PMblislwd ",

Worid SeienIifIC PllbIdhin, Co. PIe. LId.

PO 801128. Faner Road, smlapore 9121
USA DfFliu: Sui&e IB, 1060 MUI Sttcet, Rivu F.dtt. NJ 07661
UK offiu: 13 L)'MOn Mead, Toaeridle. LeMon N20 IDH

TheedilOnmdpublisbu~lrudullOlbeaudlon mdlbef~p"Ni! 101

\he vvicus journah mel bock. rorlheir _dunce aDd penninioa 10 reproduce the
.ele<ud .uclet found in diU voI._:

Ac:.dem.ic: Pm. CA-aUo{PltpiuT, Amene.o lmGMeolPhyaQ(JUv.IIDIl.PIry.f.T,

ElKVierSciena: Publilhen (JIlid. PIry.f. mdPI!;'.un.T, Pk:nIw Publithin&Corp.
(S)l'fMWtn... u.
Seioc~T, Protrus of ~ PhytiCi (f'rtIf. T1wor. PIry.f.);
Sprinaff' Vw, (C_". UtilI!. PIIy..); Americ.n PbyAcalSociecy(f''''.R.....-l
PIfy6. Ru. un).

ORlGIN OF SYMMETRIES
CopyriaJM 0 1991 by World SdmtUIC f"IIbIbbin& Co. PIe. Lad..
NI n,,," rucned. TIU boot, (X fIIIN tJwnof, - , _ be T~" ±''Xd iIlOJl,1-
(X"'lIICl_. 'LttdTDftkor-.::"""'CIIl. ~~.TI~onIiIItortuty
UV-'- lIOrap GIld TltT;"VGlI)lJklR _ .t..o- or f,o be ilfwlll4d, willw1wl
_il,... ,."..u.iotc /rotft 1M Pllhfulw.

ISBN 99'71 ·96-630-1

ISBN 9971-96-631-X(pbt)

Printed. in Slnppore by JBW Printers A. Binders Pte.. Ltd .

 ."

CONTENTS

Preface y

Chapter I. Introduction

Chapter II . Syrnmetriea From Non· Relativist ic Physics 7

Chapter III. Symmetries From t he Standard Model 16

Chapter IV. Beyond the Standard Model ~5

Chapter V. Tbe CPT Theorem 8.

ChapLer VI . The Fundamental Symmetries 92
Chapter VII. Condusion 129

REPRINTED PAPERS

[1] The Role IUld Value of the Symmetry Principles and Einstein'.
Contribution to their Recot;,nition
by E. Wiper I in S,mmetriu ill Science, eds. B. Gruber and R. S.
Millman, (Plenum Press), pp. 13 21 193

(2) Einstein and the Role of Syrnmetry in Modero Physics

by L. Radicali, in Relcz"vit" Qucznta and Cosmoloff. Vol. II, eds.
M. Pantela and F. De Finis , (J ohnson Reprint. Corporation), pp.
,23 3, 202

[3] Conceptual Foundations of the Unified Theory of Weak and Elev

tromatinetic Interactions
by S. Weinberl, Rev. Mod. Ph.,s, 52 , 515 (1980) 215
viii

[4J Zur Theorie des Wasserstotratoms

by V. Fock. Zeit. Ph,s. 98 . 145 (1935) 224

[5} On the Problem of Degeneracy in Quantum Mechanics

by J . M. Jauch and E. L. Hill. Ph" . ReI). 57. 641 (1940) 234

[5} On the Consequences of the Synunetry of the Nucle&r Hamiltonian

on the Spectroscopy of Nuclei
by E. Witjner , Ph" , Rn. 51 , 106 (1937) 239

(7) Spin and Unitary Spin Independence of Strong Interactions

by F . Gursey and L. A. Radicati , Ph" . ReI). Lett 13, 173
(1964) 253

181 Non· Abelian Gauge T heories of the Strong Inw&etions

by S. Weinberg. Ph.,•. Reo. Lelt. 3 1. 494 (1973) 255

19] Constraints Imposed by CP Conservation in the Presence of

Pseudoparticles
by R. D. Peecei and 8 . R. Quinn, Ph,•. Rev. D16 , 1791 (1977) 260

[10J The Problem of Ai ... - Festschrift in HonoW' of I. I. Rabi

by S. Weinberg. 7hllludioa,f 01 ~lt e New Yort Academy olScien f;u
Series II 38, 185 (1977) 267

[llJ Statistical Analysis or Quark and Lepton Masses

by C. O. Froggatt and 8. 8 . Ni~. Nwcl. Pit,•. B164 , 114
( 1979) 284

\12J Baryon· and Lepton-Nonconservinli Proce6See

by S. Weinbers. Phr', Rev. LeU. 43, 1566 (1979) 311

[13} Natural Conservation La,.. ror Neutral Cu:rents

by S. Glaahow and S. Weinbertj. Ph,•. ReI). DU . 1958 (1977) 316

[141 C P Violation in the Renormalizab1e Theory of Weak Interactions

by M. Kobayashi and T . Maskawa. Prof. Thear. Pia,s. 49. 652
(1973) 324

(15J Unity or AU Elementary. Particle Forces

by H. Geor,i and S. Gluhow. Pia,•. Rn. LeU. 32, 438 (1974) 330
[16] Hierarchy of Interactions in Unified Gauge Theories
by H. Georgi, H. R. Quinn and S. Weinberg, Plays. Rea. LeU. 33,
451 (1974) 334

(17) A l i n Expandable Series of Non-linear rr Models with lnstantons

by A. D'Adda, M. Luscher and P. Di Vecchia, Nucl. Plays. 8146,
63 (1978) 338

(18) The 50(8) Supergravity

by E. Cremmer and B. Julia, Nucl. Plays. 8159, 141 (1979) 352

(19) Composite Vector Mesons and String Models

by S. Mandelstam, in A PtJ$sion f or Physics: Essays in Honour of
Geoffrey Claew, eds. C. DeTar, J . Finkelstein and C. I. Tan , (World
Scientific, 1985), pp . 97-105 424

[20] Search for a Realistic Kaluza-Klein Theory

by E . Witten , Nucl. Phys. B186, 412 (1981) 433

(21) Anomaly Cancellations in Supersymmetric D = 10 Gauge Theory

and Superstring Theory
by M . B. Green and J . H. Schwarz , Ph,s. utI. 149B, 117
(1984) 450

(22) Heterotic String

by D. J. Gross , J. A. Harvey, E. Martinec and R. Rohm, Phys.
Rev. LeU. 54 , 502 (1985) 456

[23J Proof of the TCP Theorem

by G. Liiders, Ann. Ph,s. 2, 1 (1957) 460

(24] Lorentz Invariance as a Low Energy Phenomenon

by S. Chadha and H. B. Nielsen , N.cl. Ph,s. B217 , 125
(1983) 475

(25] p -Function in a Non-Covariant Yang-Mills Theory

by H. B. Nielsen and M. Ninomiya, Nucl. Phys. B141 , 153
(1978) 495

[26] Dynamical Stability of Local Gauge Symmetry: Creation of

Light. from Chaos
by D. Foerster, H. B. Nielsen and M. Ninomiya, Phys. Lett.
94B , 13.\ (1980) 520
x

[27J The Infrared-Ult raviolet Connection

by M. Veh man , Acta Phy!icG Polonica BU , 437 (1981) 526

[28) Infrared Stability or Anti-grandunification

by J . lIiopoulos, O. V . Nanopoul08 and T . N. Tomaras , Phys.
Lett. 948 , 141 (J980) 547

[29] On the Infra-red Stability of Gauge Theories

by I. Antoniadis , J. lIiopoulos and T . Tomaras , Nvc/. Phys. B227,
447 ( 1983) 551

[30J Dual Strings - Section 6. Catastrophe Theory Programme

by H. B. Nielsen , in Fundamenta13 of Qu.ark Mode18, eds. I. M.
Barbour and A. T . Davies, Scottish Universities Summer School in
Phys;cs (1976). pp. 528- 543 566
ORIGIN of SYMMETRIES
 1

Chapter I

INTRODUCTION

Symmet.ry is a vast. subject wit.h !feat sipificance in art and nature. 1 In this
book we shall confine our attention to the synunetries of the laws of nature. These
symmetries playa very important role, both in the derivation of the consequences
of the laws and in the development of physics (Papers 1 and 2]. A notable example
is the close connection between a symmetry and It. conservation law provided by
Noether'a theorem.'
It is a remarkable fact that. the laws of nature have such an arsenal of symme-
tries as translational, rot.ational and Lorentz invariance, gauge symmetries, isospin
invariance etc. The literMure abounds wit.h applications of and prop06als for sym-
metries of the laws of nature. Howe\'er we think it is fair to say that the question of
explaining why there are such symmetries bu not been so widely studied . It is the
purpose of this book to review the development in our understanding of the origin
of symmetries and to reprint a selection of papers giving derivations or explanations
of various symmetries.
We do not intend to present explanations of an essentially philOllOphical nature .
For example we do not discuss the anthropic principle,3 according La which some
symmetries have to be present because they are necessary for the existence of life
and observers. In fact we restrict ourselves La explanations whereby a symmetry
is understood as a consequence of other features of the physical laws , possibly in
some restricted regime of physical phenomena. In order to illustrate what we have
in mind, let us consider how one might imagine deriving parity conservation in a
quantum field theory with only spinless fields.
In a toy model with only spinless fundamental fields tP-(z) , (0 = 1, 2, ... , N) ,
the most general renormalisable Poincare lnvariant action is

5= Jd'.q.) (I)
2

where
N
C(z) = ~ '£, 8,4>'8"4>' - V(4)I). (2)
. ::: 1

Here V(¢') can be any polynomial up to (Ol.lrth order in the spinless fields q;'(z) . By
dimensional counting, renormaJiaability requires that there be no coefficient in the
Lagrangian ! (z) having dimension of mass to a negative power . This requirement
prevents the occurrence of terms with vadients other than the Wiual kindie energy
term !8"4i-O,,,A. Consequent.ly there is no way to make use of t.be Levi-Civita
symbol (,..,,., in It renormalisablt! action with only spinll'lflll fields. It follow8 that the
action must be invariant under the ' parity operation'
x __x

(3)
~'("x) - ~' ( ' , -x).

Under this operation we are treatinA: a ll the spinless fields as scalar (not pseu-
doscalar) fields . We can consider the above argument as an explanation or deriva.-
tion of parity symmetry in our toy world . It is a prediction of this "derivation"
that all the fundamen~al particles in the toy world are scalars under ~he derived
parity. For special forms of the polynomial V(,') ~here could, in addition, exist an
alternative conserved parity operation, under whieb some of the fields ¢' are pseu-
doscalar. Our derived symmetry (3) might not. then be called parity, hut rather
he considered as a combination of parity with a symmetry under sign shift. of the
pseudoscalar fields .
The requirement of the existence of a renormalisable field theory for the spinleas
fields played a crucial role in the above derivat.Kln of parity symmetry. If the spinless
particles were not really fundamental. but. only bound states, then our Lagrangian
(2) would become an effective Lagrangian not constrained by renormalisability. An
effective term such as
,5
A/<I
(1'''''''8 .1.18 .1.'8 .1.58
I'Y flY PY tlY
.1.<1 (4)

would then be allowed, provided there are at least five types of spinlesa fi elds. Here
M is a constant having the dimension of mass, and the parity operation (3) would
be an approximate symmetry at energies low compared to M .
The above type of argument based on renormalisability, but applied to more
realistic Yang-Mills theories. provides an explanation of many of the symmetries
of the standard SU(3)x SU(2)x U(l) model of particle physics . Thus parity, charge
conjugation , time reversal and strangeness (or, more generally, flavour) conservation
are no longer considered as fundamental synunetries, but as natural consequences
of a gauge field theory of litrong and electromagnetic interactions. The development
in the interpretation of these symmetries , from being G priori principles to being
automati c consequences of gauge theories, is discussed in [paper 3] . Thia paper
 J

also discusses tbe derivat.ion or the approximate symmetries of the st.andard model,
such as chiral symmet.ry and Gell-Mann 's SU(3) flavour symmet.ry in the limit. of
negligible quark masses. The symmetries of t.he standard model a re the subject of
Chapter III.
Encouraged by the derivation of the standard model based symmetries, it is
tempting to speculate that even the supposedly fundament.al symmetries of Poincare
invariance and gauge invariance may be derivable , in some physical limit. The
extreme point. of view , that all the symmetries of the laws of nature should be
derivable in this way, is the subject of the last chapter. It forms the nWn content
of the ambitious "random dynamics" program1 [Paper 301 .
The more conventional response to the successes of the standard model and its
symmetries is to assume that the m08t fundamental laws of nature mllSt have a
large degree of symmetry. M&IlY of these symmetries would then be broken, one
way or another, as one descends in energy to the present experimentally studied
ener!), regime. This is the point of view manifested by Grand Unified Theories and
by supersymmetry or supergravity models which we discuss in Chapter IV. In effect
this philO6Ophy amounts to p06tulating the observed gauge symmetry group , since
the symmetry is only explained by the existence of an even bigger gauge symmetry
group, which is itself not explained. However, this is t.o take t.oo negative an attitude.
For instance, the N = 8 extended supergravity model can be 'explained ' or rather
specified, by requiring t.he maximum amount. of supersymmet.ry consistent with
the absence of elementary fields having spin greater than two. Currents wit.h spin
greater t.han t.wo would give such strong restrictions on scattering processes t.hat.
they would enforce an essentially free and therefore uninteresting theory.$ Similarly
it appeared for some time that superstring theory was only consistent for an 50(32)
Of EaxEs gauge group [Paper 21], but now other possibilities are known (see chapter
4.6).
A distinction is often made between geometrical and non·geometrical symme.-
tries. However the content of such a distinction can be somewhat arbitrary. Space
uanslational and rotational invariances are clearly geometrical. But it is only since
the publication of Einstein 's theory of relativity that Lorentz invariance (or Galilean
invariance) and time t ranslational invariance have been considered geometrical. Fur·
ther it is a matter of taste whether a supersyrnmetry (should it exist) is called
geometrical, since it is inteItwined in a non·trivial way with the geometrical sym-
metries of space and time translations. Also it is questionable whether the general
cc> ordinate invariance, or diffeomorphism symmetry, of Einstein's theory of gravity
should be called a geometrical symmetry, since it is such a formal symmetry. On
the other hand the gauge symmetries of Yang·Mills theories are usually not called
geometrical, unless one takes a rather extended definition of "geometrical" within
a fibre bundle formulation. However a natural generalisation of the work of Kaluza
and Klein [Paper 20] showed how the diffeomorphism symmetry, or reparameteri·
sation invariance, of a higher than four dimensional space.-time manifold could give
rise to the gauge symmetry of a Yang· Mills theory. So, using the Kaluza· Klein idea,
4

one mAY -derive- ,;aut;c symmetries (rom the Seometrical symmetries of a hisher
dimensional space-tUm. In rad part of the diffeomorphism symmet.ry of gravity
in the hi,;her dimensions becomes a gauge symmetry in four dimenaions. In some
sense, however, "his jut meana that one Uflumed ayrrunetry is interpreted 811 or i.
transformed into another one.
Af! emphasized in Wigner's article [paper 11, we must aeparatc the symmetries
of the physicall.". from the symmetries of the whole world &8 it now e.xiau. In
general the state of the world is not invariant under the symmetries of the laws of
nature . One eays that the "initial conditiona" are not invariant.
The distinction between the laws of nature and the initial conditions becomes
rather blurred in tbe cue of spontaneous symmetry breakdown . Spontaneous sy~
metlY breakdown occurs when the vacuum state is not invariant under the symmetry
in question. In this eue, should the state of the vacuum be con.idered as part of
the laws of nature or just &S part of the initial conditions? This ambiguity plays a
role in some of the derivations of gauge .ymmetries discussed in Chapter VI.
When the vacuum is no1. invariant under a symmetry of the laws of nature,
most physical processes do not. manifest this .ymmetry. However there are .till
phy.kai consequences which make the .ymmetry experimentally testable . In the
case of a .pontaneously broken continuou. global symmetry, the Nambu· GoldsLone
theorem predicts the existence of a masaleu particle. This particle is a boson, unless
a supersymmetry is broken. The prediction can be avoided by the Higgs-.Kibble
mechanism when the spontaneoWily broken .ymmetry is a loe.al gauge .ymmetry.
The predicted Nambu·Goldstone boeon can then be shown to be unphysical, by
an appropriate gauge choice. At the same time , the otherwise massiest spin 1
gauge particle acquires a nOli-zero mass. The degree of freed om associated with the
Nambu-Goldstone bceon is transformed into the required helicity zero component
of the massive spin 1 particle. Even after "Uiggsing" "he gauge symmetry in this
way, there are still traces of the symmetry left.. For instance the coupling constant.a
of various particles to the gauge particle are related . These traces of the gauge
symmetry remain due \0 the requirement of renormalisability. RenormaJisability
forbids terms in the Lagrangian density baving coefficients of dimension mass to a
negative power. If such irrelevant (i.e. non.renormalisable) terms were allowed , tbe
last traces of the gauge symmetry would vanish . We make use of this p068ibility
in Chapter VI , when an arbitrary non-gauge invariant lattice theory is written &S a
"Biggsed" theory.
The successes achieved by tbe application of symmetry principles in physics have
led to the consideration of ~caUed dynamieaisynunetries.' Dynamieal symmetriea
are nol true symmet.ries in the sense of being exhibited by the laws of nature. For
dynamical s)'tlUlletr~, tbe Hamiltonian H i.s a function of the symmetry groop
generators but H does not commute wi"h all the generators. The multiplew of the
dynamical symmetry group are then split by H . but not mixed. Thus the dynamical
symmetry can be important for the classification of the Hamiltonian eigenstat.es,
even though the act.ual system does nol. really exhibit the symmetry in question . A
 5

CHAOS?
Chopl" W
FUHDAIIII:NTAI.
SVMIIIII!:TIUES

Chapter IV
SEYOfW nlE ST.IINOARD

""""
•, •
" .
HIGtlER
" ~

ENERGY
SMALL[A
04SU.HCE

Chapt.,. ill
THE STAH~RD
..o"

PHYSICS

Fir;. 1.1. The Quantum Sw.irusc; the upper part d of eoune .peculalive and can be chan,ted to
accbmmodate the ~ader'. favourite theory of everythin,.

simple example is the SU(3) symmetry of nuclear physics introduced by ElIiot.6 ,7

We shaU not be concerned with such dynamical symmetries in this book , since they
are not true symmetries of the Ja..... s of nature.
6

In the following chapters we give examples of symmetry derivations, which are

organised to foll ow the quantum staircase (Fig. 1.1). That is to say we first discuss
the derivation oCsymmetries in the lowest energy (or rather longest distance) regime.
We then go, successively, to higher and higher energy (smaller and smaJler distance)
physics. Thus in Chapter II we consider non-relativistic examples. In Chapter III we
consider the symmetries derived from, but not directly put into, the well-established
standard SU(3)xSU(2)x U(l) Yang-Mills model of quarks and leptons. Chapter IV
deals with symmetries derived from theories beyond the standard model. A separat e
Chapter V is devoted to the CPT theorem, which provides the paradigm for a.
symmetry derivation. In Chapter VI we present the more speculative derivations
of the most fundamental symmetries of Lorentz invariance and gauge invariance.
Finally, in Chapter VII , we present an overview of the progress made in explaining
the origin of symmetries and a discussion of random dynamics.
It should be emphasized that, in most cases, the "symmetry derivations" are
a posteriori understandings of symmetries, which have been known in advance on
more phenomenological grounds. If one takes the random dynamics program seri-
ously, however, one could hope t.o derive all the observed symmetries without having
to assume any a priori. This would represent a new philosophy of the origin of sym-
metries: the observed symmetries arise naturally even if the m06t fundamental laws
(at very high energies) do not P06Sess them.4 ,a,1I According to this philosophy, one
might attempt to read the chapters of ou r book in reverse order (i.e. coming down
the quantum staircase). Although there a re many gaps to be filled , one would then
take it as a logical derivation of first some very fundamental symmetries which are
used in earlier chapters to der ive all the experimental ly observed symmetr ies. Need.~
less to say, this ambitious program of random dynamics is far from completion in
a convincing form . It is not presented here as the mOISt compelling argument for
believing in the symmetries.

Rcfc rcJu:c.
1. H . Weyl, S ,mmtl,., (Princeton Univenity Preu, 1952).
2. E. L. Hin, Rn . Mrnl . Pit,•. 2', 253 (1957) .
3. J. Ba.rrow and F. Tipler, Tilt ,hlltl'O,i( Cum~I~lie'" Pri.ei, Ie (O:dord UnJversity Preas,
1986).
<t . H . B . Niel6en, G. .,e Tlteone. oj lite Ei,ltliCl, " AkUlompolo, Finland 1982, eds. R. RaiI.O
and J . Lincifon, p . 288 (SpriDlcr-Vt:rlq;, 1983); H. B. Niellcn, D. L. BelUu,tt and N. Brene,
~Rccen' Devclopmew in Qu.mwn Field Theory,~ P,.«cc il4fl ~J lite Nitl Bolr C.II'cn i.1
COII/er.llce, Copenh&«en. I985, u . J . Amb~, B . Durhuus and J . L . Petenen, p . 263 (North-
HoI.l&nd, I985).
5. S. Weinbers, Pit, •. Re.. B13B, 988 (196.5 ); R Behrends, B. de Wit , J . nn Holten and
P. van Nieuwmhui~en , J . Pit, • . , AU, 1643 ( 1980).
6. A . Arima and F . lacheUo, An. Pit, •. (NY) 111 , 201 (1978).
7. J . P. Elliot, P nc. Ro,. Soc. M.S, 128 and Mi2 (1958).
8. J . lliopolOUf, Pncee4i.,. oj 197!J I .'ena,liOll.l COII/ ernee o. I1i, 1t Ell e,.,., Pl.,.ie., Gene va,
Vol.. I , p . 3n (CERN, 1980).
9. F . EDJlert, ~Recent Development' in Quantwn Field Theory~. Procuci.,. oj lite Nidi B ob
Cellltlllli,I Co II/enllee, Copt:nhaJen 1985, edt. J . Ambj.m, B. Durhuus and J . L. Peter-sen,
p .39 (North-Holland, ISM).
 7

Chapter II

SYMMETRIES FROM NON-RELATIVISTIC

PHYSICS

We commence with an example from the bottom of the quantum staircase and
consider the derivation of the scaling symmetries of macroscopic physics. The laws
of physics are not themselves invariant under a change of scale, as first pointed
out 300 years ago by Galileo in his discussion of the sizes of living creatures. 1 The
physical effects of scalin,; are very often studied using the method of dimensions.
Dimensional analysis is particularly useful in tbe design of smaIl scale models for
testing an aeroplane or a ship .
~ a simple example from hydrodynamics, suppose the motion of a body in a
viscous incompressible fluid is to be modelled. The model body is geometrically
similar 1.0 the prototype, differing from it only in size. Then , provided the Reynolds
number
(1)

is the same for the model and the prototype, the flow patterns are dynamically
similar .' Here p and p are the density and coefficient of viscosit.y for the fluid , I
is a characteristic length of the object and v is some representMive velocity. For
simplicity we shaH just consider one fluid , so that p and pare const.ants. There is
then a symmetry under the operation of scaling the site of the model up by a factor
{" and , at the same time, reducing velocities by t.he same factor .
It is clear, from the atomic structure of matter, that the above scaling law cannot
be exact.. Indeed, if we scale down to a model size so small that only a fraction
of a molecule of fluid is required , the scaling law becomes meaningless. Thus we
can really only expect. the scaJing law to be approximately true, corresponding to
the situation when there are many molecules present . In practice, of course, the
number of molecules will be very large for any real model and the scaling law can
be a very good approximation.
This scaling symmetry is usually derived by applying dimensional arguments
to the hydrodynamical differentiaJ equations - Navier-Stokes equations.' Let. us
8

bere derive the scaling law directly from the underlying molecular dynamics, in
the limit when the number of fluid molecules becomes very large. The scaling law
can be stated in the following form: The time development of two flows of the
same incompressible fluid around two solid bodies are dynamically similar, when
the second flow is related t.o the first by scaling (a> all geometrical lengths by &
factor {. (b) all (non·thermal) velocities by a factor {-I and (e) aU time intervals
by a factor {~ .

FiS. ~.l. Hydrodynamic&! Ko!WJl& .ynuneuy illUfU'&ted by aame 8uid 8owin& out or two slrNlar
tube. reJ..ted by .. ~ ... Cac:tor ( :z 1/2. The lull-line &ITOW. repl'e.e:nt the di.p1a.can.ent.. 01
molecules due to the .. verase ftow of the Suid durin&: time inlervaI. 6 t and (2 A I ~vdy. The
arTO".
d ..hl:<l. r'epf'Cllent the total ddplacemenb includin& the eRects of Brownian moti«!..

The derivation of the sealing law from molecular physics proceeds along tbe
following lines. The velocity of t:ach molt:cult: of 8uid is consideroo to bt: comp06t;d of
a direded motion component Vi and a Brownian or thermal motion component VB .
Wt: now consider the development of the two flow patterns, during corresponding
small time intervals 6.t and ~2At respectively. First SUpp06t; each molecule only had
the directed motion component of its velocity. Then molecules from corresponding
regions in the two flow pattt:rns (see Fig. 2.1) would travel with vclocities Vi and
~-lVd , for time intervals 6.t and ~26.t respectively. It follows that a molecule in the
second flow travels a distance ~ times further than a corresponding molecule in the
first flow , during these time intervals. Now we must considt:r the Brownian motion
due to the thermal fluctuation8 of the molecules, which is 8uperimp06ed onto the
direet.ed motion. The root mean squlU"e displacement of a molecule in a given fluid ,
due to this thermal motion , is proportional to the square root of the time interval
considered. Thus the thermal displacement of the molecules in the second flow
is also ~ times further than that in the first flow, during the corresponding time
 9

intervals. This follows simply because we ale considering the same fluid, with ~he
same local properties, in both Row patterns, and the time jntuval for the second
flow ise times that for the first flow.
We ha.ve shown above, for both components of their molion, that the displace-
ment of molecules Crom a region of the second flow is related to the displacement
of molecules from the corresponding region of the first flow by ..he scaling fac tor { .
e
Does it now (ollow , after the time intervals ~t and 2 6t in the two floWl! , that the
average (Le. directed) velocities in corresponding small regions are still related by
a factore- 1 , as required for dynamical similarity? Yes, indeed it. does, as is easily

seen by considering the molecules in small corresponding regions of the two flow
e
patt.erns at some instant . Since the scaling factor applies to the displacements
due to both component.s of their motion, the same fraction of these molecules must
arrive in two other small corresponding regions, afler time intervals ilt and {2ilt
respectively. Thus the avera,e molecular velocities, in corresponding regions of the
two flow patterns, are still related by the scaling factor e- t.

We have therefore derived the hydrodynamicalscaling law for an incompressible

fluid from molecular dynamics, in the limit of a large number of molec!lles where
the methods of statistical mechanics apply. It should, of course, be stressed that
the above argument does not imply a general scaling symmetry of macroscopic
physics. In fact , under the scaling law derived above, the extra pressure p (relative
to some representative value in the fluid) in corresponding re,ions of the two flows
scales according to p _ e- 2p. h follows that t.he force exerted by t.he fluid on

corresponding areas of the two solid bodies is the same, and the stresses on the
body scale like e- 2 • Hence the body would break for a sufficiently small C i.e,

the strength of the solid material violates the scaling law. Also, for small e, the
velocity of the fluid and the pressure become high. In practice the velocity of a
real fluid would approach the velocity of sound and the fluid could no longer be
considered incompressible , The scaling law in the above form then breaks down,
since the dimensionless Mach number becomes relevant to the flow pattern &8 well
as the Reynold number.
As a second example from macl05copic physia, we consider the mechanical
properties of an organic compound , such as sugar, which has parity non-invariant
molecules. It has been known, since Pasteur's work in 1848, that all sugar molecules
produced naturally from living things have the same kind of thread or handedness .3
The mirror image molecule is not produced naturally. Sugar is thererore optically
active and when polarised light is passed through a solution of naturally produced
sugar its plane of polarisation is turned to the right. However when we manu-
facture sugar artificially, from substances which are not themselves asymmetrical,
both kinds of molecules are produced in equal numbers. There is thus a kind or
spontaneous breakdown or parity symmetry in naturally produced sugar, arisillg
from the common origin of life on earth right back to the completely molecular
level. Despite this parity violation in the structure of sugar molecules, the purely
mechanical properties of a susar crys~aJ or of amorphous sugar are parity invariant.
10

--~ , ,
I
I
iii"
r ____ , '

Fi,. 2.2. Suev and ill; mimw irn&&t have identical elasticity ~. delpite the parity non.
invariance 01 ill; moleculeJ.

What is the origin of this parity .ymme'r), in 'he elastic properties of flusar?
The important. point is that the parity violation in sugar ia mic.re»copic (i.e. the
handedness of single molecules) and thus can only affect the local form of \he elastic
deformation (FiS. 2.2). The local deformation of a material is described by a second
rank .ymmehic WlSOl - the atrain tensor .· The Cree cnere density of a material
thereCore only depends on the local atrain tensor, which is invariant. under puity
.imply because the atrain tensor is second rank and not a p8eUdoten80r. There is
simply no way t.he elasticity coefficienLa CAll be made complicated enough to violate
parity Iymmdry. Even if the micf08Copic parity violation mre due to truly parity
violating forces, lUI in weak interactKma, the elast.ic propert~ of augaz could not
violat.e parit.y symmet.ry.
We now proceed up t.he quantum staircase and consider symmetries in non-
relativistic quant.um mechanic.. We first consider ~he appazently accidental degen·
erades of energy levels, which occur for some dynamical ayst.el1l8 . In particular the
origins of the underlying 0(4) and SU(n) symmetries, responsible for the degenera,.-
cies in the hydrogen atom and t.he n-di mensional harmonic oscillator, are presented
(Papers" and 5] .
The best known example is that. of t.he SchrOdinger equation for the electron in
a hydrogen atom, which bas the symmetry of the four dimensional roLat.ion group
0(4). The same symmetry of course applies to any particle in a Coulomh-Hke
: potential, such as (or planet.a.ry motion. The great progress made in science
during this millennium is due in l&Z'ge part. to the tractability of the ~ potent.ial
 11

problem, which is a consequence of this very 0 (4) symmetry making tbe behaviour
of the aYlitem much simpler. The D( 4) symmetry has been very helpful in makins:
the Keplerian law simple, as .....ell as in the development. of Bohr's atomic model,
although Kepler and Bohr were unaware of the symmetry at the t ime.
The 0 (4) symmetry of the hydrogen atom is a rather hidden symmetry, which
Fad [Paper 4] first. made manifest by writing the SchrOdinger equation in momen·
tum representation . By stereographic projection onto i.he surface Sa of a unit four
dimensional Euclidean sphere (Fig. 2.3) , the points in momentum space Me ass0-
ciated with the points on S,. After a simple transformation of the wave (unction ,
Fock then obtains a new Schrooin,!\er ~uation for the hydrogen atom , in which the
Hamiltonian operator is expressed as a convolution with the fundion

constant
(2)
sin 2 ,
!!.

where w is the distance along a great circle on the 5 3 -sphere. In this way, it is
seen that the Hamiltonian for the hydrogen atom is invariant under rotations, in
4-space. of this Srspbere around ita centre.

I
,, " I
, I" /
" I /
........... - ' ' '
... - J. .......__ ..... /.

- --- ---
Fir;. 2.3. IDu.uation ol the atuoor;raphK p-oject.ion pwn by Eq. 3 of P-.per <I. One dimeru.icn
baa been auppR:Med.

The above derivation of the 0(4) symmetry for Ute hydrogen atom is exact, as
long as the finite size of the nudeus and the effects of spin and relativity theory can
be ignored . It is simply a somewhat hidden symmetry which had to be discovered .
In a similar way Jauch and 11m [Paper 5J showed that the Hamiltonian for the n-
dimensional harmonic oscillator has a symmetry under the SU(n) group. Thus the
Elliot SU(3) dynamical symmetry of nuclear physics, mentioned in the introduction ,
would be a true symmetry of the nuclear shell model Hamiltonian, if the nucleons
in the outer shell of the nucleus moved in a t hree-dimensional harmonic oscillator
12

potential wit.hout. any residual interact.ions. It has also been point.ed out. 5 that. , in
the presence of a weak magnetic field , the symmetry groups (or the hydrogen atom
and the three-dimensional harmonic oscillator both reduce to SU(2) .
An important symmetry in non-relativist.ic quant.um mechanics arises from tbe
deeoupling of spin in atomic physics (for low Z atoms) . In atoms wit h low al.omic
number, Z ..... 1, it is well known t.hat 1he order of magnit.ude of t.he electron velocity
is given by the fine atructure constant Q . Here, and in the rest of the book , we Wle
natura l units with h = c = 1. It is thus the sma11nesa of the fine structure const.ant
(1' , which makes the non-relativistic approximation applicable in a tomic physics. To

leading non-vanishing order in a it can be shown, from the Dirac equat.ion for an
atomic electron in the Coulomb field of the nucle us and the other atomic electrons ,
that the spin degree of Creedom Cor the electron completely decollples. The spin
then just act. as a label Cor two dHre rent types of electron, and its presence is only
Celt through the effects of Fermi statistics for t.he electron .
In a light atom, Z _ 1, we readily find the following orders of magnitude for the
electron variables (in the rest frame of the atom), in t erms of its mass In and the
fine structure constant Q = e2/41( .

Velocity v .... ar , r adius of orbit 0 ..... ...!...,

am
(3)

p'
momt ntum p .... O'm, tner!), E= - ..... Q ~ m . (4)
2m
Similarly the electromagnetic potential A,. , in Coulomb gauge, satisfies
(5)

The character ist ic time scale for an appret:iable variation of a dynamical variable is
given by
1 1
6t ..... - ..... -2- (6)
E Q m·
Using the above orders of magnitude, we can make a Foldy-Wouthuysen
transformation s , 7
t/I = e -is 1/J' (7)

which derouples the large and small components of the spinor 1/J' for the atomic
electron, to leading order in Q , The Dirac Hamiltonian

H = pm +0 ' (P - eA) + eAo (8)

is then transCormed into

H' = eisC H _ tOi le -iS (9)

1 .
= H + ' (S, H] - 2[S, (S, H]j - S + ... (10)
 13

The three contributions to H in Eq. (8) are of order m , crm and cr 2 m ~pectively.
To leading order we t.ake

5 = -;(ia . P/2m , (11)

which is a small quantity of the order of the fine structure constant cr , and obtain

p'
H I = pm + 2m + tAo + O(o'm). ( 12)

The second and third terms in Eq. (12) are of order crllm and are obtained by
including terms from H , in Eq. (10), up to order om in [S,H) and just t.o order
m in [5, [5, H)). It follows from Eq. (6) and Eq. (11) that S is of order a3m and
therefore does not contribute.
After this Foldy·Woutbuyaen transformation and neglecting terms of order Q3 m
and higher. there are no spin dependent t erms in the Hamiltonian H' at. all. This
leads to tbe separate conservation of the spin of each electron in tbis approximation
and, a fortiori , the total spin S of all the atomic electrons together i. conserved.
Since the total angular moment.um J is conserved, it follows that the tot.al orbital
ansular moment.um L of all the electrons together is also conserved. We have there-
fore derived , to leading order in 0', the symmetry of H' which underlies the Russel-
Saunders LS coupling scheme in atomic physics. It is an excellent approximat.ion
for light a.nd medium a.toms.
The neglected terms of order a'm include a spin-orbit. term, which violates
separate L and S conservation and ads as a pert.urbation, splitting t.he ot.herwise
degenerat.e energy levels. ~ Z increases t.he spin-orbit term grows and becomes
of the same order of magnitude as the angle dependent. part of the elect.rostat.ic
interact.ion (i.e. t.he residual Coulomb interaction between the electrons len after
subtract.ing the central Harl~Fock potential) . The LS coupling scheme then be--
comes unrealistic and other coupling schemes have been considered.s However the
alternat.ive j j coupling and jl coupling' schemes have a much more limited use,
being restricted to certain types of atomic confi&nrations.'
The Wi&ner SU(4) symmetry of nuclear physics (paper 6] is, at first. si&ht, of
a similar nat.ure to t.he spin decoupling symmetry of atomic physics. In an 5U(4)
symmetric theory all the charge-spin states of a nucleon are completely equivalent .
This would be the case if, for instance, both spin and iaospin variables were absent
from the nuclear interaction.
The Wigner SU(4) symmetry is defined as a symmetry ofthe nuclear Hamilto-
nian, under the simultaneous transformation of all the nucleons by the same group
element of 5U(4) . The transformation of each nucleon is defined by the operation
of the appropriate 5U(4) matrix on the four component column vector coDsisting
of the four spin-i.aoepin states of the nucleon p T,p 1, n Tand n 1.
A Majorana nucleon-nucleon force is proportional to the factor (1+4rl " 2) · (1 +
"'1 . "'2) times a space dependent fundKm . Here •• and ri( i = 1,2) denote tbe Pauli
14

matrices (or the spin and u;06pin of the two nucleons . Thus the Majorana force is
an exchange (orce which interchanges both the spin and isospin of the nucleons. It
is clearly 5U(4) invariant, since the 5U(4) operation simply rotates every nucleon in
5U(4) space. Majorana plus Wigner (spin-isospin independent) forces alone would
therefore give an 5U(4) symmetric model of nucleonic interactions.
However , it is not possible to derive the Wigner 5U(4) symmetry as a non-
relativistic limit. analogous to the spin decoup!ing symmetry limit in atomic physics.
The spin one vector mesons exchanged in nucleon-nucleon interactions have finite
masses and large (strong interaction) coupling eonstants. Also, aL larger range l
spin zero scalar (0' meson) and pseudoscalar (r meson) exchange are important .lo
In rad the nucleon-nucleon interaction is rather complicated and the Wigner SU(4)
symmetry is only a very approximate symmetry. NeverUtele:Sl there is a cancellation
between the teneor poten~ia.ls from p vecLor meson and ,.. meson exchanges, and
sensible result.s have been obtained l l applying the SU(4) symmetry to the ground
state energies of medium- heavy nuclei , 30< A < 110.
We have no convincing explanation for the origin of the approximat.e Wigner
5U(4) symmetry and perhaps it should not therefore have been included in this
book . However it served as an inspiration for the introduction of another approxi-
mate symmetry, the SU(6) spin and unitary spin symmetry of strong interactions
(Paper 7]. This SU(6) symmetry is interpreted 12 as the generalisation of the Wigner
SU(4) symmetry obtained by using quarks instead of nucleons . The transformation
of each quatk is obtained by multiplying the appropriate 6)(6 5U(6) matrix into a
colunm vector, with elements denoted by all six combinations of the three lowest
quark flavours u, d, s and the two spin at.at.es T and 1. The SU(6) transformation
matrix is the same for every quark. The approximate SU(6) symmetry of hadrona
is then considered to be a conaequence of the additive quark model.
In the simple additive quark model, the quarks are treated as effectively free
particles bound inside hadrona. In such an approximation, when the effective quark
masses are taken to be equal, there is no apin dependence nor 8avonr dependence
in the hadron spectrum. Consequently there is a symmetry under 5U(6) in the
simple additive quark model. It la, of course, hoped t.o derive the simple additive
quark model as a non-relativistic limit of quantum chromodynamics (QeD), the
relativistic quantum field theory of quarks. In the next chapter we consider t.he
aymmetries derived from t.his ,auge theory of stron, interactions and the Glashow-
Weinberg-Salam gau&e theory of the ele<:troweu interactions.

1. O ' An::y Wentworth Thomson, O. C.......d. n~ Fo"" (Cambrid&e Univa.ity PreM, 1917) .
2. G . K. 8atdldor, F'.i~ D, •• miCl (C.unbrida;e Univen.ity ~ , 1967) .
3. H. Weyl, S, ..""d,., (Princeton University Pnu, 1952).
4. L. I..nd... _d £ . M. Li!ah.itz, nu,., 0/ Elu/ici/, (Persamon P ........, 1859).
5. M. Modunsky, N. M~. E . Murow and J. W.B. Huches, A.a. Pio, • . 155,231 (1934).
6. L. L. Folody and S. A. WouthlJyHn, Pi, • . Rev. 7', 29 (1950).
7. C. Itzybon and J . 8 . Zuber, Q... I.m Fid~ nu,., (McGra • • HiU, 1980); J. D. 8jorken and.
S. D. OteU, Rd&ti .. u/i e Qua /.m Met;i.,.it;, (McGraw.Hill, 1964).
 15

8. H. G. Kuhn, A'oJ",i.: S,ed,.. CLon.man, 1969).

9. G. Rac.h, Pl, •. Ref. 111, 537 {1942}.
10. G. E. Brown, "Nuclear ~" , wdu'e Nol" III tAr Worbh, "'c1~ ., G./1 Ldc,
Mkbipn 1979, eda. G . F . Bertacb and D. KUTath, p . 1 (Spin,er-Vedat:. 1980).
11 . P. hlUlSini and L. A. Radiea.ti, pa,•. Left. 6, 322 (1963).
n . 8 . Sakita, Pl,•. Rn. lS6, 81756 (19&4); O. W. Greenbera:, Pit, •. Rn. Ltl'. 13, 596 (1964) .
16

Chapter TIl

SYMMETRIES FROM T HE STANDARD

M ODEL

3.1. The Standard Model

At present (1991) it seems that all the experimentally accessible interactions of
matter are described by the standard model of elementary particle physics devel-
oped durinr; the last two decades.!·2.:! The standard model combines the Gluhow-
Weinberg-Salam theory of electroweak interactions with quantum cbromodynamics
(paper 3].
The standard model is a Yang-Mills t.heory of quark and lepton interactions
based on the gauge group S(U{2)x U(3» , defined by the following set. of 5 x 5
unitary matrices U

U(2)
o o o
o o o
Uc detU ;: 1 ( I)
o o
o o U(3)
o o
This S(U(2)xU(3» group is a subgroup of SU(5) and has the same Lie alt;ebra &8
U(I)xSU(2)xSU(3). Following Michel," we here give a physical meaning to t.he Lie
group as well as to the Lie algebra, by requilin,; that all t.he particle multipleta of
the theory must. form genuine (Le. sint;le-valued) representations of the Lie gro"p.
or course the coverinr; group associated with a given Lie aJ&ebra can always be
represented in Lhis way. but. we also require the true Lie group of a theory to be the
'smallest' one that can be represented on all the particle multiplets. The standard
model group S(U(2)x U(3)) turns out' to be representable on precisely those particle
multiplets which obey the electric charge quantis&tion rule

(mod I) . (2)
 17

Here "triality" is 1 for quarks, -1 for antiquarks and zero for colourless stat.es, while
y is weak hypercharge (conventionally normalised to be int eger for leptons) and t3
the third component of weak isospin.
Let us now write down the Lagrangian densit.y for the standard model as &. sum
of fi ve terms.
(3)

The first term

(4)

contains the gauge invariant kinetic energy contributions from the U(I) , 5U (2) and
5U(3) gauge fields A,. , AJ',i and A,. .. JIi , expressed in t erms ofthe field tensors

(5)

(6)

and

F" .. ofJ =lJ,.A..,/ -lJ .. A,.o'#

i) .
- (Apo'Y A ". /' - A.. o" A .. (7)

Here Al'ii and A,... JIi are respectively 2x2 and 3x3 Hermitean traceless matrices
with vector fi elds as elements. The indices i,i = 1.2 run t.hrough a basis for the
defining representation of 5U(2) and Q , fJ = 1,2,3 run through a basis for the defin-
ing representation of 5U(3) . As usual a repeated index implies summation . The
gauge coupling constants 91 , 9, and 93 have been absorbed into t.he definition of
the corresponding U(l), SU(2) and SU(3) gauge fields and therefore do not appear
explicitly in the definitions , Eqs. (5)- (7) , of the gauge field tensors.
Next we introduce the matter fields pven in Table 3.1, making explicit. all group
and generation indices as in Paper 12. Here we have listed. the fermion (lepton
and quark) fic1d combinations i;IIL , Ge;RT etc., which are 'len- handed' in the sense
that. the parLicles which they can annihilate have negative helicities (when fermion
masses are ne&lected) . Len-handed and right-handed Weyl s pinol'S satisfy

and (8)
18

Table 3.1. The irredllCibie I'ep,......nleljon. for the leCt-handed Weyl ramlon field. and the Hiep
.e.I... field in the It.aIldanl model. I = weak .pin, 13 :: third c»mponcnt ol .w i.o.pm, , :::
weak hypCTCbN'&e and Q = electric char&e.

