero, this forces Ai + Xi to be zero for each i which in
turn forces the left-hand-side in (3.14) to have zeroes on the diagonal, from
which it follows that W is constant on the sphere.
If A cannot be diagonalized, a similar argument applied to the Jordan
canonical form leads to the same conclusion. With W constant on the sphere,
we may use a b-transformation to set it equal to the identity matrix. Many
things now simplify; by (3.4) and (3.12), A is skew-Hermitian, so we may
assume that it is diagonal; by (3.12) h is unitary; by (3.5) and (3.12) p is
skew-Hermitian; finally, the evolution equation (3.6) reduces to

- 2AV 1
00 00 p - V3 + V 2 (p, A] = o. (3.15)

Since A is diagonal, it follows rapidly from this that p is of the form

p=jA+q (3.16)

where

-' 2V
oooo! = V3
and q is a constant diagonal matrix with imaginary entries. From (3.9)
we see that q can be eliminated by a suitable b- transformation, and then
LIENARD-WIECHERT YANG-MILLS FIELDS 387

from (3.5) and (3.16) h is diagonal. From (3.1) this makes I diagonal and
the field is reduced to a product of Maxwellian Lienard-Wiechert fields.
This conclusion was also reached by Trautman (1981) but we have arrived
at it, admittedly after more labour, by beginning with a more general notion
of Lienard-Wiechert field. To find something new, we consider the case of
gauge group SL(2, R) and, for simplicity, we assume that the world-line r
is straight, so that V is 1. Define

A = (~ _0>.) , real constant >.; h = (~ ii) , real1] (3.17)

then

w - °
_ (1 2i1])
1 .

The constraint equation (3.4) reduces to

ao 00 1] + 2>'1] = ° (3.18)

while the evolution equation (3.6) becomes

(3.19)

Thus 1] is an eigenfunction of the Laplacian on the sphere, with arbitrary

time-dependence, and 4>. is the corresponding eigenvalue; for these solutions
the "charges", that is the eigenvalues of the charge-matrix A, are discrete.
The charges are conserved, but there is arbitrary time-dependence in the
radiation field.
These solutions may not be the general solution for SL(2, R), but they
are the general solution with upper-triangular wand, together with the
corresponding lower-triangular solutions, they are the only solutions with a
linearisation at fixed A.
These solutions have angular dependence and so it could be argued that
they are not Lienard-Wiechert solutions since they are not monopoles but
are higher multipoles. However, in Maxwell theory, the Maxwell spinor cor-
responding to a radiating dipole or higher multipole stationary at the origin
will not have the radially out-going spinor as a principal spinor, that is to
say it will not satisfy (2.14) (since this property precisely characterizes the
Lienard-Wiechert field, which in this case is just the Coulomb field). The
Yang-Mills fields which we have found here are purely outgoing, but have
no counterpart in Maxwell theory.
388 K.P. TOD

Appendix
In this appendix we give the spin-coefficients for the tetrad of Section 2
and the Yang-Mills equations written out in the NP formalism with the
simplifications (2.11) and (2.14). The Yang-Mills equations were given in
(Newman and Tod 1980) but unfortunately with some sign errors.
The spin coefficients for the tetrad defined by (2.7) are

K,=a=f=T=1r=A=O
1 1 V
P= --j Il = --j 1= - - (A.l)
r 2r 2V

a = -(3- = -bV - ---

1 V
cotO.
2V 2V2 r
With the definitions (2.13) and (2.15) and the conditions (2.11) and (2.14)
the Yang-Mills equations reduce to two sets. The first set define the field
from the potential;

0= Dlol' - PIOl' (A.2a)

o= D llO' - PllO'

2Xl DIu' + bioI' -

8110, - blO',IOl'] (A.2b)
2,b DIll' + 8jlO' - biOI' - [101',110']

X2 bill' + 2al ll' -

~llO' - IlllO' + b11',llO'] (A.2c)
h 8111, + 2alu' - ~101' - 1l101' + bU',I01']'
and the second set are the field equations;

DXl - 2pXl = 0 (A.3a)

Dh -2ph = 0

DX2 - hI - [XlollO']- PX2 = 0 (A.3b)

Dh - 8b - [xl,lod - PX2 = 0
LIENARD-WIECHERT YANG-MILLS FIELDS 389

8Xl + [XI,/Ol'] = 0 (A.3c)

8-h + [X}, /10'] = 0

8X2 - 6.Xl + [x2"od - [Xl,/l1'] + 2f3X2 - 2J.lXl =0 (A.3d)

8X2 - 6.Xl + [h"l0']- [Xl,/ll'] + 2!3X2 - 2J.lh =0
References
H. Arodz 1978 Phys. Lett. 78B 129
A. Held, E.T. Newman and R. Posadas 1970 J. Math. Phys. 11 3145
R.W. Lind and E.T. Newman 1974 J. Math. Phys. 15 1103
E.T. Newman 1974 J. Math. Phys. 1544
E.T. Newman and R. Penrose 1966 J. Math. Phys. 7 863
E.T. Newman and K.P. Tod 1980 in General relativity and gravitation: one hundred years
after the birth of Albert Einstein vol.2, ed. A. Held, Plenum Press: New York
R. Penrose and W. Rindler 1984 Spinors and space-time vol. 1 Cambridge: CUP
J. Tafel and A. Trautman 1983 J. Math. Phys. 24 1087
K.P. Tod 1983 Proc. Roy. Soc. (Lond.) A389 369
A. Trautman 1981a Phys. Rev. Lett. 46 875
1981b Acta Physica Austriaca Suppl. XXIII 401
THE TWIST PRESCRIPTION IN THE
TOPOLOGICAL YANG-MILLS THEORY

SORIN MARCULESCU
Fachbereich Physik, Universitat-GH-Siegen, Postfach 10 12 40, W-5900 Siegen, FRG

Abstract. The quantum properties of topological Yang-Mills theory are derived from the
N = 2 supersymmetric Yang-Mills theory in flat space via the twist prescription.

Key words: Topological Field Theory

1. Introduction

The topological Yang-Mills theory (TYM) (Witten 1988) is described by

fields of integer spin of both statistics. It includes the gauge field aJ-t and is
defined on a Riemannian manifold M. TYM has a nilpotent fermionic sym-
metry, called BRS, such that physical states form BRS-cohomology classes.
Moreover, the energy-momentum tensor is a BRS variation. The last con-
dition is satisfied because the full quantum action of TYM can be written
as a BRS variation. As a consequence, all local observables which represent
the continuous symmetries of the action are trivial, i.e., their matrix ele-
ments between physical states vanish. However, this quantum field theory
possesses local observables, the Donaldson polynomials, which have non-
vanishing physical correlation functions.
In this contribution we show that, at least in perturbation theory, all
the properties of TYM are preserved by the fully quantized theory. We
accomplish this by fully exploiting the N = 2 supersymmetry present in
flat TYM. At the quantum level the superconformal invariance of the flat
theory is broken resulting in an N = 2 anomaly multiplet. We consider
the components of the superconformal anomaly after coupling the theory
minimally to euclidean gravity.
We change now the renormalization prescription inherited from N = 2
supersymmetry to one compatible with BRS symmetry in curved space. It
turns out that the energy momentum tensor is not altered, while the BRS
current attains an additional contribution keeping it conserved.
Furthermore we study the effects of the above procedure upon the su-
perconformal anomaly. Some components, like the trace of the energy- mo-
mentum tensor, while receiving radiative corrections in each order of pertur-
bation theory, are still BRS variations and represent therefore trivial local

391
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 391-401.
© 1993 Kluwer Academic Publishers.
392 SORIN MARCULESCU

observables. The other components yield just the Donaldson polynomials

(Donaldson 1990). They are observables but obey a non-renormalization
theorem (Dahmen 1991).