Matter U(I) SU(2) SU(3)

field, repre.enw.tion repn:sentation, repr~tal ion

ir=Q- 13 ,
1,
~"L -l I
T I 0 I
CC. R
,, ,
-.,,
1
fi(uL t
C-u ... R T 0
Ca;;;;;;r "
0
"
.; t ,
! I

respectively and C is t.he antisymmetric charge conjugat.ion matrix satisfying

C11'C- 1 = _1 t1T • (9)
Correspond ing to a right-banded field ,pR, one can construct the linear combination
,pc = C~RT of the component.s ortbe bermitean conjugate field ,,1t which. formally,
is the result of a charge conjugation operation. The fie ld Ij;c is left-handed and
represents the C P conjugat.e part.icles.
T he fi elds li"L describe both neutrinos (for i=l) and lef\-banded charged, Q =
- I , leptons (for i=2) . Similarly qi 0 4L describ es left-handed quarks with two differ-
ent values of the electric charge, Q = and J -i.The right.-handed charged leptons
are described by the field e" n , the Q = ~ right.-banded quarks by U,u R and tbe
Q= -l right.-handed quarks by do"R · Rjght-handed neutrinos are not present in
the standard model. Any non· zero (Dirac or Majorana) mass for a neutrino would
have to be attributed t o effects outside tbe standard model.
The generation indices a , h = 1, 2, ... , N enumerate &et.& of isomorphic ine.-
ducible representations of the standard group on tbe fermion fielda, and will be
used to distinguish the N ~ 3 generations of physical lepton and quark staLes
e, 1' , T .•• ; tI , c, t ... ; d, . , b . . .. The '.bare' or 'mathematical' basis of Weyl fermion
fields entering the Lagrangian density C do not correspond to the physical ma.ss
eigenstates, and we introduce a prime on the lepton and quark fields Ii"" Ce~/l.T
etc . to denote tbis fact . The interaction of the fer mions witb the r;auge fielda is
then given by the second term in Eq. (3)
~ = jjii.Li"D,.iif; .. L
+ ie'aRr' D"e~R
+ Iy.~ io IIL"( "DpiI 0 , ' iP"L
,
. - ,0 II .n p,
+"11.1 " R7( lU,.o UP"R

+ id'O .R"'" D"oPctPdl . ( 10)

 19

Here the covariant derivat.ives D,. , Dl'ii I D,.,,~ and D,.i{/ are those appropriate to
the different irreducible representations in Table 3.1. For nampk, the eov&riant
derivative for the lefl.-handed quark doublet field qjP IIL becomes

1
T he fac tor of multiplying the U(l ) gauge field is just the value orihe U(I) quant um
number , !!I. 9J,.£
for the fi eld given in Table 3.1.
The final field in Table 3.1 is a complex sealar Hig&! doublet tPi . which generates
masses for t.he weak gauge bOSOM and fermiona, by Ipontaneous breakdown of t.be
electroweak gau&e,symmetry down to the electromagnetic U( I ).". gauge symmetry.
The Higgs particle is the least reliable element of the standard model and its exis-
tence has still to be established. The gauge and self interadions of the Bigs field
are given by
c, = D,. ,J ~JD+"ii41+j - V(41;fI+J ) (12)
where the scalar p ot ential is

(13)
• The Yukawa couplinl between the scalars and fermions takes the form

' h , /, i . L .......ie.R
"'4:: ..
,

+ h .,9-,i.. d.",.l+} UI.. ' R £i}

Oi

(14)
tt, e, + h ..
" - i ..
rt .LtPid'o.R
26) + Hermitian Conjugate.
The coefficients h!.> h:. and h~. are G 1'';0'; arbitrary complex Yuk.awa coupling
constan's and £i} is the antLaymmetric symbol with £12 :: 1.
Finally there are the SU(3) and SU(2) topologie&1 terms

, _ 8 1 0 ' F- I'~ ..
..., - 1611"2 r l'." "
+ 8, 1611"2
1 F iF-/A·;
/An } (15)

The dual field strength ten.&On are defined by

( 16)

F,.~ · i _ !£I'~"F .i (17)

J - 2 '"
where £I'~'. La the totally antisymmelric symbol with ,0123 :: 1. The angles 8 and
8' multiplying 'he topologica.l. terms u e ar.bitrary real paramders ol the theory.
 I •
By tbe a~d.~
I• .
n"
... Ib the pO~ ... ; -.-"- -QJ&p-Kibb
. .._.

22

3) Gall&1C invariance under the gaqe crouP S(U(2)xU(3».

4) Matter field. belong to the itteducible reptceentatioDa of Table 3.1.
The conltrainLl imposed by tbe above assumptiolll are ao powerful tha
LacaDpan density C is forced to bave a rather simple form . A. a colLIequenee
standard model Laganpan density C. exhibit. IICVUaJ other lymmeLries whic
Dot put into the theory u • ,riori principles. The.e Iymmetriel arise becau
ia limply no\ allowed to be complicated enou,h to violate them. We will die
these 'derived' lymmeLries rollowin, OUf ,metal principle of going up the qua
ItairC~. Thua we ,hall first contider the low !CDetO part of tbe atandard mod
which the weal: int.eradionl &leao weak - due to the relatively high masl, Eq.
ofLhe weak intermediate vector bosms ,..,.. and ZO - that they can be neglecte
other wordl we be&in by discUllU1& jUit the ,\fOOl and electromagnetic intera

3.2. Symmetries of tbe Strong and Electromagnetic Interactions

In order lo write down tbe reduced La&ran&ian density c'ST+£1tl for tbe.t:ron«
e1e<.&rol'J\a&nelie inter ad iou, we rnu.t identify the eleetromapetic field A,. an
vec:lor bOllOll. field Z,. . The mueleu pbolon and the ma.aive ZO particle ar
ei&enstates of the neutral vector boeon m&8I matrix, obtained by .ubstitutinS
Riga field vacuum expectattoo value, Eq. (18) into £3 . It follow. that

!A ... = 'in9w.!.2.A."II +cos9w.!.A ...

e 9' 91
",d
1 1 1
c0I8 w -Z... = c0I8 w -2A,d l - sin8 w -A ...
h "
"here the rotation an&le 9w iI known as the Weinbert; u&le. The
'1 de,,,,,,,...
, couplinl; COIllt.ant e (the mapitude of the electron charle) is related lo the
IJ ud U(l} PUle couplinS coutant.l by the equatiolll
;"
P<
e = f,.in8 w ='1 ca.9w .
The !nUl of the ZO particle iI pven by
mw
m~ = .:::: go GeV
coo w
(37)
f« the experimental value of ain' 8w = 0.23. The e1ecLro'f2lii)
"hich enters L.ST+£M ia .oven by d ..d
... &.a 4.
F ..." = B... A" - B"A,
(38)
 23

(39)
(40)

These are all the fermion fields which enter £ST+BM . We do not include the
neutrino in CST+BM because it does noL interact. The gr&viton is omitted from
I

the standard model Lagrangian density C on the same grounds. It follows tbat
CST+EM describes a vect.orlike tbeory. in which the len.-handed and right-handed
components of each fermion fieJd have the same g&use quantum numbers. The
fermion mass terms are therefore gauge invariant.
The Lagrangian density CST+BM becomes

r 1 .... ....1''' 1 I:' fJ FI'll co

-2 2"""0 "
,
Io.ST+BM=-42,rI'"r
"
~ "(8"
+ tll.1 "flO -
'A 1'12 ,
I -
.2..1'
'3 ,,110
') Up"

+ id .1"({J,.6o lJ - iAI'Q" + iiAI'6ofJ)d"a

+ te.,."(8,. + iAl')e"
+ mOl"
v.".,o"
uo• + m.IIl7d °d•..1 + m.e.e. 0-

1 I"0 1'''0'F-"" IJ co ·
+6 1611'2 (4 1)

There is of course an understood summation over the thrice repeated generation

index a on the ma.ss terms . We note that, due to the electric charge quantisation
rule ofEq. (2) , the strong plus electromagnetic gauge group is U(3) . The Lagrangian
density LST+£M is equivalent to the most general renormalisable U(3) gauge theory
of quarks and charged leptons [paper 8J.
One inunediately sees that LST+EM is invariant under independent ~:lobal phaae
transformations for each fermion flavour. For example, LST+BM is invarlant under
the transformation
(42)
for all values of the 5U(3) colour index 0' and fixed generation number index 0, all
other fields being unchanged. It follows that all the fermion flavour charges

(43)

(44)

and
(45)

are conserved, where there is a sum over the repeated colour index 0' but not oYer
the generation index Q . These 'flavour charges' are simply the number of quarks
24

or leptona of a given flavour minus the number of corresponding antipart.icles. For

example
s=-rr, (46)

LI' = Qi (47)

B = 2: ~(Q: + Q~) (48)

•
Q= DjQ: - ~Q~-Q:) (49)
•
where S = strangeness, LI' = muon lepton number, B = baryon number and Q
= electric charge . Since we are dealing with a vectorlike theory. these flavour
conservatioD laws cannot be spoiled by anomalies.T
The basic reason for the flavour conserva'Km laws is the absence of fundamental
scalar fields in the theory of IIItroog and electromagnetic mteradions. It follows that
there are only two types of allowed terms in £ST+BM : kinetic (i.e. gauge covariant
derivative terms) and mass terma. The two fermion fields in either term combine
1.0 form the trivial represent.ation of the gauge €toup. One can then perform a
simultaneous diagonaliaation of the two m&Ulces in ftavour apace associated with
the fermion kinetic and mass terma (paper 8). Had there been a scalar neutral
colourless field in the theory, there would have been a third matrix in 8avour space
corresponding to its Yubwa couplings to the fermion s. It would not, in general,
be p088ible ~o dia&onaliee the three matric.a .imult&neously and all tbe flavour
conservation laws would no1. be ob1.a.ined. If t.he scalar field instead had non-trivial
gauge quantum numbers, other combinat.ions of fermion field representations would
be allowed in the Lagran~an . For tbis reason, in the full standard model, the
Kobayuhi-Masbwa matrix K .. of Eq. (21) is not equal to the unit matrix and
t.he W+ gauge bOlOn causes transit.ioll8 between particles wi1.h different. generation
indices.
It is also easily shown 1.ba& CST+EIII is invariant under charge conjulation :

(~O)

Here ,,(x, t) is a generic symbol (or all t.he fields and ~C(x. t) denote t.be charge
conjugation transformed fields . The fermK»n fields are taken to be Grauman (anti-
commuting) variables , as in the path integral formalism. We list. below the charge
i
conjugate lauge fields and Q::; quark fields (as a typical fermion field)

. AC "o--,,~
~ _ A 0, ,- -A,
AC - (~l)

FC~-F.0 (~2)
,,"0 - - ""~ • r P" ::; - ' " ' '

FC ,,"0 ~-- - t.,,"1'.. , r"" ::; -i"" (53)

1J C ... __ CUT
1J .. . , . ....
U"" - u T C (&I)
 25

Bere T den0t6 the tranapoee and C is a " x " Dirac matrix utisf'ying

C-r~C-l = -11l'1', C- 1 = -C. (55)

The kinetic and topological terms for the gauge fields in CST+EM are clearly
invariant under the above charge conjugation transformation . Let us now explicitly
verify that. the fermion kinetic and mass terms are also invariant. We consider first
tbe kinetic term for the Q = t quarks.

~ ' (0 < .
u .1 vl'0a - I....1'''• - 1'.23..."Cl<a ' ) 'UfJ. (56)

C uT o.C..,"(8,..s; + iA,.,o + ijA,.6;)CiiT fJ . (57)

. 2 ... <.)-T,
= 'UT ..4'JT(!IoUp"'"
co
+ I"... ,./J • + 13' "'UfJ U " (58)

= .".
-U.1
,(7
v ,.0"
,0
+ I' ...I'll 0 <.) U aa
.2... ,.UfJ
+ '3 (59)

= ti!,"(8,.6; - iA pp" - i~.A,.6;)u(J• . (60)

The minus sign in Eq. (59) arises from commutin,; the order of the GraaemaD
variables ~ and U OG _ The arrow on the derivative in Eq. (59) indicates that. it
is actin, ~ t.he len.. The derivative acts to the right &8 usual in Eq. (60). which
follows from Eq. (59) modulo an irrelevant pure derivative term. Bence the kinetic
term for the Q = j quarks is charge conjugation invariant, as are thOR for the
Q= -1 quarks and charged lepLona.
The tn&8S term for the Q = j quarks transforms &5 follows

(61)

(62)
(63)

where we have used C'2 = -1 in Eq. (62) and the Grassman variable nature of u",.
in Eq. (63). The fermion mau terms are thw: seen to be invariant under charge
conjugation . So charge conjugation is a symmetry of the strong and electromagnetic
Laganpan density L.$T+BM.
In the a.baence pf the topological Lerm, the vectorlike natu~ of the theory of
strong and electromagnetic interactions would aULomaticaily enau~ tbat t. ST +BM
is invariant under parity

£ST+E,,(~P(X , t )) = £ST+EM(¢(-X,t )). (64)

The parity transformed fields t/J p(x , t) are defined as followa:

A:",' = _ (_l)f!Ap",,!t(_x, t) .A: = -(-l)·!.A,.(-x, t) (65)

26

F:"r.' = (-l)'!+':F,...o-'(-x,t) F:" = (-l)I!+I!F,. .. (-x. ,) (66)

F:",/ = -(-1)4!+': pp ,o'(-x,t) (67)

po
uP o. = ,.°Uo.( -X, t) u .. = U'" .. ( -X, tho. (68)
The topological term changes aign under parity

F.un .5 (x, t )il''',,"(x, t)

_ -F,..../( -X , t)F"" ,Q( -x, t) (69)
p

and violates parity synuneLry, Eq. (64) .

Phenomenologic.ally, the coefficient IJ of the topological Lerm must be very small
or zero. The experimental limit on the possible value of the neutron electric dipole
moment implies- that 181~ lO-·. Here we shall therefore neglect the topological
term by settinr; IJ = O. However a true derivation of parity P (and time revenlal T)
invariance for strong and electromagnetic interactions must. explain t.he smallness of
8. This requires us to go beyond the standard model and we return to the problem
later in this section, when we discu.. Paper 9.
Neglecting the topological Lerm, it is easy to verify the vaJidity of Eq. (64). We
illust.rate t.b. result., by explicitly pvint; t.he parit.y transformat.ion of t.he Q =I
quark kinetic enert;)' t.erm:

u<" .(x, thl'[81'6! - iAl'coP(x, f) - i~AI'(x, t)6~Jup. (x, t ) (70)

_ - (-l)':UO.( -x, thD1'1'

P

x [8.6! - ;A•• '(-x,') - ;~A.(-x, ')6!h·u,.(-x, ,) (71)

= U"".( -x, th"[81'.s~ - iAl'coP( -x,t) - i~AI'( -x, t).se]up.( -x, t). (72)
The anticommutation relationa oft.he Dirac l' matrices are uaed in passint; from Eq.
(71) to Eq. (72).
The vectorlike or non·chiral nat.ure of t.he fermion representat.iona under the
Itront; and eledromapdic t;aup t;rOup is, of eourae, crucial in t.he above derivation
of C and P invariance [paper 8] . So we may ask if i!Uld why this feature bas been put
into the st.andard model. Since the U(3) ,au,e goop is unbroken, the ...umpt.ion
that. all fermions have a non· zero mass requires t.he fermioDII to belon, to vedor·like
representations and have parit.y invariant pUle int.eraetioDII.II However t.here could
exist m&lJlllesa fenniona with parit.y non·invariant laule interactions. This theore~
ical pouibility can be ruled out, by requirint; suffieiently small repreaent.ations of
the U(3) t;aut;e goup and no pUle anomaly for the photon and t;luons. 10 In fact
it • sufficient to . .urne thM all fermions have elechie marIe lQl < 1 and beton,
 27

to tbe trivial L hiplet .3. or anti-triplet. r repraentation of eeloot SU(3). Then

just seven different irreducible representations of t.he gauge group U(3) are allowed .
These representations are (0,1) and three pain of mutually conjugate U(1)xSU(3)
representaLions (ll) and (-1..1), (j,.a) and (-j ,a.-) and finally the pair (-i .
.3,) and
<l.a.-). The (0,.1) representation corresponds to a particle which has no strong and
electromagnetic interaction and is therefore omitted from CST+EM -
There are three combinations of gauge particles which can ad all ext.ernallines
(or pot.entwly anomalous triangle diagrams, Fig. 3.1. Each of these three com-
binations gives a gauge anomaly proportional 10 a linear combination of the three
differeneel
y o

NI1.ll- N(-1 ,ll (73)

Nil.» - NI_I"') (74)

and
(75)
Here ND denotes the number ofleft..handed fermion species in the representation
D of the gauge group. It iI easily seen that the three equationa, l'eSuitin& from the
conditions of anomaly cancellation , enforce all three of the above differences to
be zero. This means that. the model is then vtctorlike and automatically parity
conaervin,.
Similar considerations can be applied to the full standard model ,au&e IfOUP
S(U(2)xU(3». The assumptiOlll of small representations, the absence of ,a\lle and
mixed anomalies and m.&sI protection principles &lain lead to the experimentally
observed repreaentation patt.ern of quaru and leptons. Mall protedion Dleana: t.hat
fermion repneentauona which can combine to form S(U(2)xU(3» Kau&e invariant
mass terllll will, in fact, form very heavy fermion statel. The mass seale associated
with these statea is expected to be rjven by the characteristic energo scale of .orne
new interaction, beyond the standard modeL These fermtons , which are allowed to
have masses without appealin, to the standard model Bigs mechanism, should thus
be heavy compared to the W± and ZO boson masses and unobservable at presently
available enerrjes.
Since the theory of dron, and electromagnetic intetactions involves only ,auKe
bosoDII and fermiona , it baa CP symmetry (still negledin, any topolorjcal term) .
28

The above 'explanalion' or the veetorlike nature cI. the fennion repreeentatiolUll in
l,ST+BM , therefore, implies C invariance .. well .. P invuianee.
Time reveraal. T , ayromdry of CST+EIII (with 8 = 0) now follows as a conae-
quenec of the CPT theorem [paper 23] to be discussed in Chapter V. It is however
instructive to vuify T invarianee expliciUy. Time revenal is an anti-linear operator
and the condition for invariance is

CST+EM (~T(X,t» = Cs,.+E., (¢(X, -t» (76)

where C· denotes the complex conjugate of the functional form C. In otber words,
the complex conjugation applies to the coefficients of the field combinationa occur-
ring in C but. not to tbe fielda themtelvea. The time reversal tr&ru:fonned field.
lJT(x,t) are defined as followlI:

A,T = -(-1)" 'A,(x, -t) (77)

F;"o' = -(- l)':+'~FI' ..o'(x,- t)

F;" = -(-l)':+':FI',,(x, -t) (78)

(79)

ijTo. = UO .(x, -r)T. (80)

In the discu.ion of time revenal invarianu the superscript T is used to denote the
time reversal transformed fields and not tbe transpoet. In Eq. (SO), T denotes a
4x4 Dirac malrix latiaCyin&
T=7*=r 1 • (81)
h ia now straip.tCorward to .ee that C.ST+BM fulfills Eq. (76), provided we
nepect the topological term, which chanss -1&0 under time reversal

F,. .../(x, 1)1'''·, O(x, t)

T -F,. .. o'(x , -t)F"",CI(X, -I) . (82)

A&ain we explicitly verify the invariance oC the Q = i quark kinetic term

IUO" .(x, 'h" [8,..s! - iA,../ (x, t) - i~A,.(x, t)6!1u,.(x, I) (83)

_ - IUO' .(x, -t )T,."·( -1)':

T

x [8,..s! - iA,../(x, - t) - i~A,.(x, - t).s!)Tu,.(x, -t ) (84)

 29

= tu<" <I(X, -t)'y,ll [8p cS! - iA,IIQ'(x, -t) - iiA,II(x, -t)6!]u'<I(x, -t) . (85)
Tbua we have derived the t.hree discrete symmetries C, P and T for the st.ronl
and eledromagnetic sector of the standard model. Strictly speaking, however, Lhe
'derivations' of P and T invariance are not valid, in t.he absence of an explanation
for the vanishing of the coefficient 8 of the topological term. Such an explanation
requires us to somewhat extend the standard model [paper 9].
AI. mentioned in our discussion of t.he diagonalisation of the quark mass matrices
M" and Md, chira! transformations of the quark fields Eq. (24)-(25) , induce a
change in the value of the coefficient 8. In particular, an axial phase rotation of aU
the quark fields
(86)
under which right banded quark fields are multipled by e ler and len handed ones by
e- iQ , leads to the transformat.ions

(87)

(88)
(89)
Here NF denotes the number of quark flavours under considelation. The need for
the ehange in the value of 9 arises from the axial anomaly 7 in QeD. Due to the
anomaly, the functional integral measure for the quark fields is not invariant under
Eq. (86), but undergoes the transformation'

(90)
whele
(9\)
The combination
J = 9 + arg[deL M" . det Md] (92)
is invariant under all chiral transformations. It is therefore possible to rotate away
the original topological term in CST+£M, obtaining instead an overall phase e"/N,
for the quark mass matrices. So all tbe P and T (or CP) symmetry breaking can
be arranged to come in via the quark mass matrices . ConsequenUy, if the masses
of the relevant quarks &re small, the effect of tbe oripnal 9·telm is suppressed. In
the case of a massless quark, say the up quark u , tbe effect of the topological telm
can be entirely transformed away, by an axial phase transformation

(93)
on the massless quark alone . This transformation just changes 6 to 9 + 20 , without
any change in the mass matrix if m.. = O. The ZelO mass implies chiral invari--
ance under the transformation Eq. (93) , except for the anomaly. Thus 8 can be
30

transfonned into any value , without changins the physica of the field theory. ]t
follows that P, T &nd CP are symmetries of CST+EM ' The problem of explaining
the vanishing of 8 is then replaced by explaining why m.. = O. Phenomenologically
it. seems unlikely t.hat any quark m&88 is zero [Paper 10), although mu = 0 cannot
be entirely ruled out .11
More qUMtitativeiy the effect of the 6-term can be calculated , treating the mass
term for the NF = 3 lightest. quarks lA , d and 8 as & perturbation. The 6-term is
replaced , to lowest order in e, by the CP violating m&S8 term

(94)

The contribution of Ccp to the electric dipole moment of the neut.ron can then be
calculated ,s wiinS current. algebra techniques and standard quark masses (paper 10)
to be Dn ~ 5 x 1O-lfie em. Experimentally it is known that ID.. I < 10-2<1 em and
hence 161 < to- 51 •
We have seen that chiral phase invariance is broken in the standard model,
explicitly by the quark masses, as well as by the axial anomaly. Consequently the
topological term gives P and CP violations . Peccei and Quinn [pape r 9) t.herefore
suggested embedding the sLandard model in a larger theory possessing a chiral
U(I) invariance, which is broken only by the anomaly. This global chiral U(I)
invariance is int.roduced into t.he Higgs sedor of the theory, by the addit.ion of an
extra weak isodoublet. scalar field . The U(I) symmetry is spont.aneously broken
when the t.wo Higgs fields develop vacuum expecLation values . The relat.ive phase
of the vacuum expectat.ion values is determined dynamically, by t.he anomaly and
strong interact.ion effects (instantons) . It. turns out. t.hat. instanton activit.y is very
likely to al ign the vacuum, in j ust such a direction that this relative phase precisely
cancels out the effect of the 8 term. In other words, the eft'ective parameter 8
of Eq. (92) becomes zero, in agreement with the P and CP symmetry of strong
interactions.
In fact it can be readily shown that t.he vacuum energy is invariant under the
sign change 8 - -J (Le. under parity) . The effective potential V,(¢) t.herefore haa
an extremum at J = 0 (and also at "I = r ). Dynamical calculat.ions suggest that.
'8 = 0 corresponds to a true minimum oft.he vacuum energy densit.y. So the vacuum
adjusts itself to t.ake precisely that. value of t.he '1 parameter, which ensures P and
C P invariance for the strong and electromagnetic interact.ions . These symmetries
are thereby explained.
The spont&lleous breakdown of the Peccei-Quinn chiral U(l) invariance leads to
a Nambu-Goldstone boson . Explicit symmet.ry breaking by instantons generat.es &
smaU mass for this particle, called the axion. l l Experimentally a particle with the
properties of this axion does not seem to exist . 13 However it is possible to make
t.he axion in pract.ice invisible, by furt.her extending the model. It is necessary to
introduce a third Higgs field , with a non-zero Peccei-Quinn U( l ) charge and trivial
under the standard group, which develops a very large vacuum expectation value. 1..
 31

Of coune all theee models, invoi.villl an extra chiralaymmetry, reaDy take WI oulaide
the .t.andard model.
We now tum our attention to lOme approximate symmetries of the It.rong in·
teraction sector of the standard model (papers 8 and 10]. These approximate sym-
metries arise due to the existence, for some tf!a&OD, of & set of NT quark flavours
with exceptionally small ma_ee,

(95)

Here AQcn refers to the charaderistic strong interaction mass scale of a few hundred
MeV. The quark mass m, refers to the mass parameter appearing in the L&!I'angian
density and not to the constituent quark mUll (\lied in the 'naive' qUal'&; model),
which cannot be small compared to AqCD [paper 10]. See Fi&. 3.2.

--
Constituent

7
'50

-
con1ribution f,,-
CJ AQ.c.O.

, • •
Fi,. 3.2. Rdatiomhip between tho IXJIpatitueat quuk ~ of no~rdac.iml.ic: quark IIIIOdel. aDd
the current ~bra m. oco:w'I'iAt; in the QeD Lap-anp.an. The ciiffen:r:tc:e b e t _ them i.
due to the momentum BIad"·';""· ariaiq hom CCIIlfiD_t and ;. p:&CIJKi'1'ica.al to Aqcn . Tb,e
numben indicate the c:urrent. aJ,seh.-a ma.v in MeV.

In a first approximat.ion, we can negled the masaes of the N F quarks satisfying

Eq. (9S). There is t.hen a symmet.ry under global unit.ary t.ransformations on t.he
mlWless left-handed quark Weyl fields , belonging to t.he same representation of the
colour gauge goup (i.e. t.he ~ and ,a.- s~ates respectively of Table 3.1 , restricted
to the massless subspace). Alt.ernatively, we can consider t.he NF lell.- and right.-
handed quark colour t.riplet,.a.. fields. For NF:;; 3 massless 8avours u , d and., we
have chira! invanance under 3 x 3 unitary transformations U£ and UR

(96)
32

In leneral , for Np m ...less quark flavoura, the chiral invariance goup b.

comes II priori. U(Np)LXU(Np)R . Usin, the Lie algebra decomposition U(N) ~
U(l)xSU(N), we immediately recognise that t.he diagonal (vector) U(l)v subgroup
corresponds to baryon number COll5efvat toD , Eq. (48). However , the antisymmetric
(axial) combination of the two U(l) subgroups has an anomaly, as diBeuaaed above,
Eq. (90) . The divergence of the axial baryon number curren\ is thus non-vanishin&

,. = 2NF 161..2 F1'1<0 'F~1''' ~ ..

v"',' (97)

and the axial U(OA subgroup is not a t.rue symmetry group.

The remaining chiral symmetry group is SU(Np)LXSU(NF)R. Again we can
readily recognise the diagonal (vector) s ubgroup SU(Np)v , coDsi8t.in& of t.he ele-
ments (g ,g) t SU{NF)£XSU{Np)R where 9 e SU(NF} . It. is the Gell-Mann flavour
SU(Np) synundryu , under which hadroDs are daseified into multipleta. The Car-
tan subalgebra couesponds to conserved quark flavour chargee, Eq. (43) and (44) .
The antisymmeLric combination or axialvecior subgroup SU(Nr)A is known, phe-
nomenologically, to be realised by spontaneoul synunetry breakdown . There are
good dynamical reasonsHl for believing that the QeD vacuum does, indeed, break
the global chiral symmetry SU(Nr)LX SU(Nr)Jf down to its diagonal vector sub-
group SU(Nr)y in tbis way.
Tbe successes of son pion theorems and current algebra 7 ,17 show tbat, in na-
Lure , there is an almost exact NF = 2 (u and d quark) flavour chital .ymmetry.
The Nambu-GoldsLone b08Ons, corresponding to the spontaneously broken SU(2)A
symmetry, are tbe pions 11'+ ....0 and ",-. The non-zero pion masses arise £rom the
explicit chiral syrrunetry breaking quark mass terms, m..uu and mllad, in the La-
grangian density. The low value of the pion masa, compared to tbat of a typical
hadron , is evidence t hat mIl, mol < Aqcn. I.eospin or SU(2)y symmetry is obtained
in the same approximation.
We s hould stress here that isospin symmetry is explained by the u and d quaru
both being ligbt, without requiring equality of the masses m.. and mol . In fact it i8
estimated , from mcaon masses [paper 10j , that
m,
- = 1.8. (98)
m.
Therefore it is clear tbat the accuracy of iso&pin symmetry is not due to the de-
generacy of mil and mol. It is, rather, the rna.ss difference m" - mIl which matters,
and it is small because both mol and mIl are small (approximately 7 MeV and 4
MeV respectively) . So iaospin symmeLry should nol be conaidered as a fundamen-
~ eymmetry of the etrong interactiona, brokell only by electtome.petism, It is aD
approximate symmetry, which follows dynamically from t.he existence of two quarks
satisfying Eq. (95) . Similarly the 'eightfold way ,l. of flavour SU(3)y is a eyrnmetry
of tbe strong interactions, accurate to 10 or 20%, in spite of the 8 quark being about
twenty times heavler than the d quark .
 J3

A complete derivation of chira! symmetry would, of (aurae, explain why at least

two quark flavoun are much lighter than AQCD . In the standard model quark
masses arise, via the Higgs mechanisms, from the Yukawa coupling constanLs h:.
and h:•. as in Eq. (22). These coupling const.anta are pure numbers and, a priori,
ODe would have expected them to be of order unity, in the absence of any reason
for them to be otherwise.
In principle the QeD mass scale parameter AQCD cannot be appreciably larger
than the W-J: or ZO boeon masses, since QeD would then act. as a technicolour
interaction 11 , generating W.J: and ZO masses of order AQCD . In practice AQCD is
much smaller than the W-Z mass Bcale. So it is not possible, with the order of uniLy
coupling constant philosophy, to obtain a good isoepin symmetry. In fact all the
quark and lepton masses would be of the same order of magnitude as the W::l: and
zO masses. This is just one facet. of the fermion mass problem. It follows that the
order of unity coupling philosophy must. be totally wrong or further explanations,
beyond the standard model, must be postulated .
A natural way to expillin t.he observed fermion mass hierarchy is to postulate t.he
existence of a new set of approximately conserved chiral quantum numbers [paper
11). These chiral chargee are supposed to be different for the various left.- and right-
handed quark and lepton Weyl fields . The quantum number differences between
the various left- and right-handed Weyl states lead t.o approximate selection rules
for t.he corresponding Yulcawa coupling const.ants. The fermion Yukawa coupling
constants are thereby suppressed , by factors determined by the required amounts of
chiral charge symmetry breaking. The corresponding fermion mass matrix elements
are therefore suppressed , relative to the W-Z mass scale, by different orders of
magnitude.
We have considered [paper 11] a very general model of this type, postulating a
random system of approximate selection rules. The existence of light quarks and
leptona and t.he observed hierarcbical structure of the fermion mass spectrum are
naturally reproduced . FUrthermore tbe qualitative features of the quark mixing
matrix elements K.~ of Eq. (26) an also reproduced 2o, if the s tandard model is
further unified by, for example, introducing right-handed currents and leptoquark
currents.
Another approach to the mass problem is to replace the Higgs field by a new
Yang-Mills interact.ion , t.edmicolour 19 , analogous to QeD. The spontaneous break-
down of the global chiral tech nisymmetry gives rise to the W* and ZO masses. At
first. sight, in such technicolour schemes, we seem to have too good a solution to
the problem of the existence of light fermions . All the quarks and leptons have
zero mass! It is possible to avoid tbis result, by introducing a further 'extended
technicolour' gauge interaction . However it is clear that a realistic model of this
type must be rather complicated. Nonetheless it might be relatively easy to obtain
some quarlr:s much lighter than other mass scales, such as the W-Z mass scale and
AQCD .
Again we see that. a full understanding of some strong interaction symmetries, in
this cue iaoepin and thiralsymmetry, I'1!quires WI to 10 beyond the standard model.
Let us now look at which symmetries and conservation laws are valid for the
whole standard model, and not just its low enert;)' limit .

3.3. Symmetrics of the FuU Standard Model

In the previous sectton we neglected weak interaction processes since, at low energy.
their amplitudes are of order a/mlv , where a = e'2/4y; is the fine structure constant.
Similarly weak radiative corrections, which vtolate symmetrtes of CST +EM .uch ae
strangene58 and parity, are automatically of order a/mlv rathe r than of order Q
(Paper 8}. In this section we discu!l8 the symmetries which remain, when weak
interaction effects are included.
The Raveur charges of Eqs. (43)-(45) are no longer generally conserved. How-
ever, baryon number B is conserved and we can define a conserved lepton Dumber
L. Cor eACh generation .
It is easy to see that the tetal number of quarks minus antiquarks (i .e. 38) is still
conserved , due to exact colour invariance. The covariant derivative D ...-{.J' and the
colourless Higgs field ~i must therefore connect quark fields of compensating colour
representations. It follows that the quark fields can only occur in combinations
of the type ¢ ... t/I, where t/I denotes a generic quark colour triplet field . Thus all
terms in the standard model Lagangian density C create and annihilate equally
many quarks. A priori this gives baryon number conservation , but the baryon
number current has an anomaly/ due to its interaction with the W± and ZO bOSOM.
However the amplitude for the occurrence of instantens in the SU(2) gauge field,
which could produce baryon numb er violation, is exponentially suppressed by the
factorll
exp(-8... '/g~):::::: exp(-IBO) . (99)
The predicted baryon number violation , at normal temperatures and energies, is
therefore unobul'vable .
Similarly it is easily seen that there is a lepton conservation law for eatb fermion
generation . In the standard model, neutrinos are all masaless and we are free La
define the electron neutrino v •• say, as that linear combination of neutrino species
which couples, via the W: boson , to the electron. The W± coupling then, by defi-
nition , conserves electron number - the number of electrons and electron neutrinos
minus the number of corresponding antiparticles. As we shall discuss below, the
couplings of the ZO boson to fermions do not cause flavour changing transitions, due
to the GIM mechanism [Paper 13}. It follows that there are separate conservation
laws for elect ron number Leo muon number L" and tau lept.on number L, (for N =
3 generations) . Again, there is an anomaly in each of t.he lepton number currents.
Due to the enormous suppression fact.or, Eq. (99), for SU(2) gauge field in.tantoDS,
however, there is in practice no violation of any of the lepton number conservation
laws.
It is e ven possible to construct anomaly free conservation laws, by taking dif·
ferences in which the anomalies'21 cancel out . In fact the lepton number differ-
 35

ences L,. - L. and Lr - L. are exactly conserved . FUrthermore B - L, where

L = L. + L" + Lr is the total lepton number is a conserved quantity, even with
I

anomalies taken into account.

U is interesting to note that. the quantity B -L is still conserved, when the stan-
dard model is extended to include arbitrary SU(3)xSU(2)xU(1) invariant , baryon
number violating, four-fermion interactions [paper 12]. These four-fermion interac-
tions, such as
(100)
are, of course, supposed to be an effective, non-renormalisable, low energy param-
eterisation of new physics, which occurs at a large mass scale mx > row. In
general, the coupling for an effective interaction of ma88 dimension d will be of
order (mx),t- d. The four-fermion operators have d=6 . Other baryon number vi-
olating effective interactions have higher dimensions and are suppressed by higher
powers of ..!",
In order to classify the four-fermion operators, it is useful to introduce the
concept of F parity.22 The F parity of a field is defined to be

(101)

where t is it.a weak isospin and 20 is t.he number of lefi.-handed spinor indices in the
Lorentz group representation (o , b) of the field . Due to Lorentz invariance and SU(2)
gauge invariance, F parity is multiplicatively conserved. F parity takes the values:
F=+l for quarks and leptons; F = -1 for antiquarks , antileptons, gauge fields , the
Higgs field and the differential operator 81" Baryon number violating fou r-fermion
operators must contain three quark (or antiquark) fields, coupled together to from
a colour singlet, and a lepton (or antilepton) field . It then follows, from F parity
conservation , that t.he allowed terms either annihilate three quarks and one lepton
or three antiquarks and one antilepton . Thus the 'charge' B - L is conserved by
the allowed d=6 operators .
It must be admitted, however, that there are aHowed d=5 terms , involving the
Biggs boson field fi' which violate lepton number L wit.hout changing the baryon
number B :
j'ei.L Ij'L tPl< tP/€il< €;I , (102)
1\.LljaLtPl<tP/€/;€I</ ' (103)
These terms clearly violate B - L , although they do not contribute to proton decay.
If no elementary Higgs field is available, as in technicolour models, then both tP's
in Eqs . (102) and (103) should be replaced by fermion bilinear expressions . The
dimension of these terms is thereby increased from d = 5 to d = 9. In such a
t.echnicolour-like model B - L would in fact be conserved including all effect.ive
interactions up to and including dimension d = 6. However in the pure standard
model extended to include all effective interactions up to and including dimension
d = 6, B - L is only conserved for certain processes, such as proton decay.
36

We now CODiider another c:oue~icm I•• , .hich • only valid for a certain
special type of procesa : flavour conservation for neutral currents [paper 13]. A. we
have already seen, the fermion flavour cbarges, Eqs. (43}-(45). are conserved by the
.troog and electromagnetic interactions. However strangeness and other fl.avours
are not conserved in charged current weak interactions, Eq. (28) , mediated by the
W:i: boeon. The absence, experimentally, of strangeness-changing neutral current
weak interactions was, for several yean, considered to be evidence against the ex-
istence of the neutral weak intermediate vedar boson ZO . It was then pointed out
by Glashaw, Iliopolous and Maiani (GIMf23 that the suppression of strangeRes&-
manling neutral eurrenta could be explained by the introduction of a fourth quark
flavour: ch&rm. More generally, flavour conservation for neutral currents is a con-
sequence of the fermion generations all having the sart'M!: representation , Table 3.1,
under the standard group and the existence of precisely one neutral Higgs boson in
the standard model (Paper 13].
The zO couplings cause tr&Mitions between fermions of the SArt'M!: electric charge,
belonging to the same weak isodoublet . By lUing the doublets of Eq. (24), (u , delL
i
etc. , &8 a basis for the Q = quarks (and the doublet. of Eq. 25, (ue , d) etc ., as a
buis for the Q = -1 quarks) , it is eaaily seen that the ZOb080n couples each quark
to itself only. Similarly the ZO couplings are diagonal in lepton flavour . The YukaW&
couplings of the Hig. particle are proportional to the fermion masa matrix and ,
thus, also diagonal in tbe m&88 eigenstate basis. As noted earlier, the W::t couplings
conserve the lepton number (or each generation . Bence all the intera.ctions respect
the separate lepton number coDlervation laws.
However the W::t couplings to quarks are not. diagonal in generation number
space due to the non-vanishing off-diagonal elert'M!:nts K.. of the Kobayashi-Maskawa
matrix. Quark generation number is conserved by the neutral currents, but not by
the charged currents. In addition to quark Havour conservation in second order
(single b080n exchange) neutral weak interactk>ns, the GIM mechanism suppresses
higher order induced neutral cunent effecu. in which a quark undergoes It. flavour-
changing but electric charge conserving tr&Mition . All higher order flavour-changing
neutral current intera.c:tions vanish, in the limit that all the Q = I quark muses
are equal and all the Q = -1 quark masses are equal . In pra.c:tice this limit is
u.sed to mean the approximatk>n, in which all relevant quark masses are negligible
compared to mw .
The fact that all ftavour-changing neutral current interactions vanish, in the limit
of totally degenerate quark masses, is easily seen, by analogy to the separate lep-
ton number conservation laws. In this degenerate situation , the ' Cabibbo-l'Otated'
quarks. de etc. and U e etc . are all mass eigenstates. Let lU now consider a single
quark line, correspondin~ to a Q = i quark in the initial state, passing through
a Feynman diagram, Fig . 3.3. We choose the doublets of Eq. (24) , (U ,de)L etc .,
to define a quark generation number ba&is in this mass degenerate situation. The
W::t couplings to the quark line then conserve generation number, as do all the
other standard model interactions. A similar result applies to a Q = -1 quark in
 37

F"". 3.3. The int~ of. mI.' Q='J/3 quark in an arbitrary neutral current proceu.

t.he initial ,tate. in a generation number basis defined by t.he doublets of Eq. (25).
(Uc,d)L etc. Since quuk generat.ion numbers are unchanged, it. follows t.hat quark
flavour is conserved in neutral current pr0cesse8 to all orders of perturbation t.heary.
When real non-degenerate quark masses are used, the actual degree of sup-
pression of any hi&her order flavour-changing neutral current process depen& on
the convergence properti~ of the Feynman diapams under consideration. A truly
rourtb-order weak process, in which two heavy b080DS are exchanged , suffers a power
law suppression?' 80 that t.he amplitude is of order

g~.6m2 G' A , (104)

.. - F~m .
mw
Here 6m 2 is an appropriate combination of quark mass squared differences . How-
ever if the process is second-order weak and second-crder electromagnetic, there is
a very mild GIM suppression factor and the amplitude is of order :M

(105)

The amplitude can be further suppressed by small Kobayashi-Maskawa mixing ma-

trix elements.
The full standard model, including weak interactions, is not invariant under any
of the discrete operations C, P or T . However T and CP would be symmetries, if
there had only been N = 1 or 2 generations [Paper H) (again setting the coefficient
, of the topological term equal to zero).
Af\er diagonaiisation of the fermion mass matrices , the CP violating Yukawa
eouplinp to the Higs particle disappear. Then, spart from the 8-term, the only
C P violating terms occur in the couplings of the charged current 1.0 the W.z. b08Ons:

(106)

The physical quark fields Uta_' and d 2a6 L are defined in Eqs. (24) and (25) . In
discussing the CP transforma.tion properties of the charged current, one mWlt allow
for relative phases between these quark fields . All other terma Me diagonal in the
quark 8avoura.
38

The fields hanslorm as follows under C P:

. I· i
A... I(t,x) ..... (-1) .A,,; (t,-x) (107)

111 ... £( t, x) - 'h. 'Y°CuTla aL(t, - x) (108)

,,10- et(t, x) _ "Ii. uT l ... L(t, -x)e,.o (109)
o - T2
d2,..t(t, x) - '1n1 Cd '"'L(t , -xl (llO)
d'"2a .LCt,x) ..... ,,;.JT20U(t , -X)C1°. (lll)
Here '11_ and '126 are phase radora and T denotes the t.ranspo«: . h then follows ,
using EQ. (55) and the anti-commutation properties of the Grauman fields "loe£
and d206 £ , that the first term in the Eq. (106) transforme 1M follow. under CP:

(112)

For CP invariance of the charsed current couplings, we require that the transformed
exprcs8ion of Eq. (112) should equal the &«ODd term of Eq. (106). The condition
for C P invariance thus becomee

(ll3)

or, in other words, the matrix

(ll.)

must be real.
The weak mixing matrix K•• is unitary and thuB haa N2 real degrees of freedom ,
where N is t.he number of generation•. For CP invariance, we require K~. to belon~
to the N(N - 1)/2 dimensional manifold of ortho~onal matrices. Therefore, in
~eneral , we need N' - N(N - 1)/2 independent adjuatable parameters, in order
to ensure CP symmetry. There are 2N phase (actors, '11. and ",., which can be
adjusted. However a common phaae factor for all the quark field. would leave all the
parameters '1i.'12' unchan~ed . Hence there are 2N -1 independent phases '1i.'1»
available. Thus, in order to aa\,e(y the ,enera! eondition for CP symmetry. we
require
, I
N - :iN(N - I) < 2N - I. (llS)
This condition is equivalent to
I
:i(N-I)(N-2)<0 (ll6)

which is only satisfied for N=l or 2. For N=3 ~enerations we have (N - l)(N-
2)/2 = 1 and obtain just one nontrivial CP-violatin~ phase factor in the Kobayaahl-
Maauwa matrix [paper 14J.
 39

In ihe IeCODd order (ODe boeon e:xehanp) proeeu invoIvini OI1ly enema! quark
Savou,. belonpnl t.o, eay, t.he first. two leneratiol18, thee can be no CP violat.ion.
This ronows, since the relevant quark couplings only involve a 2 x 2 sub-mat.rix
of Oat: full mixing matrix K... A CP·vioiatin, prOCe8$ mw:t either (i) be fourt.b
order in LIle weak interactions or (il) involve external quarks from three different
Je,Derat.ions.
The only place where CP violation has been observed is in the neutral bon
syat.em. The coefficient of admixture t of the CP- even state K1 • relative to the
CP-odd state K" in the long-lived eigenstate KL is given approximately by 25

t = .2..- . 1m MI2 elf (ll7)

2.;2 Re MI2

He..
o -0
M" = (K IHeffi K ) ( liS)
is the KO- KO transition matrix element of the effective fourth order weak HarnilLl>
nian densit.y Rd' changing strangeness by two units. The CP violating amplitude
1m MI2 neeesaarily involves couplings to a t.hird generation quark and is, therefore,
naturally lIuppressed relative to the CP conservin& amplitude R.e Mn. This sup-
pression is due to Lhe observed smallness of the quark mixing matrix elements K.,. ,
connecting the two lowest mass &enerations to the third gener&t.ion, compared to the
'Cabibbo' matrix element Kn.21S The empirically small value of the CP violation
parameter
1<1'" 2.3 x 10-' (1l9)
is thereby understood, without requiring the CP violatin& phase in the mixin&
matrix to be small (or dose to r).
Finally we eonsider scaling symmetry in the standard model. We use natural
unita Ii = c = 1, and measure dimensions in units of mass. By analogy with the
discussion in Chapter II of scaling symmetries in macroscopic physics, we CAn apply
dimensional analysis t o the action, S = J tJ4zC , of II. Lagrangian field theory.
The action for a field theory with no dimensional parameter (Le. IT\aSSes and
non·dimensionl_ couplin, constants aU set equal to lero) i. invariant under a .we
transCormation 27
'Z~ _ Z", -_ ..., z,. , (120)
where the transformed fields are given by

.'(%) = ~-'¢(~-'%} (12 1)

~'(%) = ~-!~(~-'%} . (122)

Here ,p{z) and ¢(z) denote generic boson and fermion fields respectively. The scale
invariance of the action is broken by particle masses. These masses should be
negligible at very high enere or very small distances, where we focmally obtain
40

uymptotie Beale invariance. However the scale symmetry of & masalMl theory is
broken by a quantum anomaly,7,27 simillU to the axial U(l) anomaly which breaks
t.he chiral invariance of I1l&S8less fermions .
The divergence of the scale current 51" for a renormalis&ble field theory. is given
by the trace of t.he cnero-moment.um t.enso! 6,.~ :

/JI's,. = 9~ . ( 123)

Formally t.he trace 9: is given by m&!lll terms, hut. another anomalous cont.ribu-
tion arises from the renormalisation of the theory. The renormalisation procedure
introduces a momentum scale, r =_p'J, at. which the renormalised dimension-
less coupling constant g(p) is defined (for simplicity we consider a theory with
a single coupling constant). The renormalisation scale p is an arbitrary parame-
ter, which breaks scale invariance and introduce8 an anomaly into the trace of the
tner&y-rnomentum un&Or. This trace is proport.ional to the renormalisat.ion group
p_function :2S
dg
P(g); "dp (124)

Scale invariance is thus restored if the physical couplin, constant corresponds to a

zero of the .8-func~ion . Howenr, if g(P) is non-zero at such a fixed point iJ=O, the
fields will in general have anomalous scaling dimensioDs,17 d 'F 1 for bC8Ol1B and
I
d'F for fermions .
Non-Abelian gauge ~heories have the property that, for small values of g(,.,), the
,8-function is pven by
(12.)

with a negative value of b. Hence it is consist.ent to &Illume that g(,.,) _ 0 and

,8(g) - 0 as IJ ...... 00. So we aped naive Kale invariance, Eq. (121)-(122), to be re-
stored asymptotically at large momenta. This phenomenon is known &I asympLotic
freed om.l,3,:»
The only dimensional parameter in the standard model L&«rangian density C is
the tachyonic mass &58Ociat.ed with the Higgs field . It seta the Kale of the vacuum
expectation value of the Hi,uJ, field , Eq. (19) , and the masses of the Wi:. and ZO
bosoDS. These masses are too large to be neglected at presently available energies.
unlike tbe quark masses which a.re suppressed by the ill-understood small values or
the Yukawa coupling constant.., The electromagnetic interaction is Abelian and not
asymptotically free . We therefore expect asymptotic scale invanance to be valid
only for the strons: interaction sector (QCD), at energies where the relevant quark
masses can be neglected . For IJ > AQCD, the QCD scale parameter , tbe strong
interaction coupling constant hu the form

(126)
 41

provided the number of quark flavours NT is leu than , . Asymptotic sealinJ

behaviour should set in for energies and momenta large compared to AqCD and the
relevant. quark masses .
Hadrons &rt extended objects on a scale of order AQhD and, t.hus, moment.a
of order AQco are relevant. even for high enere hadron beaJlll!l. It is therefore
difficult 1.0 test sealing symmetry in purely hadronic processes. Experimentally,
sealing symmetry is studied in deep inelastic lepton-nucleon scattering processes,
which involve one photon or weak vector boson exchange. The ctOSlH>ections are
de8Cl'ibed in terms of dimensionless st.ructure functiona l ,s, 1S F;(q2 . X) . Here Ill' is the
4-moment.um transfer carried by the virtual pbeLon or intermediate vector boson
exchanged, PII is the 4-momentum of the nucleon and

X=--. .'
2p · q
(127)

Invariance under the canonical scale transformations of Eqs. (121)-(122) would

imply that the strudure fundions Fj(q2 ,X) depend only upon the dimensionlesa
variable X and not on q2 . This is the famous Bjorken scalin~ behaviour (for q2
and p . q large compared to the relevant quark masses). The asymptotic se&ling
symmetry of QeD does not give exact Bjorken scaling. However the predided
scalin~ violations are only logarithmic ; the moments

M;"(q') = I.' dX X"F; (q' ,X) (128)

of the structure functions are not constants, but definite powers of logarithms. I,.,
The predided approximate Bjorken sealint;, with logarithmic violations, appears to
be in agreement with aperiment.2!I
It is interestint; to note that seale invananee implies conformal invarianee for
a renormalisable field theory.28 Conformal transformations are a four parameur
manifold of .pac&time transformations of \he form

(129)

These trans{ormationa can be writLen as an inversion

zl' - z'I' = 2 .' (130)

•
{ollowed by a translation ,
z 'I' _ Z lll' = % I' + d' ( 131)
and then {ollowed by another inversion . When combined witb Lorentz transforma,..
tions, t ranslations and scale trans{ormations, they form a 15 dimensional Lie group,
called the conformal goup .
42

Poincare

STANDARD
MODEL "'..... ---- .....,
,",,:-...r -- ~I FloYOlJf conservaTion \
\ for neulral currents :

----- -
, GIM ,/
"

Slrono and
elecTromagnetiC
interoctions
Sma ll

C
,,"
PT
CP
GO'
T

Fir;. 3.4. The auumed and derived .yrrun.e~rie& 01 the .t&ndanI model.