2. N = 2 versus singlet supersymmetry

The prescription (twist) allowing for construction of TYM consists of iden-
tifying the isospinor index of N = 2 supersymmetric Yang-Mills theory with
a dotted spinor index.
In the Wess-Zumino gauge the N = 2 Yang-Mills multiplet contains a
gauge field ao:{3' a pair of complex conjugate spinor-isospinor fields >'o:B, a
pair of complex conjugate scalars C and C and an auxiliary real isotriplet
field H AB. Spinor indices are denoted by dotted or undotted Greek letters
from the beginning of the alphabet, isospinor indices are denoted by capital
Latin letters. Whenever more indices of one sort are present full symmetry is
assumed with the exception of the anti symmetric invariant tensors fo:{3, fo{3
and gAB. All the fields belong to a representation of the (compact) gauge
group generated by the antihermitean matrices tj are normalized appropri-
ately. For TYM the twist procedure yields the following field content aJl' 'lj;Jl'
XJlll' 7], </J, >. and bJlll' as follows:
- 1
a .
0:(3
-+ a 0:(3.., >'o:B -+ -i'lj;0:/3 ; >'oB -+ i(Xo/3 + "2fo{37]) ;
1
C -+ .;2</J ; C -+ -->.. HAB -+ -2ibo{3· (1)
.;2'
The twist turns the half integer spins into integer ones without changing the
original statistics, i.e. 'lj;Jl' XJlll and 7] remain anticommuting. Similarly, the
N = 2 supersymmetry parameters (o:B, (oB are converted to anticommuting
parameters with three different SO(4) structures (Jl' (Jlll and ( presented in
spinorial notation

(2)

If TYM is minimally coupled to a curved background, generally these

anticommuting tensors become local, but constrained by the Killing condi-
tions. Obviously, only the Killing condition for the scalar ( does not restrict
the background informing us in fact, that ( remains constant.
Hence N = 2 supersymmetry is broken, leaving a symmetry parametrized
by an anticommuting scalar (. This singlet supersymmetry takes the form
(Galperin 1991)

i( 'lj;Jl ; 8'1j;Jl = (DJlcP;

0; 8>' = i(7];
THE TWIST PRESCRIPTION IN THE TOPOLOGICAL YANG-MILLS THEORY 393

1
-([</>, A]; 6XI"v = ((bl"v + 2f;:") ;
-i( (~D[I""pVl- + [</>,Xl"v]) . (3)

One verifies that two transformations (3) commute up to gauge transforma-

tion of parameter </>.
The above construction is very similar to supersymmetry transforma-
tions in Wess-Zumino gauge and suggests a superspace description for TYM
(Horne 1988). The superfields are AI"' Au, XI"V and A and depend on the
coordinates (xl", (}), (} being a Grassmann variable. From the superconnec-
tions AI" and Au one can easily construct the covariant derivatives DI"' Du
and the superfield strength Fl"v, FI"' F. The field components are defined by

AI"l al" ; FI"I = -"pI" ;

1
XI" VI XI"V ; DuXl"vl = bl"v + 2f;;v ;
AI A·, DuAl = 7]; FI = 2i</> . (4)
The action of N 2 supersymmetric Yang-Mills theory is given, up to a
total divergence, by (Grimm 1978)

J _14 - -'f3 C
a-xTr(-CDmDmC - iAc,C Dcx A(3 - riJf
. iJ
c, -"41 H A BH B A
h -
-i 2CABA cx
B
+ i h 2CA-'cx B-Ac,B - 1 -
2[C,C] ).
2
(5)

The twist prescription leads to the Witten action. We give the result in
superspace

s = 1M
[ d4 x.j9&oTr(iFl"vXl"v +~Xl"vDoXl"v
2
+iFI"DI" A+~F[A, DuA]) .(6)
4
The superspace approach is suited very well also for discussing the properties
of TYM as listed e.g. in (Birmingham 1991). For instance, the coupling
constant and the metric independence of the partition function follows by
integrating by parts with respect to (} and by assuming the BRS invariance
of the superspace path integral measure.
Moreover, all correlation functions of TYM currents vanish. We call a
TYM current any gauge invariant object which can be obtained by the
Noether procedure modulo improvements from the Lagrangian. Obviously,
it is a (}-component as the Lagrangian itself.

3. The superconformal current multiplet

An important class of TYM currents can be obtained by twist procedure
from the N = 2 superconformal current multiplet.
394 SORIN MARCULESCU

It is known (Sohnius 1979) that the R-current JOI (3' the isospin current
JOI(3CD' the supersymmetry current {jOl(FyD and its complex conjugate 8'Y o(3D
as well as the energy-momentum tensor JOI {3"r8 are conserved and belong to
an N = 2 supermultiplet. The corresponding objects bJ.L' bJ.LIIP (antiselfdual
in /l,v), AJ.LII,AJ.LlIp (antiselfdual in v,p) and BJ.L1I (symmetric in /l,v) of flat
TYM are obtained using the twist prescription

JOI(3 -+ -ib 0I{3.., JOI(3CD -+ ib"r8 01(3 ;

{j'Y ooB -+ A'Y(3oo ; 801"t8B -+ iA OI (3 "t8 ;
JOI{3 "t8 -+ B0I"t (38 . (7)

The above N = 2 currents have the following explicit form in terms of field
components

-2Tr(>'oc>'{3
- C
+ iC D{3o C) ;
-+-+

J{3oCD = 2Tr)...0(c>'{3D) ;

- 2V2 - V2-
Tr( -4f'Yo>'oB - -3->'(,,!BD8)oC + TCD(,,!0>'8)B) ;

2V2- V2-
Tr(4f"ts>'OIB + -3->'("tBD 0I8 )C - T CD OI("t>'8)B) ;

(8)

By minimally coupling flat TYM to euclidean gravity, both bJ.L and bJ.LIIP
become generally covariant and, as a consequence of the equations of motion,
covariantly conserved.
The corresponding procedure is less straightforward for AJ.L1I because N =
2 transformation properties lead to a traceless energy-momentum tensor
incompatible with general covariance. Hence AJ.L1I must be redefined such as
to allow for a non vanishing trace. There is only one quantity

ilJ.L = 2i Tr >''lj;J.L , (9)

appearing in the N = 2 transformation laws, which can be used for this

purpose. The object ilJ.L arises via the twisting procedure from the spinor-
isospinor component of the supercurrent {jaB. By imposing conservation and
THE TWIST PRESCRIPTION IN THE TOPOLOGICAL YANG-MILLS THEORY 395

selfduality of the antisymmetric part one can determine sl1" as the improved
All'" Up to an overall numerical factor we choose

(10)

and express it in terms of topological fields. After coupling to euclidean

gravity and using equations of motion it becomes covariantly conserved.
We still have to determine the singlet supersymmetry partners of bl1 , bl1 "p
and sl1'" As in the case of sl1" we need the remaining components of the
supercurrent multiplet. After twisting we call them S, All'" A, S:;" and S;;"
and present their topological field dependence in superfield form
1
S+ 20A -iTrAF; (11)
1
All" - 40S ;:" -2iTrFXI1" ; (12)

S:" - 40(sl1" - S"I1) 2iTrAF:" . (13)

Now, the supersymmetry transformation of bl1 , sl1" and bl1 "p can be written
as

i( (~A"I1" - ~O"AI1") ; (14)

i( [011" - ~Op(bl1p" + b"PI1) + ~(bl1"DS - 0110"S)] (15)

i( [~OI1A"p + bl1P( -8" - ~o"A"" + ~O"A) - (IL +-+ v)

-Apl1" + f l1"p" (s" - ~O"A) + ~OpAI1"] . (16)

We evaluate the RHS of eqs. (14) - (16) in terms of fields and couple then
minimally to curved background. In this way we get three covariantly con-
served superfield currents
....
iTr(2F" X I1 " - FI1DoA + A DI1 F) ; (17)

iTr {~FtX"P + FI1 D"A + (IL f-+ v)

-gl1,,(FP DpA + 2ADpFP)

1 } - 28oTrF;LA
1 j (18)
iTr {2X/lPF" + gl1"(2X",,F" - F"DoA) - (IL +-+ v)
+2XI1"Fp + "l11"p"F" DoA} . (19)
The O-components represent the BRS current 8 11 , the energy-momentum
tensor til" and the antiselfdual supersymmetry current 8 11"p, respectively.
396 SORIN MARCULESCU

At this point our procedure of constructing the superfield currents start-

ing from the N = 2 superconformal current is completed.
Hitherto, we encountered singlet supersymmetry only. When the flat com-
ponents of the superfield current (17) - (19) are transformed under the full
N = 2 supersymmetry group second order derivatives occur. Without af-
fecting the conservation rules such derivatives can be eliminated by further
redefini tions
(20)
(21)

The resulting improved supercurrent in flat space will transform accord-

ing to an irreducible representation of N = 2 supersymmetry, i.e. without
derivatives of higher order than the first.

4. Superconformal anomaly multiplet

It is commonly assumed that flat TYM is a renormalizable theory. How-
ever, since the supersymmetry gauge theory in Minkowski space has been
formulated in Wess-Zumino gauge there is no guarantee that both gauge in-
variance and N = 2 supersymmetry are preserved by quantum corrections.
A detailed analysis of this question is presented in (Breitenlohner 1988).
Here we shall assume that this remains true even after the twist proce-
dure, which means that TYM in curved space will be free of BRS anomalies.
Quantum corrections break superconformal invariance giving rise to an
N = 2 anomaly supermultiplet. Of course, N =
2 supersymmetry is broken
in curved space and some rearrangement of the anomaly component might
become necessary. Unlike the case of the supercurrent multiplet additional
modifications have to be made since quantizing the theory requires a certain
renormalization prescription.
The superconformal anomaly in Minkowski space leads to the following
N = 2 multiplet of TYM anomalies

Y --+ L', Y --+ ~M'

2 '
1
TfaB --+ iPa/3 ; flaB --+ fl·a, /3 = p.a /3 + -f..
2 a iJP ;
YAB --+ La/3 ; YAB --+ Ma/3 ;
Ya(3 --+ Na(3 ; Y"'/3 --+ N a/3 ;
WaB --+ - i fa(3
· ., WaB --+ w. /3 = f . iJ + f. a. /3 f ;
Q', Q'

a --+ -iA; t --+ T.

(22)
Depending on the form of singlet supersymmetry transformations one can
group the components (22) into two classes, one which transforms without
THE TWIST PRESCRIPTION IN THE TOPOLOGICAL YANG-MILLS THEORY 397

derivatives, the other one involving only the space derivative in the trans-
formation laws.
By forming the linear combinations
A
A+ = T+"2 (23)

one can organize the components of the first class into singlet superfields.
The field dependence of the relevant components can be obtained by twist
from
c ·6 - -
YAB - - 2 {Y TrA 6(AA.yB) ;

-8:
871"
2 Tr(J...(aC>"/3)C - 4V2iCfa(3);
c 1·-
- - 2 Tr(fa -y A-YB - . M C Da -y A.yB) ;
271" 2y 2
c . - 1-
- -22 Tr(j'Y aA.yB - M C D-Y aA-yB) ;
71" 2v2
a - 4~2 [Tr(faiJ ff3a - fa /3 f /3 a) + frTrC Dm C]
t -8: 2 Tr{faiJ fiJa + fa /3 f /3 a + ~(CDm DmC + CDmDmC)
(24)

The coefficient in front of the trace is taken from a one-loop N = 2 su-

perspace calculation of the superconformal anomaly in Minkowski space
(Marculescu 1987). We give here only the first class superfields we will be
concerned with

(25)

(26)

As before, the generally covariant form of RHS of (25), (26) is obtained after
minimally coupling TYM to euclidean gravity.
The singlet supersymmetry transformations of the anomaly components
of the second class can be written as

0; bPIl = -i(Y'I1L; bL_ l1v = i(Y'[I1Pv);

-i(Y'v L+ l1v ; bA = i(Y'l1fl1 (27)
where

(28)
398 SORIN MARCULESCU

The tensors L±I-'II are dual to each other. The twist prescription leads to the
following solution of (27)

w04 _c_ Trq} . lC

471"2 '
w I3 = --Trqnp'
271"2 '
2
w2 - 4: 2Tr('ljJ2 - i¢f+) ;
C 1
wI
3 - 471"2 Tr(J+'IjJ + ¢DX - 2¢ * Dry) ;
C +-+
wO
4 --2 (Trf2 + dTr).,* D ry). (29)
471"
where we introduced the space forms of various ghost numbers (upper index)
for L, PI-" L_I-'II' r I-' and A. As for the currents discussed in the previous
section we have to use the field equations of motion of TYM in order to
verify that eqs. (29) satisfy the conditions (27). Except for W6 and the wr
forms differ from Donaldson polynomials despite the fact that they satisfy
the same descent equations. The reason for this discrepancy can be traced
back to a different renormalization procedure used in TYM.
The results obtained by the twisting method presupposed an N = 2
renormalization prescription in flat space. If we require now that quantized
TYM in curved space is renormalized in agreement with BRS, the presence
of various terms entering anomaly components has to be reconsidered. For
instance, we modify the ghost number anomaly to read
n
VI-'
bl-'ren ---21rJI-'II
_ C rr: f f-I-'II + grav.conr..
t (30)
871"
On the RHS of eq. (30) we included the appropriated gravitational contribu-
tion as calculated by (Dahmen 1991). The quantum ghost number current
bl-'ren is expressed in terms of fields renormalized by a BRS prescription.
Let us define a renormalized BRS current s~en by
i(s:en = 8b:en . (31 )
On the basis of eqs. (30), (31) we introduce a modified BRS current
d C -
SIDO
I-'
=
-
sren
I-'
- W
I-' ,
w ~.. == --2Trj"
271" ~
II 'IjJ1I (32)
which is conserved. As a consequence of this, the gauge invariant polyno-
mials r I-' and A will be modified into wI-' and - ~Tr fl-'II]I-'II, respectively.
By subjecting the whole solution (29) to the change of renormalization pre-
scription one arrives at precisely the Donaldson polynomials
w.04 W4 .
o, W I3 = w1,
3 .

C Tr
__ (~'ljJ2 - i¢f)
271"2 2
C C 2
- 271"2 Tr f'IjJ ; --Trf . (33)
471"2
THE TWIST PRESCRIPTION IN THE TOPOLOGICAL YANG-MILLS THEORY 399

One can show (Dahmen 1991) that the whole set of Donaldson polynomials
as well as the dimension of the instanton moduli space (the integral version
of eq. (30)) remains non-renormalized beyond one-loop.
We turn now again to the first class superfields. To start with we assume
that one can construct a renormalized energy-momentum tensor by means
of the transformation rule

JLV 2 JLV pu VP
o(vren _ !1] ~uren) = i(tren
JLV (34)

where ~~en and v~~ are the quantum version of the quantities defined in
eqs. (9) and (21), respectively. Note that from the conservation of t~~ it
follows that vvv~~n is a BRS invariant.
JLV is symmetric , eq • (34) implies that vJLV - v v JL -1]JLV pu VP ~ uren is
Since t ren ren ren
the () component of some antisymmetric tensor. In the 'classical' TYM this
antisymmetric tensor is selfdual. The quantum theory produces a certain
antiselfdual contribution. Hence the transformation law following from (13)
is changed to

(35)

In passing to the BRS prescription we have to allow for an arbitrary

relative factor between the two superfields on the RHS of eq. (25). We may,
however, modify v~~n - v~,: such that the antiselfdual anomaly takes the
form
n'JLV-- C
411"2
T'irXJL p X vp • (36)

Let us now discuss the scale anomaly n. From eq. (34) we get
(37)
Hence stipulating (37), t~e~ cannot receive gravitational contributions while
remaining conserved. An explicit one-loop computation (Dahmen 1991) con-
firms this assumption. The scale Ward identity can be written as
onl + V JLo~JLren = i(tJLJLren . (38)
The BRS prescription means that we allow for n the form
8~2 Tr(o:FJLvXJLV - ,ADJL FJL ) • (39)