For a renormalisable field theory, t.he divergence of the four conformal currents
K~I', associated with the transformations of Eq. (129) , CAll be expreMed in terms
of the trace of the energy-momentum tensor
(132)
Apart from mass terms , Ot is given by the haee anomaly discU86ed above. Thus
the conditions for asymptotic scale invariance and for asymptotic conformal invari-
 43

anee are the same in a renonn.aJisable field theory, namely the vanishinS of the
P-function .28
It has also been pointed ou130 that, for an effective fi eld Lheory action which
is invariant under general co-ordinate transformations, local scale invariance (We)'l
invariance) implies conformal invariance in flat space.
We have seen that a rather large number of symmetries arise naturally out of the
standard model. Poincare invarianee and gauge invariance are the only fundamental
input symmehies of t.he model. The symmetries derived from the model, Logetber
with the auxiliary assumptions, are summarised in Fig. 3.4 .

RelereUCe8
1. T .-P. Chen( and L.-F . Li, G•• ,~ TIe .. ..., I;If Slem ... ,..,., P.rtid. Pl,.ie. (Oxford UniYel'Sily
PreM, I984).
'2 . K. Huanl, Q ...,..h. Lc,fou ••• a..,. FieU. (W(II"ld Scientific:, 1982).
S. C. Qui". C ...,. n.oriu oj 51"'''1, Wed u,; Efednma"uti. 1111.,...1;0., (Benjaminl
Cumming, 1983).
t. L. Michel, Gro., Tle~nl;,.. 1 c~.cc,~, ••111 Mell GIII, i. E'emn~ • .., Parlicle Pl,.., . , Le' ~.n'
GJ tle [" .... , S.mmer S,hGI GJ T4tGnliul PIt,,;u U6t, ed. F . Guney, p . W (GOI"don
and Breach, 1964).
5. L. 0' Raireartai,h, Cro., Slrach" GJ Cn,e TlteGriu (C ambridlle Uniyer-.ity P reM, 1986) .
6. K . Fujikawa, Pit, •. Ru. LeU. !!. 1195 (1979).
T. S. B. Ttoeiman, R. Jac:k.iw, B. Zumino and E. Witten, C.rr"e.1 AI,el ..... 1II A. Gm.l1u (World
Scienlific, 1985).
8. V. Baluni, PIt, • . Ru. DUI, 2227 (1979); R. Crewlher, P . di Veahia., G. Venma.o and E . Wit--
ten, PIt". Lett. 1GB, ll3 ( 19T9).
a. Ken·lchI Hlkua, PA,,;e. Scri,.a ~ lQf (l98G).
10. S Chadha and H. B. Nielsen (unpubli ..hed).
lL D. B. Kaplan. and A. V. Manahal-, Pl., •. Ru. Lell. §.l. 2004 (1986) .
1l. S. Weinbers, PIt, •. Ru. Lett tJ.. 223 (1978); F . Wilczek, Pl., •. Rn. Lell. 40, :lI9. (1978).
l.3.. S. Yamada, Pro"di." GJ J9UI.lCn&aIiG.at S'm'G,i .... G" L c,tG" a"d PhIG" 1"le ... eliG'"
at Hi,' E.er,iu, CornelI, p . S25 (Comell Uni.vtni~)', 1983); W. A. Bvdeen, R. D. Peccei and
T . Yanar;ida, Nacl. Pl." . B2T9, 4m. (1967).
It. M . Dine, W. FiKhler and M. Sredni.ck:i, Pit, •. uti. 104B, 1B9 (1981).
a. F . E. Cla.e, Aa lalrod.dio- tel Q .. r« .. III ParlG .. (Aeademic Pre&I, 1979).
1.6. C. Vala and E. Witten, N.d. Pl., •. BU", 113 (1934).
11. S. L. Adler and R. F . Duhen, C....ru. AI,cb., .. III A"I;eali n .~ Parl.ielc PIt,,;u (Benjamin,
1968); V. de AHara, S . FUbini., G . Furlan. and C. Raueui, C.nub'. HalliN. P",i" (North-
Holland., 1973) .
l8. M. Gell.Mann and Y . Ne'eman, Tlte £.,ltl/./.I Wa, (Benjamin, 1964)
l!l.. E. Fathl and R. JadOw, Dr"amical an,c S,mmd.., Bredi", (WOI"ld Scientific, 1982).
2Il. A. ConIcie, C. D . FrogaU and H. B. No.:J..en, Pl., •. LeU. 1618, 3fi (198$) .
21. G. 't 800ft, Pl., •. Rn. Left. ~ 8 (1976); V. Ku.unin, V. Rubakoy and M. Shapc.hni.koy,
Ph».. LeU. USB, 36 (1985)-. -
22. S. Wein-her" Pl,•. Rn. D22, 169-4 (1980) .
2l..
u.. M. K. G.illvd and B. W. Lee, P.,•.
S. Gluhow, J . Diopoloua and L. Maian..i, Pl., •. R~ •. D2 , 1285 (1970).
Rn. DI0, 897 (1974); M. K. G.iUud, B. W . Lee and
R. E. ScN-od:, Pit,•. R n. DU, 2674 (1976).
:u.. J. E.Ili., M . K. Gail1anl and D. V. NanopoiOWl, N.d.
and M. B. Wise, Pit, •. R n. D20, 239'2 (1979).
P.,•. 8109, 113 (1916); f . J . Gilman

lti. F. J. Gilman, K. Kleinknecht and B . Renk , Pvtide Dat.a Group, Pit, •. LeU. 2048, lIU. (1988).
:11.. S. Coleman, Aq",," oj S,mmd.., (C amhrid,., Un.ivenity Preu, 1985).
~ B. Scbroet, Naowo Dim e"!,, L eU. 2, 867 (19n) ; N. K. Nieben, N.d. P., • . BeT, 527 (1975);
N.d. Pl, •. B120, i l l (1971).
29. P . SOd... .-l O . Wolf, A . .. Ru. N.cl. P.rl.. Sci. st, 231 (1.1); F . ScnaJIi , PY"Kcdt.,••/
1985 1",1,", .li~.al S,....".i.m,II.Lqoln •• 1/ Plol~'" 1",lcradi,•• d Hi,. £.e"iu, Kyoto.,
p . 8, 1986
30. B. Zumino, Led.re. i. Eltmt.le.,., p.,.,it!e Plt."iu •• 11 Q ••• I.m Fit/II T1eorJ, P~ulli."
of 1170 B,..IIl/eu S.m",,, 1.,~i~,ft, Vol. 2, p . 4U (MIT Preu, 1970).
 4S

Chapter IV

BEYOND THE STANDARD MODEL

4.1. Grand Unification

The standard model , to&ether with Einstein'. tbeory of lI'avity, seelJlB to explain
all the fundamental interactions of pbys.ics known today. However there art several
reasotll 1 to look for theories that go beyond the standard model, in the sense of
havinl: t.he latter u a IOW-eOB«Y approximation. The st&nda.rd three generation
S(U(2)xU(3» model involves 20 arbitrary parameters; three independent gauge
coupling constanta, six quark maases, four quark mixin, paramete1't, three charged
lepton messes, lhe W.ID&S8. the Higgs particle mass and the coefficients fJ and 9' of
the 5U(3) and 5U(2) topological terma. Many of these paramete.ra are unnaturally'
small. The smaJ1ness of the quark masses are responaible for the chiral symmeuy
and the Gell-Mann 5U(3) flavour symmet.ry of t.he strong int.eradiona . Similarly
parity and CPinvarianee require t.he B·parameter 10 be effectively zero. We therefore
need a model for the parameLen, if we are to underat.and these symmetries . We
Ihould also like to understand the oripn of the puge symmetry goup S(U(2)x U(3»
it.aeIL These symmetry and other 1 consideration. strongly suggest the existence of
a more complete theory containing the sL&ndard model.
An attractive method of e.xtending the sLandard model ia 10 postulate the
unificationS,. of the strong, weak and eledromapletic inLetadions in a lIimple gauge
goup G. The simplest such gand unified theory (GUT) is the Georgi-Glaahow
G = SU(5) model [papu Hi]. Twelve new gauge boeona are introduced by the
SU(5) group, which mediate quark-lep1on transitions and must be made super-
heavy. The global baryon and lepLon number conaervation laws of the IItandard
model are thereby violated. In fact , pro1on decay and the existence of superheavy
mlllDetic monopolea are generic features of GUTs.
The grand unified gauge group G ill arranged to be spontaneously broken, at a
auperheavy ma&II scale mx I by the Higgs mechanism, down to the standard model
group. In fact precisely the standard model group S(U(2)x U(3» , and not the group
SU(3)xSU(2)xU(\),;. .sub""u. ofSU(5). The <buS" quantioa.ion rule

Q= -~ "";aIi'y· (mod 1) (1)

iAI thus automatically obeyed in the SU(5) model. Each generation of len-handed
fennions fits exactly into the anomaly-free reducible 5- +10 representation of SU(5).
However it is an exaggeration to claim that tbe standard model group has been
derived within such a GUT model. In reality the standard model synunetry is built
into t.he gauge group G 06 initio. So, strictly speaking, GUT models do not qualify
for inclusion in this book as examples of symmetry derivationa.
Nevertheless, some GUT models do have symmetries which are not. II priori
obvious and are not. imposed on the model from t.he outset. The moat important
example of this type is the eonaervation of B - L (baryon number minus lepton
number) in tbe simplest. version of the SU(5) Lbeary, despite t.he separate viola-
tion of B and L. In its simplest version, 5U(5) is broken down to the standard
S(U(2)xU(3» group by an adjoin~ representat~n of Higgs, 24H , developing asu-
permassiw: vacuum expectaLion value. The standard group is then broken by a
fund&mental representation of Bigs. 5H. which is just the SUeS) extension of the
familiar SU(2) x U( 1) Higgs doublet of the standard model. The Lagrangian for this
minima] SU(5) model posaessea a global U(l) symmetryl , with the fonowing U(l)
quantum number (called X) aaaignmenLe.

(2)

Here tbe subscript F refers to lef\...handed Weyl fermion fields . The remaining fields
namely the 24H Higgs field and the gauge fields, are assigned zero X-charge. The
SU(2)xU(1) breaking vacuum expedation value of tbe 5H Higgs field also breaks
the X -charge spontaneously. However the linear combination

4 Y
Z=X+-·- (3)
• 2 '
wbere Y denotes the weak hypercbarge leaves the 58 vacuum expectation value
invariant. The Z-eharge is therefore conserved globally and the appearance of a
Goldstone boson is avoided .s H turns out tbat for fermiona

Z=B-L . (4)

We must treat Z as a generalised B - L charge, in the sellBe that some bosons cauy
this quo.n\um number althougb neitber B nor L quantum numbers are assigned to
them.
Exact B - L conservation appears more or less acddenlally in the minimal SU(5)
model and fails if additional Higgs fields , sudt as a l5H or ION ' are introduced. For
instance, a 15H Higgs field would generate Majorana mass ter ms for t he neutrinoe,
 47

obvioualy violating L wbile B it unehanpd. Erlendiq: 5U(5) to the SO(lO) GUT,

Z becomee a generator of the xauge p'oup . ~ In order to avoid a massless gau&e boson
coupled to the Z charge, i\ must be spontaneously broken and B - L conservation
it no longer eud in the 50(10) model.
Apart from the perfect fit of quarks and leptons into SU(5) representations, the
prediction of 1.he Weinber, angle'

2
lin Bw = 0.228 ± 0.0044 (5)

is the main experimental evidence in favour of grand unification . In t.beones like

5U(5), where the compleLe spectrum ocrermions is already contained in the standard
model, the Weinherg angle is given by

sin 2 Bw = 2g~ , = !
(at the unification scale) (6)
gl + g2 8
at 8uper· higb energies, where the breaking of the grand unified group G can be
neglect.ed (paper 16). Also theories based. on larger groups, 8Ueb as SO(10) or
Ee,' in which the standard model group is embedded in an 5U(6) subgroup. pvc
the lame prediction, Eq. (6), (or ain 2 8 w at the unification aeale. Renormalisation
JOOup techniques can be used to calculate the dependence of the sause couplinS
constanu 9i(p) , and hence sin' 9w(p ), on the momentum. scale p at which they
are defined (Pllpet 16]. Below t.he unification mass scale mx, the couplinp, 9i(P)
vazy in a G non· invariant way. For the minimal SU(5) model, or any simple GUT
satisfyinS Eq. (6) with no new physics in t.he 'greal desert.' between p = mw and
p = mx, the corrected value of the Weinbers ansle at p = mw agrees with the
experiment.al value, Eq. (5), within errors . (Note adt/ed in proo/: Recent LEP
precuion meuuremenb o( ain' Sw are inconaisl.ent with the minimal SU(5) value.)
Assuming t.he &reat desert hypothesis of no new phyaia between p = mw and
p = mx , the unification mass scale is predicted to be'

mx ::: 1.3AQCD x IOU . (7)

He1:e AQco denotes the effective (our flavour QeD scale parameter (in the modified
minimal subt.raction scheme of Ref. 7) . Super heavy sallie bosoIUI of ma.ss mx
mediate proton decay at a rate proportional to mi\ leading to a predicted partial
lifetime for the decay p _ e+'II'° ot<")

(8)

nu. decay rale is c.a1culated ua;umins that all the fermion mOOns matrices, which
appear in hazyon number violating proa ma, are equal to the usual Kobayashi-
Mubwa matrix, as predicted in minimal SU(5) . The factor 10:l:o.t il an indication
olOae "heoretieal unurtainties in the ealeulation. We therefore obtain the folio_inC
theoretical boundss on prot.on decay, using present experimentallimiLs on AQCD

(9)

and
1'(P -- e+rO) < 1.4 X 10'2 yr for AQCD < 400 MeV . (\0)
These results are in contradiction with the experimental bounds

T(p __ e+rO) > 3.3 x 10'2 yr (at 90% C.L.) . (11)

The simple SU(S) model is thus ruled out by the experimental bound on proton
decay. It is neeeaaary to introduce new physiea at a maaa scale below mx, in the
form of scalars, ferrmons and/or gauge boson. , in order to obtain a GUT with

mx > 5 X 1014 GeV , (12)

compatible with the limits on proton decay. This suggests tbe existence of a larger
gauge group, such as tbe superstring motivated Ed X Sa symmetry [Papers 21, 22],
which is broken down in steps as one goes down tbe quantum staireaae. Tbe stan-
dard model group then consists of "he symmetries which survive near the bottom of
tbe s\aircase, and can be considered to arise fr om a sort of "survival of the fitteat"
mechanism.1
An appealing modification of the standard model , with new physiea below the
unification m&88 seale mx , involves the introduction of simple N 1
supersymmetry.4,lO Supersymmetry relates bosons and fermions and their interac-
tions. In the supersymme\rk version of the standard model, the spin 1 gauge bOSOM
are a.c:companied by new spin! giuinos , photinos etc., spin! fennions are accom-
panied by new spin zero squarks and sleptoOll, and spin zero Bius particles are
a.c:companied by new spin i Hiwinoe . Superaynunetric theorie8 have remarkable
renormalisation properties and in lOme cases are completely finite in perturbation
"0
theory. These properties can be exploited make the maaa hierarchy, mx »mw,
technically natural in supersynunetric pand unification , if global N = 1 superaym-
metry is broken at an enerlY around 1 Te V. In addition the &rand unification mue:
scale mx is increased, as required for consistency between experiment and proton
decay mediated by superheavy boson exchange.
The commutator of two suecessive supusymmetry transformations is equivalent
to a spac~time translation. Thus, if the supersymmetry transformations are made
local and gauged, the spin 2 graviton gauge field is automatically ineluded, as a
partner to the spin i gravitino gauge field associated with local supersymmetry
transformations. Local superaymme\ry therefore includes Einsteinian ,;ravity and
is ulually called lupergravity. Gravity can be unified with other particle interactions
in N-e.xtended superpavity, which contains N spinoria1supersymmetry geneI&tors
and N gravitinoe. The large.t of these modela i. N = 8 supergravity [paper 18). For
 49

N > 8, the model necessarily cootains massless particles with spin &reater than two,
and the corresponding conserved currents are so restrictive on scattering procC:IIIses
that the S-matrix must be trivial. 11
The N = 8 luput;ravity theory is truly unified , in the sense that all fundamental
fields of the theory appear in a single lupermultiplet. However despite the improved
I

convergence due to supersymmetry. the theory is not renormalis&ble and divergences

very likely appear at three or perhaps eight loops.12 It has a manifest 50(8) &lebal
symmetry. The lupermultiplet contains 28 vector bosoDs and , in fact , it is p08Bible
to gauge the 50(8) group. This alternative N = 8 gauged 8upelptlvity 13 haa a new
independent gauge coupling constant 9, in addition to the p'tlvitational constant
,11;(= .j'bGN in four dimensions, where GN is Newton's constant) .
The 50(8) group does not contain the minimal ~and unified sroup 5U(5) as a
subgroup. Consequently 50(8) does not contain candidate states for all the observed
elementary particles of the standard model. However , when propezly interpret.ed,
N :; 8 supergavity is found to have a hidden local 5U(8) symmetry (paper 18].
This raises the pOll8ibility that. the 5U(8) somehow sauges itself dynamically and
ada as a realistic srand unified &rOup . The N :; 8 supergavity model must then
be treated as a preon theory, in which all t.he spin 1, spin! and spin 0 particles of
the standard model are composite. Unfortunately attempts to obtain a succesaful
phenomenoioc based on this idea have been beset with difficulties. 14 Here we are
interested in e.xplainins the orisin of the local 5U(8) !lOup tosether with a Slobal
exceptional E7 grouP. as bidden syrrunetries of N :; 8 superpavity.

4.2. Hidden Local Symmetry and Dynamical Gauge B<nOns in

Non· Linear Sigma Models
A.. we shall see later, the scalar and paeudOllcalar particles of N :; 8 supersravity
form a non· linear si&ma model field theory, in which the field takes values on the
homoseneous space correspondioS to the set of caJets £7(+7)/5U(8) . In order to
set acquainted with such coset·valued field theories, we sball first consider how such
models are revealed to be sause theories.
Let us consider the seneral sipna model consistinS of a set. of spin zero fields ,
which take values on a COllet space G / H rather than in number sets as for usual
fields . By the cc.et space G/H , we understand the set

GIH = (gHI9 E G) (13)

of coset. gH = {ghlh E H), where G is a StOUp and H is a subslOup of G. We are

interested in connected Lie groups. In the case relevant for N = 8 supergavity, the
group G = £7(+1) is in fact a non·cornpad Lie Sroup. It is the non·compact version
of the exceptional group.&t , in which +7:; 70 - 63 is the sipature of G equal to
the number of non-<ampad generators minus the number of compact seneratora.
The correspondins group H ~ 5U(8) ts the maximal compact lubgroup of E7(+7) ,
havin& 63 seneratora.
50

AJ an explicit examp~ of our pneral diseu.ioD, we .hall refer to the simpler

Cl::p"-l model, which is renormali.sa.ble in two dimensions and has been studied in
detail usin& the lIn expansion [paper 17). The CORt apace for the Itpn-l model iI
obtained by taking
G = SU(.) (14)
ADd the subgroup
H = S(U(1) x U(. - 1))" U(n - 1) . (15)
Here 5(U(I)xU(n - 1» denote. tbe ~up of n )( n unitary matricea with unit
determinant of the block form
U(1) ; 0 0
---- ,
-'- - - -- ----- --
o (16)
U(. - 1)
o
having a U(I) and a U(n - 1) matrix alan, the diagonal.
The coset. space GIH Cor the CCp.. - l model is, in fad, the set of directiona, i.e.
rays, in the space C' of n complex dimensions. h is euily seen that the group
G = SU(n) acta transitively on the set of rays

(17)

obtained by letting z be any ordered set z = (%1, %, ... . I z.. ) of n complex numbers
not all being lero, z i O. This means that, for any pair of rays [%) and [yJ, tbere
exiata an element g of G IUch that

.[z) = Ill), • e SU(.) . (18)

Here t.he operation of 9 on {x] is defined to be matrix muitiplicat.k>D, treatilll Z . .

an n dimensional column vector. The 8ub,roup of G leavin, a ray {z] invariant
is isomorpbic to the goup H . For example, the ray correspondilll to the vector
z = (1,0, . . . ,0) is clearly len invariant by t.he n x n matrix: of Eq. (16) and we have

h((1.0, .. .. 0))=((1.0, ... ,0)), heH . (19)

In fact , there is a one-tc>one corr~PQndence between the ray. (z] and the coset.. gH
given by
[z)_ (. e GI.[1 . 0 .... . 0) = [zll . (20)
It follows from Eo. (19) that if. e (g e GI.(1,~, ... ,0) = [zll th,n.h e (. e
GI.[1.0 . ... . 0) = [zll and. thu., th. ." (. e G(g[1 .0.... •0) = (z)) "' of thdorm
of. rooet.H = (.h lh e H) .
It. it a remarkable fact t.hat, by a rat.her t.rivial rewriting of the t.heory, a sigma
model taking values in G I H is found to have a gauge symmetry for the group H .
 51

We may simply express lhe model in terms of & field theory taking ....Jues in the
group G itself, rat.her t.han in the coset. apace . However we must t.hen declare t.he
replacement. of a field g(z) E G by g(.r)h(z} w be a gauge transformation, where
h(z) is an arbitrary ',auge fundion' taking values in H .
The Lagrangian denaity for lIuch & sigma. model wiD, in general, contain deriv&-
lives of the coset-valued field ¢(z) = g( %)H . How should we translate tbis derivative
8,.¢(z) into the G-valued formulation? Mathelll4l.ically 8,.¢ is & laDlent vector to
the coeet space. This tangent vector space is in natural correspondence to a sub-
space of the tangent vector space of G , which as orthogonal to the t.&ngent vectors
along the coeet (viewed 88 a 8ub-manifold of the group manifold G) . Ortholonality
is defined here with respect to the Kinin, form derived metric for the Lie lWup
G. In the G-valuecl formulation , we therefore want to replace 8,.¢ by & derivative
D,.g(z) which is paranteed to be a t&DIent vector to G orthor;onal to the tan-
lent vectors alan, g(z)H. Formally this is achieved by introducing the covariant
derivative
D,,(z) = 8,,(z) + ',(z)A,(z) (21)
where A,,(z) is constructed from .(z) and 8,.9(Z), usiD, the projection operator PH
from the Lie al&ebra • of G onto the Lie al&ebra JI of H :

(22)

These equatiom are moet easily uDderstood by tbinking of them aa expreesed in

IIOme representation; abstractly Eq. (21) may be rewritten as a relationship between
Lie algebra elements:

,(Z)-1 D,g(z) = ,(z)-18,,(z) + 'A,(z) . (23)

We should now like to include AI'(z) as an auxiliary field, in aucb a way that
varYln& the ac"OO with reepeet to AI'(z) pvee u.s Eq. (22) . It is easily teen, Fi,. 4.1,
that we require A,.(z) to lie in tbe Lie a1J:ebra 11 and to m.inimiae the expression

(24)

Here the norm of a mat.rix M is defined by

IMI'=M+M . (25)

More abltractly, we minlmiae the Killing form for G of g(z)-18/1-9(% ) +iAI'(z) with
ii8elf.
We can immediately recopile tbe expression, Eq. (24), after minimisation , as
the kinetic enerQ term
(26)
foc the GIH valued sigma model. Hue cPa(er = 1,2, ... ,dim(GIH» is a set of
CCM>rdinatee enumerating the points on the sigma model tar,et apace GIH, and
 53

.br.
rIC. (,2. SdI_u
val~
d.iacr- for the A"ls)
.:alar field ..,(s).
lidd. The eoIKIliD. mdicaie propaaaton Cor the Lie

In order to perform ordinary perturbation theory for this interaction, we lineariae

the grou~vaJued field g(z) , by writin,

(33)

where ,.(z) is a Lie alt;ebra valued field . A perturbation expansion of the act.ion,
Eq. (28), 'hon P"'"

1
= 2/ Tr«8,,(.) + A,(.»')

•
- 2/ Tr(8,,(·)h(·),8',(.)])

• Tr(A'(')h('), 8',(.)]) + ....

- 2/ (34)

The existence of & term of t.he form t'n-(A.. (z)8P'Y(z» in the action means
that., in t;eneral, the A.. {z) and 1(%) fields mix. However we choose a ,.use in
which this mixin& vanishes. [n t.he Lorenb r;auge, lJI'A.. (z) :::;: 0, the intepal
f dl'zTr(AJt(z)lJI',.(z» becomes a pure bounda.ry term. In !.he -unitary gaule".
PH(-y(z» = 0, the term Tr(A,.(z)81',),{z)) eimply vanishea. The propelaLora for
A.. (z} and r(z) do not. then mix. The hare proPAIator for ,.(z) becomee tht of
& maaste.. scalar particle, while t.he bare prop&&aLor for the AI'(:) field ia -12/6:

and does not. have a pacticle pole.

There is a one loop eelf enere diqram, FiS. 4.2, for the A,. field, where the
vertex wrreeponds to t.he term -f.rTr(A,.h,a"1D in Eq. (34) . The computat.ion
of t.his Feynman diqram in moment.um space pves an expression oC t.he form

(35)
54

where ,..(p') turns out to be diver,ent in &eneral Now we reeognise (pI'POl _ p'6:>
as the usual gauge field kindic enerlY term written in momentum apace. So, the
one loop correction haa generated a kinetic energy (or AI' Md, to this order the A"
I

bilinear part of the effective Lagrangian density becomes

_ '(P' )'I\(F F")+ 2..'1\(,1') (36)

2 "l' 2/"
where
F",,, = 8,.A .. - 8.. A,. . (37)
Thus, in the PH(r(Z» = 0 lauge, quantum effects have lenerated a term corre-
sponding to a vector A,. particle of mass mA given by

(38)

In this sense, the dynamically leoerated A,II laule field is • priori Higsed.
We note tbat an n~loop FeynID&D diagram corresponds to a term of order 'An, in
a power series expa.n.aion in Planck'. constant h. Hence T(P') is of order h and lobus
mA is proportional to ,,-1/2. So, in the classical limit ,,- 0, the mass rnA diver,u
and t.here is really no A,. field particle. This is reasonable , since the A,. field is &
purely formal object initially.
We now consider whether the A,,(z) gauge pa.rticle. can obtain zero ma.as by
quantum 8uctuations of ~he g(z) field . Surpriain,ly enou&h, it is indeed possible
to generate an unbroken local laUle .ymmetry dynamically in this way. This is
iUustrat.ed by our example of the Cp,.-l model.
In the G::p,.-l model, the IallIe particle associated with the Abelian U(l) factor
&roup of H is shown to be tnaallea in d = 2 dimensions, usinl the lIn expansion
(paper 17]. A masalesa gauge field in two dimenaions haa no real physical pa.rtide
assoc.iated with it, .ince it has d - 2 = 0 transverse components. However it lives
rise to a linear confining Coulomb potential, V(r) ::::=: r . On the other hand , a
ma88ive vector field in two dimensions p,ves riat to a Yukawa potential of the form
V(r) ~ ,.e.xp(-mAr). Thus, even in d = 2 dimensions, there is some phy.ical
content in a gauge field beinl ffia&8Ie&8. n follows that the cp.. - 1 model provides an
e.xample of the reveree of the HiUl' effect, dynamically generating a gauge symmetry
[Pap., 19J.
It is a detailed dynamical question whether or not quantum fluduationa can
l enerat!! a phy.ical gaule .ymmetry, corresponding to the fOl'mal gallic &roup H
(or to some subgroup of H) . When an H gauge &roup .ynunetry is leneral.ed, it
appe&ra thaL the quantum ftuduations cause tbe probability distribution of the field
g(z) over G to be invariant, under the .ymmetry operation of rigbt multiplication
by elements of H, at. any point in the va.cuum. For instance, in tbe c:.aae wben
the fundament.&! model has a &Iobal G invarianee, the distribution of g(z) could be
completely evenly distributed over G, like tbe Bur measure (see Fi&. 4.3). When
 55

the probability dinrihution of g(z) over G in tbe vacuum is invariant under H,

there is no longer any spontaneous breakdown of the ,Iobal part of the H gauge
symmetry. Thus the formal gauge bosolU of the ,auge group H are not Higgsed. We
should then expect the presence of massless lauge particles, due to the dynamical
generation of a kinetic energy wm like Eq. (35) .

Fie- 4.3. ~ ol ". v.aNm apllRllClJy in • _11_ t.ipMo model. The licwe illub"ates
.ymboUall)' bow quNltum Buctuationa in ~ field ,(~) caD. be invariUlt under ri&ht mwtiplieati_
by element, 01 H, """" the ~ u-.npk of .." 51 ~et spfC't .

The other p08libility is that the probability distribution or g(z) over G in the
vacuum is not invariant under H transformat.ions, but rather is peaked somewhere
in G. Then we sbould expect the theory to remain in a Hiw-lilr.e phase with
massive , or perhaps non-existent, gauge particle..
ICC (or G/ H) is a non-compact. space, it seems unreasonable to expect the quan-
tum fluct.uations to spread tbe field g( z) over the whole non-compa.d apa.ce. In fa.d ,
a G-invarianl distribulion cannot be normalised in a non-compact space and, &0,
cannot be used u a probability distribution. There(ore we do nol expect lhe in-
verse Hig:I effect to work in a non-compact group space. Indeed, detailed analysis l5
aeem8 to confirm thal the phenomenon, o( dynamical generalion o( maaa1esa gauge
hoeoos, does not occur in non-compact sigma models.
The idea., behind lhe analysis of Ref. 15, is to consider the effect 01. a renor-
mali..tion group block-spinning on the vacuum values of the GI H valued sigma
model field ;(z). This block-spinning corresponds to integrating out the momenta
in an internl between, say, p and pIn , where JJ is the cut-otr energy acale. There
is, thereby, a transition 10 a more coacse-grained field formulalion , in which n" val-
ues of the field are blocked together. However I in order to take averages over the
non-linear si&ma model field ;(z), we must allow it to take values in a larger linear
'pace, conesponding to the convex closure of lhe GI H manifold. The model must
therefore be apresaed in tertn8 of a field 4t(z), which lakes n1ues aD the target
,pace of the coneaponding linear sigma model (i.e. tbe convex. closure of GIH).
At the re¢arisaiion scale Il of the model, 4t(z) is constrained to lie in GIH. This
is .ymbol&.ed in Fig. 4.4 by (a) a full-line circle for a compact target apace, and
(b) a full-line hyperbola for a non-compact target spate. A typical block of field
values is shown and the mean value is denoted by a star. The target. space of the
56

' I.-, - t'; I :

lot

Fi,. 4 .4. ill.ration of thoe dl'ed of ~-.pinniD.J the vacuum v.aJ.ws of dw! a>pu. model field
o)( r) . (_) FOf"" COttlpKt lipna~. w tared. apace become. amaI.Icr _d unall.er iD tM infra.
red. (b) For .. nOl>-co~ 'i&ma model, the 'VIet apace becom.lqer ia the inCr..nd..

blodc-spinned field values is symbolised by a dashed curve .

As sugested by Fig. 4.4 we find, on the average, that the mean vaJue of a block:
of n ~ field values lies closer to the ori.s;in, in the linearised spa.ce, for a compact.
G I H manifold. On the other hand, we lee that the ~an value is very likely 1.0
He furth er away from the oripn for a non-compact. GI H manifold. We therefore
have the possibility, in the compact group case, of obtaining a phase in the infra-red
corresponding lo tere vacuum expectation value of 41(%). Fo.. a noo-compact group
G, however, the vacuum expectation value of ~(z ) will not become zero. In this
laUer non-<:ompad case, we still have the Higgs mechanism a t work and tbe gauge
bOSOM remain massive. It is only in the compact group G case tbat the gauge
particles (or the croup H can become ma8llese.
The 70 spin zero fields of the N = 8 aupergavity model are described by a
=
non·compacL GjH sigma model, wjth G E7(+7) and H:::::I SU(8) . Since H is the
maximal compact subgroup of G, the theory bas no ghoats and the HamilLonian is
positive definite [Paper 18]. However, the discu88ion ofthe previous paragraphl sug·
geats tbat quantum flu ctuations wiD not dynamically generate an unbroken SUeS)
Yang· Mills interaction in the model. For 'his and other reasons,t4 it 8eeITl8 unlikely
tha, N = 8 supergravity can provide a realistic theory o( particle interactions.
The occurrence of hidden e;lobal Ert+T} and local SU(8) symmetries in N = S
supergravity remains a s uprising result. We now turn to a discussion of tbe origin of
these symmetries , which is the aspect of supergravity t heory relevant to tbia book.

4.3 . Hidden Symmetries in N = 8 Supergravity

T he N = 8 supergravity model [paper IS] baa an SUeS) gauge symmetry, which
arises formally in much t he same way as (or the Gj H valued non· linea: sigma model
described in the previous section . A more truly hidden symmetry is the invariance
of the equations of motion under the global non·compact exceptional uoup E 7(+7) ,
which has SU(S) jZl a8 a maximal compact subgroup. 1.0 fact the 70 scalar and
 57

pseudoscaJar fields of N = 8 superpavity can be shown [paper 18J to parameterise

the coset. space E7(+7)/(SU(8)/Z2)' The ma.in features of the above syrrunet.ries are
illustrated in [Paper 191, by a qualitative discussion of the local 50(7) and global
5L(7, It) subpoups.
The 70 spin zero fields of N = 8 super gravity caD be packed into a 56x56 matrix,
called V , in such a way that these physical fields are unchanged, if V is multiplied
on the left by a matrix belonging to the 56 dimensional representation of SU(8).
Multiplication on the right, by a matrix belonging to the fundamental 56 dimen·
sienal representation of E1(+7). transforms V into another matrix corresponding,
in general, to another set of spin zero fields. This transformation of the spin-zero
fields is non-linear.
The simplest. way of understanding the construction of the matrix V is to use
a classification oC the spin zero fie~ds , resulting from the dimensional reduction of
the N = 1 supergravity model in 11 dimensions [paper 18]. Dimensional reduction
involves compactifying seven of the dimensions to a point and assuming that the
fields do not depend on the seven extra cc>ordinates. The 11 dimensional N = 1
supergavity Lagranpan depends on these gauge fields: an elfbein e",A, a Rarita--
Schwinger field .pM and a rank 3 antisymmetric tensor field AMNP . The index M on
the elfbein denotes a 'curved' vector index, which transforms according to the rules
of general relativity under a reparameterisation of the 11 dimensional co-ordinates .
On the other hand, the index A is a 'flat ' tangent-space index, which transforms as
a vector under local 50(1,10) Lorentz transformations. The flat metric has the 11
dimensional Minkowski form rJAB = (+1 , -1 , -1, ... , -I) . In general, letters at the
beginning of the alphabets are used to denote flat indices, and the later characters
are used for curved indices. Since we intend to rewrite d = 11 supergravity under
a " + 7 split of co-ordinates and indicu, we let A, Q, a (respectively M ,p, m) run
from 0 to 10, a to 3 and 4. to 10 respectively.
The elfbein eM A describes a graviton with 9 x 10/ 2 - 1 = 44 massless spin
states. The Rarita--Schwinser field .pM satisfies a Majorana condition and describes
a gravitino with (11 - 3) x 16 = 128 spin states. Finally, the antisymmetric tensor
(Kalb.Ramond) field A MNP provides the extra 9 x 8 x 7/(3 x 2) = 84 boson spin
states, required for supersymmetry. By dimensional reduction, the above 256 states
become the 256 staLes of the CPT self·conjugate field representation of N = 8
supergavity in d = 4 dimensions: 2 forming a spin 2 graviton , 16 forming 8 spin
i sravitin06, 56 forming 28 spin 1 vector bosons, 112 forming 56 Majorana spin!
fermioos and the rest forming 70 spin zero particles.
The spin zero particles are made up of 35 scalars and 35 p&eudoecalars. The
scalar particles arise in two ways. There are 28 scalar stat.es described by the 7x7
internal metric tensor,

•
gmn=em • en rJd , m , n=4,5 .. . , 10 (39)
associated with the seven compactified dimensions. Here the internal flat metric
'1" = -6•• . The remaining seven scalar fields ¢m(m = 4,5 , ... 10) are obtained from
58

the eomponent.t A,...... (p ,v = 0, 1,2,3) cA the ant.ieymmet.rie I.ensor field AJINP.

These component. A,IIf'rn form 7 second order antis),mmetric tensor puge fielda in
the (our dimensional space. A simple analysis ,iven below shows that, by a duality
transformation , any second order antisymmetric gauge field B,." in four dimensi.ona
is equivalent to a ,pin zero field . This field is pseudOfK.alar if B,.., is an ordinary
teZUOf , and sc&l.ar if B,. .. is a. pseudotensor as A,. ..m is taken to be.

The gauge invariant field Itrengt.h

F....."... (z) = B(pA .... jm (40)

obeys the equation of motion

lJI'F,. ...... (z) = 0 (41)

in the (ree field approximation. Here and subsequently, we use square bracket.
[ I to denote the antisymmetrized sum over all permutations of bracketed indices,
divided by the number of these permutatiou. It followa t.hat the dual field atrengt.h

- , ",%=2"
F () 1 ""F,. ....... ()
1: (42)

obeys the equation

- -
8" F'm(z) - 8'F" ",(ot) = 0 . (43)
The Levi·Civit.a tensor e···· , in any dimension, with upper indices is defined to be
equal to +1 for even permutations of the natural order . From Eq. (43), nes1ectinl
all Lopo!olical effects, we see that. the dual field stren&th must be of the form

(44)

It then foll ows from the Bianchi identity

that 'm. behaves like a scalar field , obeyint;

D'.(z) = 0 (45)

in the free approxim&tion.