However, the scale anomaly is prescription independent to one-loop (Gross

1975). By using (26) one finds 0:(1) = ,(I) = 1, where the superscript refers
to the one-loop approximation.
At least in the background field quantization scheme, it follows that all
one-loop wave-function renormalization constants are equal as for the N = 2
400 SORIN MARCULESCU

supersymmetric Yang-Mills theory in Minkowski space. This however, means

that the normalization factors of the antiselfdual anomaly n/LII, the trace
anomaly n and the Donaldson polynomials are related by N = 2 supersym-
metry even after changing to BRS renormalizing prescriptions. This fact has
been already taken into account in formulae (30), (32), and (36).
Also the one-loop /3-function of TYM coincides with that of N = 2 Yang-
Mills in Minkowski space (Dahmen 1990). This is not in conflict with the
properties of TYM as discussed above. Indeed, we have shown that both
the radiatively corrected currents s~od and t~~ are BRS variations. Thus,
their correlation functions vanish and the renormalized partition function
is metric independent. Finally, its gauge coupling constant independence
follows trivially from the absence of a genuine BRS anomaly.

5. Conclusions

In this work we attempted to explain all the properties of TYM in the light
of the N = 2 supersymmetry observed for a flat metric. To this end we
constructed a system of currents conserved in curved space which forms an
N = 2 supermultiplet in the limit of flat space.
In passing to the quantum theory, superconformal invariance is broken
and the system of currents develops anomalous Ward identities. As a con-
sequence the BRS current and the energy-momentum tensor receive quan-
tum corrections which can be represented as BRS variations. On this basis
one can understand the metric independence of the partition function and
the vanishing of correlation functions of the BRS current and the energy-
momentum tensor.
By imposing BRS invariant renormalization prescriptions we were able to
derive the Donaldson polynomials from the one-loop N = 2 superconformal
anomaly.
The N = 2 supersymmetry of flat TYM is still present at the one-loop
level showing in a common normalization factor of various anomalies. This
explains why the /3-function of TYM coincides in this approximation with
that of N = 2 supersymmetric Yang-Mills in Minkowski space.

Acknowledgements

This paper is a somewhat simplified version of work done in collaboration

with Hans Dahmen and Lech Szymanowski. I would like to thank them for
their valuable contribution to all the results presented above. I am also par-
ticularly grateful to John Klein presently at the Mathematical Department
of the University of Bielefeld for teaching me modern differential geometry.
THE TWIST PRESCRIPTION IN THE TOPOLOGICAL YANG-MILLS THEORY 401

References
Birmingham, D., Blau, M., Rakowski, M. and Thompson, G.: 1991 Phys. Rep. 209, 129.
Breitenlohner, P.: 1988 in Renormalization of quantum field theories with non-linear field
transformations, Lecture Notes in Physics 303 P. Breitenlohner, D. Maison and K. Si-
bold eds., Springer-Verlag: Heidelberg, p. 64.
Dahmen, H.D., Marculescu, S. and Szymanowski, 1.: 1990 Phys. Lett. B 252, 591.
Dahmen, H.D., Marculescu, S. and Szymanowski, 1.: 1991, 'Currents and anomalies in
topological Yang-Mills theory', Preprint Sf-09.
Donaldson, S.: 1990 Topology 29, 257.
Galperin, A. and Ogievetsky, 0.: 1991 Commun. Math. Phys. 139, 377.
Grimm, R., Sohnius, M. and Wess, J.: 1978 Nucl. Phys. B 133, 275.
Gross, D.J.: 1976, in Methods in field theory R. Balian and J. Zinn-Justin eds., North
Holland:Amsterdam, p. 141.
Horne, J.H.: 1988 Nucl. Phys. B 318, 22.
Marculescu, S.: 1987 Phys. Lett. B 188, 203.
Sohnius, M.: 1979 Phys. Lett. B 81,8.
Witten, E.: 1988 Commun. Math. Phys. 117, 353.
ON SYMMETRY PROPERTIES OF CERTAIN
CLASSICAL LAGRANGE FUNCTIONS UNDER
ROTATIONS
PETER STICHEL
University of Bielefeld,
Germany

and
JAN T. LOPUSZANSKI
Institute of Theoretical Physics, University of Wroclaw,
pl. Maxa Borna 9, 50.204 Wroclaw,
Poland

Preliminary version, presented by J.T. Lopuszanski

Abstract. It is shown that for a one-particle Lagrange function in a 3-dimensional Eu-

clidean space of the standard type, giving rise to Euler variation which is supposed to
be a vector with respect to rotations, the difference between this Lagrange function and
rotated one is equal to a time derivative. We discuss also the problem how to recover a
rotationally invariant Lagrange function from a non-symmetric one, which however, gives
rise to rotationally covariant Euler variations.

The aim of this talk is to convey to you two remarks related to the work
done currently by Peter Stichel and myself. This work is still in progress.
The first remark refers to the following statement. We are going to show
that for a one-particle Lagrange function in a 3-dimensional Euclidean space
of the type

L(x,x) = ~X2 - V(x,x) X=(Xl,X2,X3) etc.

giving rise to Euler variation

f ( .) d oV oV
j = 1,2,3, (1)
j x, X == dt ox'J ox J.'
which is supposed to be a vector with respect to the rotations, we have

L(Rx, Rx) - L (x,x) = ~~ .

Here R stands for a 3-dimensional matrix representing an element of the
rotation group, viz.
R = fl, R- 1 = RT and det R = 1,

403
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 403-412.
© 1993 Kluwer Academic Publishers.
404 P. STICHEL AND J. LOPUSZANSKI

c) is a function of R and ~ (it does not depend on x), viz.

c) = c) (Rx-, R) ,.

it may also depend on other parameters built in into the model. Notice that
we do not require that V should be rotationally invariant, i.e. that V is a
scalar. Since the Euler variation is linear with respect to L we may restrict
ourselves to V only, as !x
2 is for sure rotationally invariant.
This assertion can be immediately extended to rotations in an n-dimensional
Euclidean space n = 2,4, ....
The proof we are going to present here has the merit to be so elementary
that it can be used in regular classes on classical mechanics for beginners at
the University.
Before we enter, however, into the proof let me say few words about the
setting of the story.
It is well known [1] that the necessary and sufficient condition for any
trajectory to be an optimal one is that the Lagrange function is just a
time derivative. But having two Lagrange functions which differ from each
other and yield the same set of solutions of the Euler-Lagrange Equations
(so called s-equivalence) or even yield the same Euler-Lagrange Equations,
this does not yet imply that these two Lagrange functions, say,

L(~,x,t) and L' (~,x, t)

differ by a time derivative of a certain function, viz.

L - L' ¢ ~~ (~, t) .

To see that take the example of the 1-dimensional harmonic oscillator [2]

(,2
L -_ -1 x - x 2) and
2
A, a real constant, can e.g. be viewed as a scaling transformation x -+ x e- a ,
-0 - a real parameter (e- a = A). The Euler variations

x+x and
lead to the same Euler-Lagrange Equations. Nevertheless,

What, of course, is true is that on the so called "mass shell" (i.e. where the
equation of motion is satisfied) we have

x =-x
ON SYMMETRY PROPERTIES 405

and consequently

There is a lore that for Land L' yielding the same Euler-Lagrange Equa-
tions or the same set of solutions [3] we have

, diP
L -L = aL+-
dt '

0: being a constant.
But this conjecture is also not true as shown by the following example [4].
Let us take
(.2
L = 2"1 Xl + x 2 .2) and

in a 2-dimensional Euclidean space. Both Lagrange functions yield identical

Euler-Lagrange Equations, viz.

but
1 diP
¥ aL + dt .
2
L - L' = 2" (Xl - X2)

What is essential in this example is that Land L' give rise to different set
of Poisson' Brackets, as we have

aL
PI == -.- =
.
Xl but
,
PI =
.
X2,
,
P2 =
.
Xl •
aXI

It was shown by Henneaux [4] that if Land L' yield the same Euler-Lagrange
Equations, as well as the same Poisson Brackets up to a multiplicative con-
stant, viz. [Xi,Xj]L = a[xi,xjlL', [Xi,Xj]L = a[xi,Xj]U a a real constant,
¥ 0 (assuming that canonical Poisson brackets are satisfied) then really,
diP
L' = aL + dt (2)

and vice versa, if (2) holds then Land L' are equivalent in the above sense.
Going back to the proof of our assertion, announced before, let us first
present the following Lemma:
Given are three functions

j = 1,2,3.
406 P. STICHEL AND J. LOPUSZANSKI

If there exists a Lagrange function V(O)(~,~) such that fj coincide with the
Euler variations of V(O), then the most general expression for the Lagrange
function V(~, ~), which yields the same Euler variations, is

-
V - V(O) + diP
dt
. (3)

Proof: The function h do not, by assumption, depend on :R and also, by

assumption, are supposed to be Euler variations relatively V(O). If there are
other Lagrange functions V, leading to the same Euler variations, then these
V's have to satisfy also the relations

(4)

Hence
= 0
or
3
V(~) = L Aj(~) Xj + B(~) . (5)
j=1

Relation (5) concerns, of course, also the Lagrange function

L
3
V(O) = A}O) Xj + B(O) .
j=1

If we insert (5) into (4) we get

3
h (~,~) = L ajk(~)Xk + bj(~) (6)
k=l

where
= ajk (7)

and
8B
bj = - 8x'J '
(8)

By assumption ajk as well as bj (j,k = 1,2,3) are given. The most general
solution of (7) and (8) reads

8iP'
A-
J
= A(O)
J
+ 8xj
(9)
ON SYMMETRY PROPERTIES 407

and
B = B(O) + B' , B' being a constant.
Then (3) follows immediately from (5) where

<I> = <1>' + B't .

This proves the Lemma.

This result can be immediately generalized for the case of an n-dimensional
Euclidean space n = 2,4, ... ,.
Let us now make a digression taking into account the assumptions of
our main assertion as well as the just proven Lemma. According to our
hypothesis as well as (1) and (6) ajk behaves like a tensor and bj - like
a vector under rotations. Relation (7) is then similar to that obtained in
electrodynamics. There too the so called vector potential A( x) does not
need to form a vector, in contradiction to the field strength, which forms
a genuine, skew-symmetric tensor. The relation (9) resembles the gauge
transformation. We have also the Hamilton function
3
H = L:Pk:h - L = ~ (E + .AY + B
k=l
where
aL
Pk == aXk
analogous to the Hamilton function of a charged particle in an external
electromagnetic field. B plays here the role of the so called scalar potential.
After this digression let us tackle the problem of rotations.
Let us denote
3
xj = L: RjkXk .
k=l
We have also, according to (5), for a Lagrange function V(~,i:) arbitrarily
chosen among the set (3)

3 3
V(4',i:') = L: L: A j(4')Rjk Xk + B(x') .
j=lk=l

Notice that the Euler-Lagrange Equations

3
fJ(4',i:') = L: Rjdk(~' X) 0
k=l
408 P. STICHEL AND J. LOPUSZANSKI

remain unchanged, as fj is a vector with respect to the rotations. Further

we have

3 3
L: L: RjIRkmalm(-K.)
1=1 m=1

-z= Rjkbk(~)
3
oB()£')
ox'-
J k=1
or
(10)

(11)

Then by virtue of our Lemma

3 3
V(~') = L: L: Aj(~')RjrXr + B(~')
j=1k=1

V(x) + ~~ (x', R)

This proves the assertion.

Turning to our second remark, it is well known that the symmetry of
the Euler-Lagrange Equations can exceed the symmetry content of the La-
grange function from which they originate. Our second remark is related to
this problem, namely how to recover a rotationally invariant Lagrange func-
tion from a non-symmetric Lagrange function, which, however, gives rise to
rotationally covariant Euler variations.
As the rotational group is compact and consequently has a finite volume,
the simplest way to solve this problem seems to be to integrate the non-
symmetric Lagrange function over the group [2]. This procedure should not
apparently affect the Euler variations, which were supposed to be covariant
under rotations.
This procedure, although seemingly logically well founded, is, however,
not as straightforward as can be seen on the following example in the 3-
dimensional Euclidean space [5].
Let us inspect the Lagrange function

(b~)(b (~A x))

V(x,x,b) = 2 (12)
I~I (b A~)
ON SYMMETRY PROPERTIES 409

where b is an arbitrary constant vector; we may, of course, restrict ourselves

to

Ibl 1.

3
Here (ab) stands for L aibi and
i=1

3 3
LL fijkajbk .
j=1 k=1

Strange enough the Euler variations

33·
d aV aV
LL
XjXk
fijk -lxl
3
(13)
aXi
J=1k=1

form a genuine vector with respect to rotations and consequently do not

depend on the direction of b.
Let us try to integrate V over the group of rotations, to obtain the ro-
tationally invariant part of it, or, what turns out to lead to the same goal,
average over all directions of b.
To this aim we introduce the following notation

bX = r cos B

b1 L sin Bcos cf>

Ib A xl = rsinB
r = lxi, L == 111

Ibl = 1
410 P. STICHEL AND J. LOPUSZANSKI

...........•........
".
....\
....········0 ,
,/ • .1C _
_2
0 -'l -----,,------_. x
i
1\ L
¢
......................
..... _-!...
---
--- .!

-,
L

Then

~
n
JJ 11" 211"

L cos () sin () cos ¢ sin ()d()d¢

rsin 2 ()
= 0.
o 0

This Lagrange function, for sure, does not give rise to the Euler variation

°
(13). The reason for the failure of the method of averaging, presented above,
is the singular behavior of V for r = and/or for b being parallel to x.
This singularity prevents the interchanging of integration over the group
with differentiation with respect to X and X.
We close our considerations with exemplifying our assertion on the model
(13). For
b =(0,0,1)
we have
V = X3(XIX2 - XIX2)
(14)
r (xi + x~)
As the Euler variation is a genuine vector and so transforms covariantly
under the rotations, we should have, taking into account our assertion

dq,
V(Rx) - V(x) = dt (15)
ON SYMMETRY PROPERTIES 411

Let us check (15) by inspection for a special case of transformation, namely

the rotation
X~ = Xl.

Then we have

Xl (X2X3 - X2 X3) X3 (XIX2 - XIX2)

V (x') - V (x)
r (X~ + X~) r (xi + xD

Notice that for this particular model we have

del> (Rx, b, 4>, n) del> (R'x, b, 4>', n')

dt dt
(16)
del> (R' Rx, b, 4>', n') dr(Rx) dr(x)
dt dt dt
where
R = R(4),IJ) Inl = 1

and R' = R(4)',n'), In'l = 1 is any rotation which connects band n, viz.
b = R( 4>', n')n. (17)

Such a rotation is a product of an arbitrary fixed rotation transforming n into

b and any element of the little group of b or/and n. There are good reasons
to expect that the relation of the type (16) does not only conforms to the
considered model but is a general property of Lagrange functions subjected
to rotations. To justify (16) let us start from the formula established before,
namely (15), viz.

L(Rx,Rx;b) - L(x,x;b) =
(18)
del>
dt (Rx, b, 4>, n)
Since
L(x,xn) (19)
as
R(4),n)n = n,
412 P. STICHEL AND J. LOPUSZANSKI

we may write (18), by virtue of (19) and (17), as

L(Rx,R~;b) - L (Rx,R~;R'-lb) + L (x,~;R'-lb)

-L(x,~;b)

= -[L(RIRx,R'R~;b) - L(Rx,R~;b)l +
+ [L(RIX,R'~;b) - L(x,~;b)l

which proves the assertion (16).