The pRudotmaor nature of AuN p in four dimensions implies that the compo--
nents Am.", describe 7 x 6 x 5/(3 x 2) = 35 paeudOlwar fields . We will rdum to
the paeudOlcalar fields , after tint showing that. the 35 scalar fields apan t.he coeet.
.pace 5L(8, 1R)/50(8).
It. is relatively easy to extend the 7x7 inktnal 'curved' met-ric tensor gij into
an 8x8 matrix Si'i ,(i', j' = 4, 5, . . . , 11), which describes all the 28 + 7 = 3S scalar
 59

Selda. Thia matrix S"J', with primed indicel tuMin, over 8 values each, is con-
.trueted (rom a field V·',I, which ia an extension of the aiebenbein e·, = ei· ' A.
indiea.ted here, the order of the upper and lower indices on the sieben-bein is irrel-

..,
evant. The upper index is '6at' on the siebenbein e·, and 'curved' on ita inverse
'
The extended siehenbeio, or acbtbein, Vo'i' haa unit determinant and i. con·
strueted from the siebenhein e·; and the 1.calar fields' .... as follows:

.' , _ e";
II .,_u... -1/8 ( 0 ... 0 (46)

where
t.. = (dete"Sl = -detgij . (47)
The extended metric tensor Si'J' is then constructed from V·';"

(4S)

where ~J = 9jl" and ".'" = -6.,,, . It follows that the rea.laynunetrie 8x8 matrix
St'P satisfies det s.'j' = 1, and has 8x9/2 - 1 = 35 degrees offreedom to represent
the 35 scalar fields .
It is not. difficult to see that the values of the S;'i' tensor are in one-tcKme
correspondence with 1-he cosd space SL(8, 1R}/SO(8). The achtbein v·' f' is an 8x8
matrix with det v·' i' = 1 and can, therefore, be considered to be an element of
the poup SL(8. JR.) of special linear maps of an 8 dimensional real vector space.
However v·' i ' and the transformed achtbein JtI" .,v·' i ' (with det R'"'., = 1, so that
the transformed achtbein haa unit determinant) correspond to the same tensor
~'j" when R'"' .. belong. to tbe group SO(8). Thus a scalar field configuration, or
equivalenUy a particular tensor Sf'i', corresponds to the set ofSL(8,1R) matrices

{v.'i' E SL (8, It)1·'·'

" i' V ; ' '1.'.' = Si'i ' )
= (ll"'(OW
.,' "Ill" ' _ E SOtS)} (49)

where ,,(O)i ' ;, is some representative achtbein, obeyin, v(o),' i'V(O~'l'""c'

= Sf'i' . We
immediately see that this set is a coed 9H, where 9 E SL(8, JR.) and H =SO(8).
It follows, from tbe ,eneral considerations of the previous seetion, that the scalar
field sector formally has an SO(8) ,au!:e symmetry. In fact , it alao has a hidden
SL(8, It) ,lobal symmetry. However we will postpone a discussion of the global
symmetry, until we have included the p&eudO$calar sector.
The 35 pseudoscalar fields are described by tbe components Am", (m , n,P =
4,5, ... ,10) of the antisymmetric pseudotensor field . By contradion with the
aiebenbein , we introduce pseudoscalar fields

(50)
60

with 'flat.' indices. Then we {annally add an extra indu, set equal to 11, by definiDl
'he totally antisymmetric field

(51)

wit.b primed indices taking 8 values (4,5, . . . , 11). Using the Levi-Civita symbol
with 8 indices, we can also define the dual field

(52)

"hich is z:~o if one of its indices is equal to 11. It is now pOl.8ibJe to pack t.he
pseud06calar fields into a 56x56 matrix

(53)

where we consider the fields A.,. ,c' ''' and • A·I.' c' ot' as 28x28 matrices with roWI
and column. labelled by non-ordered pairs, (a'lI) and (c'd") respectively, of unequal
indices. The matrix V_ can be recognised as an element of the E1 group in its
fundamental 56 dimensional representation (Appendix B of Paper 18).
We must now introduce a 56x56 E7 IJUltrix V+ to describe the scalar fielda ,
80 that we can eonstrud the 56)(56 matrix V =
V_ V+ referred to earlier, which
describes
, all 70 spin zero field.. The matrix V+ iI formed by ·.quar~· the a.chtbcin
v· i'
(54)

where
., t' ·, .,
v.'' = '1.'t'V ;oS'1 . (55)
The rowl of V+ are labelled by a non-ordered pair or unequal Sal indicea (a'b') -
upper indices for the first 28 and lower indices Cor the lut 28 rowl. The columna
are labelled, in an analoloUl way. by curwd indices (i',l) .
Under an 50(8) lau«e tranafonnation on the ac.htbein, we have

v
.' i' - ~
" ,
t'll " (56)

(57)

where
R.," = (R _1 ) " . ' = R' , .' . (58)
It Collowl that the correspondinl loc.al 50(8) transformation on V+ it pven by
 61

wh...
(59)

Under this 50(8) transformation on the 8at indices, V_ transforms &8 follows

(60)

Thus the matrix of spin zero fields

(61)

transforms in the S&Olt way as V+.

V_RV - (62)

The 50(8) transformation matrix R can actually be extended to an SU(8) (or

rather SU(B)/Z2) transformation matrix U. Since V+ and V are transformed by
left multiplication, we &l't really only interested here in the transformation of a

(;:,.,'.' )
56-column
(63)

obtained by selecting specific values for the curved indices (i' ,l) . In order to
consider the general form of the SUeS) transformation matrix U , it. is convenient to
pack the above 56-column into a complex 28-column

Z II'" . .. ,.,
+ IY
(64)
J2
The 50(8) group corresponding to the transformation matrix Ris not embedded
in SU(8) in the standard trivial way, but is related to the trivial embedding by one
of the outer autorn01'phisms of SO(8). In otber words, tbe components of the 28-
column in Eq. (64) do not transform simply &8 the antiaymmetric product ZA8 of
two fundamental 8 dimensional representatioIUI of SU(8). However, &8 dillcuased in
the Appendices of Paper 18, they can be carried into such a representation by the
linear transformation

+ i YI .,,) (r""') A8 .
zAS -_!4 (4.'"/2 (65)

am
(66)
is the antisymmetric product of two 8x81'5_ proje<:ted 1'-matrices , appropriate to a
Weyt spinor in 8 dimensions.
62

Calculation t.hen .hOWl t.bat. an infinitesimal SU(8) transformat.ion on ZAB tram-

latee into the (ol1owin, infinitesimal transformation on V

(67)

The 28 antisymmetric infinitesimal parameters A"I... = -A~'., descJibe the SO(8)

transformations. They are relMed to the parameters A~ B = -NSA, associated
with the SO(8) generators in the defining representation of SU(8), by the automor-
phism
(68)
We note that the metric 11.. " - :::; -6.,._ is used La lower indices and that the position
of the indices A , B on r e ' " is of no significance . The traceless set. of symmeLric
infinitesimal parameters A " A B :::; Alt SA . corresponding t.o the other 35 generators
of SUeS) in the defining 8 dimensional representation, is given by

(69)

Here r· ' · '~' ''' is the antisymmetric product

(70)

IlIld the symmetry property A" AS = A" SA hanslates into the anti-self-duality
condition
'<" _ ."",,"'c',t (71 )
""0"','. ' - - "" .
We now wish to establish the advertised resu)t that the sealar and paeud06Calar
fields parameterise the homogeneous space Er(+7)/(SU(8)/Z2), by showing:
I) The SU(8) transformations U acting on the set of matrices V obtained from the
spin zero fields,
UV = UV_V+ , (72)
generate all the 56x56 representation matrices of & ( +1) '
and
II) For no matrix V = V_ V+ does the coset {UV IU E SU(8)) contain &not.her V,
representing a different spin zero field configuration .
We shall also see below that the action for the spin zero fi eld sector is invariant
under global E7(+7) transformations, but this will not be relevant to our discussion
of local SU(8) invariance.
The group E7( +7) may be defined. in the fundamental 56 dimensional represen·
tation , as the aet of matrices

(73)
 63

Rere, 1':.'''c'.' is any real totally antilymmetric fourth tank teD.Ior (70 parameters)
and A-'._ = -A,,-' iI any real traeeleu second order tensor (63 parameters) . It.
SU(8) subgroup is obtained by restricting AG'" to be antiaymmetric and E.,••c'.'
to be anti-selfdual. The symmetric part of A·\. and the selfdual put. of E.,t'c'.'
parameterise Er transformationa , corresponding to the generatora orthogonal to
SU(8) . Property I above requires that these transformations can be constructed by
multiplying an SU(8) matrix, U, by a matrix of spin zero fields V ::: V_ V. o
We have already seen that the IIcalar field confilUrations parameterise the ho-
1llOIeneoua space SL(8, lR)/SO(8) . with the general achtbein field 11-' i ' being any
element 0(SL(8, R) . Thus V+ parameterises the SL(8, it) subgroup of E7(+T)" Fol-
lowing Clemmer and Julia, we shall be satisfied to show that UV_ V+ parameterises
the £7(+7) goup in the neighbourhood of the identity I .
An infinitesimal scalar field configuration pves an achtbein of the form

(74)

where aVo'i' is an arbitrary infinitesimal traceless second order tensor. An infinites.-

imal pseudoscalar field configuration is described by an infinitesimal fourth order
antisymmetric tensor Ao' ''c.'d'. having 35 independent components wit.h d' = 11.
The corresponding matrices V+ and V_ take t.he forms

(75)

and
(76)

An infinil.esimal SUeS) transformation takes the form

(77)

where
-AO .=
.. -A-.' 0' • (78)
It follows that, La leading order, UV_ V+ is an Er<+7) matrix of the form of Eq. (73)
(multiplied by a unit matrix converting flat t.o curved indice. ) with

(79)

and
-
E.,. ,c' ''' = E .. •••c,,,· + A.,.,c'''' . (80)
In order to derive property I, we must now show that any set. of E7 parame-
ters (A·' . " E., .. .,..,.) can be obtained in this way. F'tom Eq . (74) and (79) • .6.v·'. ,
and A·'., are arbitrary in~nitesimal traceless second order tensors. The infinites-
imal anti-selfdual tensor E.,.,c'. ' has 35 independent components, which can be
chosen to be the components with no index equal to 11 (Cor which Eq. (5 1) and
(80) give E.,. ,c'.' = E.,.,c'~)' The infinitesimal p&Cudoscalar fields A.,.,c'~ can,
then , be chosen to give any required infinitesimal value to the otber 35 independent
components oC E.,.,,'.' = =
E."'e'.' + A."'e'.' with d' 11. Thus we generate ar-
bitrary infinitesimal values Cor the E 7 (+7) parameters (A·' " , E."'c'eI' ) and establish
statement I in the neighbourhood oC the unit element .
Let us now verify statement II above, but again to lowest order in the pseu-
dosealar fields. We first notice that the matrix V+V is invariant under the SU(8)
transformation V - UV. In order to establish property 11, we must show that the
invariant matrix R = V+V(= vTv when V is simply V = V_ V+ and not tran&-
Conned with U), which has only curved indices, specifies the values of the physical
scalar Si'i' and pseud06calar Am", fields .
We find , to linear order in the paeud08calar fields , ~hat

S"m,Si',., •A m',.'.',' + Ai'/""""'S'" '.' S,."' ) (81)

!-CoSi'·' Si" , - Si" Si'.')
where the rows and columns are labelled by the non-ordered pairs (i' ,l) and (1:' ,I') .
The achtbein v·' i ' has been used to change from flat to curved indices, on the four
index antisymmetric pseud06calar field A"j'."' . It Collows from Eq. (46) and (51)
that Ai'i'.'" = 0 iC all indices are different from 11. The inverse metric t.elUlOr SO'i'
is pven by
.; )
- 1 . (82)

It is not difficult to see that the diagonal 28x28 blocks of R completely speciCy
the metric tensor Si'i', because the blocks have a s uperOuous number of componenLs,
compared to the 35 independent scalar field components of Si'i" Similarly, knowing
Si'i' , the olT-diagonal blow of R determine the 35 independent p&eud06calar field
components of Ai'i'.'" (or equivalently of its dual· Ai'j'.·"). We have thereCore
verified statement II , for all scalar fields Si'i' and infinitesimal pseudoscalar fields
A .....p .
We have thus established that the 70 spin zero fields of N = 8 supergravity pa-
rameterise the coset space E7 (+7)/(SU(8)/Z2), in the neishbourhood of the identi~1.
We now simply asaume, as in [Paper 18J. that the manifold of the spin zero field. if
the full coset space. Accordins to the general discussion of the previous section, the
theory should then have a formal SU(8) gauge invariance, with a set of auxiliary
gauge fields Q,.A B . Here it is convenient to use the basis of Eq. (65), with capital
letters, A,B = 4, 5, .. . 11 , which features the SU(8) subgroup of £? In ~his basis,
the 56x56 maw describin& tbe .pin zero fielde tues the block form
u aMN
V' = ( ijlBMN (83)

"here tbe bar denotes complex conju,ation

uABMN = (UABMN)- TjABMN;: (VABAINr. (84)
The local SUeS) Iymmetry ad.. on the left of V' (Rat indic.es A, B) and the global
E7 symmelry, considered below, acts on the right (curved indices M,N) .
Under an infinitesimal SU(8) transformation , we have
, ' - 1 (2A(A(C6B]DI 0 )
(lsu(.)V)V = 0 2:.1A(C68)0) (85)

where the traceleae antihermitian 8x8 matrix AAC(AA c) acLa on an 8(1r) repre-
sentation. The set of (tracelesa antihermitian) gauge fields QI'A B is defined, by
projectin, the E7 Lie al~bra elements (8I'V')V'-1 onto the SU(8) subalgebr&.,
"hich we: have just seen, Eq. (85), corresponds to the block diagonal elements :
Q I'(A(C6 S }D)
(8I' V')V'- l -_ (2 PI'ABCD PI'ABCD)
2QI'[A[c6 BIDI .
(86)

Using QI'A S as SU(8) g&.uge fields, the cov&.riant derivative of the spin zero fields is
given by
(DI'VI)V ' -l ;;; (pI'A~CD PI'A~CD) (87)

which is clearly ortho~nal to the SU(8) Lie algebra. The complex totally antisym-
metric rank-4 tensor PI'ASCD satisfies &. se:lf.du&lity condition and, therefore, haa
70 degreee of freedom. We note tht Q,.A S and PI'ABCD &re: inv&ria.nt under the
global E7 tra.nsformation., which act on the right of V'.
We conclude tb&.t, at leut formally, N = 8 supersr&vity contains a composite
SU(8) Yang Mills field QI'AB . However, whether or not this ,ause ,roup appears
with zero mass ,au,e oo.oIll is a dyn&mical question . We now turn to the global
~+7) symmetry, which is not purely formal: it is a true, but hidden, synune:'ry of
N;: 8 supergravity.
As we have: already seen, detailed calculations e8~abliBh that the 70 scalar and
paeud06Calar fields or the N ;: 8 supergravity multiplet take values in the coset
Ipace E7(+7)/5U(8) . This means that 'hen: is a way in which the group E1(+7) acts
on the spin zero fidds , but doell not in it.aelf require the action to be invariant under
,lobal E 7 (+7) tranlformatKlRB. In fact , direct calculations {Paper 18] show that t he
Lagangian demity for the spin-zero fields can be written in the form

L 5 -- ..!-.'I'<[(D
48K2 I'
V'V' - ')'] (88)

_ 'p P ABCD 1''' (89)

- 2411:2 I'ABCD" 9
66

J-
where e == det e,. 0" == defog,." is the determinant olt.he vierbein , and 1e 2 == 4.'rG",
is the t;ravitational coupling constant. The Lagranpan density Ls is mllPif5tiy
invariant under pobal E1(+7) transformations, as well as under loc&!. SU(8) tr~
formations. The local SU(8) and global Er(+1) groups ad non-line&rly 00 the spin
zero fields . However there it a global SU(8) symmetry ading linearly on the fields,
which correeponda to combining a constant. SU(8) tr&llBformation of loca.l SU(8)
wit.h the same SU(8) transformation of E7(+7) - When restricted to SO(8), this is
the usual global 50(8) invazianee of N == 8 supergravity.
We have restricted our above discussion of the local SU(8) and global Er(+7)
symmetrielll to the spin zero &ector of the theory. These symmet.ries are aleo estab-
lished for the fermion and vector fields in Paper 18. The fermion fields (8 gravit.inoe
and 56 Majorana IIpin-! fields) are invariant. under E7(+7) , but transform conva-
riantly under local SU(8) . The 28 vector fields of the supersravity multiplet are
Abelian and invariant under SU(8). The only term in the N = 8 super,ravity
Lagrangian density, which is not invariant under E7(+7) . is the kinetic term for
these spin- l vector fields . However, the ,lobalsymnletry Er(+7) is realised on the
field equations - exchan,in, the equations of mot.ion of the vector fi elds with their
Bianchi identities. When both field equations and Bianchi identities are salisfied
(on shell), there is invariance under Er( +T) adin, linearly on the 28 vector field
strengths and their 28 generalised dual tensors , which together form a fundamental
56 dimensional representa.tion of Er<+7) . Thus the global ~+7) ,roup is a symme-
try of the equationa of motion, although not of the complete action. The complete
action c.an be made manifestly invariant under the local SU(8) symmetry.
Finally, we mention tbe symmetrtea of the alternative gau,ed N = 8 super-
gavity model,1' in which the SO(8) subgroup of E7(+7) is gauged . This SO(8)
subgroup is realised on the 28 vector fields of the supergravity multiplet, which
become the SO(8) Yang-Mills fields . The global E 7(+7) invariance of the field equ~
tioos is thereby broken . Nevertheless, the ET/SU(8) coset structure of the spin-zero
fields remains intact, and the local SU(8) invariance is not affected by the laug-
ing of SO(8). Consequently, the gauged N = 8 supergravity model has a local
SO(8)xSU(8) symmetry.
Unfortunately, it is difficult to reconcile either version of N = 8 supeJ'lI'avity
with the well-est.ablished standard model phenomenoJoc. At least some of the
SU(8) Yang-Milia fields QI'AB must dynamically develop poles corresponding to
physical gauge particles, such &8 tbe W± bosona. Quarks and leptons must also
appear as bound states of the fundamental supergravity multiplet . Attempts to
obtain realistic particle spectra in t.his way are beset with unresolved dynamical
problems.l4.,l&

4.4. Kaluza-Klein Theories

In the previous section, we saw how N = 8 supergavity is constructed, by the
dimensional reduction of a simpler eleven dimensional theory to four dimensional
space-time. The extra ieven dimensions were introduced 'here as a purely mathe-
 67

matical device. However it is temptin, to attach phyaicalaipifieaoce to the simpler

t.heory and to t.he higher dimensional apace. The idea or int.roducing ext.ra phys-
ical dimensions at the fundamental level ,oes back to the 1920's and the work of
Kaluza l • and Klein. 17 They sbowed bow electrodynamics could be obtained , by
starlin! with Einstein'. general t.heory of relativity in five dimensions and compact.
ifying the extra spatial dimension into a very small cirde. The extra dimension
is then not directly ob~vable, but t.he associated components of t.he met.ric ten·
IIOr field can be interpreted as the Maxwell electromap1etic field together with a
Brans-Dicke scalar field [paper 20).
Einstein', equations, in the five dimensional Kaluza--Klein theory, contain not
only tbe four dimensional gravity equations, but also Maxwell's equations for the
electromagnetic field and the Klein-Gordon equation for the scalar field . Electro-
magnetic gauge invariance is a remnant of reparamel.erisation invanance in Iohe extra
C(K)rdinate. A few years ago, there was a renewed intereslo in generalised Kaluza-
Klein theories, motivated by the hope of explaining the origin of the non-Abelian
gauge invariances of the standard model in a similar way [paper 201.
In higher dimensional Kaluza-Klein theories,l' we start with a general relativistic
theory in 4 + n dimensions, having a metric tensor gAB together with additional
matter fields. The 4 + n dimensional manifold is assumed to have the product
form Mi x B , where Mi is the usu al Minkowski space and B is some compact
Riemannian manifold of dimension n . More precisely, we assume the vacuum state
to be Mi x B , described by the ground stal.e meloric,

( (~ • "'» = (g,.(z.)
gAB , 0;- 0
0) '
"rij (tP'") (90)

Here g",,(.%O ) is the ordinary metric tensor of the four-dimensional world , and
1iJ (tP'") is the metric \ensor of the int.ernal space B havins co-ordinateB 41'" , i =
1,2, ... , n o
Symmetries of the compact inl.ernal space B , which generically has dimensions
of the order of the Planck lensth GN 1 /2,.., 10- 33 em, appear as gauge synunetries
in the effective four-dimensional world . If the space B has a continuous group G
ofisometries (transformations leaviIll the metric tensor l'ij(tP'") form-invariant) . we
obtain a four-dimen sional Yang Mills theory with gauge gro up G. The condition
that the infinitesimal C(K)rdinate transformation
., -., +,KM) (91)
should leave ")'ij(IP.I:) form-invariant leads to the Killing vector equation
D;KJ + DjKi = 0 , (92)
where Di denotes the eovariant derivative for the space B. For an N dimensional
symmetry group G, there are N Killing vector fields Ki (,), a = 1,2, ... , N , which
obey the Lie algebra

[K:a~,K:8:J =r·<Kl a:. (93)

68

r
where k are the structure constants for the group G.
We now consider small ftuctuations cOflel!lponding to maasleat excitations in the
off-diagonal components of the metric tensor

(94)
The four dimensional fields A,II-(zO') are massless gauge fieldt of the gauge group
G. This may be seen by noting that. there is a symmet.ry under the infinitesimal
gauge transformation

A:(z) -0 A:(z) + 8"c a(z) + /dc£'(;J;)A~(z) (95)

which results from the special co-ordinate transformation (see Fi,. 4.5) :

(%0' ,ft) _ (zO', ¢; + t:(- )(z}Ki (¢» . (96)

Fie . 4.S. The ddormatioD 01 the ~ ,,.--.. Eq. (86). co.. liP c.Jdina to' . pup tramCar..
matiOll., Eq. (95). in. Kal ....... Klein model.

This diffeomorphism transformation leaves 'Yij(4I) and , in the linuriaed approx-

imation , g..... (z) form invariant. In order to obtain exact form-invariance of 9,11"(%)'
the linear ftuduation ansatz of Eq. (94) it on.eo generalised to include quadratic
terms in A:(z) as follows :

(97)

Substitution of this Kaluza-Klein &Dsatz into the gravitation&! action,

...
s = -......!...-, jatzrf' ¢lJldet9AB IR
or into some other diffeomorphism invariant action, must dearly give an action for
(98)

the A~ ( z ) fields which is invariant under the gauge symmetry of Eq. (95). The
effective action , obtained after int.egration over the compactified co-ordinatea ¢" I
therefore contains the usual Yang Mills kinetic term

(99)
 69

for the case of a simple goup; if the symmetry poop or the internal "Pace B is Dot
simple a different coefficient. 1/(491) is needed [or each basic invariant subpoup.
In the above discussion we neskcted the effects of any ma.tter fields. In partie-
ula: , we implicit.ly assumed that matLer fields with nOD-zero vacuum expedation
values transform as singlets under G. The gauge symmetry group G is then not
spont.aneously broken by the matter fields. However the value of the gauge cou-
pling constant. 9' and thestrudure oftbe non-Iineaz terms in the massless excitation
anAatz (or gAB depend on the detailed form of the maUe r-field interactions, as do
the spin-zero massless modes.
At first sigM we seem to have an example of a derived symmetry, namely Yang
Mills gauge symmetry for t.he gauge &roup G of isometries for the space B . However
this symmetry is simply a ~pression of the diffeomorphism symmetry under the
special transCormation of Eq. (96). tn rea1ity we do not have a derivation of a
symmetry. but rather a metamorphosis of one already present. Even to obtain
such a metamorphosis of the diffeomorphism symmetry into a gauge symmetry, it
is crucial to assume that the eompactified Riemannian space B has an isometry
group G. Otherwise the produced gauge fields get BiUsed .
The symmetries of the apace B should arise as a result of minimising the vacuum
energy. The expression for the vacuum energy of a space B bas a diffeomorphism
symmetry, and its minimisation can easily lead to the space B being symmetric
under some isometry group G. It is really a question of whether or not there is
a spontaneous breakdown or the diffeomorphism symmetry. In practice the form
of the eompactified space B and the consequent continuous isometry group G are
usually just assumed, which is tantamount to putting in the gauge symmetry by
band.
As illustrations of the Kaluza-Klein mechanism for obtaining gauge symmetries,
we brie8y reconsider the two versions or N = 8 supergravity discussed in the ~llevi·
OU8 section as different compactifications of eleven dimensional supergravity.lS The
Cremmer-J ulia dimensional reduction corresponds to taking the seven dimensional
compact space B to be the seven·tOrll8 TT, with isometry group U(I)7. There are
thus seven Ka1uza-Klein U(I) gauge bosons in tbe Crernmer·Julia N = 8 superpav·
ity. In addition 21 Abelian spin· 1 gauge b080DS arise from the components Allmn
of the third rank antisymmetric tensor present in superp-avity. The deWit-Nicolai
gauged SO(8} version of N = 8 supergravity corresponds to taking B to be the
seven sphere 51, which haa 50(8) as ita isometry group. In this case all 28 of the
SO(8) gauge bOSOM arise as Kaluza·Klein gauge fields, and the four index gauge
invariant curl of the antisymmetric tensor ANPQ

(100)
develops a non·zero vacuum expectation value, in a necessarily four dimensional
space. The existence of the four index field strength FMN PQ here provides a dynam-
icaJ reason for the compactification to four space· time dimensions. 19 The potential
spin-l «a.uge bosons of the hidden local SU(8) symmetry of N = 8 supergravity
70

would be dynamic.&l bound states of the spin-O fields and nol Kaluz..Klein gaup
fields.
There are many other p08Sible compadificationa of eleven dimensional super-
gravity, with some or all of the N = 8 lIupenymmetries broken in four dimen-
sions. In Paper 20, compadifications to a space B having SU(3)xSU(2)xU(1) as
ita isometry group are considered. However the fermion spectra obtained from these
models are vedorlike, in contradiction to the complex representations of the stan-
dard model. This is & general problem for Kaluza-Klein theories, in which all t.he
gauge fieJd, arae aa components of the metric tenaor in higher dimenaions. 20 A tbe-
orem due to Atiyah and Hirzebruch shows that, in t.he absence of elementary ,aule
fields , the Dirac equation in 4 + n dimension. alway. leads to non-chiral fermion
quantum numbers in "dimensions. There is 8. similar chirality problem for the
Ratita-Sc.hwinger field .
In order to circumvent the chiral fermion problem in Kaluza-Klein theory, both
(i) chiral fermions and (ii) explicit gauge field. must be present already in 4 + n
dimensions. The extra number of dimensions n is required to be even , since it is
not possible to define a generalised "'Ys Dirac matrix, and thus chirality, in an odd
number of dimensions. Also to ensure the survival of chirality on compactifica'ion
to 4-dimen.sions, tbe gauge fields must assume a topologically non-trivial structure
on the compaetified space B.
As a simple example 21 with n = 2 , the space B could be an S2 sphere having
a constant magnetic field FS8 on it, corresponding to a Dirac monopole in the
embedding space for the S2 sphere. The integal of the m&8oetic field over the
52 sphere is, apart from a normalisation factor , the Pontryagin index for the two-
dimensionallpace. The Atiyah-Singer index theorem then ensurea that a fermion
field , chiral in 6 dimensions, will have one or more lero modes of the Dirac operator
for the two-dimensional space B . These zero modes are observed as masale88 chiral
fermions in four dimensions. Here the natural mass scale is given by the inverse of
tbe radius of the 52 sphere, which is generically of the order of tbe Planck mass
GN- t/2 _ 101IIGeV. The 'massless fermions' can obtain small masses from low
energy gauge symmet.ry breaking, u in the st.andard modeL
Inclusion of explicit gauge fields in the higher dimensions is fundamentally dif-
ferent in spirit to the Kaluza-Klein idea. The oripnal idea oC getting gauge fields
from hieber dimensional gravity is given up , and the gauge fields are put in by band.
The presence of chiral fermions coupled to gauge fields in 4+ n dimensions means
tbat tbe problem of anomalies must be faced.

4.5. Anoma1y Cancellation

In higher dimensional theoriea with an e Yen number D = 2m dimensions, it is
possible to have a parity violating gauge theory, in which the leCt..-handed fermions
transform differently under the gauge group from the way the ri,bt..-handed fermions
transform. Fermion loop diagrams with m + 1 external gauge particles are poten-
tially anomalous in such a chir&! theory.~ Anomalies thus arise from triangle dia-
 71

Clams, sueb u Fig. 3.1, in D = 4 dimensiona and from hex8lon diag&mB, FiS. ".5,
in D = 10 dimensioDI . These quantum anomalies violate gauge invanance and uni-
tarity of the S-matTix. Consistency then requires the cancellation of all anomaliea.

Fi&. ".6. Hexacon diap"am 'llll'ith alunU .......41 b<»ON« sr_vi.t.c.u, whkb can live PUle, g .....i-
lational or mixed anonwlis in ten dimemiON.

FUrthermore, in D = 4.1: + 2 dimensions , the particle and antiparticle associated

with a Weyl field have the lame helicity or chirality. Thil is in contrast to the
situation in D = 4.1: dimensiona, where the helicities of particle and antiparticle
are opposiLe. The grav,t.ational coupling of a Weyl fermion in 4.1: + 2 dimensions is
therefore chiral, and we have the possibility of gravitational anomalies.23 Gravita-
tional anomalies violate diffeomorphism symmetry and lead to non-conservation of
the energy-momentum tensor. In addition to purely gauge and purely gravitational
anomalies , there is the p068ibility of mixed anomalies, arising from fermion loop
diagrams wit.b both gauge part.ides and gravitons as external lines.
Higher dimensional chira! theories in 4!-+2 dimensions are severely constrained,
by the requirement of vavitational and gauge anomaly cancellation.23,24 This is
illustrated by N = 1 supergravity in D = 10 dimensions, which is the highest di-
mensionality consistent with a chiral supergravity theory having no particles with
spin greater than 2. The pure D = 10 supergravity multiplet coDsistsoftbe massless
partides listed in Table 4.1. The numbers of physical degrees of freedom associ-
ated with each of the massless fields are indicated in brackets. The gravitational
anomalies, due to the chira) vavitino 1/1# and spinor A fields in the supergravit.y
nrultiplet, do not cancel. Thus pure N = 1 supergravity, without matter fields , is
not consistent quantum mechanicaJly in 10 dimensions. Local Lorentz invariance
and hence energy momentum conservation are violated . However the gravitational
anomalies can be cancelled (paper 21] by including as matt.er fields, a superaym-
metric Yang-Milla nwhiplet, Table 4.2, provided the gauge group G is 50(32) or
Ea x E, . Moreover the gauge and mixed anomalies can also be cancelled for these
t.wo groups.
n
Table 4.1 . The N = I lNpupvity mukipW in 10 eIi- a"-,

Field (Oqn:.:s of Freedom) Oaaiption

,,. .. (3S) tp'avitoD

t/J,.(56) left-handed. Majorana-Weylvavitino
8 ... ,,(28) anwymmetrk teNor
A(.) ri&ttt.-banded Majoxana.-Weyl .piJKll"
¢(l) ..,.w

Table 4.'2. Supasymmetric Yan,·Mill. multiplet in 10 dirnenaion. ror a.allle goup G 01 dimerW_
".
Field (Oqne. of &eecbn) De.c:ripti __

A.II°(h) PUce particle

X·{s.) tefl-banded Major_ Weyl .pinon
in adjoint rep tMS'Iwion 01 G

The requirement of a non-trivial cancellation of anomalies, bdween the super-

, ravity and s uper Yang-Milia multiplets in 10 dimensions, thus determines the gauge
y oup to be 50(32) or Ea x Es (apart from the relatively trivial cases 0( U(I)49O and
Ea x U(I)"IS) . In a sense, therefore, t he ,auge group G can be said to be derived
and it is for this reason that Paper 21 is included in t his book. However it is more
correct to say that tbe gauge group is selected, rather than derived, by t.he anomaly
cancellation requirement .
Le t us now examine how the anomaly cancellation requirement specifiC!t the
super Yang MiD, group [Paper 21]. The anomaly is defined as the variation 6.\r of
the one loop fermion effective action

(10 1)

under an infinit.esima1 gauge or general cl>Ordina.te transformation .\(z) . Here lJ

is the inverse propagator for a Weyl fermion, in a background of both Yang Mms
and gravita.tional fields. The variation of a formally gauge invariant action rinD
dimensions must be an integral , local in .\(Z'), of tbe fo rm

(102)

Here aD is a D form . as is required to make the integral sensible.

It is very convenient to use t he language of differential forms to construct the
anomaly D form .\aD . in terms of a formal gauge invariant D + 2 form ID+2 . The
construction is given by the descent equations

(103)

and
(104)
 73

'I1:Ua construction is ambilUoua up to a dosed form. but the int.e&ral or Eq. (102)
specifies a unique anomaly corresponding to JD + 2 •
The gravitational, gauge and mixed anomalies in D = 10 super Yang Mills plWl
supergravity theory can be constructed from a 12 form, which is p roportional to

ITrF' + 24
I 12 = - 15 ITrF' t r R'

- 9!0 Tr F'[4l, R' + 5(" R')'l

tf ~ tr R'tr R" (t.r Rl )3 ]
+ (n - 496) [ 7560 + 5760 + 13824
1
+ ~tr R2 tr gt + 32 (tor R' )' . (105)

Here F is the gauge field strength two form

F= ~F,."d;r"dz" (106)

in the adjoint. representation of the gauge algebra, and R is the curvature two form

R= ~R,. .. dz"dZ" . (107)

The Riemann curvature tensor RI''' is expressed in the zehnhein formalism , with
two suppressed flat indica. Thus t.be two form R is a 10x l0 matrix, belonging
to the fundamenta1 representation of the Lorentz algebra SO(9 ,1). In the above
formula we suppress the wedge product symbol so that , for example, F 2 denotes
F AF. The symbola Tt and t.r are used to denote t.races in adjoint and fundamental
represent.atiollB respectively. The number n of mat.ter Weyl Cermion species is equal
to the dimension of the Yang Milia gauge group G.
The anomalies associated with 112 are clearly non~ vanishing for any gauge group
G . However Green and Schwarz pointed out the existence of a non &&uge invariant
local counterterm that cancels these anomalies, whenever 112 faclorises int.o an
expression of the form
(108)
where k is a constant and Xs is 8. gauge invariant eight Corm made out of F and R.
The two necessary and sufficient conditions Cor this Cactorisation to be p06Sible are

n = dimG= 496 (109)

and
Tr 1" = ~Tr F'Tr F' _ 1 (Tr F' )' (110)
48 14400
witb .c uniquely determined to be
1
k = -30. (lll)
74

It .. these two conditions "bieb characterise the aU~d poupe 50(32) and Et x Ea.
In addition there are just. two other groups, U(I)498 and E. x U(1)2-M, "bich satisfy
the conditioOJ rather trivially.
When the In fadorisation condition is satisfied, the key to the anomaly can-
cellation lies in the addition to the action of II higher derivative counterwm of the
form
Sc - J BX,. (112)

Here B is the two form

B = B,."d:z:"dzO' (113)
and 8,. .. is the antisymmetrie tensor field belonging to the supergravity multiplet.,
which is requ ired to transform non trivially under both local Lorentz and Yang
Mills gauge transformations

1 1 1
6B = Wn - 30W,y (114)

The t.wo forms W~L and ""~y are obtained by descent Crom tr R' and Tr F2 respec-
tively. The superscript 1 indicates that they are linear in t.he infinit.esim&l gaule
parameters J.(z). They are constructed by applying infinitesimal transformations
to the Lorentz and Yang·~1il18 Chern·Simonl three forms WSL and W3Y

(115)

&.ISY = dw~Y . (116)

The Chern·Simonl three forms themselves satisfy

(117)

dw3Y =TrF2 . ( 118)

For example , the Yang·MiUs Chern-Simons form is given explicitly by

(119)

The field strength HI''''' lUlSOciated with B"" is made invariant under local Lorentz
and Yang-Mills r;auge transformations, by the definition

1
H = dB +WlL - 30WlY (120)

of the three form field strength H .

 75

The basic point in the Green·Schwarz mechanism is t.hat t.he new terms intro-
duced into the action generate anomalous B exchange diagrams, Fig. 4.7, which
just eanceJ the anomaly from hexagon diagrams, Fig. 4.6.
It is thus the Green-Schwan factorisation condition , Eq. (108) , for anomaly
cancellation which selects 80(32) and Es x Es as the allowed gravity-coupled super
Yang-Mills gauge groups in 10 dimensions . However what. is the motivation for
considering N = 1 8upergravity in 10 dimellBioDa? It is possible to have nOD-
trivial anomaly cancellations between particles of different spin in higher than 10
dimensiona,'8,27 but these theories suffer from incorWstencies. They either contain
I
a massless spin particle (a "avitino) without being supersymmetric, or ma.saless
particles with spin greater than 2. Chiral supergravity theories in (even) dimensions
higher than 10 have more than N = 8 supersymmetry generators when reduced to 4
dimensions, as is easily verified by counting the number of spinor component.s for the
gravitino. The graviton multipld for chiral N = 1 supergravity therefore contains
maaslese particles with spin greater than 2, unless the number of dimension. D < 10.
So the maximum number of dimensions in which it is p068ible to construct a
consistent N = 1 supergravity model with ehira] fermions is D = 10. Howe ver
in it.selC thia is not a convincing reason Cor seleding a ten dimensional model. In
Cad, at present, the main interest in N = 1. D = 10 super gravity coupled to super
Yang-Mills theory lies in it.s identification as the low energy effective field theory
limit oC a consistent superstring theory. Indeed the anomalous B exchange diagram,
Fig. 4.6, and the associated higher derivative counterterms, discussed in Paper 21 ,
were sunealed by the low energy expansion of type I superstring theory. The
critical dimension, D = 10, of superstring theory is determined by the uquirement
oC invariance under infinitesimal reparameLerisations of the two-dimensional world
sheet of the string (confonnal invariance) .2S
In .tring IMpage the above anomaly analysis means that the unorient.ed closed
superstring,(25) which has the N = 1, D = 10 8upergravity multiplet of Table 4.1 as
its 'pectrum of massless states, suffers from anomalies. For consistency it is neces-
sary to introduce an open superstring, with Chan Paton facton(2$) corresponding to
the unique gauge group 50(32). The massless states of the open string correspond
to the Yang-Mills supermultiplct of Table 4.2. Thus we have the rt:markable resul~
that anoma1y cancellation, which rcaJly means quantum meehanicaJ consistency,
implies that once we consider non-orientable dosed strings then open strings are
forced to exist also.
The realisation of E, x E" as a string theory gauge group , required the con-
76

at.rudion of the oriented dOlled heterotic string [paper 22]. It is just the Yan,·Milll
I,"ule groups 50(32) or Es x E" which un iquely em~ r,e ~ required, but rather
hidden, symmetries of the D = 10 heterotic IItrin& with N = 1 spacdime super--
symmetry. The origin of these specific groups is discussed in the next section. The
crucial inl!:redient, which lUarll.nLees the anomaly factorisation and cancellation by
the Green Schwarz: mechanisms, is the invariance of the two-dimensional string
world sheet under co-ordinate toran.formations not continuously connected to the
identit.y (modular invariance).27

4.6. Strings
Superstring &beory 25 has recently aroused &reat interest, as a candidate theory for
unifying all fundamental forces in II. 'theory of evetythin,'. In particular quantum
string theories requiu gravity in mod casea. 21 lt also appears, under 'Very resuictive
conditions though, tht.t they could be consistent. a.nd finite . Thi, ia in cont.rast. to
qua.ntum poiot part.icle theories or gavity, such as supergravity models, which all
have non-renormalist.ble infinit.ies.
In the critical space-time dlmension D = 10, there is t. near uniqueness of the
superstring t.beory. There are just three models wi~h N = 1 apace-time supenym-
metry and a Ya.ng Mills gt.uge group: the SO(32) type I.uperst.ring a.nd the E, x E.
and SO(32) heterotic strings. The gt.uge groups, SO(32) and Es X Ea , au: the &&rne
u those suggested by the anoma.ly canedlation requirement. from ten dimensiont.l
N = 1 lu persravity. Of these models, the E. x E. heterotic string .cerna the mo.t
promising phenomenologically.
On reduction to four space-t.ime dimensions, however, the uniqueness of t.he het.-
erotic string essentially disappears." Generalised 'compactifications' to four dimen-
sions can be mt.de, which treat the cl>Ordinatea for left..-moving and right.-moving
modes on the closed string differently. It. is tben difficult to interpret tbete co-
ordinates conventionally, as describing six extra space dimensions compactified on
a scale of the order of t.he Planck length. It is more appropriate to formulate
.ueh model. directly in .. space-time dimensions, and t.o interpret t.he remainins
string world-shed degree. of freedom (which a re required to avoid an anomaly in
conformal invariance) u internal degrees of freedom . There are very many four
dimensiont.l superstring models, having gauge groups of rank 22 or less. These may
be constructed, in a fermionic formuiation 2Sl , by modifying t.he periodicit.y condi-
tions (boundary conditions in t.he ease of open strings). By t.aking eombint.tions: of
different sets of boundt.ry (periodicity) conditiona, in the two string co-ordinates tr
and T, it is p06Sible to construct many modular invariant and , thereby, consistent
models.
In thia section we consider (1) the Ftenkd- Kac-Segt.l mechanism responsible for
the origin of a no n-Abelian symmetry in the heterotic string, and (2) the general
appet.ranee of gauge symmet.ry in string theories.
 77

4.6.1. Tle Hderotic Stn"f uti K(Jc-M~, Alfdru

Here we shall discuss the by now classical example of the lO-dimensional het-
erotic string [Paper 22J . Its charaderistic property is that it haa 16 boaonic fields
X[(~. 'r), I = 1,2, ... 16 defined over & cIOfJed orie.ntabJe strin,; and taking values
on a certain somewhat skew torus TIll . The co-ordinate& (T and r paramet.erise
the 1 + 1 dimensional world sheet of the string and the 16 fields Xl(cr, r) are thus
described by 8. 2-dimcnsional field theory. These fields X [ are peculiar in having,
when quantised, & symmetry under either E, x £8 or Spin(32 )jZ2 (which bcu the
same Lie algebra as 50(32)). The classical Kaluza-Klein symmetry, obtained from
8uch a toroidal compadification, is just the Abelian subgroup U(l}lll ; the isometry
group of the torus Ttll .
An important ingredient in obtainin,; the symmetries E. x E. or Spin (32)/ Z2
is t.he assumpt.ion t.hat. t.he fields XI(O' ,T) are purely lefl-moving. That is to say
that the fields Xl are assumed to be only functions of T + 0', where T is the time
co-ordinate and 0' the space co-ordinat.e, in the 1 + 1 dimensional field t.heory. Thus
any dist.urbance in t.hese fields moves with the velocity of light to tbe len and is
massless, in a 2-dimensional world sheet sense, when viewed aa a quasiparticle on
the string.
Now a toru.valued field theory haa the p08Sibility of having the field!: wound
up around t.he torus . These 'topological' winding modes ofthe field!: Xl are known
aa solitons. They exist. in addition to the usual oscillator modes or phonons on the
string (F~. 4.8). ....

--

The torus-valued fields Xl can take aa values all real numben , where we then
ident.ify seta of Xl values congruent modulo a lattice A in the RIS space.3D This
means we regard the torus aa a quotient space TIs:::; Rls/A . It is coovenient
to write A :::; dr, where tbe fundamental string length scale I :::; (flO)-1/2 will
hencefortb be set equal to unity. We t.hus chooee uoita in which the string tension
To:::; 1/211a' = 1/11.
From the requirement that the disturbances in the Xl variables corresponding to
soliton states be purely left...movinr;, it follo.,s tbat tbe soliton momenta pi should
78

lie Oft th~ l&ttiu r . Quantum meehanieaDy, the .,liton momenta pI tntlSt aleo lie
r,
on the dual lattice since the alrin,; has II. sinr;le-valued wave function on the tofWl .
Physical sLates are coJUltructed by combinint; ri&bt.- and lef\-mO\'int; oeci.I.lator and
soliton modes, which satisfy the condition

N = N _ 1+ ~ ~)pl)' • (121)
I

and the ten-dimensional mass operator is

(mass)' = 8N . (122)

Here Nand N are the normal ordered number operators for the rir;ht.- and left..-
moving 06cillator modes. Since N and N au integer-valued , it follows that. the
squares of the allowed soliton momenh. LJ(pl)2 must be even integers:... Finally "he
consistency of strinr; t.heory requires that. the lattice be self-dual r = r , otherwise -
modular invariance, and thus unitariLy, breaD down. This means that one-loop
amplitudes develop anomalies, which_dest.roy £Iabal reparameteriaation invariance
of the string world sheet, unless r = r .
The above requiremenLs can only be sat.isfied if t.be lattice r is even and self-
dual. A lattice is said t.o be even if the 1en3t.h .quared of any latt.ice vector is; an
even integer. There arejuat two even, self-dual, 16-dimensionallatiices, r. x ra and
ru; . The first possibility is t.he IM~ice of roots for tbe group Sa x EI; the second is
a certain subset. of the possible weir;hLs of t.he Lie r;roup Spin(32) , in fact. the wei&bt
lattice of the factor group Spin(32)j Z,. So remarkably the same groups, E. x Ea
and Spin(32)/Z" are .elected .. t.hose con. istent. with anomaly cancellat.ion in 10
dimensions. The crucial requirement in aeiedinr; these two POOP', rather than any
ot.her simply-laced Lie group of rank 16, is modular invariance. As mentioned in
section 4.5, it. is alao modular invariance which &narant.ees anomaly cancellation, by
t.he Green-Schwarz mechanism, for the heterotic .t.rinr;.(21)
However, II priori, the .tring running on the torus T t " defined by t.he lattice
r. r.
x r
or llt , has no symmetry under t.he goup. Es x E. or Spin(32)/Z2 . The
X [ degrees of freedom correspond to the Cartan sub-algebra of the Lie group,
but there are no fields corresponding to the other directions in the Lie algebra.
Nonetheless it turns out - and here we have an example of a hidden symmetry -
tbat the quantised system of the Xl variables, running on t.he torus Ttl, indeed
does have the full symmetry under the r;roup E. x Ea or Spin(32)/ Z2 - The groups
are represented, in this 1 + 1 dimensional field tbeory, by transforminr; XI-phonona
and XI_solitons into each other (Fir;. 4.8) . It is, of course, a prerequisite (or such
a symmet.ry to occur that the two-dimensional world sbeet enert;)' and momentum
should he the same for t.he phonons and .olitons. Classically this is; not. possible,
since then the soliton energy is qU&n\iaed but the phonon energy is not. Quantum
mechanically, the equality is arranged by a fine-tuning oUhe radii and angles oUhe
torus . Theaixteen Xl-phonons with N = 1 can then lie in the same rrwltiplet as the
 79

480 IOlitona "it.b E,{,pl)2 = 2. Neverthelml it. stiR seerm 8Omewh&\ miraculOQl
t.hat. such fine-tunings are sufficient to make the quantum system invariant under
the group E, x E, or Spin(32)/Z2_
We will now demonstrate the presence of the symmetry by writing down t he
group generators as quantum operators,30 using the Frenkel-Kac-Segal
construction.' l - 54 These generators commute with the mass squared operator and,
thus, the physical string st.at.es provide a representation of the synunetry group at
each m&88 level.
The commuting generators HI, of the Cartan sub--algebra U(1)18, Me simply
represented by the momentum operators along the torus directions

(123)
The basis generaton EX. not belonging to the Carlan lub-algebra, are each repre-
sented by an integral over a normal-ordered vertex operator

V(K. , ) =, oxp[2iL K I Xl (,)] , ( 124)

I

where Kl is a root vector, of length squared two, and

, = .xp[2i(. +<)] . (125)

The operator EK is required to create (or annihilate) a soliton , characterised by a
shin, lyin& from the identity element to the lattice point Kl and , hence, also by
momentum KI ,
The &enera.t.ors EK are actually taken to be

EK =AKCK(P) ( 126)

where

AK = - I[ V(K,
• 0
Z)dT ( 127)

f d,
=
."
pV(K.,)

and CK(P) is a sip factor . The factors CK(P) are functiollJ of the 16-momentum
( 128)

operators, which must be constructed to obey the equation

CK(P - K')CK'( P) = (_I)K .K· CdP - K )CK(P) ( 129)

in order that the leDerators EK have simple commutation rules. The scalar product
K . K ' is of. course defined to be
..
K . K'= L:K1K'/ . (130)
1=1
80

In order to cloee the at,ebra and obtain tbe eotted eornmutation rules, we aJ.o
require
(131)
where £( K, K') = ±1.
There are many solutions to Eq. (129H131) . A particular explicit 501ution2$
is expressed in terms of an ordered * product, which i. defined by chOO8in& some
definite order ins of the 16 basis vectors e{, i = 1, 2, ... 16, Ci • ei = 2, of the lanice
r . x fa or fill . Any two lattice vectors K and K' can be expanded in tums of Ci
with integer coefficients

KJ = "
Ernie! (132)
;=1

K 'I " m~e[

= L: (133)
;=1

and the * product is defined by

K • K' = E m;mjt.; • C; . (134)

'>j
Then the sign factors
(135)
..",fy Eq. (129)-(131) fo<
,(K, K ') = (_I)KoK' . (136)

Wi~h the definitions Eq. (123) and (126) of the generat.ora, the Lie alsebra
commutation relations are readily verified . In Cad Frenkel and Kac!!1 and Se,alS2
constructed not just the Lie algebra but a larger infinite-dimensional KavMoody
algebra. The Ka.c.Moody algebra, in string theory, ad.. as a partial spectrum
generating algebra, linking states of different maae. The extra generators e: and
H~ , of t.he affine E,)( E, or affine 50(32) algebru, are cOD8truded by inttoducins
a factor zn, for integer n , under the intet;ralsigns in the expressions , Eq. (126) and
(128) , definint; eK , and in a similar expresaion for HI as an integral over ~ .
By the above construdion of the generators, we have demonstrated tbe ~oba1
Es x Ea or 5pin(32)/ Z2 symmetry of the ten-dimensional heterotic string. Among
the physical zero mass, N = 0, string states are vector particles belonpng to the
adjoint representation of Ea x E. Of 5pin(32)/Z, . These string states are gause
particles, in the 9 + 1 dimensional space-time, and would thus .isnify the presence
of a gauge symmetry, if the string theory wa.s written a.s an infinite component field
theory.
 81

The sause particle Itrins dates are or the (orm

(137)

for t.he neutral vector bosona, arising from the Carlan l ub-algebrA, and

IQ. X !p' ; 2:(pl)' = 2) (138)

I

for the other 480 charged vector boson!_ Here li)R are tbe ten-dimensional spin 1
&round states of the ri,ht-handed sedor, which are singlets under the symmetry
goup E. x E. or Spin(32)/Z:z_ The leR-moving sector Xl-phonon modes ii ~110}L ,
having if = 1, and the Xl-soliton modes 1pJ' ;E[(pl)'2 = 2) together make up the
adjoint. representation of the group E, x E. or Spin(32)/Z2 _ These len-moving
modes are seaJan in 100space.
Tbis leads us into a general discussion of gauge symmetries resulting from string
theory.

-I.6.! . GUft: Srmmdry from String"

An interestin, example of a 'derived' symmetry, or rather the metamorphosis of one
symmetry int.o another one, is the appearance of gause particles and thus of gauge
symmetry in string theory. For simplicity, we shall restrict our explicit discuaaion
to Abelian. sause symmetry, although it is true that Yang-MiDs gauge symmetry
appears in open strin( models with Chan-Paton factors and in the d06ed heterotic
string, where the charges are distributed along the length of the string.
The above gauge symmetry was in fact a problem for the orwnal dual models
of hadronic physics. The dual model was shown to be COlIsistent with unitarity only
in D = 26 dimenaiona and when it contained massless partides having spin 1.
Since gauge symmetry is really a formal symmetry, it will naturally depend upon
the formalism whether or not a second quantised string theory will exhibit gauge
invanance. For instance, writing string theory in the light.-cone formalism, with
only physica.l degees of freedom ,3S leads to a formalism without gause symmetry.
This, of course, may just mean that, in &orne way, the sause hu thereby been fixed .
In fact , even in the lighkone formalism, there are massless vector bosons with only
D - 2 components, rather than the D - 1 components of a massive vector boson.
In this sense, there is the physical imprint of a gau,;e symmetry.
When the low energy or ~ero ReUe slope a_ 0 limit is taken, string theories
can be described by an effective quantum field theory, which pOlJ5eSSe8 YanS-MiIIs
gauge invariance. In this sense , we can say that gauge symmet ry is contained in
string theory. What now is the origin of this gauge symmetry; is it put directly into
the theory or does it appear dynamically?
It haa been known for a long time that the gauge symmetry in string theory is
connected with 2-dimensional diffeomorphism symmetry; more precisely with the
suhaet of conformal transformations generated by the Virasoro algebra. In order
82

to see this connection, it is convenient to chooee a parameteriaation of the atrin,

world s heet, in which the world sheet metric is the flat 2-dimensional Minkow8ki
metric. The s tring action is still invariant under conformal transformations in this
parameterisation , whi ch is characterised classically by the conditions

(139)

The quantum creation operators Q~ , n < 0, for the 08C.iIIator modes, i.e. the
phonona, on the open bosonic 26-dimensionalstrinl, are then defined by t.he Founer
expansion of t.he p08ition variables in tbe 26-dimeJUlional Minkowski space-t.ime
manifold
X"(D", r) = Ill'
.,. "
+ Q'~.,. + i E .!.o:e- iftT eoancr . (140)

Here q" and ~ = pi' are the centre of mass p08ition and momentum of the IttinS.
P hysical states rib) in the string theory formalism must obey the conditions

L..I¢)=O " = 1, 2, ... ( 141 )

and
(L. - I)I ~) =0. (142)
The Viruoro operators L" are defined as the Fourier component.e of the two-
dimensional cnerQ-momentum tensor Tap and are pven by the equations

(143)

The massless pUKe boson, which we shall call the phoLon, is described by the
string states

(144)

(145)
Here ,~ is an arbitrary polarisation vector, and we consider strin« states havin«
a definite centre of mass momentum t,. . The photon staLes must satisfy the non-
trivial physical condition
(146)
where
~

L, = (ltp,. +L ~+1 (1-,,1' . (147)

,,:: 1
 83

Thit condition impliea that the polarisation vector satisfies the equa.tion

( 1(8)

What then, in tbe string theory, corresponds to the gauge t.ransCormat.ton

A,(r) ~ A,(r) + 8,A(r) (149)

of the photon field? For a photon wave Cunction

(150)

the gauge transformation of Eq. (149) becomes

~I' _ £,. + i.a,. (lIil)

where.\ = .\(t) it the appropriate Fourier component of A(%) . In string language,

this means the addition of the state

(152)

10 the photon state It) ofEq. (145) . The state aL_tIO) represenLl the change in the
state IO}. under & repacamet.criALion of !.he ~"ordiD&1.ee (11, r) of the atring WOtld-
sheet. It is an example or 8. nun physical state, being orthosooal 10 all physical
string states including itaelf. In (ad for any physical.tate ItP), satisfying Eq. (141) ,
we have
(153)
Thus the string .tates Ic} and 1£ + i'\J:} = It) + iAL_1IO) have the same physical
significance. The unphysical nature of 8. contribution to £,. of the form i.\.l:" and
Eq. (148) together ensure t.he disappearance, as physical states, of both time-like
and longitudinally polarised vector part.icles, just as in quantum electrodynamics.
It foUows that the ,au,e symmetry of the photon, in strin, theory, is just part. of
the reparamet.erisat.ion symmet.ry of the strin, formalism .
We should also emphasize here that the diffeomorphism symmetry of ~neral
relativity arises, in atrin, theory, in a similar way to sau,e symmet.ry. There is a
musle811 Ipin 2 l\.aloe in the dosed Ityin& lector. The existence of zero norm ltatea,
which decouple from the S-matrix, ensures ,eneral covariance and the identification
of this striQl state with the Uaviton.
Finally we ment.ion an • priori different way of obtaining gauge particles in
at.rio, theory, propoeed by Virasoro.3$ He considered a model having & condensate
of cloaed atrinp in the v&Cuum.
84

. . . . . . . .CtlS

I. H . Hu.n, Fundamental FGroeIt, Procnii.,. 0/ lie t1/Jr, Scotti,l U..i.crliliu S.""ncr Scl .. l
j" Pl,.ie •• St, ADdrew., p . 3.57 (SUSSP Publications, EdiJIbur&h Ulliftnity PreM, 1985), M.
Duff, P"cuii'll,' oj IAt 1I,'cMldin&l £.,..,1"j" Co-Jerncc "fI H i,ft E.c", Pl,.iu, Bui,
198.5, p . 611iI (European Ph¥akal Society, 1986).
2. G. 't Hoolt , Rec",1 Dnti",mc." i. G•• ,e rlr.uriu , p. 135 (Plenwn PreM, 1980).
3. P. ~,PI., •. Re,. nc, 1&5 (1981); P. Lancadlu, Proucii.,. 0/1I8S 1.lc...... tio ••l
S,m,o,i.m •• £C,lo. .,,; Phi ... I"'erod;,., at Hi,l E.r"iu, Kyoto, p . 186. 1986; P. Lan.
&&dtu, Ni.,l W.rJ:do, n G....; U.ific.1ioa. Ai.·L-B.h., 1988, p . 3 (World Sciem.ific,
1988).
4. G . G . Rou, Grn; U.ific. nconu (8eQ,jamin/Cwnminp, 1985).
S. F . WikzeJr. and A. Zee, Pl." , LeU. ItB, 311 (1979).
6. G . 't. Hooft, NuL Pl,•. B:U , 49 (1971).
7. W . A. Banleeo, A. J . Bur.. , D. W. Dub azul T . Muta, Pl,., Rn. D18, 3998 (1978).
8. Y. Totaub, Proccc;i." oJ 118S 1.'enl"i••• 1 S'III' .....'" •• Lc,lo. •• ~ PIa.'n I.tt,..eti ...
• 1 Hi,i E.rrrc#, K,-oto, p . 120. 1988.
9. H . B . Nm.m and N . BJ"me, N.d. Pi, •. B22.... 396 (1983).
10. J . We. and J . B-uer. S.,rr.,...rfFJ .. l S.,cfJ,.. ...". (prinoet~ UniYenity p.-.. 1983),
p. O. O. ~d, I"trol.clin 10 S.,n"""''''FJ (Cunhridp Uniwnity P ....... , 1986), P . Wett,
Iatrol.,I;,. I, S.,rr6fmmrlrp •• l S.'r",.. ...,' (World ScientiIic:, 1986).
11 . F. A. Berend., J . W . van HolLen. 8 . IU Wit and P. van Nieu-.-huaen, J. Pia, •. AU. 1643
(1960).
n . P. Howe and U. LindstrOm, N.d. Pia, •• Bill , 481 (1981); N . Muew: and A. Sapotti, N.d.
Pi, •. B250 77 (198$).
13. B . IU Wit and H . Nicolai, N.,I. P.,•. B20l, :ttl (11iMn).
14. J . EUl&, M. K. Oaillard and B. Z.. m ....., Pia, .. utL I14B, 343 (1i80); J . Elli&, M. K. Oaillud
_d B . ZumiDO. Ad. P.,.;" Pol•• ;" BU, 2$.3 (1982).
15 . A. C. Dayill, M. D. fho,em ... and A. J. ~,N-u:L Pia, .. 825',393 (19&$).
16. Th. Kaluu., SitsW!pbu. rn.-. Abel Wo.. Bedln, AI.,.. P.,•.
11. O . Klein. Z. Pl., •• n, §5 (ltnCl); O. KIeia. Nat." 111, 516 (lInG).
Xl . 066 (lint).

18. M. J . Duff, B. E . W. NiJ.oa. aDd C. N. Pope, Pl,•. Re,n1. UO, 1 (1986); T . AppelquU4,
A . Chodo. and P. G . O. Fr-euDd, M.ler. X.lou.·Xir;. nClFJ ••• A"iir.,i... (BatiIlQl.i.Q.
Cwnrninp. 1987) .
19. P. G . O . Freund and M . A. Rubin, Pl,• . UIL 9T8, 233 (1980) .
20. E . WilLen, Sirhrr 1.1... /1; P",uui.,••, I.r 118J Shhrr 1,1••• C.,.jr".u .. Q••• , • •
Fid. n~orp ••• t.h Fu •• m,.tal P"'/eIM
P ........ 1965).
0' P1,nel,n. R. Jacki. If .1. p . n7 (MIT

21. S. Raocijbat.-Oaerni. A. Salun..d J . Sw.lbdec, Nul. Pia, • . B214, 491 (1983).

23. L. AlyKa-G.ume and E. Wil'-o N.eL P.,•.

22. P . H. FramptQll and T. W. Kaphart, Pl,• . Rn. LrU. 60, l343, 1847 (1983).
B234. 168 (1983).
24. L. Alvara-Gaume and P . Gimpwa, An. Pl., •. tel . 423 (198li).
25. M. B. G..-, J. H. Schwan aDd E. Witlal., S.,crnn., nun, .. Volumm 1-..1 l (Cambridp
Uniynt)' p~. 1987}.
26.
:l7.
J. Th.ieny·M'" Pl.,•. LeU. lTlD, 163 (19l16).
A . N. 5dwIkk'" and N. P. W. . . . N.eL Pl., .. BUT, 317 (1987).
28. A . Bit..! and J . L . Oen-Ue, Pl,• • uU. lITB, 39 (1987).
29.
30.
J . H. Schwan. 1.,. J. MH. Pl., •. .u, $93 (1981) .
D. J . GI'OM, J . A. H_)', E. Martinec and R. Robm. Nul. P., •• B2~e . 253 (1985); N_eL
Pl." . B2eT, 75 (1988) .
31. t. B. FraUul and V. G . Kac. 1...,.,. MatA . 82. 23 (1980).
32. G. Sepl, Com",. Moll. Pl,•. 10,301 (1981).
33. P. Goddani and D. Olive. V,"" o,r... l.r... M.dllra.liu ... PI.,nu. py· .........inp of. CoD-
ference , Novanher 1().17, 1983, D . J. Lepo...k,. S. MaDdcktam and t. M. Siqer (SpriDcer.-
VerU.c. 198$), p. 51-
34. P. Goddani and O. Olive. 1.,. J. MI4. Pl, •. Al , '303 (1986).
 as

GoIdI,_.
35. P. Gocldud, J . C . Rebbi and C. B. Thom., N.eL Pl., •. lUe, 109 (1973) .
38. M. A. Vn-a, Pl,•. Lett uB. 436 (1979).
86

Chapter V

THE CPT THEOREM

The dearest example of a derived symmetry is provided by the famous CPT

theorem [Paper 23]. This theorem slates \hat any quantum field theory is invariant
under the operation CPT corresponding to the combined action of the three discrete
I

operators: charge conjugation C, parity P and time reversal T .

The CP T theorem has been proved in great generality' , and under very mild
assumptions. These assumptions, from axiomatic field theory, are:

(1) Poincare invariance

(2) MiclOStopie causality as expresaed by local commuLativity
(3) Continuity of quantum field operators - their matrix elements are "alten to be
tempered distributioJUl .

In fad only a weak form of the second assumption is required. This weak local
commutativity condition is the statement that the vacuum expectation value of any
product offi eld operators w(%) satisfies the equation
(Ol~(_,) ... ~(_.)IO) = ;F (01"(-. ) ... "'(_,)10) (\ )
when all the CCH>rdinate differences Xj - Xj are spacelike. Here F is the total number
of half-odd integer spin fields in 'he product.
The CPT theorem may also be proved diredly, in the more restrictive framework
of Lagrangian field theory. In this approach (Paper 23) , the most general form of
Lorentz invariant local hermitian Layangian density is constructed from the field
operatora. Aner appropriate normal orderin, and quantization with the normal
spin-statistics relation, CPT invariance is readily verified .
We will now outline the ideas underlying the general argument for CPT invar""
ance. Then , as an example, we will verify it, by explicit construction of the CPT
operation , for t he standard model Lagrangian density discussed in Chapter Ill.
One way of understanding the CPT theorem is via the related concepts of cr05&-
ing symmetry and the S-matrix 2 , which involves introducing Lhe asymptotic cond""
Lion necessary Cor the existence of in-stat.es and out-states. Any S-maiJix element,
 87

or scattering amplitude, is 3. function of the aternal momenta with analyticity prop-

erties following from assumptions (2) and (3) . We imagine continuing a scattering
amplitude analytic:ally in one or more of the external momenta PI" but keeping
the external particles on mass-shell, i.e. P; : ;: m', Initially we consider spinless
particl~ . By going into the complex plane, it is possible to continue from values
of P", having positive energy, Po > 0, to negative energy, Po < O. The resulting
new on-shell amplitude can be related to the residue of poles in a vacuum expec-
tation value of a product of fields . It is then , in general, possible to argue that
these poles must correspond to physical particles, whose scattering is described ,
kinematics permitting, by the new analytically continued amplitude. The particles
with Po < 0 have to be interpreted as incomin& instead of outgoing (or vice-versa),
and consequently must carry opposite additive-conserved quantum numbers such
as electric charge. Hence we are led to ~he existence of antiparticles.
We have thus crudely argued for crossing symmetry in the following form. By
analytically continuing a scattering amplitude in the external momen~a , so that
some of tbe particles have their energies Po turned ne&ative, a scatterin& amplitude
is obtained for a new correspondin& set of particles. In this corresponding set,
those incoming (outgoing) particles having negative Po are replaced by outgoing
(incomin&) antiparticles with opposite values of the four-momentum.
Once we accept the above crossin& relation , we may obtain the CPT-transform
of a given amplitude by taking a special analytic continuation, which corresponds
to a continuation in the parameters of a proper Lorentz transformation . By going
through a sequence of complex Lorentz transformations cOQneded to the identity,
it is possible to &enerate the transformation :t~ -- -%~ or p~ __ -p~ for all the
particles . Any Lorentz invariant amplitude is of course unaltered by such an analytic
continuation . It has been proved 1 tbat the analytic continuation, p~ - -p~ , is
possible for any scatterin& amplitude, if wealc: local commutativity holds.
We are thereby led to the CPT theorem for spinless particles : The scattering
amplitude for a set of particles is equal to the scattering amplitude for the corre-
spondin& set of antiparticles, but with all incoming particles replaced by outgoing
particles and vice versa.
Let us now consider the generalisation to particles with spin . We must take
into account that, under a Lorentz transformation , the various spin componenLs of
a particle are transformed into linear combinations of each other. The analytically
continued proper Lorentz transformation A, Jeadin& to p~ - -p~ , can be COD-
sidered as composed of a boost in, say, the %t-direction, by an imaginary rapidity
i7r, followed by a rolll.tion about the %l-axis by an angle 7r . The Lorentz boost is
illustrated in Fi&. 5.1 , where we show the effect on a unit time-like 4-vector (e.g.
the 4-velocity of a particle) , VP = (1 , 0,0, 0), of a boost by a rapidity 4J in the
%l-direction. For any real value of 4J , the boosted 4-vector VP lies on the upper
branch of the hyperbola drawn in the figure . However if 4J is allowed to take on
complex values, then the components Vo and V 1 are continued into the complex
plane and can end up on the lower branch of the hyperbola. In particular for purely
88

imaginary values of ¢ . the boost actually carresponda to 8. rotation in the (Re V o,

1m Vi ).plane by an angle ?/i. The {-vector Y" then runa tbrou,h the dashed circle
drawn in perspective in Fig. 5.1. It follows that a boos\ by a rapidity? = i. in the
%l-diredion changes the signs of Po and PI. leaving P2 and P3 unchanged. If this
boost is followed by & rotation of". about the zl-axis, ~ and P3 a.lso change sign
and the full transformation A lew to PI' - -P" .

-+-_R.V'

Fil. S.I. Uh..'ralKm of the .ipo", _ _ clthe yO_component cI. unit time-lib +YedOl" Vii, by
analytic c:oQtmu,'j,m 01 .. ~. t.oo.t to oompMlI ...a- 01 the rapidity ,_

The above complex Lorentz transformation A it easily shown t.o leave the spin
&nlUlar momentum of a particle invariant. For inalance, the result immediately
follows if we choose the zl-axii along the direction of the spin component under
considerat.ion . So, under the complex transformation A, the spin of & parti.cLe it
unchanged wbile ita 4-momentum PI' changes sign. Hence t.be beHcity of a particle
also changes si&n, l. _ -l., under tbe transformaLion A. By crouint;, we now
replace incoming (outgoing) particles by outgoing (incoming) antipartieie. with the
oppOflite sign of 4-momentum and spin angular momentum. So each antiparticle
carries the same: 4-mome:ntum PI" but the oppcait.e helicity -l., La that. which the
corresponding particle had before applying the transformation A. Thus, ignoring the
phase factors which arise: in a precise definition of particle and antiparticle he:licity
states, we are led La the equality of tbe scatterin, amplitudes for two prOCf"'u.
related by croesing all the particles.
For example, tbe sUUerin5 amplitude for the process

(A , P A, 'A) + (B, p., ,,) - (C, P c , ' c) + (D , PD , 'D ) + (E , P B,'B) (2)

 89

it t.he lame as t.he scattering amplit.ude for the process

(C, Pc, -'c) + (lJ, PD, -'D ) + (E, PE, -'E)

-(A,PA ,-'A) + (Jl",p. , -,.) . (3)

Here we have used the notation (X, P X , ~x) for a particle X having J...rnomentum
P x &nd helicity AX and X denota the antiparticle of X. This result ea.n be written
in terms of elemcnt.ll of the S·matrix

(4)

up to a helic.ity dependent phaae factor, as fonows :

(C, PC. AC; D,PD • .lD i E , PE • .l&tTIA,PA • .l A ; B,PB • ..\S)

= (A,PA,-AA i B,PB ,-.\sITI<;,pc.->'c; ~, PD'->'D ; ~, PE ' -).B) '
(5)

Simply writing the right.·band side of Eq. (5) in terms of its complex conjugate , we
can express the result &I a relaliooship between amplitudel, for which the antipar.
tides corresponding to incoming (outgoing) particles are also incoming (out.going):

(C,Pe . le ; D,PD,AD ; E ,PB' , ASITIA,PA ,lA ; B ,PB,ls)

= (e,Pe,-lei D, PD , -AD; E,Ps,-l.slrtl
A,PA ,-'" Jl",p.,-,.)' . (6)

This is the expression ahbe CPT theorem for 1-he proc:ml (2) and (3) .
By definition, t.he CPT operat.or

S=CPT (7)

has t.he following act.ion, up to a phase factor , on a lingle particle helicity state

(8)

The moment.um PAis left. invariant, because it is inverted by bot.h P and T . The
spin is inverted by T and unchanged by C and P . Bence the helicity .\A is inverted
by 9 . H ls, of coune, C that. transforms t.he particle A into ita antiparticle A. The
combined operation, 9 = CPT, may exist even when the individual operatora C,P
and T are not defined.
The .ynuneuy derived by applying the complex Lorentz transformation A and
crouing to a scattering amplitude, for general multip&rticle states Ie} and 16), can
t.herefore be writ.ten in the form

(bITIQ) = (Sbl:r+ISQ)"
= (SQITISb) . (0)
90

This invariance property

ere-I =7* (10)
of the S-matrix is the general expression of the CPT theorem. The hermitian con-
jugation, which appears on the ri&ht-hand side of Eq . (10), Ls due to the antiunitary
nature of e.
In place of tbe above general argument, based on the analyticity properties of
axiomatic fi eld theory, the CPT theorem may abo be derived by considering t.he
most general form, a Lorentz inY8J'iant local hermitian Lagrangian density .c( ",,(x , t))
can take. It is then verified [Paper 23] that .c(~(x , m satisfies
C(¢(x, t» = C (¢·(-x, -t» ( 11)
and the quantised field theory described by £ (¢(x , t» is therefore CPT invariant.
w
Dere tP(x , t) is a generic symbol for all the fi elds and e ( -x , -t) denotes the CPT
transformed fields. The. on C! denotes complex conjugation of all the constants
in the Lagrangian density, and its presence is a consequence of the antiunitary
nature of the CPT transformation . The half-odd integer spin fields are 1.aken to be
anti-commuting Grassman variables, as in the path integral formalism .
As an illustration of the second approach, we now brieBy consider the standard
model Lagrangian density, which consists of five terms.
(12)
as discussed in Chapter III. Using the no1.ation ofCbapter 111 , we list below the CPT
transformations oft.ypical Bose and Fermi fields selected from Table 3.1. The CPT
transforms of the SU(2) gauge fi eld A"hx, l) , its field strength tensor FI'''ii(x ,t)
and dual field strength tensor FI'" ,J(x, l ) are given by
A:/ (x , t) = -A,./( -x , -t) (13)

i!'../(x, t) = P,. ../( - x, -t) (14)

-e . - .
F,. .. /(x ,t) = F,. ..;'(-x, -t ) . (15)
The left-handed quark field ql iO'aL(X, t) and the right-handed Q = 2/3 quark field
u~aR have the following CPT transforms:

q' 8 i .. a(.(X , I) = i1'°1'6q'TiO ad -x, -I ) (16)

_'eia
q aL (x, t) = Iq
. 'r ;O'aL ( -x , - t) l' , l' , (17)

U '. ( t ) = 11'
..aR X , ."'Y U- 'To aR ( - x , - t) (18)

U 90' aR ( X, t)
_' . ' T O'aR ( -X,- t )1'1'
= IU , , . (19)
 91

Here the superscript T on a spiner denotes its transpose . We note that (/8 i",,,L(X, t)
is a risht.-handed antiqulU'k field and u'e Q.R(X ,t) a lefi·handed antiquark field , as

.
required by Eq. (8). As a final example, tbe CPT transformed Higgs field tPi(X, t)

(20)

.""(x,') = .;( -x, -.) . (21)

The CPT transFormations of the other K&uge, quark and lepton fi elds in Table 3.1
follow , from these examples, by analogy.
It i. now readily seen that, under the above CPT transformations of the fields
and the reversal of the space-time co-ordinates ,

(22)

the standard model Lagrangian density, Eq. (12) , satisfies the condition, Eq. (11),
for CPT invariance. The CPT invanance of a r;eneral Lorentz invariant local ber-
mitian field theory is similarly proved , by explicit construction, in Paper 23 .

R.eCeteDCn
1. R.F. StnI!ater and A.S. Wiptman, PCT, S,ill .. J $Iat;";,,, ... Ii All Tl., (8erUamin/
Cwmninp, 1964; -=ond primin" with addition. and QOQ'CC;t;CIN, 1978).
2. R.J. Eden, P.V. Landthotf, 0.1. Olive and J.e. Polkingaome, Tlc A •• I,I;c S.M.lriz (C.....
hrid&e Univen.ity Preu, 1966).
92

Chapter VI

THE FUNDAMENTAL SYMMETRIES

6.1. Introduction
Poincare invariance and &.ule invariance are the basic input symmetries of the
standard model. These syrrundries au usually considered to be t.ruly fundame.n1al
and not derivable, even in theories beyond the standard model. For example, they
are both input UIIumptiol1l in Grand Unified Theoriet and Poineare invariance
is needed in the proof of the CPT theorem. Therefore, Crom the conventional
standpoint, Yang· Milla gause invariance, Lorentz and traJlllation invariancc are all
expected to be valid at the top of the quantumatairC:Ne (Fig. 1.1) . In thi.. chapter we
consider the alternative point of view I according to wbich even these "fundamental-
symmetries can be derived, albeit at a rather hi&h point on the quantum staircase.
We have already touched on the possibility of the dynamical ,eneration or a
gauge symmetry in our discu.ion or Papers 17-19 in Chapter IV. We have abo
diac.uued , in Chapter IV, the oripn of gause invar1anee in Itrio, theory. A third
approach to derivin, ,ause invarianee w the renormalilalion goup method, in which
the symmetry becomes a better and better approximation u one ,Deldown the
enere aeale (or quantum staircase). Theee three approachel have alao been appl~
to the derivation or Poincare invariance.
We claaaify the attempts to derive Poincare invariance and ,au,e invariance in
Table 6.1, accordin, to tbe IOllowin, three methods:
1. Formal Appearance. Derivations or this type reveal the synunelry to be a purely
formal one, which nevertheleu attainl physicalaipificance, in some phase or the
vacuum, due to quantum fluctuations.
2. Renormaliaation Group. In thw method a symmeirY. which is not . . umed
valid at ahod disla.nces, becomea more and more acc!urate at lu,er and larser
distances. In other worda, the renormaliaation UOUP ,8-funct.ioaa imply the
luppreaaion or Iymmetry breakin, terDll towarda the infr..red.
3. Stria, Theory. This method applies to Iymmetries which arise, without bein,
 93

Table 6.1 . C1aai6ca.tion of methods for cleriviq poiftQft in¥ariaDc:e and pup ~.

~
Poin.caR lnvariam:e Lou!
Lo~l:l Translational G .....
M.<hod In¥Ni-.no;e Invariance Invari&Dce

Special caM or Special caM of

Fonnol P."...
repwametmaation repuameteriaation
AppearaDCe 11,1',lhnd"
InVanulce m~

Ra.onn.alisation P....,. Chapter VI P....,.

G~p 24,2S and 29 Sea.... 2.3 28 and 29

Stein, Chapter VI CMpte:r IV

T....." Rererence 8 Section 6.2

explicitly put. in, as a eonsequence of other symmetry properties of string theory.

We now consider, in tUtn , these derivations of Poincare invariance and cauee
invariance. This chapter concludes with a brief discuS&ion of attempts to derive
8upersynunetry by the renormalisation group method .

6.2. Poincare Invariance

In this section we discuss the various methods proposed to derive Poincare in·
vanance.

6.! .1. Formal ClppeolUnce of Poincare invanance

Poincare invariance can b e considered to be a consequence of general cc;Qrdinate
invariance, and the absence of "prior r;eometry" ,I in gravitation theory. By ·prior
geometry" one means! any aspect of the geometry of space-time that is fixed im-
mutably, i.e. that cannot be changed by changing the distribution of gavitating
sources. A "prior geometry" is an example of an "absolute object" , which is defined 2
to be a function , with one or more components, of space-time that is not dependent
on the state of matter.
An example of an absolute object is the concept of absolute time, as introduced
in Newtonian physics. In this example, the physical theory contains a real number
funct ion [O (z') of space-time, whkh specifies the lime in an absolute sense. The
value of the absolute time [0 (:rJ') is not affected by the presence ofmatt.er, but mat-
ter can be influenced by absolute time through the physiea.llaws. For instance, one
could imagine constructing a clock, which displayed the absolute time independent
of its state of motion. Relativistic physics, of course, does not allow the introduction
of absolute time. Any absolute object in relativity theory must be invariant under
Poincare tranBformations.
The action for any physical theory can be written in a reparamet.erisation invui--
ant way. This formal diffeomorphism symmetry is, in general, obtained at the ex-
pense of introducing absolute objects into the theory. The special diffeomorphisms,
eoneaponding to Poincare tranaformations, only generate a true physical Poincare
symmetry if the absolute objects are invariant under these transformations. The
basic idea, for deriving Poincare invariance from diffeomorphism symmetry, is then
to find dynamical , quantum mechanical argumentl ror potential absolute objecta
being Poincare invariant or non-existent.
Hiatorically, of eourse, general relativity wa.s cOnltructed with Poincare aymme-
try in mind. However, in the apiri" of the preaen' book, we can take the point of
view that general relativity arises from some pregeometry model,l,3 without assum-
ing apecial relativity from the outset. For the purposes of our initial discussion, we
temporarily aaorne the absence of non-Poincare invariant absolute objectl. We can
then take either of the fonowing two attitudes towarda Poincare symmetry:
1 Poincare invariance is just a aubgroup of the formal reparamet.erisation in-
variance group of general relativity.
II. Poincare invariance is an approximate symmetry, vaJid when the gravita-
tional field is neglected.
Aceording to attitude I, Poincare invariancc ill a purdy formal symmet.ry of
general relativity under the special cCKlrdinate transformation

(1)

The two co-ordinate ayatellll zI/I and zl' are, by definiti on, related to each other by
a Lorent.a transformation A: followed by a tranalation 0./1 . Thil iI illustrated, fo r an
arbitrary co-ordinate synem z/l , in Fip. 6.1 and 6.2. The cCKlrdinate systems zl'
and z'/I are in each case denoted by full and dashed curves respectively, while the
eo-ordinate axes are drawn in heavier print . In this attitude of Poincare symme-
try being just part of diffeomorphism symmetry, one only ha.s Poincare invariance
provided the gravitational field is tran.sformed as well a.s the matter fields .
Accordin& to attitude lI, 'he maner fic.ld.!i Me transformed under the Poincare
traJUlformation, but the gravitational fieJd is left unchanged. There iI then, of
couree, only Poinca:e symmetry to the extent ~hat the gravitational fie1d can be ne-
glected, i.e. in the approximation of flat apace-time. If, however, space-time is flat.
and we choc:.e e.;ordinates z/l 80 that the metric tensor is constant as a fund ion
of zl', then diffeomorphiun ayDll1letry can be used to derive Poincare symmetry.
This means that, once we have a reparameterisation invariant theory, the question
of deriving Poincare syJlllIletry, in attitude II, has been transformed into the c0s-
mological question of why space-time is almost flat nowadays. This appraximate
flatness is due to the Hubble expansion, having scaled up the size of the Universe
by a big factor.
Let ua now consider rotational invariance, as observed in daily life. It would seem
that the up-down axis is singled out, and that there is only rotat.ional invariance
around a vertical axis. For instance, a pendulum suspended (rom a support will
direct it.eelf downwards, after it haa had time to lose any kinetic en er~ via fri ction,
independent of the orientation of the support. So rotational invariance is broken,
 95

Fi,. 6.2. T",o co-ord.inale ')lItem. relI.ted by • tl'anaWion.

with the up-down direction being special, in attitude 11. It is possible to formally
restore exact rotational symmetry by taking attitude I. In att.itude I, when rotating
the apparatus, we must also :remember to rotate the gravitational field and the earth
creating it.. This is achieved by a diff'eomorphism of t.he form z'I' = A~z" which
rot.atea everything physical relative to the c(>-ordinate system.
The above remarks are illustrated in Fig. 6.3 by di&8:rams of a simple pendu.
96

lurn experiment. When the supporting stand is rotated about ill. horizontal axis,
rotational invariance appears k) be broken as shown in the top two diagrama. It
is easy to introduce ill. formal rotational symmetry; ODe invent.s the existence of a
gravitational field - symbolised by a little demon in the bottom two diagrams -
and rotates it as well as the lupportin, stand. Rotational invariance of the law8 of
nature is thereby formally restored.

Fie. 6.3. The top "'''0 dlacr-- illu.tr&te the lack ci rotational. invari&nce ill daily life, by meNU
01 a pendulum upmmf!I'J.t. The bottom two diacraml ah.ow how l'OC.at.ianal invariaDa! • _W'ed
when the pavitational fi.o!ld, ~t.ed by • little d _ •• rotat.ed .. well .. the pendulum.

AJ long as we aruy cORsider the ullual daily life situation, we ean ignore the
idea of asaodating the sravitational field with the earth . The gravitational field of
everyday experience can be treated as an immutable property of our region of space-
time , giving a constant downward acceleration. The introduction of a gravitational
field of th~ type, uninftuenced by the presence of matter, means that we now bave a
non-rotational invariant absolute object in the theory. Rotational invariance would
tben indeed only be a formal symmetry, introduced basically by definition . One
simply postulates a field , or little demon , giving tbe rotational symmetry breaking
effect. Tbe existence of rotational symmetry becomes very mucb a question of
semantics. By identifying a symmetry breaking effed with a field in this way, one
can formally introduce synunetries wherever one wishes.
However if we view the situation from a larger distance scale, we immediately
recognise tbat the gravitational field is due to the earth. The gravitational field
would be modified by moving the eartb. So the demon, i.e. the gravitational field, ~
not really an absolute object but a dynamical object, determined by the distribution
of matter . We can tberefore conclude that the formal rotational invariance, in fact,
eorresponds to a true physical symmetry.
Analogowly it is quite easy to introduce Poincare invariance formallY i it simply
 97

corresponds to ~he invariance or the action , under the special diffeomorphism of

Eq. 1. However irtbe theory was not truly Poincare invariant to begin with, there
would now exist a non-Poincare invariant absolute object in the formalism . In
this case some relians of apace-time would have specific properties, which could be
described by introducin, a set of II priori physically si&nificant co-ordinates, [II aay,
as aeet offour sealar fields . These scalar fields 1-(%'), a = 0 , 1, 2, 3, would depend on
the co-ordinates z, in an arbitrary co-ordinate system. Then any dependence of the
Lagrangian density upon the fundamental co-ordinates 1-, which in the fundamental
lansuage breaks diffeomorphism symmetry, just becomes a dependence on the fields
I·(z,) . We thereby formally construct a diffeomorphism symmetric Lagran,ian
density.
The pretence, or ahsence , oCphy1Jical Poincare invariance now be<:ome. the que..
tion of whether, or not, the absolute obje<:t 1·(z') is in a Poincare invariant state
in the vacuum. Here we asaume the vanishinf!; of the cosmological constant and
take the f!;ravitational vacuum to be 8at Riemannian space-time. If 1·(z,) is not
Poincare invariant in the vacuum, it will act as a source of spontaneous symmetry
breakdown . So the question is really whether or not the formal Poincare symmetry
is spontaneously broken . A natural way for such a spontaneous breakdown to occur
would be for the scalar fields 1·(%'} to take on vacuum values

(2)

in some co-ordinate system, where the metric tensor g,. .. (z'} is approximately con-
stant. This does not occur in standard f!;eneral relativity theory, where any scalar
fields present take on constant values in the vacuum.
A true derivation of Poincare invariance must therefore explain the absence of
fields, like 1·(z') , taking space-time dependent vacuum values, such as in Eq. 2.
There appear, at first , to be two p088ible ways in which this could happen :
Ca> The fields 1-Cz'> may take on Poincare invariant (i.e. constant) value. in
the vacuum.
(b) The fields J-Cz'} may quantum fluctuate with the same distribution for all
.'.
The degrees of freedom correspondinf!; to the fi~ds 1-(2:') have to be interpreted as
some matter def!;ree5 of freedom , e.g. four , as yet to be discovered, scalar particles.
P088ibility <a> would seem attractive, since scalar fields can easily take on con-
stant value. in the vacuum. However, if the fields 1-(z' } are to be interpreted as
"fundamental ~ordinates" at some level, the ordered set 1-(z'} should not take
on the same value more than once:

(3)

One could imagine that 1-(z'} fluctuates quantum mechanically over a broad-
peaked distribution, without taking the same value twice. Then the average could
98

both exist and be constant. This would require the existence of weak conel.Liolll
between the valuee of ]"(z') at different pointa. There could be a problem with
topological obstructions, in constructing such a probability distribution on a set
of continuous (undions 1"(%'). Thia problem with topolosy can be avoided , by
introducing a fundamental space-time lattice instead of fundamental apace-time
co-ordinates. S
The above picture for possibility (al relies 10 much on quantum fluctuations
that it differs very little, physically, from possibility (b) . However the quantum
fluctuations , in possibility (b), may be so large that it is not possible to assign
well-defined average values to the fields ]"(z') in the vacuum. This essentially
occurs in the lattidsed quantum Vavity model of Ref. 3, in which the supposedly
fundamental lattice lites (and linu) are dynamical variables. The sites fluctuate
quantum mechanically over a Riemannian space-time manifold, in the functional
integration, like tbe molecules of a gas . The fields [4(%') must be interpreted, in
this model, as numbers specifying tbe name of tbe site, which happens t.o be present
near tbe point %' on tbe manifold. One of the ma.in conclusions of Ref. 3 is the
uistence of a phase, in which the sites fluctuate , with a flat distribution, allover
.pace-time. This means that the [ 4(%') take on extremely uncertain values in the
vacuum, but the probability distribution for the [4(%') is the lame for all %' . Thus
Poincare invariance is derived in this pregeometry model.
It is now natural to ask why only Poincare invariance is derived in this way,
and not any other part of d iffeomorphism synunetry. The vacuum values of the
fields [4(:1") may indeed be fuUy reparameterisation invariant, but the vacuum
state it.self is not. The metric 91''' of the flat space-time vacuum of general rela-
tivity is invariant under precisely the diffeomorphisms corresponding to Poincare
transformations. The full diffeomorphism symmetry of the general relativity action
is thereby spontaneously broken down to Poincare symmetry. Indeed the graviton
may be considered to be the Goldstone boson associated with the broken GL(4 , IR.)
group of transformations .·
Recently an attempt was made to give a more explicit derivation of time transla-
tional invariance as a formal symmetry.' In this approach the original time vaziable
of the time non-translational invariant model is made into a dynamical variable,
which can essentially be identified with the variable 1 4 (%') discussed above . How-
ever, a severe problem arises with this treatment of time as a dynamical variable
since, as pointed out by Pauli , there is then no lower bound to the Hamiltonian.
This problem is not manifest, and therefore easily forgotten , in a Euclidean formu-
lation.

6.!. ! . Lorentz invariance from tAe renormalisation 9ro.P

In this subsection we consider the derivation of Lorent% invariance as an approx-
imate symmetry which becomes more and more accurate t.owards the infrared , in
contrast to the exact formal Lorentz invariance of general relativity. Lorentz invari-
ance can be shown to arise as such a low ener&y symmetry for non-covariant quan-
 99

tum electrodynamics (paper 24] and non-covariant Yang-Mills theory [paper 25J,
The main assumptions made in these DOD- Lorentz invariant field theory models are
translational invariance, PUJe invariancc, chiral (massless) fermions and renormal-
isability. Scalar fields are ignored on the &rounds that they generically become very
massive and therefore irrelevant Lo low cnerO' physics, unless protected by some
additional syrrunetry. A similar mass protection principle, for Iilht fermion8 , was
used in Chapter 3.2 to exclude observable fermion slates which are vectorlike under
the standard model gauge group S(U(2)xU(3)) .
Under the above assumptions, tbe most &enual allowed form of local Lagrangian
density has terms of the following type.
a) The kinetic term for gaule fields in an anisotropic vacuum, baving mapletic
susceptibility and dielectric properties depending on direction and Lorentz
Crame:
c. = -i'1""""'F" .. CZ)F,u(z) . (4)

Here F" .. (z) is the &au&e field stren&tb tensor and the 256 componenLs of
Jf"'" are dimensionless couplinS constanLe, wbich do no~ transform under
Lorentz transformation. There ace just 20 independent componenLe of rf"'"
which effectively contribute to C• .
b) The gauge covariant kinetic term for each (lefl.-handed) Weyl fermion field
tPP = !(l- ls)¢ p :
C. = i¢Pl"VJP)"D",pp . (5)
Here P denotes the fermion flavour and, for Vans-Min. theory, tPP is a1ao
an irreducible representation of tbe &au&e &roup. The vierbein vJ P )" for tbe
fermion of flavour P i. just a llet of 16 dimensionless couplin& constanu. The
&au&e covariant derivative is

D" = 8,. - iA" (6)

-
where A" is the &auge field in the appropriate representation of the Yang-
Mills uoup or, for the Abelian case,.4,. = qA,. and q is tbe U(l) chac&e of
"'p . A motivation for usin& IUch Weyl fields ",p , without Lorentz invariance,
is pven in [Paper 30).
c) Momentum shift terms for the fermions :

£ - .,. roy(P)"eP.I.p (7)

c - <;'P " "" <;' •

The four coupling constants {: here have the dimension of mass. There
could also be flavour mixing terms of thiB type.
d) A topological term, oripnally IU"ested by C .H. WOO, of the Chern-Simons
type:
(8)
100

The four couplin& constants q,. also have the dimension of mass.