The results presented here can be generalized to the case of systems of
several particles as well as to the Galilei group symmetry.
The problems considered here are not new and numerous papers were
devoted to them [6]. It is not excluded that the results reported here were
obtained earlier by other authors by using a more sophisticated mathemat-
ical apparatus. As mentioned before, the main reason to present our results
here was to emphasize the pedestrian way these results were derived by us.
We are grateful to Professor Jan Rzewuski and Dr. Marek Mozrzymas
for comments made after my talk.

References
[1) see e.g. R Courant and D. Hilbert "Methoden der Mathematischen Physik" 2-nd edi-
tion, Berlin, 1931
[2) We are grateful to Mr. Jerzy Cislo for making this remark
[3) E.L. Hill, Revs. Mod. Phys. 23 (1951) 253
[4) M. Henneaux, Am. Phys. N.Y., 140 (1982) 45
[5) J.C. Houard, J. Math. Phys. 18 (1977) 502
[6) J .-M. Levy-Leblond, Commun. Math. Phys. 12 (1969) 64,
J.-M. Levy-Leblond, "Galilei-Group and Galilean Invariance" in " Group Theory and
Its Applications", vol. II (edited E.M. Loebl), Academic Press, 1971
J.-M. Levy-Leblond, Am. J. Phys. 39/5 (1971) 502
M. Henneaux and J.C. Shepley, J. Math. Phys. 23 (1982) 2101
TUNNELLING OF NEUTRAL PARTICLE WITH
SPIN 1/2 THROUGH MAGNETIC FIELD

M. MIJATOVIC, G. IVANOVSKI, B. VELJANOSKI AND G. APOSTOLOVSKA

Institute of Physics,
Faculty of Natural Sciences and Mathematics,
Skopje, Macedonia

Abstract. The main part ofthis lecture concerns work announced in Ref. [1]. The last part
contains some new results, not previously reported. We have investigated the scattering
and bound states of a nonrelativistic and relativistic spin-1/2 particle in the system of N
magnetic barriers (or magnetic wells). We have studied two types of problems: tunnelling
with spin and band structure.

1. Introduction

The motion of a neutral spin-l/2 particle through a magnetic field has been
extensively studied in recent years, first motivated by the measurement of
the final state polarization in the neutron-spin echo experiments, and also by
the measurement of the final state of the neutron wave function in neutron
interferometry.
The second approach to this consideration is a study of the anomalous
magnetic moment of the neutrino in connection with the solar neutrino
problem. Namely, it was argued that a neutrino magnetic moment of the
order of 10- 10 JlB would be sufficient to flip a large number of the left-
handed neutrinos into right-handed ones over length L in a magnetic field
B such that BL "" l07Tm, e.q. L", 108 m, B '" lO- l T.
The one dimensional treatment of the motion of a quantum particle
through the field of potential barriers is the simplest approximation. Since
the historical paper of Kronig and Penney on electron motion in an infinite
periodic chain, this model has served as a valuable tool in explaining several
interesting properties of real materials as forbidden energy gaps.
Our paper has several aims. In Section 2 we will make a generalization of
the paper [2) i.e. we will study the tunneling of a neutral spin-l/2 particle
through a finite number of magnetic square wells (or barriers, depending on
spin polarization).
In Section 3 we consider the bound states of an infinite chain of identical
magnetic square wells.
The Section 4 contains the relativistic generalization of previous results.

413
Z. Oziewicz et al. (eds.), Spinors, Twistors, Clifford Algebras and Quantum Deformations, 413--420.
© 1993 Kluwer Academic Publishers.
414 M. MIJATOVIC AND AL

2. Tunnelling with Spin

Let us study the scattering states of a spinless (for simplicity) particle which
incidents from the left and transmits through N rectangular potential bar-
riers of width a, separated by the gap b (the period has length I = a + b)
(Fig. 1).
vex)
Vo
E- - - - - - -- - - - - - -
I II III
x
o 3CV2+b (N-1)l-Q.Iz (N-1)l+0I2
Fig. 1. Multiple magnetic potential barrier
In the beginning we restricted our consideration to two barriers. Then
the solution of the Schrodinger equation can be written in the form (for
E < Vo = /LBB)
Aeikox + Be- ikox , x < -a/2

Ce- vx + De vx , -a/2 < x < a/2

w(x) = Ale ikox + Ble-ikox, a/2 < x < a/2 + b (1)

C1e- vx + D1e vx , a/2 + b < x < 3a/2 + b

Feikox + Ge- ikox , x> 3a/2 + b.
The connection between incoming and out coming data for the scattering
on two barriers follows from the relations

(2)

where

(3)

and the matrix elements of so-called M -matrix are given by

Mu = e i J1- cosh .x, M12 = i sinh .x,

TUNNELLING OF NEUTRAL PARTICLE WITH SPIN 1/2 415

M21 = -i sinh.x,
c
p = koa - arctan( 2" tan va), (4)

v ko v ko
c = ---,
ko v
1] = -ko +-,
v

ko = v'2mE,
n

v = Jvo2 - ko2.
It is not difficult to show that for N barriers the expression for the M-
matrix becomes

MN II
= N-l (
M l1 . (5)
j=O M 21 C)

The matrix product (5) is calculated in the Appendix of Ref. [1] with the
result

M= N (Xl)N-l (AF DC-(N-l))

HC-(N-l) + ( X2 )N - 1 (Bg £C-(N-l))
:fC-(N-l) , (6)

where

A = C + x l M l1 , B = -C + x2Mn
x2 - Xl X2 - Xl

D = _ x l M 12 , £ = X2 M 12
X2 - Xl X2 - Xl

(7)

H = 1 - x l M 22 ,
X2 - Xl
and Xl and X2 are the roots of the equation

(8)
For the calculation of transmission coefficient T we need the matrix ele-
ment M{i only
416 M. MIJATOVIC AND AL

M{i = (Xl)N-l A + (X2)N- 1B. (9)

Using the notation (4) the equation (8) becomes

2x cosh A cos(f..L - kot)e iko1 + e2ikol = o.

,*
X2 - (10)
Introducing the new variables, and by

if L~l
L = cosh ACOS(f..L - kot) = { cos
cosh' , *
L ~ 1 '
l'f (11)
the roots of the equation (10) can be written in the form

xl/2 -_ e ikol
,*
({ cos,} 1= l
.{.sin, }) ,
,* (12)

,*
cosh smh
where the variables, and are used according the condition (11). Then,
transmission coefficient can be found in the form

T _ _1_ _ _ _ _ _1-;--..,....,,-:-:-~
(13)
- IM{i 12 - 1 + sinh 2 A{ Si?;2J~' }.
sinh' 'Y'
The above method can be simply generalized to the cases E > Vo and also
to scattering of a spin-1/2 particle by a rectangular magnetic barriers. Here
we present only the final formula for the transmission coefficient if k ~ 1,
(k = kola, a = avo).