On purely dimensional ,rounds we should expect t.he constants e:

and '1,. to be
proportional to lOme fundamental mass, presumably the Planck mass. Therefore,
unlese we can find arcumenta for edtin, them to zero, the terms Cc and £~ would
completely dominate £. and .c. in the low enerto)' regime and Lorentz invariance
would not. be obtained . In fact, the fermion momentum ahin terms C.C are easily
removed, by an additive renormaliaation of momentum achieved with simple field
redefi.nitiona of the type

(9)

The fermion kinetic enetK)' term C•• when written in terms of ¢'p , develops a term
which cancels £c o Thus C c can be simply transformed away and ipored. The
term C~ can also be forbidden by appealing to gause inva.riance, provided that we
aaaume the exil~ce of mq:netic monopoles. ThiB arfOUmenL is explained below.
Thus we should really include the exiatence of m&«netic monopoles as one of our
input assumption.; otherwise an alternative method of excluding Ctl is called for ,
We now explicitly show, for the Abelian model, that the non-eovarian.. Lapologi-
cal term £. is not laule invariant in the presence of a non-zero malnetic monopole
current deDJ!lity J&.-c ' The dual field tensor

(10)

then satisfies the equation

(11)

Let us coDJ!lider the variation of the action

(12)

corresponding La the topological term

~
~. -- '1,.'"""A •F,,, (13)

under the gauge transformation

6A,. = lJ,.A , (14)

 101

We obt.ain

(IS)

( 16)

(17)

(18)

(19)

So t.he act.ion 54 is not gauge invariant and is therefore forbidd en if t.he t.heory
contain. mapet.ic monopoles.
We the~fore drop the terms C c and £4 . leaving us with t.he model non-covariant
Lar;ranpan density

t:. = -itt'''. . FI'"F.... + E itP p.,oVJP )pD,.t/.>p

p
(20)

whith is formal ly scale invariant. However, as discussed in ChapLet 3.3, quantum

anomalies cause a breakdown of scale invariance and require the renormalised cou-
pling constants to depend on the enerl)' scale ~ . Thus we introduced renormalised
or effective coupling constanta ~;'''P) and v~2"p). which depend on the renor·
malisation point A. There is a Jarge degree of freedom in the precise definitions of
these effective coupling constanta. However, to leading order in tbe gauge coupling
constant. g2(l.), the expressions (or

IJI""" = drf~"'(~) (2 1)
• d logl

and t.he corresponding ,8-functions (or the seale dependent vierbeina v~l"(.\) are
unambiguous. Explicit calculation of the p..fun ctions (Papers 24 and 25) showl that
the effective coupling cooltants approach Lorentz covariant values more and more
osely, as the energy scale .\ is lowered.
The physically significant content of the vierbeios ia provided by the associated
ormalised metric tensors

(22)
102

"hue

-~ j)
I 0
o~ _ 0-1
1J - ( 0 0 (23)
o 0
Note that each Weyl fermion field ~p(z ) has i'.
own scale dependent metric
9~)"¥P.). It is also possible to define a mdric g~)"'''(A) for the gauge field, by
requiring it to minimise the expression

(24)

where II ... 11 is • norm on fourth rank tensors and

'I'''';:-(l); 29,'(l ) [g~ )"(l)g~)"(l) - g~~)"(l)g~)" (l)1 (25)

The metric g!~)""(.\) is normalised so that

detg~)""( ).) =-1 (26)

and the gauge coupling constant g2p.} is free to be adjusLe<i in tbe minimisat.ion. We
then denote the deviation from covariance, with resped to this metric, of rf~"'(.\ )
by
(27)
In t he cue oC massless non-covariant electrodynamics there are three metrics:
one each for the photon field and the lefl.-handed electron and pOIitrco Weyl fields.
Introducing the notation of [paper 24J, we have the photon metric

(28)
the electron metric
(29)
and the positron metric
(30)
For Lorentz invariance, of course, we require the three metrics to be identical and
the non-covariant part of the photon coupling constant.s 6!f"' '' (l ) to vanish. 1\ is
therefore convenient to introduce the deviation metria

5g~'(l); g~'( l ) - ';'(l) . (31)

In this notation, the renormalisation group equations become

8g'"(l)
P'"(l);
h
::a-
' -¥
8 Jogl
(32)

; -·i:) [6gnl) +5g~'(l)1 (33)

 103

and

fJ"~ = 8g~"(l) (3')

U (Hog).
_ '.(l)! "(') (35)
- 3r 9:1: A •

Hm
1
.(l) = •• .'(l) (36)
is the running fine structure constant in non-covariant eledrodynanUca .
It is easily seen, from the signs in the expressions above for tbe p..functioD, that
the various metrics approach each other as the renormalisation point goes towards
the infrared , Le . as the energy seale>. - O. This is illustrated in Fig. 6.4 , where
the planc symbolizes the space of all normalised metric tensors. The development
of the photon, electron and positron metria as >. decreases, due to renormalisation
group effects, is indicated by arrows. In the infrared limit all three metria meet at
the weighted averat;e
~(',nl) + ,~'(l) + ,~'(l)J . (37)

Fi,. 6.4. Approach 01 the thHoe IMtne. , ~ " (1), ,':" (A) and .:" CA) towU'da the Loren"
invariant in!r&ftld fj"od point, in DOn-<:ovariant electrodynamic:..

Also it follows , from the expression for 11:"'14 in (paper 24] , that the non-covariant
part 6'1""'" (A) of tbe coupling conatants tends to zero relative to tbe covariant part
1fc!!'''P) in the infrared limit. Thus the ,D-fundiona show that Lorentz invariance
is simulated more and more accurately as the energy scale is lowered.
When various non-covariant Yang-Milia fielda are included in the model (pa-
per 2:)), different metrics are introduced for each simple gauge group, by minimisinl
104

expressioru analogous to Eq. (24). For example, the metria g~~ t::t\P.) and g;~ 3 p.)
would be derived (rom the (ouplinK constanLl '1!';~;'(2)(.\) and 1J~~U(3)(.\) for ~a(2)
and SU (3) gauge groups. Provided that all the various gauge groups couple to some
common Weyt fermion , all the metria approach a common limit in the infrared .
The ,8-fuDctions also show that the non-covariant part 61'/~"~(.\) of the coupling
constants '1:';:~(A ), for ea.ch simple gauge group G, becomes smaller and smaller
towards the infrared , relative t o the covariant part C~(.\) .
We have thus shown that if Lorentz invariante breaking is appreciable at. some
very high energy scale, the breaking will be much smaller at the low enerIY scale
where present-day experiment.a are performed . We would, of course, like to assume
that the coupling constants 1/;"'(1 and yJP)" of the non-covariant model are all of
order unity at t.he Planck m.au seale. The renormalisation group development of
the couplinS constants must then ensure Lorents invariance, to the observed hi&h
accuracy, at low energy. This requires the ,8-functions discussed above to be of order
unity, 80 that they can act effectively over a larse ranse of energy scm. In other
words we require a strongly interacting theory over many orders of magnitude of
energy, 80 that Lorentz invariance improves a~ an appreciable rate 118 l goes down
the energy scale. The gauge group fine structure constant O'G(J.) mu.d therefore be
oCorder unity over a similarly large range oCenergy scales. This in turn requires
the Yang-Mills p-function , determining the l dependence of O'G(l), to be small,
which can be arranged by introducing an appropriate number of Cermion field. .
There can then be an almost exact cancellation between gauge boson and fermion
contributions to the Yang-Mills ,8-function . However it is not p088ible to achieve
the required degree of Lorents invarianee in this way in the standard model, Cor all
three gauge groups U( I), SU(2) and SU(3) simultaneously, by simply introducing
extra generations of quarks and ieptons .e In particular, for the U(I) gauge particle,
we would then expect Lorentz invarianee breaking effects to be of order 0' ~ lS"
in obvious conflict with experiment. It seems that extra structure, in the form of
new particles and new interactions, is required between presently available energies
of 100 GeY and the Planck mass of 1011 GeY, if the hi&h accuracy of Lorents
invariance is to be explained 118 a renormalisation group effed .
The renormalisation p-oup method derivel Lorentz invariance as an approximate
low energy symmetry, in contrast to the exact formal Lorentz invariance obtained
Crom general relativity. It is thereCore natural to uk why non-covariant coupling
constants, such as rf'''''', do not occur in gravitation theory. Their absence from
the sravitational Lagrangian density, in fact, follows Crom the assumption that
they should be comtant over space-time, i.e. translational invariance. In general
relativity, this means that they should be covariantly constant and obey

D, rf"'~ = 0 . (38)

The Christoffel symbols r~", occurring in the ,ravitational covariant derivative D,. ,
are defin ed to make the metric tensor g"" covariantly constant. It is then, in general,
 lOS

impoeaible to satisfy Eq. (38) in .. curved space-time, unless rf"" takes the usual
COVUlant form proportional to (9"'9" - g"" g .').

6.!.9. 7hln61olional invariant!!! from dimcfuioAai .naiy"i"

In this subsection we describe the rudiment.a of an argument fOf deriving transla-
tional invariance, as a low enerl)' symmetry, analogous t.o the renormalisation group
derivation of Lorentz invariante. The argument is r ..tbet formal. but we have in
mind an amorphous or glasa-like model for the translation ally non-invariant vacuum
state . A glasa-like spacdime, of course, is only chaotic at sbort distances and has
t ranslational invariance built into its long distance structure. So our "derivation"
really only abows that it is possible for ITIACf06C.Opic physics to be translation ally
invariant even though microscopic physic. is not.
In a real &lass, it is wen known that light propagates, to a very good approxima-
tion, as if there was no b~akins oftranalational invariance. Any lack oftran alational
invariance manifests itself as a breakdown of momentum conservation. The essen-
tial ~int in our arsument is that couplinS constants associat.ed with interactions
which break translational invariance have, in natural units , a mass dimension lower
than the correspondins momentum eonserving int.eraet ions.
Let us now outline the argument, by writing the Lagrangian density C(z) as
the sum of a translational invariant term C'i (Z) and a t.elm breaJc:ing tr3IUiational
invariance C'6(Z):
C(=) = C,i ( : ) + .c u (:) . (39)

We imagine constructins vertices for Feynman diasrams, in momentum space, from

the tenns C'i (Z) and C'6(Z) . A Feynman vertex constructed from .e'i(Z) will con·
tain a momentum conservins factor 64 (EiP;) . This 6-functton will be absent from
a Feynman verLe.x constructed from C,,(z) . The momentum conserving factor
64 (E;P;;) has dimension

(40)

We now consider two interaction vertices 9,; • 64 (Eop!) and 9" . arising from
.e'i(Z) and .e,.(z) re!pectiveiy, which have the same form. apart from the m0-
mentum conserving 6· function . The coefficients 9'i and 9,. must have different
dimensions. In fact we have

(41)

and hence
(42)
106

We now assume that the interactions derived from eli and Cu are of the same order
of magnitude at sorne fundamental JI1lLII8 !Kale nlf..t>d, such as the Planck mass. We
then have the simple dimensional result

(43)

For a low energy experiment, conducted at a momentum scale P , the translation

symmetry breaking interactions will be suppressed, relative to the translation&!
invariant interactions, by a factor of order p4 /m1und in amplitude. Taking mrund ::
10 111 Ge V and P = 100 Ge V, we obtain a violation of trarulalional invariance of
order
(44)

in amplitude, and thus a suppression in the rate of a genuine momentum violating

process by a factor of IO- 1S6 .
The above remarks are based purely on engineering dimensiona, without even
considering the elfec.t.a of the renormalisation &rOUp. There could be a logarithmic
dependence on the energy scale coming fr om the «normalisation group, but this
effect is completely dominated by the dependence given by Eq. (44). We therefore
conclude that translational invariance will be accurately aatisfied in a low energy
scattering process , in which all the external particlea have small momenta of order
P . However once we allow enerA)' non-conservation, typically of tbe order dE:::
mruMl , an initial ~tate containing only low enerA)' particlea may lead to a final
slate containing some high enerA)' particles. The amplitude for such an energy
non-conserving process does not contain the suppression factor of Eq. (44), because
some of the external momenta are of order mrund rather than P . Thus scattering
experiments with low ener!)' particle beams would not necessarily conserve energy
and momentum.
It would be natural for all final state particles to have small momenta, in experi-
ments with low energy particle beams, if we poetulate exact energy conservation. 'Ve
thereby assume time translational invariance and derive just space translational in-
variance at low energy. In such an energy conserving but momentum non-conserving
theory, the interaction vertices arising from the translational symmetry breaking in-
teraction £" must contain an energy conserving !Ii-function. The suppression factor ,
Eq. (44) , in the amplitude for a 3-momentum non-conserving process then becomes

= lO- s1 . (45)

Any remnant of translational invariance breaking at low ener!), 's expected

to manifest it.self by a very small fraction of scattering events violating momen-
tum conservation strongiy, rather than by a tiny violation in all events. Actually
such momentum non-conserving events are analogous to the Umklapp process in
a crystat.1 The amount of momentum non-conservation in an Umklapp process
 107

must correspond to an inverse lattice vector for the crystal. Thus for a crystalline
model of the vacuum, having a lattice constant of order mfu~' the amount of m0-
mentum non-conservation would have to be of order mtund - Since we have really
failed to derive time translation invariance from dimensional analysis, let us take
exact energy conservation as one of our assumptions. So we only need to latticise
three-dimensional space and can take time to be continuous. For a regular lattice
momentum conservation is then exact at low energy. This follows because all the
partielefJ have momenta P small compared to rnrund . due t.o enerC' conservation,
and any violation of momentum conservation must be of order mtund .
In an amorphous or glasa-like model of the vacuum, the amount of momentum
non-conservation in an "Umklapp reaction" could be of any size and magnitude.
The phase space for momentum violation by an amount IPI < Rp is proportional
to RJ,. We therefore end up wi~h the strong suppression factor of Eq. (45), in the
amplitude for a momentum non-conserving process at low energy.
In conclusion the method of dimensional analysis does not really s ucceed in
deriving energy conservation as a long distance symmetry. \Ve have only shown
that the energy non-conserving amplitude is suppressed , provided we assume that
the 4-momenta for both incoming and outgoing particles are small. However, if
exact energy conservation is assumed , J..momentum conservation results as a low
energy symmetry, without explicitly assuming that the final state 4-momenta are
small. Thus the method of dimensional analysis is successful in deriving space
translational invariance as a low energy symmetry.

6.! ..{ Lorenl: iUdrian ce from Jlrin9 lheory

Lorentz invariance can be derived from string theory, in the sense that it is a
consequence of the other postulated symmetries of the underlying two dimensional
conformal field theory.4 In particular, the following fou r assumptions are made for
the bosonic string:

(i) The 2 dimensional fields describing the string consist of a synunetric second
order metric tensor 90'(0') and a set of real scalar fields X P (/1), which are
interpreted as position coordinates in the exterior space-time.
(ii) Reparameterisation invariance under transformations of the two coor-
dinates 0' = (/1°,0' 1); this is the usual general relativity diffeomorphism sym-
metry in two dimensions.
(iii) Translational invariance in t he exterior space-time under the following trans-
formations:

(46)

90,(/1) - 90, (0") . (47)

108

(iv) We)'l invariance under the local rescaling of 1en&ths:

g.,(-) - f (-)g.,(-) , (48)

X'(_) - X'(_) . (49)

Under the reparamet.erisation corresponding Lo a scaling of the two (f eo-ordinatee

(50)
we have the transformation properties
d2 (T _ l.',/l(T (51)
and
~_).-1~ (52)
8tr' IJrro .
So the string action can only remain iDvariant under the Kaling transformation
Eq. (50), while satisfying the combined requirements of reparamet.erisation and
We)'l invariance, if the Lagrangian density contains precisely two derivatives.
The above arpment, for the presence of just two derivativeI , "aciUy uaumes
the absence of derivatives in the denominator of any term in the Lagrangian den-
sity. The justification for disallowing 8uch denominators, consisting of a non-trivial
homogeneous polynomial in derivatives, is as follows . A positive definite term in
the Hamiltonian density havior; derivatives in the denominator, 8uch as

( X' )' -_ OX" OX" _

{Jrtl {Jrt 1 11" .. (53)

with ij" .. some constant real symmetric matrix, gives an infinite contribution t.o the
energy for X' = O. The pres~ce of such a term makes it energetically unfavourable
to have :tetO derivatives, X' = 0, in the ground state. So the scalar fields X"(c:r)
take on IT-dependent values X:ac(lT) in the vacuum. The Hamiltonian deJUlity can
then be re-expressed in terms of new shifted scalar fields.
X"(IT) = X"(c:r) - X!'K(IT) (54)
which describe the deviation from the !found state. This new form of the Hamilto-
nian density no longer containl a denominator which il a homogeneoua polynomial
in derivatives. So either lcale invariance is IpontaneoUily broken violating &88ump-
tion (ii), or any denominator just becomes equal t.o a constant.
It followl that the mo&t general form of action, allowed under the above &88ump-
tions, is

l (g, X)=- 2~ Jd21T [Vldetgl (g(O~(IT)~~:~~; ~" ..

+ lR + g(OI' (IT)D(OD~X"e,.)
+t
(01' {JX" {JX"
{Jc:rO {Jail
1
x"" . (55)
 109

Here R is the two-dimeru!lional curvature scalar, Do is the tWc>dimeruionaI gravita-

tional covariant derivative and t,0/J is the aniisymmetric symbol. The rul number
seta '1,.", k, ~I' and XI''' are constant paramet.ers in the tw~dimelUional field theory.
For a Lorentz invariant action , 'I,,,, would be the metric tensor and the parameten
{I' and XI'" would be zero.
We now argue that the non-Lorentz invariant terms in the action are of no eoo-
sequence. In fad all the terms in Eq. (55), apart from the first , are total divergences
and can be transformed into boundary terms by partial integration . The l&IIt i.eI'm
eives rise to the boundary term

1
- 2W"
J ,8X" ..
dvt Drill x,. .. X (v) (56)

which can be interpreted as tbe action (or a string having electric charges at it. ends
and moving in a constant background electromagnetic field . The boundary term
arising from the third I.erm in the action leads to inconsistent equatton.8 of motion,
and this term is therefore forbidd en. The boundary term arisin,; from the second
term, which is just the two dimensional Einstein action, is a topological invariant
proportional t.o the Euler characteristic X of the strin,; world sheet. It contributes
a factor to the functional inte,;ral for a string &mplitude equal to a constant raised
t.o the X-th power and, therefore, it effectively renormalises the string interaction
couplin,; constant.
We have now euentia.lly reduced the action to the single term

(57)

where '1"" is a constant metric t.enaor in the exterior space-time. By a linear chlUl,;e
of buis for the aealar fielda X "'(C7) , the metric tensor '1,." can be brousht into a
standard dia,sonal form

+1

+1
o
'1"" = (58)
o
-I

-I

with just +1's , zeros and -1 'I alons the dia,sonal. The zeros alonS the diasona.l are
easily eliminated, by discarding those linear combinatiolUl of X"' {(7) fields that do
not occur in the action . The last step is to &rsue that, in order to avoid nesative
norm modes on the strins, there can be at most one +1 on the diasonal .
110

So, using ,wo-dimensional reparameterisation and Weyl invariance and exterior

space-time translational invariancc, WeinbergS derives Lorentz invariance for the
Itlin,. There remainsjult the single ambiguity in tbe metric, which must be either
Euclidean or Minkowlkian with one time cc>ordinat.e and several space cc>ordinates.

6.3. Local Gauge Invariance

We have already discussed. in Chapter 4.6.2, the origin of sauge symmetry in
string theory. In this section we consider two other method. of deriving local gauge
. .
mv&raanee.

6.3. 1. Formal oppearonce 0/ !latlgc .,mmd,.,

The gauge symmetry of Yang-Mills theory and of electrodynamica is very rem-
iniscent of diffeomorphism symmetry in general relativity. It should therefore not
he too surprising to learn that gauge invariance can be introduced as a formal sym-
metry, purely by definition. Quantum 8uctuations may then reveal it to be a real,
physically significant, gauge synunetry, for some finite region in the space or coupling
constants [paper 26]. We have already discussed examplea of this phenomenon, in
Chapter 4.2 and 4.3, Cor the GIR valued non-linear sigma model (papers 11- 19J.
For the type of gauge symmetry derivation to be discussed in the present subsection,
one starta with a Cundament al laUice action," which contains the aame degrees of
freedom as a gauge theory but is not gauge invariant. There can, for instance, be a
latticised gauge boson mass term in the action.
As a simple example, we consider a Euclidean action for U(l) lattice eleetrody-
narruc:!l of the form
s=PL:R<Uo+0L: R<U(-) . (59)
o
The action S is defined on & regular hypercubic Euclidean space-time lattice , with
lattice spacing o. The fundamental variables are the link variables U( - ) which are
defined on each link - of the laUice and take complex values of unit norm

U( -) E U(I) = 1* E <c, 1'1 = 1) . (60)

The pJaqueUe or ftux variable Uo is n~ an independent variable, but is given by

the product..
Uo = U(O)U(G)U(Cl)U(G) (6 1)
of four link variables associated with the pJaquette 0, made up of four neighbouring
links . The summations Eo and E
respectively, run over all plaquettcs and aJllinu
on the lattice.
An exactly invariant pure U(l) gauge theory corresponds to retaining just the
first term in Eq. (59), where II is twice the inverse gauge coupling constant squared .
In fact this first term is invariant under the lattice gauge transformation

U(;---i + 06,.) _ A(z)U(;---i + 06,.)A- 1(z + 06,. ) . (62)

 111

Bere JJ refell to the direction of the link ;----z + 0.61' connecting the .ites with c0-
ordinates r.' and z ' + 06: . The other term in the action S corresponds to a photon
mass term, which is not gauge invariant. The remarkable result of {paper 261 is tbat,
provided P is sufficiently large and Q is sufficiently small but still nOR-zero, quantum
fluctuations can generate an effectively exact gauge symmetry in this model. This
gauge invariance is obtained via an inverse Higgs mechanism (Paper 19]: the massive
photon is converted , by quantum effects , into a ma.ssless phoLon plus a m&88ive ICaIa.r
particle.
Gauge invariance is initially introduced into the model as a purely formal sym-
metry, obtained by expressing t.he action in terms of a new super8uously large set
of field variables (U.(-) , H(-». These new "human" fields , UiI( ........... ) and H(o ),
afC defined on the set of links and sit.es respectively and take values in the t;roup
U(I) '
u.(-) E U(I) , H(-) E U(I) . (63)
They are related to the oripnal fundamental, or "God 's", variables U( - ) by the
definition

U(z---i: + 06,.) = ul! (zz + 06,. ) (64)

= H- 1 (x)U.(;-i + a6,. )H(x + 06,.) . (65)

he link variables U(-) and UIIo(-) are strictly speaking functions of oriented
inks satisfying

UIIo(Z"Z + a6,.) = U. 1(;-+"i + 06,.) (66)

= U. (;+-"i + a61') . (67)

r e new site variable H(o) may be regarded lIS a Higgs fi eld.

Clearly we have introduced more "human" field variables (U,,(-), H(o » than
f here were "God 's" variables U(-). There is thus the possibility of transforming
f he "human" variables around without changing U(-) . In ract the field vari-
~bles (Ullo (--) , H(0» can be transformed under the artificial gauge transformation
~(z) E U(I) '

u.(n) -- ut(G)
= A- '(z)U.(n)A(y) , (68)
H(z) _ A-'(z)H(z ) (69)
y=x + a61' ' (70)

ithout chanpng the oripnal field

U(n) = W'(z )U. (n) H(y) (71)

- (A-'(z)H(zW' A-'(z)U.(n)A(y)A - ' (y)H (y) (72)

= U(n) . (73)
112

It. foUow. that any action S[U(-n. which ill a functional of the fundamental
field U(-) only, is automatically invariant under the formal ga~e transformation
of Eqs. (68)-(70). In particular the act.ion of Eq. (59), for U(l) latt.ice electrody-
namics, manifests this formal gaule invariance when expressed in terms of the new
variables:

S[U• •H] =SIU =UrI (74) .

= PLR<U,"
•
+ L Re{H-1(:r)UA(Z---Z + o6,.)H(z: + 06... )] .
0' (75)

Here we have used the fact. that.

U"" = U.(<;J)U.(O)U.(O)U.(O) (76)

=Uo (77)

which is invariant under the U{l) gauge traruformation of Eq. (62).

The surprising result is that the apparent.ly purely formal gauge symmetry,
Eqs. (68}-(69), is physical enough to give rise 1.0 a phase having a maseless
photon.10,1I Monte Carlo studies of the lattice gauge-.Bigs action, Eq. (75). have
mapped out the phase boundaries. Bere we shall he content to demonstrate the 6:-
iatenu of the Coulomb phase, aDd ita massless photon , in a mean field appraxima.-
tion. l l In mean field theory,' link and aiLe variables are assumed Lo have expectation
values, which are tben determined by selC-couistency conditiou. These conditiou
are obtained by aliowiDJ a single link, or site, variable La ftuctuaLe quantum me-
chanically, while replacing all fielda on neighbouring linb and aita! by their average
values. Since only fluetuatiou of one link or siLe are treaLed at a time, it is ~ible
to obtain noo-zuo averase values and La buu local sauSe symmetry spontaneously,
without the necessity of working in a fixed ,auge.1'
We use the translational and rotational invariance of the vacuum Lo make the
,aule-Bigs mean field anaaiz;

(78)

(H(.)) = VH (79)

where VuJo and VH take on the same real value for each link and siLe respectively.
The real expectation value VuJo is independent of the orientation of the link, as
follows from Eq. (61) . The real expectation value of H-l(zo) is VH . as follows from
the equality of H-l(zo ) and H· (zo) . It also follows that VuJo and VH should both
have a norm less than or equal to unity.
 Il3

We now introduce an eft'eetive single-link action by replacing all the variable.,

except for just the single-link variable U.(;---z. + 06,.) it.se1f, by their average val·
nes in ~he action of Eq. (75) . In d spacetime dimensions, each link baa 2(d - 1)
neichbouring plaquetLes and the effective single-link action

i. easily obtained . Sim.ilacly each site liaa 2d uei&hboucing linb twd we oLt.ain t.he
effective .in&ie-sit.e action

(81)

Usmg the elrecLive single-link action in the defining functional inLe!!,al expreaaion
for VU., we obtain the self-consistency condition for the mean value of the link
variable

Vi _ f DU1(~ + a61')e.xp{Scff(U.(;---% + (6 )J}U.(;---% + 06,.)

11
(82)
u~ - f DU,,(;---;' + 06,.) exp{Sctr[U,,(%""% + a6... ))} .

We now make the aub.Litutiou

(83)

for the link variable aDd

(84)

for the Haar meuure in the U(l) group integration.' The self-consistency condition
then take. the form
Vo, = I,[2(d-I)PVJ, +o~il (85)
Io[2(d-l)PVJ. +aVHl
"hue 1.. (X) denotes the modified Be.el function

I.. (X) = " _e

d8 Xcm ' coe" 9 (88)
/', 2.
and d ill the dimension of Ipa.eetime (d = 4) . Analo~oual)' the aelf-consistenc),
condition for the mean value of the HiU;S field beeomea

VH = II(odVu • VH ) . (87)
10(odVu.VH)
We ITlUIL now look for aelf-wnsilltent solutionl of Eqs. (85) and (87) in the plane
= =
01' the two eouplin~ constanta (o,P). Since 11(0) 0 and 10(0) I , it immediately
follows that. "he trivw solution

(88)
114

Fie. 605. Grapbicaholution of EQI. (89) and (90) for (a) o. VUa.< I , (b) ... VU" "" I and (e) ...
VUII;> I .

aatisfiea t.he equations for any values of 0' and p. In lOme tel\:ions of the (cr. P )
plane, another non-triviallOlution can be oMained graphically. This is iIIustrat.ed
in Fig. 6.5 f<x the second seif--consist.ency condition, Eq. (87), written in the form
of two simultaneous equations
(89)

and ,
VH = "";7-
odVu.
(90)

The function 11 (:)/10(:) ::: z for ,mall z and its slope then decreases monotonically
to zero as z - (X) , where the function approaches its asympt.otic value of unity.

Clearly there is a 5OIution of the simultaneous equations satisfying

VH '# 0 iff lodVu.l> 1 . (91)

In particular, since d = 4 and !Yu.1 < I, only the solution VH = 0 exista when
a< t.
Similarly, when VH = 0 and fJ is small enoush, the first self. consistency condi-
1
tion , Eq. (85), only has the solution VUII = O. Thus, (or a < and small p, there
is only the trivial self-consiatent solution of Eqa . (85) and (87) with VII = 0 and
VUl = O. For VUl ¥- 0, the right-hand side of Eq. (85) tends to unity tI8 fJ __ 00 .
It follows that, for sufficiently large p, there is a solution with VUII> a little below
unity, VUl < I . This non-trivial, VUII> :F 0, solution of Eqs. (85) and (87) will have
VH = 0 for 0 < iand VII ¥- 0 for sufficiently large o .
For each combination (0, (J) of the action parameters, there is jUlIt one physical
solution of the self-consistency conditions. This physical solution is the one having
 115

the lowest free energy, i.e. having the largest value oClog Z where Z is the partition
Cunction
z= J DU,DHexp[S(U" H )). (92)

The value of log Z , corresponding to a certain self·consistent solution (Vu", VH ). is

obtained by estimating the "entropy corrections" to the action S(VUIto. VH) . These
"entropy-corrections" correspond to the logarithm of the DUItDH-vo!ume. which
contributes signifi cantly to Z in the mean field approximation. It turns out that
the "entropy corn:ctions" essent ially depend logarithmically on cr and p. For large
Q or P values, the act ion S( Vu •• VH ) (or a non-trivial seIr-consistent solution will

dominate the entropy contribution to lOS Z. since S{VUII, VH) depends linearly on
Q or /J. The non-trivial solution will therefore be realised physically for iar,;t a or

(J, rather than the trivia1 80lution with zero action S(O,O) . In particular, for large
!
fJ the physical mean field solution has VUh #- 0, with VH = 0 (or Q < and VH #- 0
for large Q. For small Q and fJ there is only the trivial mean field solution, VUh = 0
and VH :::: O.
The complete mean field phase diaAtanl (or the gauge-Higgs action can be de-
termi ned numerically.u It turns out that there are three phases in mean fi eld ap-
proximation:
I. 5t1'O"g co.pling or confin~d phau, which includes the region of small 0" and fJ
values and corresponds to the trivial mean field solution with VH :::: VUa = O.
2. Higgs phGse, which includes the region of large 0" and fJ values and corre-
sponds to a solution with VH #- 0 and VUA #- O.
3. Co./om6 phGse, which includes the region of large fJ values with a < and l
corresponds to a solution with V H = 0 but VUh #- O.
These phases are illustrated in the phase diagram of Fig. 6.6. In reality tbe fillit two
phases are analytically connecled ;IO the boundary separating the confinement and
DiW phases, shown u a dashed line in Fig. 6.6, terminates in tbe interior of the
diagram and does not extend down to fJ :::: 0 in Monte Carlo lattice calculations. 11
However our main point is the existence of the Coulomb phase: a finite region in
the (o , fJ) plane with VH = 0 and VUllo #- O.
We now want to argue that, in the Coulomb phase, there exist long range cor·
relations corresponding to a massless photon, due to the formal gauge invariance.
The appearance of a massless gauge particle in a theory is determined by the gauge
properties of the vacuum. For the discussion of these propert ies, it is necessary to
fix the gaugel l and we choose the Lorentz gauge (81'A" :::: 0) . In this gauge, we
consider the behaviour of the vacuum expectation values VUllo and VH under (a)
global gauge transformations with a constant gauge function 0

(93)

and (b) loc.al gauge transformations with a linear gauge function O'''z"

(94)
116

" ...... " Hi9qs phose

"-
'\
Slrong \
couplinc;l
p.....

1\ follow, from Eqa, (68}-{70) ~ba\ UDder lbe PoOal ,ause \rauCormatioD
(95)
and under the Cauge transformation with a linear lauge funelion
A =AL : VVIl _ eiO'··"VVA . VH _ e-f""··VH . (96)
Here all is the appropriate lattice link vector.
The laule symmetry properties of the vacuum, under Ao and AL, are readily
deduced from Eqe. (95 )-{96) (or the phMet (wnd in mean field approxima\ion, and
ate presented in Table 6.2.

6.2. Oa... Irmmeuy prapatiel of vaeuurnllh- in Lhe U(I ) IMI&.HicP model.

~ '"""......... CguhJlt u....

Gaqe F\mctioa
0.,... """"""
ph- AO(.) "" i Ol
AL(.) "" io"··
In...... """"-
V....._
""""-' V~

VH .. 0 YU,\ = VB =0
Noa-iA.....-iant NOft-la~
V__
HiIP V .... p.m
VH'" 0 VUIIJl 0, VH '" 0

In...... Noa-inftliaD1
C......... V~ V~

VH = 0 VUUlO
 117

M&88Jess gauge particle. (photon,) can he eOlUlidered as Nambu-Goldstone

bosoM, accampaoyins the spontaneous breakdown of the Iymmetry under gause
transformations with a linear gauge functi on.13 However the proof of the Nambu-
Goldstone ~heorem maket essential use of translation invarianee . It requires the
spontaneously broken symmetry generator to effectively commute with the momen·
tum operator PI' : the commutator must annihilate the vacuum. The eommutaLor
in question is
[P,. ,Q.. ] = -igl'"Q (97)
where Q .. is the generator of gauge transformations with a linur gauge function
A£(z) and Q is the generator of pabal gauge transformations AG - The genera-
tors Q and 0" can be defined in terms of an expansion of the generator Q[~l of a
general infinitesimal gauge transformation; Q[).] is a functional of the correspond-
ing infinitesimal gauge function .\(z). Expanding the arbitrary infinitesimal gauge
function .\(z) in a Taylor series

.\(z) = 0' + O'I'ZI' + QI'''ZI'Z" + ... (98)

we obtain the required expansion of Q[.\):

Q[.\} = QQ + o:l'QI' + QI'''QI''' + ... (99)

The existence of a massless photon as a Nambu-Goldstone boson thus requirel

the generator Q to annihilate the vacuum. The following two conditions must
therefore be latisfied by a phase containing a massless photon:
(i) The gauge symmetry for linear gauge functions must be spontaneously br~
ken, i.e. the vacuum should not be invariant under gauge transformations
with a linear gauge function AL{z) .
(ii) The global part of the gauge symmetry must not be spontaneously broken,
i.e. the vacuum should be invariant under gauge transformations with a
constant gauge fundion Aa(z) .
It is immediately seen from Table 6.2 that the Coulomb pbase , with VUh #: 0 and
VH = 0, satisfies both condhions and therefore must contain a masslesa photon .
We bave thus seen that the dynamics of the gauge non-invariant adion , Eq. (59),
for U(l) lattice electrodynamics can generate an exact massless gauge symmetric
theory, without fine-tuning of parameters. We will now briefly discusa the gener-
alisation of this result to a non-Abelian lattice gauge theory, with explicit gauge
symmetry breaking terms in the action. A very weakly coupled non-Abelian gauge
theory is approximately equivalent to a field theory with several Abelian types of
gauge particle. So, in this approximation , we should exped a phase of tbe theory
to exist possessing an exact gauge symmetry. However we can hardly expect the
existence of a Coulomb phase with physical massless gauge particles; a non-Abelian
theory should exhibit confinement," unless the number of matter fields is sufficiently
118

large to violate asymptotic freedom. In such a eonfi.ninl phase there will, iporinl
matter fi elds, be no massless st.ates but only glueball states, i.e. gauge group sin-
glet bound states of t.he gauge field . The glueball maB8 seale arises by dimensional
t.ransmutation,i in an analogouB way to t.he strong int.eradion mass aea1e AQco in
the standard model. This mass scale is exponentially suppressed as a functi on of
t.he inverse gauge coupling constant squared , p, defined at some fundamental lattice
scale.
A non-Abelian generalisation of the gauge symmetry violating lattice action of
Eq. (59) is ea:!iiy cons'ructed, by requiring the link yuiablc U( -) w take values in
the non-Abelian group . As for the U(l) case, a new set of "human" variables can be
introduced, including a Hig.like field H(·) which takes values in the group mani-
fold . Block-spinningl4 leads to a more coarse-grained field t.heory, where H(o) can
presumably be approximated by a field taking values in a small fait.hful repreaenta--
tion of the group , such as t.he fundamental represent.ation . The action expressed in
terms ofthe "human" gauge-Higgs variables automatia1ly P06SesSes a formal gauge
.Invanance.
.
In the Don·Abdian theory there is no region , in (0' , .8) puameter space, cor-
responding to a Coulomb phase. The phase diagram for a latt.ice non-Abelian
gauge-Higgs model is expected to have the form shown in Fig. 6.7, which has been
verified by Monte Carlo studies for SU(2) and SU(3). At. large value. of .8, there is
a phase boundary bet .....een t.he ,mall a confinement. phase and t.be large a spont.a.-
neously broken Higgs phase. The phase boundary terminates in the interior of the
diagram: tbe confinement and Biggs phasea ue analytically connected l O and really
form just one big phase without any maaslesa: putides. However sa .8 becomes
luger and larger , for small enough a , we enter what would be the Coulomb phase
in t.he Abeli&n case . In this region, shaded in Fig. 6.7, the glueball mass tends
towards zero in an exponent.ial way:

m, == -a1 exp( - c.8) ( 100)

as .8 tends to infinity, where a is the lat.tice spacing and c is a p06it.ive const.ant of

order unity. Long distance physics in the lattice theory is then well approximated
by a conventional gauge symmetric continuum Yang. Mills theory. In this sense t.he
appearance of gauge symmetry "out ofnot.hing" also works for a non-Abelian group.
The assumption of a regular lal.t.ice is not. essent.ial for the success of the inverse
Higgs mechanism. We may equally well consider a random discretised model, with a
basic set. of fixed sites distributed randomly in four-dimensional space-time. Indeed
the mechanism i8 expect.ed to work for a very general t.ype of field t.heory. For
instance it should work in a field theory glassl5 where the parameters , such sa
coupling constants, and even the type and number of degrees of freed om vary from
place to place in space-time in a quenched random way: they are chosen once and
for all and not varied in the functional integral implementing quantum mechanics.
It is because of these frozen·in parameters , bein« analogous to the atomic binding
 119

a
" Higgs"

I-
.3
,

structure in a real glass, that we use the term "field theory ,lass" . The action
for a field theory glass is not even strictly local at the fundamental lattice scale,
which we take to be the Planck length. The field theory glass idea is used in the
forrrwiation of the random dynamics project discussed in Chapter 7. Here we use
it to demonstrate the very general nature of the inverse lIiggs mechanism and the
spontaneous appearance of gauge symmetry.
The construction of a field theory glass model is most easily visualised in terms
of an imagined MenLe Carlo computer simulation. The computer program first
selecl.s a randomly distr ibuted set of points, or sites, {i} in four-dimcnsional spa.ce-
time . Next it follows an algorithm for constructing, by means of random numbers,
a manifold Ali at each of the points i on the random lattice. The manifold M;,
somewhat unusually, changes randomly from point to point . The fundamental dy.
namical variable is a generalised quantum field f( i) which maps each point i onto
the corresponding manifold M;. The parameten of a very general semilocal action
S lU'e in tUrD chosen as random numbers. Thereby one finally has a field theory
glass model, with random degrees of freedom ~(i) and random action, whose long
wavelength properties can be studied by Monte Carlo methods. Our claim is that
the resulting physical degrees of freedom, wit h the most long range correlations,
are similar to the gauge degrees of freedom obtained in the translational invariant
lattice models discussed above. This means that it should be possible to introduce
some new "human- variables, analogous to U" and H of Eqs. (64) and (65), and
hence an exact formal gauge invariance . Then some of this formal gauge symmetry
is realised physically, in the sense that it gives rise t.o massless gauge particles or to
low energy confinement. \Ve outline below how such an exact gauge symmetry may
be generated in a very general type of fie ld theory glass .
We take the field theory glass semilocal action to be a sum over contributions
S,. of quenched random form , from very many overlapping small regions r of space-
120

time, of the order of the Planck size:

5[.) = Es.(Ou)) , i E• . (101)

•
This action is of course no' Poinco.re invariant, We Meume here that P oincare
symmetry &rises as a consequence of one of the mechanisms discussed in Sed. 6.2,
and concentrate on the origin of gause symmetry in the field theory glass. For this
purpose, we imagine that a small demon goes through the random theory, small
reKion aRer small regioD , looking Cor approximate local symmetries of the action.
These accidenta.llocalsymmetries correspond to transformations of just the degrees
of {reedom in some limited part of space-time, in general not coinciding with a region
r of Eq. (101) , which leave all contributions to the adioo approximately invariant,
Such a small part or "ball" of space-time, which in general is not spherical and is of
the order of the Planck size, will be called a "gauge ball". The demon selects a site a
as the centre of the gauge ball and identifies the approximate local gauge symmetry
group G(a) within the gaup ban B{a). He passes through all of apace-time and ,
in general, finda a large density of strongly overlapping gauge balls; at least let WI
imagine &0 .
The approximate local symmetry groups G(.). within e&(h gauge ball B(a),
are now converted into formally exact gauge symmetries by our procedure of in·
troducing a superfluously large set of "human- variables (tP/o,H) . The generalised
Higgs field H is defined on the centres. oftbe gauge balls and H(a) belongs to the
local symmetry group G(a) . The field ¢" is a new human variable, with the same
structure as the fundamental field tP, and is defined by the equation

¢Ci) = .~(i). (\02)

Here ¢o dellote. the rellult of transforming ¢ with an element n fcolII the dircd
product oC all the local symmetry groups G(a). At a particular site i, of course, the
field tPlI.(i) can only transform non· trivially under those groups G(s) for which the
gauge ball B(a) contains i .
The original action 5[¢) can he written in terms of the new variables

S[•• ,H] = Sr. = .:'] . (103)

This new action 5[tPlI., H] is automatically invariant under the formal gauge trans.-
Cormation

tPlI. -- tPr (104)

H __ n- 1H
(105)

since the fundamental field tP is unaltered by this artificial transformation

.I. _.l.H (_n)n-1H _.l.H _.I.
(106)
... - "'" - - V'/o - "'/0 - ... .
 121

The field t.heory glasa, expressed in terrm of (fJlI. H), thus manifests a formal gauge
symmetry of a thaotic nature, in t.hat t.he symmet.ry group cbanges randomly fr om
place to place.
In general, for any group K. we expect to find gauge balls dist.ributed th roughout
space-time, within which K is an approximate gauge symmetry of the fundamental
action 5[4'1. We are, of course, interested in the situation where the associated Biggs
field H , in the new action S(¢It, H] , has no long range correlations but shong local
flu ctuations; so that ita "average value" is zero in the vacuum. (In order to define
such an "average value" , the group K has to be embedded in its convex closure.)
Thus we restrict our attention to gauge balls within which the syIllJ1'letry breaking
terms of the fundamental action are rather small, corresponding to the small Q'
situation in our earlier lattice Saulte.-Higs examples. The symmetry breaking terma
in the fundamental notation are transformed into correlation terms, between the
Higgs field at different points , in the human variable notation. For sufficiently
small symmetry breakins, it follows from Iteneral decay correlation theorems l6 that
the correlation function between the Higgs field at two points, H(z} and H(y) ,
does indeed fan off exponentially with the separation. So we now assume that the
vacuum expectation value (H) = 0, as is needed Cor the invariance of the vacuum
under global transformations.
We now wish to introduce a continuum gauge field A:(z) for the group K ,
with generators ~· /2 , describing small deviatioI15 from the globally gauge invariant
vacuum state. Therefore we need a rule to implement the modification of the
vacuum state described by a smoothly varyinlt continuum Yang-Mills field A:(z) .
So let us consider the human field variable f.(i) at an arbitrary site i in the vacuum
stal.e. We now locate all the gauge balls which contain i and belong to the restricl.ed
set , with sufficiently small K-symmetry violating l.erms in the fundamental action
La satisfy the conditions of the decay correlation theorems. We may consider the
continuum field A:(z) to be essentially constant over all the Itauge balls containins
the site i , having coordinates zf . The modification of the field 'theil , relative to its
vacuum value, is then constructed from the vacuum value , by applying the linear
gauge transformation which would set up a constant continuum field with the value
A:(Zi) . In other words, we apply a series of gauge transformatioru;

( 107)

tof. (i), one afier the other; one for each gaulte ball containing Z; . The coordinates
z: are those of the sauge ball centre s . In this way we modify the field flo at each
site and set up a new configuration corresponding to the continuum Itauge fi eld
A:(z) .
The gauge transformation A,(Zi) depends on the choice of gauge ball centre
z~ . However, this dependence is simply equivalent to a gauge transformation corre.-
sponding to the gauge ball at 8 independenl of Zi , i.e. it is the same for every field
variable fA(i). Here again we make use of the fad that the continuum Yang-Mills
122

field A:(~) is smoothly varying. So, in the abaenee of any physical dependence
upon the choice of z" the modification corresponding to the continuum gauge field
A:(z) , as given by Eq. (107) , is well defined up to a microscopic sauge transforma-
tion. The latter has only formal significance. In any event the centre z, should of
course be d06e to the sites involved in the gauge transformation A,(z/) and is there-
fore rather well defined. It can happen that two or more isomorphic groups K are
found as subgroups in G(.). In order to avoid ambiguity in this case, it is necessary
to apply the same gauge transformation A.(Zi} for them all; thus we only consider
a continuum gauge field for the diagonal subgroup. There is a similar ambiguity in
setting up a continuum gauge field for any group with an outer automorphism;l7
complex conjugation is an outer automorphism for SUCH) groups with H > 2. Our
construction of a continuum gauge field fails for a uoup with an ouLer automor·
phism j it can however be rescued if the field theory glass representations of the
group are not symmetric under the automorphism. We noLe that, although the
standard model group S(U(2)xU(3» is symmetric under complex conjugation, the
chiral quark and lepton representations are not .
By the above construction, for each smoothly varying continuum field configu·
ration A:(z) , we assign a new configuration of the field theory glass. The action
can be evaluated for each new configuration and it must be invariant under the
gauge group K . This follows from the natural correspondence between a gauge
transformation on the continuum field A~ ( z) and a gauge transformation on the
field Lheory glass variables ;.(i). The effective action for the continuum fi eld A~{z)
is therefore invarilUlt under K gauge group transformatioll5. In this way we have
apparently set up a Yang· Mills field theory for an arbitrary group K without outer
automorphisms. With outer automorphisms, our "demon" would have to make ar·
bitrary choices allover space· t ime, or else one would have to rely on matter fields
to specify which way to implement the group.
In order to avoid spontaneous symmetry breaking of the gauge group K by the
Higgs field H , it was necessary to make a careful selection of gauge balls; so there
may be only rather few gauge balls selected. 1n this case the majority of sites i do
not lie inside any gauge ball, let alone inside several. It follows that mosL of the field
variables f lo(i) transform trivially under gauge transformations and are unch&nged
by setting up a continuum gauge fi eld A:(z) . The action is then very slowly varying
&S a fun ction of A:(z) j this means that the vacuum is in a strong coupling phase,
corresponding to the small fJ situation in our earlier lattice gauge-Higgs examples.
The gauge field A:(z) is thus confined at the Planck scale and the gauge symmetry
is not realised physically in the long wavelength properties of the theory.
Clearly several conditions must be satisfied before a formal gauge symmetry
group K of a field theory glass manifests itself physically, as a conventional gauge
theory at large distances. In particular the fundamental action S[~l must mani·
fest the symmetry to a r;ood approximation over a large fraction , locally, of the
field theory glass, and the coefficients of ~he non·trivial symmetry co nserving terms
should be large . The chances of a fi eld theory glass satisfying these requirements,
 123

by accident, are better the smaller the number of degrees of fr~om involved . Thus
inv&I'iance groups K with a few generators and small representations are favoured
to show up with massless gauge boson. or low energy confinement.. The represent ....
tions ofthe standard model gauge group S(U(2)xU(3)) ate indeed small. For the
purposes of this discussion, we may consider the W:t: and ZO particles to be massless
gauge bosoM, since their masses are negligible at the Planck scale. We postpone, to
Chapter 7, further discussion of the conditions which a gauge group should satisfy,
in order to survive from the Cundamentalscale down to the low energy scale.
We conclude that low ener!), gauge symmetry can arise spontaneously in a very
general class of theories, although we do not have a complete proof of the effed. In
principle we sbould make real Monte Carlo computer calculations for a field theory
glass , io order to verify our speculations and to find out which gauge groups survive
at long wavelengths.
It has been emphasized (Paper 27) that the gauge symmetries of a theory in
the infrared must also be symmetries in tbe ultraviolet, in order to avoid noo-
renormalisable ultraviolet divergences. If a gauge symmetry is broken at the funda.-
mental cu!,.off seale, the divergences arising fr om loop di&&rams with internal vector
bosons will not caned. For instance, the phenomenological successes of the stan-
dard model would be destroyed, if the gauge invariance of the model were broken
at the fundamental lattice scale. At first sight , this infrllIed.-ultraviolet connection
is a disaster for the inverse Higgs mechanism u the origin of low energy gauge
symmetries. However the physical gauge symmetry, arising from the inverse Higgs
mechanism, is a low eners,y manifestation of a formal gaule invanance, which is
exact in both the infrared and the ultraviolet ; the formal gauge invariance is purely
a consequence of the definition of the human variables (tP/o, H) . Tbua tbe formal
appearance of gauge symmetry it not in conflict with Veltma.n's result .
The infrared-ultraviolet connection does, however, pose a threat to a renOI-
malisation group derivation of gauge invariance at an infrared fixed point . The
renormalisation group approach to the origin of gauge symmetry is the subject of
the next subsection.

6.S.!. Gauge sJ'mmdry from the rcnormoiisation group

The perturbative renormalisation group method has already been used to derive
Lorentz invariance u a low energy symmetry in Sect. 6.2 .2. Here we consider its
application to gauge invariance [Paper 28J and also to combined gauge and Lorentz
invariance (paper 29J .
It is clear, on dimensional grounds, that the vanishing of the photon mass term
cannot be derived perturbatively in the infrared limit. The coefficient, !p]
say,
of any photon mass term til] A; in the Lagrangian density hu the dimension of
(masa)2, and actually geu more and more important in the infrared limit . It it
necessary to appeal to a non-perturbative renormalisation group calculation, in
order to derive the zero photon mass . Indeed the U(l) lattice electrodynamics
Monte Carlo calculation,ll mentioned in the previous subsection, can be interpreted,
124

in terms of blockspinnin&.14 as successively producing variables at longer and longer

distance Kales. In this subsection we shaH, therefore, set the photon mau equal to
aero and study other gauge symmetry violating terms perturbatively. Actually we
shall lid all mass lenni to uro, in order to investir;ate the far infrared
Abelian gauge symmetry is found to be an infrared attractive fixed point for
quant-urn electrodynamiCi (QED) with a spiDor field (the eledron say); while for
QED with a sea1ar field it is not (paper 28] . In order to avoid the difficulty with
renormalisability (paper 271, the gauge invariant theory is approached from a class
of theories quantised in a Hilbert space with an indefinite metric. This use of an in-
definite mdric Hilbert space is analogous to that in the Gupta-Bleuler formulation II
of gauge invariant QED. However it is introduced at the cost of losing the p hysical
interpretation of the theories away from the gauge symmetric limit, since the nega.-
Live norm squared stata do not then decouple. An alternative way of avoiding the
problem with renormalisability, without introducing negative norm squared states,
is to eonsider a ellLSS of theories which violate Lorentz invariance as well as gauge
invariance [paper 29). It can then be arranged that Ao = 0, 80 that the theories are
formulated in II. Hilbert space with a positive definite metric.
The Lagranpan density for gauge symmetry violating, but Poincare, parity and
charge conj1llation synunetry conserving, spinor QED is

C, =¢(ilJ- m)y.. - i(lJlIA w - 8wAII)2 + ~p2(A"A")

+ { (aIl A,,)2 _ e""''' "A" _ ±g(A II A" )2 . (108)

AI mentioned ahove, we ut m = po = 0 and consider the behaviour of the di-

mensionless coupling parameters e, 9 and { in the infrared. Gauge invariant QED
corresponds to the limit 9 = ~ = 0, and so we work to lowest order in the gauge
I)'nunetry breaking parameters 9 and ~ . The one-k>op renormalisation group fJ
funetioD5 then come entirely from the wave function renorma1isation of the electro-
R\&Inetic potential A" and are given by
, 1 ,
... P.> = 5e . (109)
, 1 ,
... P, = "3e 9 , (110)

... 2A
1
= 6_e4e
1-'( ... ( III )

These expressions imply that the running couplir.g parameters 9 and ~ get smaller
and smaller towards the infrared ; gauge invariant spinor QED is an infrared at-
tractive fixed point. The dimensionless coupling parameters approach their gauge
symmetric values logarithmically ; the approach is t.oo slow to realistically c.laim that
it may account for the observed high accuracy of gauge symmetry in quantum elec-
trodynamics. This problem with tbe accuracy of the derived low energy symmetry,
in the renormalisation group method, is discussed in Sect. 6.2.2.
 125

We note here that in (Paper 28) the term {(8,.A"r1 is ignored, on the grounds or
it jU6t fixing the gauge in the symmetric limit , and a term of order 9 2 is included in
p, . The term {(8,.A")' is eonsidered in [Paper 29), using a Lorentz non-covariant
<
formalism . It is pointed out that { and the coefficient oC a Lorentz symmetry
breaking term go to zero in the infrared; gauge symmetry and Lorentz symmetry
are simultaneously recovered at the infrared attractive fixed point.
Scalar field QED however does not have a gauge symmetric infrared fixed point.
In fact it is sbown in [paper 28] t.hat the running coefficient Q of the seagull term,
O'A,.A"'¢+ 41- , does not go towards the gauge symmetric: value, e2 , in the infrared
limit. Rather the parameter Ih = a/e 2 -1 , measuring the deviation from this gauge
symmetric value, has the one-loop fJ function

(112)

in the extreme infrared limit. Here ~ is the measure of anot.her sause symme'ry
violating term which does go to zero in the infrared, as does the coefficient ~ of
the term -l~(,+r)~ in the Lagrangian density. It follows , from the sign of the
fJ functi on in Eq. (112), that 61 does not go to zero and that the gauge symmetric
theory is infrared unstable.
It is well known that non-Abelian gauge theories are asympLotically free ,18 unless
a sufficiently large number of matter fields are included. Asymptotic freedom means
that the wave functi on renormalisation group effect of the precursor gauge field, in
a non-Abelian theory with gauge symmetry breaking terms, has the opposite sign
to that in an Abelian theory. The vector b060n wave function renormalisation is the
dominant effect, responsible for achieving gauge symmetry in the infrared for spinor
QED. So it should not be surprising that gauge symmetry is an infrared repulsor
for all non-Abelian groups.
A general non-Abelian theory, with both gauge and Lorentz symmetry violating
terms, is considered in [paper 29) . The Lorentz and gauge invariant fixed point
is explicitly shown to be infrared repulsive, in the absence of matter fi elds. If
sufficiently many fermion species are introduced that the fJ function for the precursor
Yang-Mills sauge coupling constant changes sign, then both Lorentz and gauge
symmetry are obtained in the low energy limit .
By approaching the infrared fixed point via a class of theories which violate
Lorentz invariance, the threat posed to renormalisability by the infrared-ultraviolet
connection is avoided. However asymptotic freedom is a property or QeD in the
standard model (with up to 16 quark flavours). So we must conclude that t.he renOf-
malisation group approach fails to explain the origin or Yang-Mills gauge symmet.ry
in the standard model.
(Note Gdded in proof: We should also mention an alternative auempt~l t.o derive
gauge symmetry based on the idea t.hat some scalar fields - the so-called "spoiler
fields" - would adjust t.hemselves so as to make the effective gauge symmetry
breaking terms vanish .)
126

6 ..... Superlymmetry
The renormalisation group method has also been used to study the stability of
superaynunetry as a low energy symmetry. The infrared stability of N = 1 global
auperaynunetry was firs~ investigated by Curtright and Gandour,I' who found that
in general a supersymmetrie infrared attractive fixed point does not exist.
A simple example, for which global N = 1 8uperaymmdry is an infrared attrao-
tor, is presented in [paper 28). The model involves a Majorana spinor VJ(z), a scalar
A(z) and a pseud08ular B(z). The class of theories considered is described by the
Lagrangian density

1 ,121.- .-
£ = - ,(O,A) - ,(O,B) - ,'''''' - 'I¢¢A
_ i/.,r,I/JB - ~l(A2 _ 8')2 _ gA'B' . ( U3)

This, of course, is not the most general form possible for the Lagrangian den-
sity; there could have been different coefficienLl multiplying the scalar "'t/JA and
peeudoscalar "'1,;'8 Yukawa terms, and also different coefficient. on the At and
84. terms. M in our previous applications of the perturbative renormalisation
,roup, we have i~ored mass terms. The supersymmetric limit is then obtained for
~=g=f2.
It is convenient to use the parameters

>-g
6,= 7' (U4)

and
6, = (~ - 1) + ~ (;, - 1) (lIS)

~ meuure the deviations from lupersynunetry, The correspondin& {J fun ctions

behave like:
(116)

and
(1l7)

cloae to the lupenymmetric situation. The ligns of the fJ functions imply that the
lupenymmetrie fixed point is &ll infrared attractor. Thus we have here an example
of derivint; N = 1 t;lobalsupersymmetry all a low energy approximation .
The renormalisation t;roup method cannot be applied 1.0 loeally supersymmetric
1heories, due to the requirement of renormalisability. The largest number N of
Ipinorial supersymmetry t;eneral.on, for which a globally supersymmetric theory
C&ll be cons\ructed is N = 4. It is shown in [Paper 29] how an N = 4 global

lupersymmetry may appear aa an infra.red attractor, once N = 18upersymmetry

is impoaed. A t;eneric N = l supersymmetric Yant;-MiJls theory, with t;auge group
 127

G t is investi!;ated , which has the foUowin! field content: a lange superfield V and
three chita! Buperfields ~j 1 i = 1, 2, 3, each being in the adjoint representation of G.
If the gauge group G is a SOotalled safe group, for which

(118)

is an identity, then N = 4 supersynunetric Yan,;-MilJ.s theory is an infrared at.-

tractor. If, however, G is onc of the groups SU(n) , D > 3, Of ESt which are not
safe, then in general N = 4 supersymmetry is not infrared stable. Infrared stability
can be obtained by imposing an 50(3) symmetry. under which the chit.! super-
fi elds rotate into each other. This 50(3) symmetry serves to prevent the presence
of a 'fi ( ~;{~ , ~a,) term in the Lagrangian density. However it seems unreason-
able to introduce an 50(3) symmetry solely for this purpose, without any physical
justification .
It is perhaps satisfactory that N = 4 8upersymmetry is infrared unstable, since
its presence as a low energy symmetry would forbid the existence of the chiral quark
and lepton representations of the .standard modepo lnd~d there is no direct phe--
nomenological evidence for the existence of even N = 1 supersymmetry, although
it might be technically UBeful in undentanding why the mass scale of the standard
model is so much leu than the Planck mass.20 The example derivation of N = 1 su-
penymmetry given above is really only a very partial derivation, due to the special
form of Lagrangian density assumed . It is tempting to interpret this failure of the
renormalisation group method , to really derive any supersymmetry in the infrared
limit, as evidence against the existence of supersymmetry as a low energy symme-
try. However this would be dangerous, si nce the renormaliaation group method also
failed to derive non-Abelian gauge symmetry as a low energy symmetry.

References
1. C .W . Misner, K .S . Thome and J.A. Wheeler, Gr...ffaff"., (W .H . Freeman, 1973).
2. J .L. Andenon. Pri.ci,lu "J R~lafi_i', P",,;c" (Ao;Memic Pre., 1961); H.C . Ohanian, G,n-
ila li". a." S,acdi"lC, (W.W . Norton, 1916).
3. M . Lem.o, H.B. Nielaen and M. Ninomiy&, H.d. P .. , •• B212, 213 and 228 (1986).
4. A.B. Borisov and V.I. O&ie~t ..kii , Teor. Mal. Fi,. 21 ,329 (1914).
5. M. Lehto. H .B. NieJ.ez. and M. Ninomiy&, P.. , •. Lell. 219B, 87 (1989).
6. J . EIlia, M.K. Gaillani, O.V. N&I'IOpI.>Iow and S. Rud,q, Hacl. PA" . BlTO , 61 (1980).
7. C . Kittel, [.'ro"_diu ' " Soli" SI.le P",.ie., (Wiley, 1986).
8 . S. Weinbcra:, Proc. "J tie XXIII /.,_ Cu.J. o. Hil" EurfJ P Ar,ieo, Berkeley, 1986. (World
Scientific, 1981), p . 211.
9. M. Creuta , Quri., 01.0'" .... L...ou;cu , (Ca.mbridae University P ..... 1983) .
10. E. Fradkin and S.H. Shenker, PIo..,. Rn. 019, 3682 (19'19).
II . J. R.anft, J. Kripf&ana and G. Ranft, P..., . Ret. 028, J60 (1983).
12. S. Elit"",r, PA". Ret. D12 , 3918 (1975) .
13. R. Ferrari and L.E. Pk.auo, N.. d. P..... B31 , 316 (1971); E.A. Iv&nO¥ and V.I. O&ievet.tUi,
JETP Lell. 23, 606 (1916) .
1<1. K .G. Wilson _d J. Ko",t, Pl" . Re,orll12, 15 ( 1914).
15. H.B. Niet.en and N. Bnne, W"rl.A . , ". Sl:,..... ioa ••• " A•• malin, (World Scientific, 1981),
p. <193.
16. L. Grou, C.mm .... M.IA. PA" . 08. 9 ( 1979); M. Lehto, H.8. Nichen and N. Ninomiy&,
C"mm .... M . ti . PA ••. 93 , 483 (1984).
128

11. R. Gilmtn, Li. Gru,., Lif Af,dr&ll •• 11 ".c ./ I.h ir A"N,tlio ••, (Wiley. 1974).
la. O. Itaya- aDd J .B. Zuber. q.","Y1l Fid" ntol.." (McGraw.HiIl. 1980).
19. T . Cu:rt.ri&ht aDd G . GhuMIolil', A ••. Pl" . 1041, 209 (1977); 112, 231 (1978).
30. E. Witten. N.d . Pi, •. BII. , 513 (1981).
U. K. Cahill aDd P. DeneI, Nu .. aM, LIII. n , 184 (1982).
 129

Chapter VD

CONCLUSION

In this book we have collected together several examples of cases in which sym-
metries of the laws of nature can be derived ; the symmetries are deduced under
some aMumptions not involving, too openly at. leut, the assumpt.ion of the .ym-
metry in question. We begin this chapter with & status report on how many of
the known symmetries can be derived . This provides a motivation (or t.he random
dynamics project reviewed in the second section . In the final section we attempt to
classify the mechanisms responsible for producint; symmetries.

T.1. Conclusion on the Origin of Symmetries

We review here our pr0t;ress in understanding the origin of symmetries in (i)
macf06copic physics and (n) the standard S(U(2)xU(3» model of microscopic
physics.
The well-known conservation laws of classical dynamics are associated with the
symmetries of Euclidean ~metry and time translationa! iovariance. These 'geo-
metrical' .ymmetries arc rather fundamental and any derivation mUlt be somewhat
apeculativc, appearing near the top of the quantum staircase. Rotational symmetry
and space-time translationalsyrnrnetry are jUlt part of the diffeomorphism Iymme-
try of general relativity, which sugests that they have their origin in gravitation
theory. The theory of general relativity wa.s of course cOlUtruct.ed to bave these sym-
metries, &S part of an overall Poincare symmetry, built into it. So it is necessary to
go beyond general relativity and cOMider a pregeometric tbeory, in order to obtain
a proper derivation of physieal Poincare invariance . It is then possible to introduce
a purely formal diffeomorphism symmetry; the crucial point is to show that the
corresponding formal Poincare symmetry is not spontaneoualy broken. Quantum
fluctuations have been proposed , in Chapter 6.2.1, &S a mechanism for stabilising
the Poincare symmetry.
A m or e convincingly derived macroscopic symmetry i. that of .cale .ymmetry,
since we know it is not present at the atomic scale. Scaling symmetry arises when
a very large number of molecules is present.
The parity symmetry of the mac.r06copic elasticity properties of, say, sugar is
another good example of a derived symmetry. In fact, for this system, parity aym-
130

metry appears and disappears twice &8 a function of scale. At very smallsub-nuelear
distances, where weak interactions are relevant, there is no parity syrrunetrY i then
at nuclear and atomic sca.le!, where only sttong and electromagnetic interactions
are relevant, parity symmetry is valid . In the case or sugar parity symmetry is spon-
taneously broken at the molecular level, as a consequence of biological effects ; only
species producing one handedness of sugar have survived. Finally, at macroscopic
scales, parity symmdry reappears 88 a consequence of the presence of a very large
number of lugar molecule:tl.
Parity symmetry is now understood &8 arising naturally from the low ener~
limit, or long distance behaviour, of the standard model. The same is true of time
reversal invariance; however the initial conditions in tbe world violate time reversal
symmetry so much that the second law of thermodynamics is valid .
The macroscopic effects of electrodynamic. manifest gauge symmetry and elec-
tric charge conservation. Any derivation of gauge symmetry has its o~in beyond
the standard model and must be considered speculative; sugested derivations are
discussed in Chapter 6.3.
At the macroscopic level, we have considered symmetries under the following
seven types of transformation: I) Rotation, 2) Space Translation, 3) Time Tran&-
lation, 4) Parity, 5) Time Reversal, 6) Scale and 7) Gauge. Although we have
presented derivations for all seven symmetries, four of the derivations are rather
speculative while three of the symmetries (parity, time reversal and scale symme-
try) have rather well-established origins. In these statistics we only include sym-
metries valid for the whole field of macro&copic physic. and not those specific to
particular dynamicalsituatiofl.5, such as the 0(4) symmetry or planet.ary motion in
a gavitational I j r potential.
At the standard model level, we can also ask how many symmetries are derived
sarely, i.e. rrom the standard model it.sel£, and how many are derived more specula..
tively, i.e. higher up on the quantum staircase. We consider eight symmetries at the
standard model level: 1) Space-Time Translational Invariance, 2) Lorentz Invari-
anee, 3) Gauge Invariance, 4) Charge Conjugation Invariance, 5) Parity Invarianee,
6) Time Reversal Invariance, 7) Flavour Conservation and 8) Chiral Symmetry. In
this list, baryon and lepton number conservation is subsumed under 7) and Gell-
Mann SU(3) flavour symmetry is subsumed under 8) . As summarised in Fill:. 3.4,
only the first three or the listed symmetries require a speculative explanation out.-
side the model itselr. The remaininll: five symmelJies are essentially derived within
the standard model, although with the help of some mild auxiliary assumptions
(see Fig. 3.4); it is necessary to assume the existence of small quark muses and
the absenee or the QCD topological 9-term. It is reasonable to suppose that these
auxiliary assumptions will be understood in t erms of physics beyond the standard
model. We can thererore claim that. five out or eight symmetries have been derived.
It may even be p08llible to explain the other symmetries, Poincare invariance and
gauge invarianee, near the top or the quantum staircase.
So what do we now learn rrom the above s'atistics? There appear La be two
main conclusions:
I. An experimentally well-established symmetry is not necessarily fundamental .
II. Approximate symmetries appear naturally as derived symmetries.
 IJI

From conclusKln I, we are warned of the danger of imposing synunetries on a

fundamental theory. The development of wealc. interaction theory Will delayed by
the general assumption , prior to 1956, that parity invanance was a fundamental
symmetry of nature . Another example is provided by Heisenber,'s unified field
theory of elementary particles l ; this theory assumed isospin ayrrunetry (and also
scale symmetry) to be fundamental, but spontaneously broken by an asymmetric
vacuum. Isospin symmetry is now known not. to be fundamental, see Chapter 3.2,
but to be just a consequence of SU(3) colour symmetry and of the ~tence of
two light quarks. Similarly the five symmetries, listed above and derived from the
standard model, should no lonler be considered Cundamental.
Conclusion II follows from the observation that many of the derived syrrunetries
considered in this book only appear u approximate symmetries. For instance,
of the symmetries derived from tbe standard model, just CPT invarianee, baryon
number conservation and the conservation of the lepton numbers for each seneration
are exa.ct (apart from the effects of the weak anomaly, which are neslisible at
normal temperatures 2 ) . C, P and T invariance, chiral symmetry (includins Gell-
Mann SU(3) symmetry) and flavour conservation are all approximate symmetries .
The examples taken from macroscopic physics and atomic physics are also only
approximate symmetries. The scale symmetry of macroscopic physics is valid to
the extent that the number of molecules is very large; the decouplin!!: of spin in
atomic physics is only true ignorin!!: relativistic effects; the 0(4) symmetry of the
hydrogen atom, or of the sun-planet system, is only valid in the presence of an exact
non-relativistic l/r potential. The SU(n) symmetry for an n-dimensional harmonic
oscillator is exact but does not eorrespond to a real phYlical system.
We have also considered, in Chapter IV, the derivation of some symmetries which
have been suggested as relevant somewhere beyond the standard model . These sym-
metries turn out to be exact within the models considered, but many of them must
be spontaneously broken if they are to be relevant for nature. For the speculative
derivations, in Chapter VI, of Poincare invariance and gause invariance, we have
considered both exact and approximate derivations (the formal approach and the
renormalisation group method).
So we see, from the above examples, that weakly broken symmetries have a
natural understandin!!: when symmetries are derived . As long as the Iymmetries
were lupposed to be a fundamental property of nature and treated &I input to a
theory, it seemed rather mysterious why, say, parity or Itrangeness conservation was
violated by just the weak interactions. Now that parity and stranseness conservation
are derived as consequences of the theory of stron!!: and electromagnetic interactions,
there is no II priori reason why tbey should be conserved by the weak interactions.
There are essentially two views on the role of symmetries in physics. As sum-
marized by Weinberr : it is a question of whether or not you think it is the job
or physics to explain symmetries or to explain their absence. Weinberg's point of
view is that 'ynunetries ha"'e to be explained and if there is no good explanation
for them you should not believe them. The alternative view, expressed by Pati 4 at
the same meetin,;, is that you should look for the maximum p068ible symmetry and
then, whatever you do not see in nature, explain how it got broken. In this latter
approach, symmetries are provided with a kind of "holiness'". An example is the
132

aeathetic desire to restore parity invariance u a fundamental symmetry at short.

distances, despite its violation by the weak intera.ction, in the len-right symmetric
SU(2)LX SU(2).x SU(4), mod,\.'
We consider that the above statiatiCi on symmetry derivations ahow that .Y~
metries are really not so holy. The origin of so many symmetries ia now understood
that tbey should not be considered as fundamental or mysterious. So we support
the point of view, expressed by Weinberg, that symmetries ought to be explained
and not just imposed as • priori fundamental principles. We therefore conclude
that there is no good philosophical basis for introducing ten-right symmetric mod-
els purely on the puunds of impOling parity invariance as a fundament.al Iynunetry;
there may of course be other motivations for considering IUch models.
Induct, encouraged by the lueceu in undentandin& symmetries, we have gone
further ~han Weinbers and suuetted tha~ all the symmetries of the laws of nature
may be derivable; in Chaptu VI we questioned the rundamental nature of transl..
tional invarianee, Lorents invarianee and Sau&e invariance. These three symmetries
are conventionally &Illumed to be fundamental, reftectins the seometrical structure
or space-tUne and the dynamics needed for a consistent relativistic quantum field
theory.
We have understood five out of eight of the symmetries of micrOlGOpic physics
within the standard model, which in all likelihood is the low energy limit of some
more fundamental theory. Thue is therefore a hope that the three rem&ining sym-
metries may be explained by new physics on the way up the quantum staircase, from
the eledroweak W: , zO scale or 10- 18 m to the presumably fundamental Planck
seale of 10- 35 m. The seventeen orden of mApitude change in scale is similar La
that in p&llling from macroscopic physics, at the 1 m scale, to the electroweak scale.
We see, from Fig. 1.1, that the latter change in scale corresponds to around six
Iteps, or the development of six fields of physics, on the quantum staircase . There
is thus plenty of room, up to the Planck IIl&II or even beyond, for the development
of new physics and the derivation of the fundamental symmetries of Poincare in-
variance and sause invariance, even if the speculations of Chapter VI are not taken
seriously.
This scenario, where all the Iymmetries of the lawl of nature can be derived,
leads us naturally to a discUllton of random dynamics.
1.2. Random Dynamics
As discussed in the previous sedion, the progress made in derivins 110 many of
the observed symmetries of nature is very suggestive of the idea that, as we learn
more about physics beyond the standard model , we shall learn the oriPn of all the
known symmetries. It is then even natural to go one step further and speculate that
all the regularities, i.e. all the laws of physics known today, should similarly have
an explanation. This is the random dynamics hypothesis: all phenomenologically
found regularities have lOme explanatton, i.e. there is some mechanism responsible
for making the known laws of nature valid in at least some domain . Typically
this domain is the re,;ime of low energy physics defined as, say, the experimentally
accessible energies below 1 Te V.
The existence of a mechanism for deriving lOme symmetry or other regularity,
such as locality or the principle of superposition , impliet that almost any theory
 133

propoeed at random will pOllse. the regularity in, say, the low ener&), domain. If
indeed there is such a mecllaniam (or each of the known regularities, one could take
a1moe~ any model and all these replaritiea would appear in one or more limite, such
as at. low energy. However it would tben hardly matter what the fundamental 'the-
ory' is; it. would anyway give rile to the observed re«'liarities. This would mean that
almost any theory could explain 'everything' known and thus be good enough as a
theory of everything (TOE) . There would then be no reason to accept any special
model or TOE as the truth; it would be better to imagine tbat the most funda-.
mental physic. is chaotic - a random model - or noL to a.saume anything about
it. This is really the content of the random dynamics hypothesis: the fundamental
physics, or TOE, does not matter, since almost all models at the fundamental level
will have sufficient structure that they agree with the phenomenologically observed
regularities.
We may take the above random dynamics hypothesis as a replacement for a
fundamental TOE . In practice all one would gel out of a fundamental theory would
be the low energy regularities; the rest of a TOE cannot be realistically tested
anyway. If it is indeed true that we do not need any special fundamental model,
in order to obtain the low ener«y regularities, the existence of such a TOE might
just be a superfluous hypothesis . We could just as well chOO8e a random theory
from a very large class of dynamical models. This is really a rather simple ides and
random dynamics provides a realistic alternstive La superstring theory or any other
TOE .
If, for any reason, the fundsmental dynamical model was very complicated ,
random dynamics would seem to be the only practical approach. We would have
to rely on the hypothesis that the observed physical laws are stable under changes
of the fundamental model. This basic philosophy' was proposed1 several years
ago [Paper 301. A similar philoeophy can be found in some other approaches to
the laws of nature.' In particular, recent developments in quantum gravity lead
to a remarkably similar picture; this baby universe theory is discussed in the next
subsection.

1.t.1. 806, wniver,e theory ,ugge,ting random dynamic,

The idea that physics is chaotic at the fundamental level is supported by recent
progress in quantum gravity: baby universe theory.' When quantising Einstein'.
theory oC gravity, using the Feynman path integral prescription,IO one functionally
integrates over all possible four-dimensional Riemann spsces. It is then natural , at
least in the Euclidean Connulation where a topology change is not accompanied by
a singularity in the metrie, to include Riemann spaces of all possible Lapologies;
quantum mechanically even the topology of space-time is expected to fluctuate .
In this way one can argue for the existence of a 'spacetime foam', which has a
complicated topological structure at the Planck scale C;Jl .... 10- 33 em.
The topological structures relevant to baby universe theory are wormholes, which
provide a microscopic connectwn between two different, otherwise weakly curved,
,pace-time manifolds; alternatively a wormhole may connect two well separated
regions of tbe same manifold. Such a baby universe wormhole may be visualised,
by imagining that two small4-volumel of space-time are removed from the _-eakly
134

curved Riemann manifold(s) and a tube of the topological shape 1 x 53 attached

in their place. Here I is an interval of t.he real axis and 53 is the sphere in a four·
dimensional space (but S, could be replaced by another compad 3-dimensional
dosed space). It is the 5 3 -shaped 3-spacc which is known as the baby universe and
it is assumed to have a size of order the Planck length. The wormhole corresponds
to the exchange of a baby universe from one 'footprint' to the other on ~he weakly
curved space-.time manifold(s) .
So we are led to consider eontributioD8 to the Euclidean path inleuaJ from eon-
figurations consisting of many weakly curved space-time manifolds, interconnected
by a large number of wormholes, as in Fig. 1.1; one of these manifold. represents
the space-time development of our own universe . This mt:ans that 'ou r universe'
is being bombarded by baby universes from 'outside space-time'. Baby universes ,
just like our own universe, have no position ; position only makes sense inside our
universe or inside a baby universe . In order to describe, quantum Jne(;hanically,
the effects of these baby universe deUees of freedom , it is convenient to introduce
operators, a,+ and a.. for c~at.ing and annihilating whole baby universes. Here the
label i denotes the interior state of a baby universe , specified by ita size, geome-
try, contents etc. These 'third quantised ' degrees of freedom, ot and 0" are not
functions of the usual quantum fields 1/1(%) defined at space-time points %: they are
degrees of freedom from 'outa.ide space-time'. We shallllC'C, in Sec. 7.2.2, that the
more speculative ideas in random dynamics also suggest the existence of degrees of
freedom outside of space-time.
There is an interaction between the usual quantum field theory degrees of free-
dom and the third qUlI.Dtised degrees of freedom, where II. baby universe unites with
our own universe or with another weakly eurved universe . Since the third quantised
degrees of freedom have no location in space-time, it is not surprising that the effect
of this interaction turns out to be the same over all of space-time . In fact it turlUl
out to renormalise the parameters, i.e. the coupling colUltanls ~i ' of the quantum
field theory in our universe.
The general form of a 'fundamental' field theontical Lagrangian density for our
. .
uDlverse 15
(I)

The summation is over all possible local functions c." constructed from the metric
g(%) , the fields 1/1(%) and their derivatives, which are consistent with the gauge sym-
metries of the theory. The label i runs over infinitely many terms, since we include
non-renormalisable interactions which are irmevanL to low energy physics; of course
many of the coupling constants ~i could vanish . This fundamental Lagrangian den-
sity C.O is modified, by the inclusion of the effects of baby universe emission and
absorption, into tbe effective Lagrangian density

C• .(z) = £o(z) + ~)at + a;.)C;(;(Z) ,9(Z» (2)

;

for distances large compared with the wormhole scale. We may ben think of the
operator 0;+ + OJ. , where i· denotes tbe CPT conjugate of a baby universe of type
i , as just some dynamical variable representing a third quantised degree offreedom.
 135

Fi,. 7.1. A typical m~tion 0{ ...onnbole QClJ\neCtio..... between weakly curved .pace-time
manifolds.

The dynamical variable 4;+ +0;. is independent of space and time; once its physical
value a j is meaaurea, it will have the measured value over aU space forever . Thus the
effeet of baby universes is just to renormali&e all the fundamental coupling constants
A; . so that the observed coupling constanta become Ai + OJ.
The blatant nonlocal behaviour of wormholes is therefore effectively absorbed
into ·.hift.a of the fundamental coupling constants; this is because the Einstein field
equations guarantee that a baby universe cannot carry away any 4-momentum from
a weakly curved universe. For example, the Newtonian gravitational field emanating
from a region keeps track of the amount of ener&), within it; 80 if some energy
suddenly disappeared down a wormhole , the corresponding Newtonian field would
disappear violatins the Einstein equations. Since a baby universe does not take
up any 4-momentum, it must enter our universe with the same amplitude at any
space-time point, as a consequence of Heisenberg'S uncertainty relation. So a baby
universe couples with the same amplitude all over space-time, and this is simulated
by a change in the coupling constants in the effedive Lagrangian density Cetf.
The faet tbat the wormhole corrections 0'/ to the coupling eonstants Ai appear
as quantum mechanical dynamical variables means that, in I\:eneral, their measured
values are unpredictable. Before a measurement of the coupling constanta, the
initial state of the universe is not expected to be an eigenstate of the operators
136

~+ + Cli' ,
For instance the Hartle-Hawkins initi&l. ataLe condition lO corresponds to
lS8uminS that there are no baby universes in the initial state; this means that the
third quantised degrees of freedom are M first in harmonic OIICilIaLor ground states
10), satisfying a;lO) = O. The effective coupling constants Ai + 0; beeome a prio,;
random variables.
This predicted randomness of the effective coupling constants.\; +Oi haa &everal
interesting consequences:
1. Random dynamics may b e cOn5idered to be a consequence of baby universe
theory. A theory baving random couplinp i. a lood step in the direction
of being a random theory. According to t.his attitude, random dyn&miea
would be an effective theory applicable bdow the wormhole ~ersy sale.
It .....ould not be the fundamental underlyin, theory ; the liause poup and
matter degrees or freedom may not be random.
2. Purely global conservation lawl are violated by baby universea carrying a
non-zero global charge. Then 0 priori vanishing constanta, >'J = 0, which
measure the stren&th of the gobalsymmet.ry violating interactioDll, are renor-
maHsed by the impact of such charged baby universes; the effective symmetry
violating coupling constanta >./ +0'1 = O'J are non-vanishing in general. How-
ever gauged charges remain conserved even after wormhole renormaliAation;
a gauged charge cannot escape into a baby universe without upaetting the
Coulomb field len behind . Thus a baby universe is forbidden to carry a
gauged-charge just as it is forbidden to earry 4-momentum. It follows that,
according to quantum gravity, we do not expect any of the observed global
conservation laws to be fundamental. A &lobal conservation law must. have
some explanation at an energy lower than the wormhole scale, as the baryon
and lepton number conservation laws, discussd in Chapter 3.3, are under-
stood in the standard model.
3. Discrete symmetries are similarly violated by the effects of baby universes.
As an example we may consider parity in a lell-rjpt. Iymmetric model, which
I priori has no parity violation:
Let us imagine we bve one Higgs field for t.he right.-handed gauge field and
another Biggs field for the ieft.-handed gallle field, each havin& the aame
Bigs potential. This meana that one superposition tP.(z) of Biggl fielda
is a scalar and the orthogonal superposition ';,.(z) is a pseudOlcalar. The
branching off of a baby universe, containing one scalar and one pseudOlcalar
Higgs (or anti-Hias) particle, induces a term of the form O',..41,.(z)I/I.(z)
in the effective Lagrangian density 4.ff. This effective interaction violates
left-right Iymmetry; the 'tachyonic' maas degeneracy of the Higgs part.icles
is broken and parity is not conserved . So baby universes will, in general,
lead to violation of all discrete symmetries. Any obaerved discrete symmetry
must be understood as the result of some mechanism, operating below the
wormhole energy scale, which forbids the presence of synunetry violating
couplings. An example is the explanat.ion of charge conjulat.ion and other
discrete symmetries in the strong and elect.romagnetic interactions provided
by t.he standard model, as discussed in Chapter 3.2.
4. Any t.heory of everything, TOE, disappears down the wormhole, in the sense
 137

that it is impouible to teat ita ccuplin, constant prediction. due to their

random renormalisation, ..\i _ ..\; + OJ . It may .till be possible to partially
telL, say, some superstring theory via. iLs gauge group and matter field gauge
SlouP repreuntatKm.s. However the inability to predict the constanta of
nature is a severe blow to any theory claiming to be a fundamental TOE.
5. Random dynamica may be considered to be the fundamental theory beyond
Ott wormhole enetQ seale. According to point 4 the theory beyond t.be
wormhole Kale becomes rather insignificant . Here we go further and 8uggeet
that, fOl' physics observed at experimental BCaies, it does not matler at all
what the Cundamental theory may be; we might a.I well imagine it to be a
random model. According to this attitude, random dynamics would be more
fundamental than baby universe theory. This should be contrasted to the
attitude of point I, where random dynamiCi is just an etl'ective model approx-
imatin& the effects of quantum &ravity. The derivations of &aUJe symmetry
and the field theory &Iass scenario, discussed in Chapters 6.3.1 and 7.2.3,
are only relevant in this second attitude, where random dynamics is oper-
ative beyond the wormhole scale. Of course diffeomorphism symmetry and
quantum ,ravity must then he hoped to arise automatically (rom a random
pre,eometric model, as speculated in Chapter 6.2.1.
The renormalisation and consequent intrinsic uncertainty of all phyaical param-
eters seems to be the InOfIt reliable result derived from baby universe theory. A more
controversial reault is the daim ll •12 tbat the probability diatributions for the values
of IIOme, and posaibly all, of the physical parameters ~ + OJ are very sharply peaked
and thereby effectively fixed . In particular the cosmological constant is predicted
10 vanish and potentialsoiutioDs are sugested for other fine tuning problema: the
value of the topolopcal 8 parameter in QC 0 1S and tbe smallness of the weak inter-
action mass scale, and Higs mass, compared to the Planck maaa.12 If successful in
predicting the correct value of tbe 8 parameter, baby universe theory would explain
the orilin of CP symmetry in strona interactions without an axion .
The probability diatributions for the values of the physical parameters are calcu-
lated usina tbe euclidean path integral approach to quantum gravity. The summa-
tion of wormhole effects introduces a sharply peaked atra factor into the probability
distribution for the parameters aj , in addition to the factor arising from the ansatz
for the third quantized wave function of the universe. This extra factor in the mea-
sure for the integration over the third quantised variables 0' = {O';} dominates the
partition function

Z(a) '" oxp U

[ox p "'z,/ii(g ,a)) 1 (3)

p
= oxp [ex (SGN'(:)A(a)) 1. (4)

Here £(g, a) is tbe effective Lagranpan density as seen at very large, i.e. astronom-
ical, distances where the geometry of space-time is measured. £ (g, a) is supposed
to include quantum corrections from all the different interactions to all orders in a
loop expansion. Expandin, t(g,a) in powers of the Riemann curvature, the two
JJ8

leading terms are

(5)

where A(a) is the cosmological constant and ON(a) is Newton ', gravitational con·
stant. (We note that the usual cosmologic.aJ. constant of ,eneral relativity is defined
to be ~ = bONA .) The integration in Eq. (3) is over the space-time of a large
smooth universe, with the topology of a 4-sphere, conneded to our own universe
by wormholes.
The partition function Zeal clearly blows up when G~(Q)A(Q) = O. Thus
values of the OJ for which the long-distance or 'dressed' cosmological constant A(o)
vanishes are overwhelmins:ly favoured . This is how baby universe theory solves
tbe c06mological COnstant problem: why is space-4ime 80 flat. tha~ the to6ffiological
constant satisfies the astronomical bound G1A < 1O- 120 ,? The wormhole solution i.
to provide a distribution of 'dynamical ' coupling constants Ai +Oi , which is strongly
p~ak~d for Gk (a)A(a) v~ry small and pOl!litiv~ ; so th~ bound Gk(a)A(o) < 10- 120
is exceedingly Iik~ly to b~ satisfi~d .
In addition t.o a strongly p~ak~d distribution for C'iv-A, the rapid variation of
Z(o) should in principl~ d~termin~ all th~ oth~r physical couplinS constants. This
'big fix' of the physical parameters would effectiv~ly remove their randomness; the
version of random dynamics considered under point 1 above would disappear down
the wormhole. However the attitude t.o random dynamics taken in point 5 would still
survive; random dynamica at the fundamental level could lead to the appearance of
an effective diffeomorphism gauge group and hence of quantum gravity and baby
universe theory. Random dynamica misht also predict the gauge groups and their
matter field representations.
The above 'biS fix ' of the cOl!lmological constant and the other physical cou-
pling constants has been subjected to a number of serious criticisms. A C&r~ful
re-analysis 14 of the calculation of Zeal reveals a missing minus sign, SO that the
previously dominant fact.or in the partition fundion, Eq. (4), now becomes

(6)

spoiling the argument for G~A = O. It is important for the derivation of this result
that the inner exponent

(7)

should be real . This exponent is a large number and just a small phase, from say
a tiny CP violatins 8-term, might easily compensate the minus sign in Eq. (6) .
In fact such terms would occur in complex conjugate pairs in the path integral,
corresponding to the two orientations of the 4_sphere,lt and would introduce a
real prefactor / (0) multiplying the inner expo nential in Eq. (6) . If /(0) could b~
negative , the values of 0 for which l ea) is negative would be favoured and the
argument for G~A = 0 reinstated . However there is a general argument ,14 based
on the hermiticity of the Hamiltonian, that matter fields can only give a positive
 139

prefaet.or I(er) . The vavitational field haa to evade this general argument , in order
to justify the introduction of the minus sign itself in Eq. (6) , and can only do so
due to the instability of the Euclidean action .
Another criticism is the claimed presence of wormholes of arbitrarily large scale,1$
which would imply strong non local interactions over arbitrarily large distances.
There are arguments t.haL this catastrophe could be averted .1$
We musLconclude that the status of the predictioIUI of the values of physical pa·
rameters from baby universe theory is very uncertain at. present. Indeed , apart from
the v&l.ue of the cosmological constant, the predicted values are not quite correct.
The effective topological 8-parameLer (sec Eq. (92) of Chapter 3) is predict.ed 13 to
=
take the CP-conser ving value "9 'If , up to weak interaction corrections, rather than
the phenomenologically preferred value 8 ;: O. Again although scalars by themselves
may become massless according to ba by universe theory,12 the origin of the sma11
eledroweak scale relative to the wormhole scale is not yet understood. Nonetheless
baby universe theory may well earry a good deal of truth, since it should be a con·
sequence of almost any theory of quantum gravity. Here we mainly eonsider it , see
point 5 above, as a motivation for random dynamics as the fundamental underlying
theory.