T=

(14)

+ 1 . 2 r-;:2-; { sin2 2N'Y2 }

sin 'Y2
1+ 2 2 sm ay k - 1 sinh 2 N'Yi
4k (k - 1) 'nh2 ,
Sl 'Y2

where we have introduced the variables ,j, ,] (j = 1,2) in the following

manner

LI = cos(aVk2 + l)coska~-

= {COS'
h I'*
-2
- 2k + 1 . ( ~k21)' -k'
~ sm ay k + 1 sm a",
- -2 cos 'I'
2k k +1
TUNNELLING OF NEUTRAL PARTICLE WITH SPIN 1/2 417

L2 = cos(aj-k2 - 1) cos -;;;a~-

-2
2k - 1 . ( ~k21)' -k I: { cos 12, (15)
- -
~
-2
sm ay Ii; - 1 sm a<" = *
cos h 12,
2k k - 1

where ~ = bja,
On the other hand if 0 < -;;; ~ 1, the following formula is satisfied

T=

1+ 2
4k (k
1
2
+ 1)
sin
2
a~
-2
{ sinh~}+
sin
N"I° 2
sinh2i
"11

(16)
1)301 2
+ ------------~~----~~~~~
1+ 2 2 2a\h - -;;;2 { ~~;2J:* }
1
4k (1 - k )
sin
sinh2 "1*

where

L = cosh(aVl- -;;;2)cos-;;;a~+

-2
1 - 2k . ~. - { cOSJ, L ~ 1
+ _ ~smh(aYl- k)smka~ = h * L> 1 . (17)
2ky 1 _ k 2 cos 1 , -

The plot of the transmission coefficient obtained from the equations (14)
and (16) is shown in Fig. 2. On the x-axis is the variable k (impact momen-
tum ko divided by Vo = v'2m/-lBBjh). The white and black bands in the
bottom of the graph describes the bound states of a infinite chain, which
will be considered in Section 3 and 4.
418 M. MIJATOVIC AND AL

N=5 .,{,=1
5 =1 .,{,o=1

o 1 2 3
a >17//////////////Im V////////i//II/Ii/11111T/i1
b )1....1_ _ ---ILe.L.
./h~%LJ.7/-'_LZL£.72_L.Vuh.L.I.?AL.....J~IL.4'/1.~ZL.t.7t..L....'/i-~'/I-LLZL.t.7i-..L....'Iu...ZLJ.71'
/ J.....7/'u...JJZ}

c) 1 - . 1_ _ -----Lf%£...,Lh....L,.%L..J,'a:....L.2.L..J.7J1~-----Lf':.L..J.21L--J:~'_L2~71....L..7i-LL%L.J.7t'....L..2t....L'//....L..Zt....L7/....L..II.{.LZLJ./Z:...LJ/J

d) V1AWI///fl//////1 Vl/71/////fl/Tfl/IIJII//IJ
Fig. 2. Above. The transmission coefficient of the system of rectangular
magnetic potential barriers as a function of impact momentum for
fixed values: N = 5 (number of barriers), ~ = 1 (ratio between band
a), a = 1 (a = avo) and ao = 1 (spin polarization).
Below. The band structure of an infinite chain of rectangular magnetic
potential barriers for fixed values: ~ = 1 and a = 1 which correspond
to (a) up-projection of spin; (b) down-projection of spin; (c) solution
of the equations (18), (19) and (d) relativistical up-projection II = 1.

3. Band Structure

The band structure of the infinite one-dimensional magnetic chain can be

expressed quite simply in terms of the properties of a magnetic moment in
the presence of a single magnetic barrier [1]. Here we present the final result
for band energy equations only:

-2
~ 2k +1 ~-
coS! = cos(ay k + 1) cos -kcl:~ - ~ sin(ay k- + 1) sin ka~, (18a)
2JiyJi2 + 1
TUNNELLING OF NEUTRAL PARTICLE WITH SPIN 1/2 419

-2
~ - 1-2k ~-
COS"'( = cosh( o:y 1 - k-) cos ko:~ + ~ sinh(o:y 1 - k) sin ko:~,
2ky 1- k 2
(18b)
where 0 ~ k ~ 1. The first equation corresponds to the up-projection of
spin (first band in Fig. 2.). The second equation corresponds to the down-
projection of spin (second band in Fig. 2.). The third band in Fig. 2. in fact
is folded over the second band and it corresponds to the common solutions
of the equations (18a) and (18b).
In the case k ?: 1 the equation (18b) transforms into

-2
~ - 2k -1 ~-
coS"'( = cos(o:y k - 1) cos ko:~ - ~ sin(o:y k - 1) sin ko:~, (19)
2kyk2 -1
while the equation (18a) remins the same. The folded over permitted bands
for a different spin projection mean that a flip of the spin is possible. How-
ever, this process is forbidden according to the law of conservation of angular
momentum.

4. Dirac Particle in Magnetic Field

Let us study the Dirac particle of mass m which incidents from the left
and transmits through N magnetic barriers (Fig. 1.). In the case N =1 the
eigenvectors are

(20a)

(20b)

(20c)
420 M. MIJATOVIC AND AL

The joining conditions at x = -a/2 and x = a/2 determine the M-matrix

which elements are same as (4), with a new variables

v
ko
(V~_~)2
w 21?
iT
1
112 '
(21)

v e ko eo ve koeo
E:- - - - - -
- ko eo v e'
TJ= --+--,
ko eo v e
(22)

where

e II 1 + (23)
eo - II + 1- ~'
2k

and II = kofi/mc.
References
M. Mijatovie, G. Ivanovski, B. Veljanoski and K. Trencevski, to be published
Barut, A. 0., Bozic, M., Marie, Z. and Rauch, H., Z. Phys. A328, 1 (1987)
 LIST of PARTICIPANTS
1. Viktor ABRAMOV
Department of Applied Mathematics
Tartu University
Liivi 2-417
EE-2400 Tartu, ESTONIA
e-mail: [email protected]

2. William E. BAYLIS
Physics Department
University of Windsor
Windsor, Ontario N9B 3P4, CANADA
e-mail [email protected]

3. Ingemar BENGTSSON
Institute of Theoretical Physics
Chalmers University of Technology
S-412 96 Goteborg, SWEDEN
e-mail: [email protected]

4. Andrzej BOROWIEC
Instytut Fizyki Teoretycznej
Uniwersytet Wroclawski
pI. Maksa Borna 9
PL-50-204 Wroclaw, POLAND
e-mail: borowiec@plwruwll

5. Krystyna BUGAJSKA
Department of Mathematics
York University
4700 Keele Street
North York, Ontario, CANADA M3J 1P3
e-mail: [email protected], for Dr. Bugajska

6. Ursula CAROW-WATAMURA
Department of Physics
Tohoku University
Aramaki, Aobaku
Sendai 980, JAPAN
e-mail watamura@jpntuvmO

7. Jerzy CISLO
Instytut Fizyki Teoretycznej
Uniwersytet Wroclawski

421
422

pi. Maksa Borna 9

PL-50-204 Wrodaw, POLAND

8. Ludwik Di}BROWSKI
Interdisciplinary Laboratory, SISSA
Strada Costiera 11
34014 Trieste, ITALY
e-mail: [email protected]; fax (+39)40-3787528

9. Geoffrey DIXON
Mathematics and Physics Department
Brandeis University
Waltham, Massachusetts 02254, USA
e-mail: [email protected]

10. Chris DORAN

Department of Applied Mathematics and Theoretical Physics
University of Cambridge
10 Silver Street
Cambridge CB3 9EW, GREAT BRITAIN
e-mail: [email protected]

11. Michel DUBOIS-VIOLETTE

Laboratoire de Physique Theorique et Hautes Energies
Universite de Paris-Sud
Batiment 211
91400 Orsay Cedex, FRANCE
e-mail: [email protected]

12. Andrzej FRYDRYSZAK

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pi. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

13. Gary W. GIBBONS

Department of Applied Mathematics and Theoretical Physics
University of Cambridge
10 Silver Street
Cambridge CB3 9EW, GREAT BRITAIN
e-mail: [email protected]

14. Dan Radu GRIGORE

Department of Theoretical Physics
Institute of Atomic Physics
 423

P.O.Box MG 06 Bucharest-Magurele, ROMANIA

e-mail:[email protected] telex: 11350 ifa r

15. Ralf GRUNEWALD

Humboldt Universitiit zu Berlin,
Sektion Mathematik
Unter den Linden 6, PSF - 1297,
0-1086 Berlin, GERMANY

16. Steve GULL

Department of Physics, Cavendish Laboratory
University of Cambridge
Madingsley Road
Cambridge CB3 OHE, GREAT BRITAIN
e-mail: [email protected]. uk

17. Dmitri GUREVICH

Max-Planck-Institut fur Mathematik
Gottfried-Claren-Strasse 26
D-5300 Bonn 3, GERMANY
e-mail: [email protected]

18. Leopold HALPERN

Department of Physics
Florida State University
Tallahassee, Florida 32306-3016, USA

19. Jerzy HANCKOWIAK

Instytut Fizyki
Wyzsza Szkola Pedagogiczna
pI. Slowianski 6
PL-65-069 Zielona Gora, POLAND

20. David HESTENES

Physics Department
Arizona State University
Tempe AZ 85287, USA
e-mail: [email protected]

21. Alexey P. ISAEV

Laboratory of Theoretical Physics
Joint Institute of Nuclear Research
Dubna
Moscow, SU 101000, RUSSIA
e-mail: [email protected]
424

22. Arkadiusz JADCZYK

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND
e-mail: ajad@plwruwll

23. Bernard JANCEWICZ

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

24. Cezary JUSZCZAK

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

25. Andrzej KARPIO

Katedra Fizyki
Filia Uniwersytetu Warszawskiego
ul. Lipowa 4
15-441 Bialystok, POLAND

26. Jaime KELLER

Div. de Ciencias Basicas, F.Q.,
Universidad Nacional Autonoma de Mexico (U.N. A.M.)
Apartado 70-528
04510 Mexico, D.F., MEXICO
e-mail: [email protected]

27. Richard KERNER

Laboratoire de Physique Theorique
Institut Henri Poincare
11, Rue Pierre et Marie Curie
F-75231 Paris, FRANCE
e-mail: rk@frunip62

28. Malgorzata KLIMEK

Instytut Matematyki
Politechnika Cz~stochowska
pl. Di}browskiego 73
PL-42-200 Cz~stochowa, POLAND
 425

29. Wojciech KOPCZYNSKI

Instytut Fizyki Teoretycznej
Uniwersytet Warszawski
ulica Hoza 69
00-681 Warszawa, POLAND
e-mail: [email protected]

30. Anthony LASENBY

Department of Physics, Cavendish Laboratory
University of Cambridge
Madingsley Road
Cambridge CB3 OHE, GREAT BRITAIN
e-mail: [email protected]. uk;
fax (+44)0223-354599

31. Jan LOPUSZANSKI

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pl. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

32. Pertti LOUNESTO

Department of Mathematics
Helsinki University of Technology
SF-02150 Espoo, FINLAND
e-mail: [email protected]; fax 4871-36848

33. Jerzy LUKIERSKI

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pl. Maksa Borna 9
PL-50-204 Wrodaw, POLAND
lukier@plwrtu11

34. Vladimir Dmitrievich LYAKHOVSKY

Department of Theoretical Physics
Sankt Petersburg State University
198904 Sankt Petersburg, RUSSIA
e-mail: [email protected]
or: [email protected]

35. John MADORE

Laboratoire de Physique Theorique et Hautes Energies
Universite de Paris-Sud
Batiment 211
91400 Orsay Cedex, FRANCE
426

36. Shahn MAJID

Dept. of Applied Mathematics and Theoretical Physics
University of Cambridge
Silver Street
Cambridge, CB3 9EW, GREAT BRITAIN
e-mail: [email protected]

37. Wladyslaw MARCINEK

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pi. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

38. Sorin MARCULESCU

Fachbereich Physik
Universitat - Gesamt Hohschule Siegen
Postfach 10-12-40
D-5900 Siegen, GERMANY
e-mail: [email protected]

39. Sergey A. MERKULOV

Department of Mathematics and Computer Sciences
Odense University
Campuswej 55
DK-5230 Odense, DENMARK
e-mail: [email protected]

40. Mijat MIJATOVIC

Institute of Physics
University "Cyril and Methodius"
P.O.Box 162
Skopje, MACEDONIA
fax: 38-91-228-141

41. Jan MILEWSKI

Instytut Fizyki
Uniwersytet Adama Mickiewicza
ul. Matejki 48/49
PL-60-769 Poznan, POLAND

42. Jan MOZRZYMAS

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pi. Maksa Borna 9
PL-50-204 Wrodaw, POLAND
 427

43. Marek MOZRZYMAS

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

44. Zbigniew OZIEWICZ

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND
e-mail: oziewicz@plwruwll

45. Josep Manel PARRA

Departament de Fisica Fonamental
Universitat de Barcelona
Diagonal, 647
E-08028 Barcelona, SPAIN
e-mail: ubaffpOl@ebcesca1

46. William M. PEZZAGLIA

Department of Physics and Astronomy
San Francisco State University
1600 Holloway Avenue
San Francisco, California 94132, USA
e-mail: [email protected]

47. Ziemowit POPOWICZ

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND
e-mail: ziemek@plwruwll

48. Krzysztof RAPCEWICZ

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pl. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

49. Jerzy RAYSKI

Instytut Fizyki
Uniwersytet J agiellonski
ul. Reymonta 4
PL-30-059 Krakow, POLAND
428

50. Jakub REMBIELINSKI

Instytut Fizyki
Uniwersytet L6dzki
ul. Nowotki 149/153
PL-90-236 L6di, POLAND

51. Jerzy R6zANSKI

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

52. Jan RZEWUSKI

Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

53. Michael SCHLIEKER

Institute for Theoretical Physics
Karlsruhe University
Kaiserstrasse 12
D-7500 Karlsruhe, GERMANY

54. Pavel V. SEMENOV

Moscow State Pedagogical University
Kozmodemjanski 9A kv. 27
125130 Moscow, RUSSIA
e-mail: [email protected] (with subject to Semenov P.)

55. Uwe SEMMELMANN

Humboldt Universitat zu Berlin,
Sektion Mathematik
Unter den Linden 6, PSF - 1297,
0-1086 Berlin, GERMANY
56. Przemyslaw SIEMION
Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND

57. Garret SOBCZYK

Facultad de Estudios Superiores Cuautitlan
Universidad Nacional Autonoma de Mexico (U.N. A.M.)
 429

Apartado Postal # 25
Cuautitlan Izcalli 54700
Edo. de MEXICO
e-mail: [email protected]
58. Jan SOBCZYK
Instytut Fizyki Teoretycznej
Uniwersytet Wrodawski
pI. Maksa Borna 9
PL-50-204 Wrodaw, POLAND
59. Vladimir SOUCEK
Mathematical Institute
Charles University
Sokolovska 83
18600 Praha, CZECHO-SLOVAKIA
e-mail: vsoucek@cspgukll
60. Wolfgang SPROSSIG
Fachbereich Mathematik
Bergakademie Freiberg
Bernhard-von-Cotta Str. 2
0-9200 Freiberg, GERMANY
61. Jan TARSKI
492 Michigan Avenue
Berkeley, CA 94707, USA
62. Ireneusz TOBIJASZEWSKI
Instytut Fizyki
Uniwersytet Adama Mickiewicza
ul. Matejki 48/49
PL-60-769 Poznan, POLAND
63. K. Paul TOD
Department of Mathematics
University of Oxford
St. John's College
Oxford OX1 3JP, GREAT BRITAIN
e-mail: [email protected]
64. Valeriy TOLSTOY
Institute of Nuclear Physics
Moscow State Unversity
119899 Moscow, RUSSIA
e-mail: [email protected]
430

65. Helmut URBANTKE

Institute of Theoretical Physics
Universitat Wien
Boltzmanngasse 5
A-1090 Wien, AUSTRIA

66. Dmitry V. VOLKOV

Kharkov Institute of Physics and Technology
ul. Akamicheskaya 1
31010S Kharkov, UKRAINE
telex: 311052 dekan su

67. Satoshi WATAMURA

Department of Physics
College of General Education
Tohoku University
Kawauchi, Sendai 9S0, JAPAN
e-mail: watamura@jpntuvmO; fax (+Sl )22-262-2429

6S. Julius WESS

Max-Planck-Institut fur Physik und Astrophysik
Fahringer Ring 6, Postfach 40-12-12
D-8000 Munchen 40, GERMANY
e-mail: jup@dmOmpill

69. Aleksandr A. ZHELTUKHIN

Kharkov Institute of Physics and Technology
ul. Akademicheskaya 1
Kharkov 3101OS, UKRAINE
telex: 311052 dekan su