7.2.2. The fir81 ,'ep. in nlndom dyno:miC6

The s tarting point of random dynamics should be the consideration of as large a
class of conceivable 'tbeories' or ' models ' as possible. A probability measure should
then be chosen on the set of models. The random dynamics hypothesis (Paper 30J is
that a model chosen at random from this set will almost alwaYI contain the empiri.
cally known lawl of nature; i.e. all the presently known laws of physics follow fr om
almost any model, which is complicated enough , provided we go to some limit, gen-
erally the low energy limit. In order to verify this hypothesis, one shou ld in principle
progress by developing a series of physical models, successively approximating one
another, as one goes down the quantum staircase of Fig. 1.1. One I Uch intermediate
model is the so-called quantum field theory glass introduced in Chapter 6.3.1. In
this subsection we discuss how the quantum field theory glass might be derived from
random dynamics, while we consider the consequences of the quantum field tbeory
glass in the next subsection. The content of this subsection is t.herefore the part
of random dynamics furthest away from experiment and, consequently, the least
studied and most speculative part of the subject. So it contains difficult material
and may safely be omitted by the reader.
At first , it may seem rather hopeless t.o try to derive low energy physics, or even
the quantum field theory glass, out of a totally random mathematical model. How·
ever we believe some concepts are so general that we can hope to find them relevant
in almost any sufficiently rich and complicated 'model'. For example it should be
possible to introduce a crude concept of distance, or of some sort of topolo~, on
almost any mathematical structure S which contains a huge number of 'elements'
having relations between themselves: the longer the chains of intermediate elements
needed to establish a rdation between two 'elements', the further apart the two 'el-
ements' are defined t o be. For sueh a mathematical structure, rich in the sense
of baving a huge number of elements and also being rather repetitive, it must be
140

poaible to define the concept of small modifications Il, h . .. of the structure S. This
should be a modification which only chanties the number of repeated i uhltructurea
by a relatively small amount. Assuminr; that the random model obeys some axioms,
which are then of course also random, the system of small modifications will apprO'f-
imately make up some Abelian sroup; indeed we expect that the most important
small modifications could be naturally organised as a vector space V .
U.inr; the approximate topology suggested above, we may consider the restric-
t.ion of an allowed small modification to some neighbourhood with respect to the
Lopo)oK)'. We want to interpret IUch small modifications as wave functioDs of the
Wheeler-DeWitt type,IG which describe the situation for all times, acting as follows:

(8)
Here N is an integer which p06llibly depends on where we are in the topological
space Sr of the atrudure S . The component. of ..p measure the VafioUi amount.
of sm&ll modifications of different type in the 'region ' in question. The axioms will
impose linear relations b etween the simultaneously allowed small modifications in
neighhouring regions of the structure.
We define a formal restriction biG of an allowed and meaningful modification b
to a neighbourhood G £ S as follows : biG is that small modification which equals
b for the part of the structure inside G and equals no modification outaide G . This
formal restriction of a small modification will of course introduce problems, e.g.
disageement with the axioms OIl the boundary of G. Nonetheless these formal r&
s,"rietions biG may be used to generate an extended vector space "Ii. This extended
V«WI apace V _nlially has the (.haradeI oC a spac.e of CUD(.tioDS 1/1 on the wpo-
logical apace Sr . The truly allowed small modifications now appear as a subspace
V allowed by the 'axioms'; the lubspace can be charaderized by a set of corulraint
equations

where the .i are linear Cunctionals.

We required above that the modifications be small enough to guarantee linear-
(9)

ity and the existence of a vector space "Ii. There may therefore be a problem in
restricting suc.h a modification to a very small region G: if G is itself supposed to
be a relatively simple substructure the effect of any modification on it is likely to
be quite severe. A remedy for this problem is to choose regions , such as G, that
are sufficiently large as to contain a fair amount of repetitive structure. The region
G could still be extremely small compa:ed to, say, all of S and be approximately
a point in 5T. The physically relevant states '" should be superp06itions over 50
many local contributions, like biG, that the need 1.0 modiCy the same region G twice
practically never arises.
\Ve have now a:gued that one can construct, out of almost any random mathe-
matical structure S, a vector spa.c.e or Abelian group of 'wave Cunctions' l/i. These
wave functions " are identified with small modifications and constraints ensuring
compatibility with the axioms not only locally but also globally. The wave fun c-
tions are defined on a topological space Sr constructed Crom the structure S . The
constraints are local, since the I.opology is defined by means oC the relations between
the elements.
 141

Although we have not yet introduced space-time, we may obtain a very gen-
era.! form of quantum mechanics from the mathematical structure provided we can
interpret some of the constraint equations, ¢Ii'" = 0, as a sort of Whceler-deWiU
equation. The Wheeler-deWitt equation of quantum gravity is noL an equation of
time evolution in the usual aense; in Cad time does not appear at all and it takes
the form of a Hamiltonian constraint equation. IS

'I/~ =0. (10)

Time must be interpreted as a description of the correlations between variables

in the Wheeler-deWitt wave fun ction. In this int.erpretation 17 one postulates the
existence of a weakly interacting, more or less classical, variable which plays the
role of a physic&i clock. One then essentially takes the clock Hamiltonian density
?ldodt to be conjugate to the variable h describing the position of the 'hand' of the
clod:
I 8
1{clodt = Pit = i lJh . (II )
The Hamiltonian constraint operator is decomposed into the form

'H = 'Hclodr. + 1{ral (12)

and the Wheeler-deWitt equation becomes

(13)

The constraint Eq. (13) can now be interpreted as the SchrOdin«er equation for the
rest of the variables. In the same way we want to interpret some of the constraint
equatiollJl of our system, ;j1P = 0, to be a generalised Schrooinger equation.
The next step in such a random mathematical model is to introduce a space-t ime
concept. The identification of physical objects in the general mathematical structure
S is of course essential , if there is to be a physical interpretation of tbe structure.
These interpretive assumptions are to a high degree arbitrary, particularly in the
first st.epa of random dynamics. We bave already assumed that the vector space of
small modifications of the mathematical structure is to be identified with the Hilbert
space of a generalised quantum mechanics. As a further tentative interpretation,
let us now assume tbat the tQpolo«ical space ST is the configuration space for
a somewhat extended quantum field theory describing nature. In other words, we
imagine that eaeh point , i.e. each element, in Sr corresponds to a field configuration.
We would now like to consider the quantum field as a fu nction mapping all of
space-time into a target space. In this way we treat space and time on an equal
footin« and the equation or motion fot the field will take the fOlm of a constraint .
The interpretation of the configuration space as a space of (unctions requires
that ST is a very large, or infini te, dimensional space. The field values over all
space-time are then coordinates for the topological space ST , which must therefore
essentia1ly be a differentiable manifold. So we assume that it is possible to construct
a tangent vector space to Sr . For this purp06e, it is necessary to have an additive
142

composition law on the tiny neighbourhood of elements in Sr near to any element

A of the mathematical structure; this neighbourhood is then approximately the
tangent plane. The composition law AB + AC = AD is naturally defined by the
requirement that the same small operation is needed to go from A to B as from C
to D . When the points A , B ,e and Dare clo6e t.o each other, the operation needed
to go fr om onc of these point.s to another has so little effect that the order of the
operations should nol matter . This means that the composition law is commutative.
We now need to introduce a topological structure on the basis vectors or 'di-
rections' in the tangent space, 80 that we can identify the coordinates in ST with
points or small regions in a pregeometrical space P . We can think of the set of all
the various 'diredioDs' Of ' vectors', in the tangent space of the confi,;uration space
Sr, as another example of & complicated mathematical structure. So we argue, as
we did for ST , that an approximate topology can be introduced , by saying that two
'elements' or 'directions' in the tan,;ent space are close together when there is a
relatively simple operation connecting one to the other. Once we have introduced
such a topology, the 'directions' in the tangent space can be ,;rouped into neigh-
bourhoods or small overlapping subspaces. These selected subspaces are then to
be associated with small re!ions in a pregeometrieal space-time P . If the above
construction is possible, we have associated the field values, i.e. the coordinates
of 5-r , with small regions of space-time, ensuring a 1o<:al field theory in the long
wavelength limit .
The next step is to assume that it is possible to identify some configuration as the
vacuum state and find an approximate symmetry, under some ,;roup of ,;auge trans--
formation s, in the configuration space close to the vacuum. It is then hoped that
the mechanism described in Chapter 6.3.1 can be used to promote this approximate
symmetry group to an exact formal symmetry which , via quantum Huctuations,
is realised physically as a local gauge symmetry. Among these gauge symmetries
there may be some which have the structure of a diffeomorphism group, transform-
ing some of the points of the pregeometric space P into each other. In addition we
might hope to find some Yang-Mills type gauge symmetry as a subgroup at each
point in P.
We can only give a rather weak argument for the presence of the diffeomorph ism
group. It is based on the remark, in Chapter 6.3.1 and the following subsection,
that the field theory glass transformations, which survive 88 low energy symme-
tries, should be rather uniquely implemented with as few outer automorphisms as
p068ible. We suppose the Yang-Mills gauge transformation voup appears in the
pregeometrical space as a cross product of extremely many isomorphic factors, one
for each point in P . The presence of the diffeomorphism uoup is then favoured , in
the sense that it converts many of the permutations of these factors from outer to
inner automorphisms.
If the diffeomorphism part of the gauge group transforms the pregeometric points
in the usual way, we can identify these points with the points ofspace--time obser ved
phenomenologically. The orbits of these points under the diffeomorphism group
would then be identified as space-time manifolds.
It is not necessary that all the points in the pregeometric space-.time are ~
ciated with an orbit for the diffeomorphism part of the group of gauge transforma-
 143

ions; it eould happen that some of the points are invariant under diffeomorphisms.
Such trivially transforming points would have their associated degrees of freedom
Jut.side of space-time, in the &arne sense as the third quantised degrees of freedom in
~aby universe theory. We therefore expect these diffeomorphism invariant degrees
of freedom to act in the same wayan aU space.time points and, for practical pur·
~oses, just to influence the values of the parameters in the effective laws of nature.
~ this feature of exterior degreell of freedom is e.xpeded in random dynamics even
il, for some reason, genuine wormholes should not form. or course we expect that
~andom dynamics should lead to quantum gravity and hence to the existence of
I)aby universes. As discussed in Chapter 6.2.1 , general relativity result.s from the
~ontaneou8 breakdown of the above diffeomorphism gauge symmetry to an exac~
~oincare gauge symmetry.
t We have given above some preliminary ideas on how physically relevant conclu-
SlOns might be deduced ft om a random complicated mathematical model. There
ke clearly many gaps still to be filled in the argument . Our main motivation , in
~resenting such an incomplete discussion, was to convince the reader that it is not
~ forlorn hope to derive physical laws from a general mathematical model. Indeed
if something like the above construction is made to work on a general mathematical
s}ructure, we would be close to deriving the fi eld theory glass model of Chapter
q.3 .1 , with some gauge symmetries and some extra degrees of freedom outside of

~
ace-time . However we also introduced a diffeomorphism gauge symmetry and
us, in the sense of Chapter 6.2.1, gravity and translational invariance. Of course
e field theory glass is not translational invariant and we thereCore seem to have
derived too much symmetry at this stage. But this is presumably harmless since,
ib the end , the field theory glass is supposed to be smoothed out , by quantum flue-
thations, into a continuum model with reparameterisation symmetry. We had to
~ppeal to similar quantum ftu ctuations to obtai n diffeomorphism gauge symmetry,
+ d hence translational invariance, from the pregeometrical space-time P . 50 we
lually expect to obtain the continuum version of the field theory glass and the
iffeomorphism gauge symmetry Cra m quantum flu ctuations , on the same sLeps of
e quantum staircase .
The above discussion of the origin of space-t ime did not add ress the question of
1 dimensionality and how the diffeomorphism group of 3+ 1 dimensions is selected
i a random theory. We speculate below that fermionic degrees of freedom might
~lay an important role in singling out 3 + 1 dimensions .
. 1 So Car our discussion has been very abstract and we have only outlined how ,
in a very general model, we might identify some fundamental physical notions:
~ linearity in quantum mechanics, b) time, c) space- time and d) locality. We
'Y0uld now like to present some slightly more concrete derivations of some of these
~otions. These derivations will be based on different assumptions, which deviate
i detail from the above ex~ingly speculative discussion. This is consistent with
t e spirit of random dynamics, according t.o which the final physical results should
e insensitive t.o the input assumptions .
. QUClntum },ftchanic~

First we shall consider a derivation of quantum mechanics or rather the linearity

the 5chrodinger equation. A linearity rule is generically explained by assuming
144

analyticity and uain& a Taylor expansion; it is then ncceesary to IlflUe that the
const.ant term in the expansion is zero or should be subtracted away. and that all
the terrns beyond the linear term are ncr;li«ibly small. It was precisely t.he smallness
of the small modifications, which was crucial for t.he speculat.ive identification of
quantum mechanics in the random gcncral mathematical structure discussed above.
A. the starting point (or our derivation of quantum mechanica, we assume the
existence of a time variable t and that the evolution of the world is described by a
general autonomous differential equation

dX(' ) _ F(X(I)) (14)

dt - .

Here X(t) has a buge number of components describing the state of the physical
world. Time-translational invariance has heen assumed 80 that the velocity field
F(X(£)) only depends on X(t }. In the spirit of random dynamica, F (X (t» can now
be taken to be a randomly chosen analytic field, in a computer simulation which
looks for the typical behaviour of the solution to Eq. (14). With a measure chosen
80 that a random number generaLor creates fields F{X{t » with many zeros, the
typical behaviour for the time development of X(t) turns out lS to be the approach
to a fixed point Xo :
X(f) _ Xo as t _ 00 . (15)
It follows that, after a long time, X (t) is very close to the fixed point Xo and
the Taylor expansion

dX
dI = F(X(' ll
8FI
= F(X,) + (X(.) - X,) 8X x. + ... (16)

is a good approximation , retaining just the fint ~wo terms. At the fixed point Xo

F (X,) = 0 (17)

and we can rewrite Eq. (16) in the form

( 18)

when we define
w(' ) = X(') - X, ( 19)
and
H = .8FI
Iax x•. (20)

This notMion suggests that we should interpret !/J(t) as the wave function for the
world and H as the Hamiltonian operator, so that Eq. (18) can be identified as the
SehrOdinger equation .
 145

However there is no reason why the matrix H should be hermitean and ita
eigenvalues ~ will in general be complex. The sLability of the fixed point Xo requires
that
Im.\$O . (21)
The time development of the wave function can be expanded in terms of the eigen 4

vectors ¢" of H:
0(1) = L, ,-lA'., . (22)

AI time P&SSeI , all the components in the expansion wiU die out, except (or th()8e
with eigenvalues having a numerically very small imaginary part 1m >. ~ O. It
follows that only the almost real eigenvalues ate relevant after a long time and then
the Hamiltonian, restricted to this subspace of wave fun ctions, ma.y appea.r to be
hermite&n to a very ,ood approximation . However, in addition to the reality of the
eigenvalues, hermiticity requires the eigenvectors to form a complete orthonormal
basi•. It may therefore be necessary to adju.st the definition of the Hilbert space
inner product, so as to make the Hamiltonian H more closely hermit.ean with respect
to it.
The &«e of the universe is indeed very large in Planck units. Thus we can
claim to have derived some of the general principles , i.e. linearity and hermiticity,
of quantum mechanics from random dynamics; although hermiticity is not fully
derived. It is actually a prediction of random dynamics that the hermiticity of
the Hamiltonian is not aut, and thereby ' the probability of living at different
moments of time' is not exactly constant. Experimental tests looking for a p068ible
small antihermitean part of the Hamiltonian ,I ,lt and also for small deviations from
Iinearity,2° are therefore important.

B . J + 1 DimelUions of Space-Time
We have already suggested that space-.time could result from the appearance of a
diffeomorphism .yrometry in a rather general pregeometric space. This mechanism,
however, would Dot. priori explain why the dimensionality of space-time should be
3 + 1. In a slightly different scheme, there is a random dynamics argument for space-
time having precisely 3 + 1 dimensions21 [paper 30]. This argument presumes the
existence of very general fermion quantum fields in a random non· Lorentz invariant,
but space-.time translational invariant , model. This model would therefore Dot be
viable if indeed Lorentz invariance and translational invariance appeared together,
as special diffeomorphism symmetries in a pregeometrie model.
Assuming space-.tirne translational invarianc:e. but not rotational or Lorentz in-
variance, the general free fermion field equation in d-dimensions takes the form

(23)

in momentum representation. The assumption analogous to that of bermiticity

for the Hamilt.onian, in the case of a single particle, is that , say, the ener&y Po is
required to be real if the other components of the d-momentum vect.or PI' are real.
This suggests that D{P) should be a hermit.ean matrix, when the components of PI'
are taken to be real. In the spirit of random dynamics, we assume D(p) to be a
146

random or senene matrix fundion of PI" In other words we look (or properties of
D(P) that would be true for whole regions in the space of matrix fUDctions aatisfyinS
D(P) = D(P)+ for real p".
The condition
d, tD(p) = 0 (24)
must be satisfied in order 10 have a solution of the free fermion field equation,
Eq. (23). This condition is generically satisfied on 8ubmanifolds in momentum space
of codimension 1. In d-dimensional space a aubmanifold of dimension d - c is said
t.o have a codimension c. We shall be interested in degenerate solutions of Eq. (23) .
Generically ODC finds q solutiona, with the aame value of PI" on submanifolds of
codimeD.ion 92. This can be seen by taking a basis for ~(p)-8p&Ce , in which the q
degen era~ solutions are taken as basis vectors. In order to stay on the q-degeneracy
submanifold, the q x q submatrix of D(P) corresponding w the q soluttons must
be zero. Thus q2 real conditions must be satisfied, since D(P) is assumed to be
hermitean.
For a. low-energy physicist , the only accessible fermion states are those near the
Fermi surface, which separates empty and filled level.!!l . We now want to consider the
possible relationships between the Fermi surface and tbe q degeneracy surfaces. We
will first show that the Fermi surface cannot consist solely of points in momentum
space on degenerACY surfaces with q ;;:: 3. Around a point in p,.-space with q ;;:: 3
degenerate solutions of Eq. (23) , there will ge.nericaJly be solution branches with
lower q-~uetl . In the nei&hhourhood of a point in p,. -,p ~ with q sol u~ion8 , tbe
relevant part of D(p) is a q x q bermitean matrix. The manifold in this neighbour-
hood with q solutions corresponds to such matrices havins rank q - q. If q - q = I ,
the 9-solution manifold is cut into two disconnected pieces when the q-flOlution sub-
manifold is removed . This result follows beca.use the single non-zero eigenvalue, of
the rank 1 q x q matrix, cannot change sign without passing through zero. However
if (q - q) > 1, tbere are at least two non-zero eisenvalueB and they can easily change
sign, without passing through a p,.-point having two of them zero simultaneously.
So, for example, removing the 3-solution points splits the 2-solution manifold into
pieces, but the I-solution manifold remains connected . We therefore conclude that
the Fermi surface, which must separate empty and filled states, cannot conslst of
q ;;:: 3 points alone.
The Fermi su rface might , however, consist solely of points in momentllm space
with q = 2 solutions . Indeed this seems to be the situation in nature when we
COlUider the fermions of the standard model to be massless Weyl particles, as is
suggested by the fact that the len-handed and right-handed fermions have different
gauge quantum numbers. In nature it appears that the Dirac sea of negative energy
states is just filled and the Fermi surface for a Weyl particle lies at the isolated
zero ener&y-momentum point p,. = 0. For all Weyl stales "ith PI' 'I- 0, the helicity
constraint that the particle should be, say. left-handed allows only one polarisation
state, satisfying the condition
Po = -~ ' f' (25)
Elowever for the zero energy-momentum state, with Po p= = 0, the helicity con-
straint, Eq. (25), does not restrict the spin £. and two states of polarisat ion are
allowed . The fact that the Fermi energies of quacks and leptons now essentially
I IQ

~;ncid' ~
w;Ih Ih, douhl, d,."",,,y po;nl, p, = 0, ;n I"., p.,1 du, \0 Ih,
Hubble expansion of the universe attenuatin~ the fermion number densities. The
k:tual values of the baryon and lepton numbers of the universe are of course not
~et understood .
I Let us now assume that, even in the hypothetical d-dimensional non-Lorentz
invariant theory, the Hubble expansion drives the Fermi su rface towards the su})..
dtanifold of d-momentum space having q = 2 solutions of Eq. (23) . A physicist
~errorming experiments at energies low compared to the Planck mau (here mean-
fug the fundamental mas&) , would on ly have access to states very dose to the Fermi
,furface. So the observed eners,y-momentum dispersion relation should be obtained
from a Taylor expansion of D(p) around values of PI' on the Fermi surface , which
* e have assumed are point.s with q = 2 solutions. Near a q = 2 point we may
~ffectively take D(p) t o be a 2 x 2 matrix , since the solutions in this neighbourhood
k dominantly composed from the two degenerate states. Denoting the q = 2 point
~bout which we expand by P(O) ... , the expansion of the effective 2 x 2 matrix takes
the form
D(P) = (~ ~) + (p, _ P,O),)· 8~(p) I ~ + .... (26)
PI' ". -"("~
The requirement thM D(P} be hermit.ean for real values of PI' implies that t~
I Eq. (26) is hermitean. Sinee there are only four independent 2 x 2 hermitean
Pauli matrices 0'0 = [ ,vI = (f~ , 0'2 = 0', and r
= o'~ , we may write

(27)

Thus the equation of motion for the free fermion field , Eq. (23) , in the neighbour-
.hood of the q = 2 point becomes
I (p, - p'O),)V:.·~(p) = O . (28)
e now renormalise the d-momentum and define the practical momentum to be
PI' = PI' - 1>(0) .... (29)
The equation of motion then becomes the Weyl equation
p"V:(f<>,,(p) = 0 . (30)
The vierbein V,: can easily be transformed away, by definin,; an affine basis in which
' he momentum components are p~ = V,:p" .
I The signature of the metric associated with the Weyl equation, Eq. (30) , may
be deduced from the dispenion relation condition
0= detD(p) (31 )
= det(p... V:O'°) (32)

=det
p,,(Vo" + Va") p,,(V.1' - iV2"») (33)
(
PI'(V1" + iV21') PI'(Vo" - V3 ")
= (pI'VO ... )2 - (p" V.I')2 - (p" V21')2 - (PI' V3" f2 (34)
= PI'P" Vo"V.6 "'1°~ . (35)
148

H.",
"Q ~ = diag(l , -1 , -1 , -1,0, 0, ... 0) (36)
is a metric with the usual (1.-1,-1,-1) signature, except for some degeneracy.
We conclude that the rank of the effective metric

(37)
for the fermion is only four I even when the dimension d of the enerC'·momentum
space is higher than four! The fermion "moves" with zero velocity eomponent.s in
the directions for which t.he met.ric vanishes. Any maLerial made from such fermions
cannot move in those directions. So the directions in which the metric vanishes are
effedively unobservable for purely fermlonic maUer . In this way the appearance of
3 + }·dimensional space-time is favoured.
The boson fields might still , a priori move in all d-dimensions. Similarly different
types of fermion field might move in different seta of 3 + 1 dimelUions. However
there is It. mechanism for aeill!ctingjult one 3 + l.dimensionalspace--time aa relevant
at low energy, if the fermion fields couple to a sause field. The fermion fields then
contribute to t.he renormaliaation of the non· Lorentz invariant couplinS parameters
'1"""" in the d·dimensional non-covariant Yanl-Milla action, analogous to Eq. (20)
of Chapter 6.2.2. The contribution of each Weyl field to the renormalisation sroup
bet.a function P:""" is non·\'aIlishing in the appropriate 3 + 1 dimensions, and the
correspondins componenLl of .". ..,.., in these 3+ 1 dimensionssrow relatively stronger
towards the infra.red. Provided the 3+ I-dimensional con~ributions from t.he various
Weyl fields are not too unparallel in the d-dimensional space--time, they should add
up t.o give a major term in the renormalisation group equat.ion along some 'average'
3 + I-dimensional space and minor terms in t.he other dimensions. The components
of 11"""" in this average 3 + i-dimensional space will then grow large in tbe infrared,
compared to the components in the other d - 4 spaee--time directions.
The coupling pa.rameters rf""" are essentially inversely proportional to the
square of the effective gauge coupling constant. Hence the above renormalisation
group behaviour implies that the effective Yang-Milt. coupling strength will be--
come greater in the extra d - 4 dimensions than in the average 3 + I-dimensional
space. This favours the formation of a ~called layer phase 22 with confinement,
at relatively high energy, in all directions except in those of the average 3 + 1-
dimensional space. In such a situation , there would be fou r·dimensional layers in
the higher d-dimensional spaee--time and particles with non-trivial gauge quantum
numbers would be confined to their layer. Thus the d - 4 excess dimensiOnB would
be unobservable (or gauge--charged matter .
As diseusaed in Chapter 6.2.2, the renormalisation group acale depend~ce will
also make the effective 3 + I-dimensional metrles gP" for the different Weyl fi elds
approach each other, and a Lorentz covariant theory, t.owards the infrared .
Under the influence ofthe low energy BiW vacuum for the e1ectroweak interac-
tions, neutrinos and ZO particles might be able to escape from a 3 + I-dimensional
space--time layer out into the d·dimensional world. However this would be avoided
if the diffeomorphism ~auge invariant gravity theory is also in a layer phase; en-
ergy and momentum are then confined t.o the 3 + I-dimensionallayel$ and particles
would not be observed disappearin~ from our 3 + I-dimensional space-time.
 149

Fia. 7.2. A diffeocnorphWn whio;h only lr.... ronn. po.ihl.& iD • am.tJ.I neishbourbood.

We conclude it is poeaible to obtain 3 + 1 effective apace-time dimensions

~hat
in a natural way from random dynamics.
I
O. Loc41it,
r Aa a third example we consider bow the principle of locality mipt be derived
in, random dynamica. For this purpose we assume that general relativity in 3 + 1-
drmensionalspace-time bas already been derived, to&ethu with ita characteristic
ffeomorphism symmetry and a space-time metric g,. ... (z).
The simplest type efnon-local, but stili reparameterisation invariant, action 5",
tike the form of a lurn of products of ordinary allowed local a.ction expressions:

S., = f: J/.(0,)d' •• C;(0( •• ),8.0(•• »·J/.(",)d'.,C;(0(",),8.0(",»

"=1
.. -j /g(0~)d'.~C=(0(0~), 8.0(.~» . (38)

ere 9 = I del 9,. ... 1and .c:. (¢(y;), tI,.'(II.», for i = 1, .. . ,m, are ordinary reparame-
tf:risation invariant Lagranpan density expreSlionl, dependent on a set of fields
;:(lIi) and their derivatives. Terms formed from an in~eA:ran d dependinA: on ~wo
~iDta z: and 11 in a combined way are Dot diffeomorphism invarian~> because i~ is
J:-ible to make a reparamet.e.risation which translates one of the points but not ~he
other. A diffeomorphism that only makes a chan&e close to one point is illustrated
i Fi&. 7.2.
ISO

In order to obtain a diffeomorphism invariant action, involving two or more

points in a combined way, it is necessary to consider constructions involving the
metric field gp .. (=) along curves connecting the two points. An example of such a
non~local term is

(39)

This integral over pairs of space-time points % and ." has an integrand which itself
contains a functional integral over the set of all curves X(r) connecting the points
% = X(r;) and 11 = X(TJ) . The functi onal int.eual measure is denoted by DX(r)
and 6 is a constant. The expressioDs C1 and C2 are ordinary reparameterisation
invariant Lagrangian density Cunctions of the fields at. z and y.
Let us now consider the physical eonsequences of the actions S .. , and S~, re-
spectively.
A non-local action of the form 5 .. , does not give any truly non-Joeal physical
effects. All the separate inte,ual5

(40)

in Eq. (38) must, when evaluated , have some numerical valuCl; they ace constant in
time and over space . The equat ions of motion derived from 5 .., can thus be exactly
simulated by a local action

(4 1)

Here the effective Lagrangian density is given by the double sum

00 m

£ ('(z ), 8.,,(zn = E E £~("(z) , 8.,,(zne: (42)

,, =1 ;:1

and the constants c:. are given in terms of the integrals I j" of Eq. (40) by

(43)

The action S", ean therefore be replaced by S and only l!,:ives rise to local effects.
Non·local actions of t.he form S~, involve the path lengths of space-t ime curves
between the interacting points z and y. Inter;rating out. the fun ctional intep-al in
euc.lidean space, one obtains the approximate distanee .behaviour up( - b~n . d~,)
(or the action term s:.,.
Here bren is the ' renorma1ised ' value of the constant 6 in
Eq. (39) and dr, is the space·time separation of the points z and y. On dimensional
grounds the constant b~n is expected to be of order the inverse Planc.k length (again
meaning the fundamental le8l;th) or zero. If bren vanisha the action 5;., degcDuat.es
 151

to the product form 5", of Eq. (38), which we have just seen only gives rise to local
etrect.s, If bRn is of order the inver&e Pland, length, the non-local effects fall off
exponentially for dist.ances greater than the Planck iencth . So we would effectively
obtain locality at experimentally accessible scales.

7.t.3. Field theofJ glaS8 Gnti gauge gltJ.J.f

We have indicated , in the previous subsection , how tbe quantum field theory
glass model of Chapter 6.3.1, with its rather chaotic ad of degrees of freedom. might.
arise from a random complicated mathematical structure . In particular we a.ssume
here that. the rudimentary physical concept.s of a 3 + 1 dimensional space-time,
locality and linearity in quantum mechanics bave been derived at some fundamental
scale lp I which we usually take t.o be the Planck length. More specifically we &88ume:
1. The existence of an underlying 3 + 1 dimensional space-time.
2. Translational and Lorentz invariance are broken at the fundamental scale i p ,
but are present statistically over larger distances.
3. The action and degrees of freedom , in each amall region of space-time. are
chosen in a quenched random way ; 80 the model has an amorpbous structure,
like a glasa.
4. The action is semi·local: there is no direct interaction between degrees of
freedom associated with points separated by distancetl greater than a few
times Ip .
a. The action is taken r.o be that of a discretized. lattice-like, quantum field
theory, rather than a genuine continuum theory.
6. The resulting quantum field theory glass is formulated as a Feynman path
integral , in which the quenched random parameters are kept constant.
Quantum fluctuations are expected to smooth out the field theory gl&S8 model
into a continuum theory, as part of the presumed mechanism for the appearance of
diffeomorphism symmetry and quantum gravity at distances somewhat larger than
i p . The resulting quantum gravity theory would resemble a "molten" field theory
glass, containing small field theory glass pieces locally: it would be analogous to a
ftuid containing small piecetl of approximate crystals but, on the large scale , having
statistical mechanical ftuctuations everywhere. This continuum theory hu not been
developed , but it is expected that the more intuitive discrete fidd theory gla.ss model
will give similar results in the long wavelength limit.
The construction of a field theory glass model was described in Chapter 6.3.1. Its
degrees of freedom are incorporated in a generalised quantum fi eld q,( i), defined on
a set of quenched random space-time points til , where it takes values in quenched
random manifolds Mi . Quenched random parameters are chosen randomly hut are
then not varied in the Feynman path integral. The fi eld theory glass action

(44)
,
is semilocal , being a sum over contributions S~h~(i» from very many small over·
lapping regions ,. in space· time.
We argued in Chapter 6.3.1 that the inverse Higgs mechanism, promoting ap-
proximate gauge Iymmetrie!l to exact phYlica1 gauge symmetries, should work even
152

roc a field theocy Sl.... Wit.h the action and field value manifold varying from point
to point, it is likely tbat some de~ of freedom will, by accident, approximate
onc or another diacreti.sed p.u,;e theory in various regions. We t.herefore int.roduced
a formal gauge symmetry

(45)
(46)
by defining a new so-called "human- field variable fll, and the Sauge group valued
'"«ius field" H , related to the oriA:inal Cundamental field by the equation
(47)
We the~by formally constructed a field theory ,lUll endowed with sause symmetries
of a random nuun, i.e. (or which tbe saule symmetry sroup varies randomly from
upon to repon. This is what we call a gauge gl. ..(23)
We also described how continuum gaule field deUees offreedom A:(z) might be
implemented on the lauge gI.... by a suitable modification of the formal gauge glass
dcpee8 of freedom ,,,(i). A priori we mi&ht attempt to implement a Yang-Mills
theory (or an arbitrary «au«e «roup K; we would probably find ,.uge ,lass degrees
of freedom correspondin« to it, here and there, throughout space-time. However it
may only be pouible to implement the ,au,e field A:(%) , for the group K , very
sparsely in space-time. Theefl'ective continuum,au«e field Lagrangian density then
acquires a term of the form
I
Cetr. = - 49' FIji, -(% )Flln(%) (48)

with a very small coefficient, _1/4g 2, because A:(%) represents relatively few gauge
glaae degreee of freedom. Thil means the gauge coupling constant, g2 , is strong;
10 the sause de,;reetl of freedom are confined close to the Planck scale and become
irrelevant to experimentally acceAible low enerl>' physics. In order for a gauge group
K to survive at low enerpe. , it iI necessary for the group to be richly represented,
as a,;ood approximate symmetry of the fundamental field theory &lass action S[I,6},
throut;hout space-time . Thil is more likely to happen, by chance, when the number
of degrees of freedom involved is arnall, Le. for sause ,;roups of small dimension and
for maUer fields belont;ins to small representationa. There are also technical reasons,
which we discuu below, why many «au«e symmetry groups would spontaneously
break down near the Planck scale.
There is a particular problem in implementing a continuum Yang-Mills field
A:(%) 00 a gause «lass, when the gauge sroup K has outer automorphiams. 24
This technical difficulty seel1lll to be of BOme importance, in random dynamia,
for uoderstandins the on,;in of the standard model group and of the number of
quark and lepton seneralioos. The group of outer automorphisms of a group K is
defined 24 as the factor sroup of all automorphisms I , calculated modulo the inner
automorphiama h which are juet the similarity transformations:
{/ : K _ KIf automorphism}
Ou«K) = {h ' K _ K I3bVg E K[h(g ) =bgb 'J) (49)
 153

.Jt.he gauge group K has a nontrivial outer automorphism, there is a problem

iJ deciding, locally, which degrees of freedom of the gauge glaas to identify with
.,Jhich degrees of freedom of the cont.inuum field A:(z) . For example, suppose a
s{ven shirt .61/111(i) of t.he human gauge glass field , from the vacuum configuration,
i.J made to correspond to a continuum gauge field configuration A:(z) . Then we
J uld equally well make the correspondence in which the shirt .6.q,,,(i) is associated
v.lith the continuum field A~·(%), obtained by applying the automorphism f to
A:(:r). So, in every small space-time region, there is an ambiguity in deciding how
I represent the continuum gauge field .
In the case of an inner automorphism,

f(g) = 696-1 , bE K and 9 E K , (.0)

t ~e transformed field Alo(z) is related to A:(z) by a gauge transformation, with a

constant gauge function :

(.1)

+,here U(z) = 6 and ~ are group generators. So we recognise the ambiguity in

i~entifying the continuum field, due to inner automorphisms, as the usual gauge
ibbiguity in defining the gauge field . It does not give r ise to any ambiguity in the
hysica, but just adds a background fidd .
For a nontrivial outer automorphism howev~r , the id~ntifit;&tion ambisuity ill
fore than just a gaug~ ambiguity. How can this ambiguity be resolved? The re-
quirements of continuity and of gauge invariance for the effective action of the con-
~nuum gauge field A:(z) will, in general , enforce a definite extension of the identifi-
Gation convention from one neighbourhood to the next. The semilocal contribution

.iIl
$r['., HJ to the action, from a small region r overlapping both neighbourhoods,
not be invariant under a convention shift in only one of the neighbourhoods.
fu. this sense a relative convention is fixed . So it is possible to extend the identi-
~cation convention along some closed curved r in space-time. But now we have &
CjOosiste.ncy condition: the convention on returning t.o the starting point must be
,'he same as the starting convention . In a gauge glass model it seems inevitable
~hat, among the many dosed curves, some of them will lead to an inconsistency
between the initial convention and the final convention . The gauge group on return
" said to get 'confused ' with its own automorphit; image. 13
I The fact that inconsistencies are lik~ly to arise, in the identification convention
lor the gauge glass degrees of freedom, may be seen by considering a change in the
gauge glass structure, in some small region r alonr; a closed curve r . A change in the
quenched random parameters , occurrinr; in the semilocal contribution Sr [ifJlt., H] to
(he action, may simulate the effect of making a relative convention shin between two
~uceessive neighbourhoods in r along r . The interaction between the human fi ehla
t .(i) and tPIIU) , in two successive neir;hbourhoods, and the Higgs field H(.) is then
~ffectivel y replaced by a similar interaction between the fields f(~II(i)),tPlI(j) and
fI(')' where ! CtPlt.(i» is the appropriate automorphic imace of tP.(i) . This change
JR ~he relative convention, between two auccessive neighbourhoods along r , requires
154

a correspondin~ change in convention along the rest of r . So a switch in convention

is needed on return to the atarting point of r . Since the quenched parameters are
randomly chosen in all of the small regions along a closed curve r, the probability
for any specific convention being needed, on return to the starting point, should be
the lame (or all identification conventions. There are many possible closed curves
f ; so it seems overwhelmingly likely that some of them will lead to an inconsistency
in the choice of com"entian.
The inconsistency can be gauge transformed away in the case of ~ inner au\o-
morphism, but not in the case of an outer automorphism. An inconsistency corre-
sponding to the inner automorphism l eg) = bgb- 1 is t ransformed away, by a gauge
transformation with a gauge function varying from I to ~ along the dosed loop r .
This transformation is achieved at the cost of setting up a magnetic flux through
the loop: the "inner" inconsistencies are transformed into background gauge fields .
The lowest energy state will, of course, tend to adjust iudf 50 lUI to cancel any
background fields if it can.
It appears that a continuum Yang-Mills theory, for a gauge group with outer
automorphisms, is unlikely 1.0 be implemented consistently on a gauge glass. There
is one way of avoiding this conclusion if fermion degrees of freedom, or other matter
degrees of freedom, are present throughout the gauge glass: the matter degrees of
freedom can be used to specify a consistent identification convention. For example
the cr088 product group 5U(2»(5U(2) has an outer automorphism, corresponding
to a permutation of the two isomorphic 5U(2) groups. It is possible to characterise
one 5U(2) factor , in an 5U(2»(5U(2) gauge group, by having a particle in , say, its
doublet representation , but no puticle in the doublet representation of the other
5U(2) factor . Confusion of the two isomorphic 5U(2) groups is tbereby avoided.
Random dynamics therefore predicts that the continu um gauge group must not
bave outer automorphisms 'hat can be extended to true discrete symmetries. For
instance 50rne left-right symmetric models ...•5 have a discrete symmetry of the gauge
group and matter field representations, corresponding to a parity symmetry. These

"..
leCt-right symmetr ic models do not arise as effective theories out of random dy nam-

Some simple groups have outer automorphisms , corresponding to the permuta-

tion symmetries of the Dynkin diagram for the associated Lie algebra.24 The 5U(N)
group, for N > 2, has an outer automorphism corresponding to complex conj uga-
tion. Similarly the standard model group 5( U(2»(U(3», discussed in C hapter III ,
h aa complex conjugation as an ouier automorphism: complex conjugation of the
elements of the 5 )( 5 matrix U defined in Eq. ( I) of Chapter 3.1, corresponds to
charge conjugation. Charge conjugation symmetry is broken in the standard model,
since only left-handed fermions and right-handed anti-fermions couple to the Wi:
intermediate vector boson . If the Cermion degrees of freedom that break charge con-
jugation symmetry are present in all the considered small regions of the gauge glass,
they may be used to specify a convention for the identification or the continuum
gauge field A~(%) with the gauge glass degrees of freedom.
Non-semi-simple groups have a U(I) factor in their Lie algebra and at least
one outer automorphism: a complex conjugation, which changes the sign on the
U(I) subalgebra. The standard model group thus has the smallest number of
 ISS

outcr automorpbisms pouible for a non-serni-simple group. It is also distinguished,

among 'be non-semi-simple groups up to dimension 18 with only complex eonju-
cation aa an outcr automorphism, by having the fewest so-called generalised outer
aut.omorphisms,l$ Generalised ouLer automorphisms of a non-sernl-simple group are
isomorphisms between two of its (actor groups, obtained by dividing out invariant
subgroups with a small number of elements. In order to find a generalised outer
automorphism for the standard model p'oup S(U(2)xU(3)) , one of the invariant
subgroups divided. out has to have at least live elements; whereas for the other
abovementioned non-semi.impJe groups of dimension less than 19, the invariant
subgroups can have just t.wo or three elements. The generalised outcr automor-
phism oflhe standard model r;roup , obtained by dividinr; out the invariant subgroup
with five elements, corresponds to scaling the U(l) bypercharge up by a factor five
combined with complex conjur;ation.
Generalised outer automorphisms act as ordinary outer automorphisms of the
Lie algebra and, thus, of the continuum gauge fields A~(z) , which we seek to im-
plement on the gauge giMS. There is again a danger or inconsi.stencies arising in
the identification convention for the gauge glass degrees of freedom , when the gauge
group K has generalised outer automorphisms. It seems that inconsistencies, aris-
ing from such an outer automorphism of the Lie algebra, are more likely the cle&er
this automorphism is to being a symmetry of the group structure and of the matter
fi elds. Inconsistencies may be avoided , if the group structure and the matter fields
can be used to keep order in the identification convention , connecting the gauge
glass variables to the continuum fields A:(z).
An inconsil5tency in the identification convention for the gauge glass degrees of
freedom, corresponding to a generalised outer automorphism, involves the identifi·
cation of a factor group of the gauge group K with the continuum gauge field . The
identification of the factor group'· with the continuum field means that the set of
r;roup elementa, belonging to a given coset of the factor group, must oorrespond to
the same continuum field A~(z) . This requires that these group elemenLs should
all correspond to approximat.ely the same value of the gauge glass action . In other
words, there must be an approximate symmetry of the action under the permu·
tation of the group elements inside a coset . The larger the number of elements
inside each coset, the Jess likely it is for the gauge glass to manifest such a sym-
metry by accident. Thus the threat of inconsistencies, due to a generalised outer
automorphism, is expected to beoome Jess significant, as the number of elements in
the associated invariant subgroup, divided out of the gauge group K , increases. As
remarked above, this number of elements has to be at least five in the case of the
standard model group .
Identification ambiguities , due to generalised outer automorphisms, may also be
avoided in ~he presence of gauge glass fermion degrees of freed om. If the fermion
spectrum present in the gauge glass is not invariant under the generalised outer
automorphism, it may be used to specify an identification convention. In general
the fermion spectrum would have to consist of infinitely many particles, in order
to be invariant under the sealing of a U(l) "hypercharge" subalgebra by a factor
S . If there exists a particle with hypercharge Yo, invariance of the spectrum under
scaling requires the existence of particles with hypercharge Y = Yo$", where n runs
156

t.brou,h all intepr value. b~h pOIIitive and nelative; thus the Dumber of particles i.
required to be infinite, with accumulation points at bypercharge values of zero and
infinity, unless 5 = -1. However one can say \here is an approximate symmetry
under acaline by a factor S, if there exist some fermion! 'hat transform into other
existinS fermion! under luch a scaling; i.e. if there exist lOme pairs of fermions with
bypercharges differing by a factor S,hui otherwise having identical gauge quantum
numbers.
The (ermion spectrum of tbe standard modd is almOlt completely devoid of IUch
an approximate symmetry under Kaling of the weak hypercharge. In mOl!lt Clllies,
Weyl £errruolli wit.h a siven combination of non-Abelian gauge quantum numbers
beloIll to exactly the lame weak hyperchar&e representation. The fermion represen-
tations are repeated, in each of the three quark..tepton generations, without vary-
in, the weak byperchar,e nlues. There is only one combination of non-Abelian
quantum numbers, for which more than one weak hypercharge value is found phe-
nomenologically: the lefl.-handed antiquarb, see Table 3.1, have b i
= - and
ill != respectively. So there u. a very approximate symmetry under swnS weak
hypercharge by a factor S = -2 in the standard model. Apart from this exception,
which has to be present in order to avoid sause anomalies, nature appears to avoid
even approximate symmetries under weak hypercharge scaling. Again tbe stan-
dard model seems well equipped to avoid inconsistencies due to generalised outer
automorphiams.
The confusion mechanism, causing the collapse of a gauge group K with outer
automorphiams near the Planck scale, can. be further clarified, by considering the
aimulation of a gauge &lass wing a generalisation of ordinary lattice gau&e theory.23
This approach also allows us to identify other collapse mechanisms,20 which may
prevent a lIaulle group K surviving from the Planck scale down to the relatively
low energy scale of presen~day experimental particle physics. The modifications of
ordinary lattice lIauge theory are made in a quenched random way, i.e. they are
chosen randomly but then kept fixed in the Feynman path integral or its Monte
Carlo simulation. We list the let of modifications below:
i) The regular hypercubic lattice is replaced by an amorphous glass-like struc-
ture, by letting the la.ttice sites take random positions separated by distancetl
of order the Planck len&th lp .
ii) The gause sroup i. allowed to vary from site to site.
iii) The link variables U( ......... ) for the gauge group K are coupled together to
form plaqueue variables Uo , defined in Eq. (61) of Chapter 6.3, and pJaquette
adion So cont.ributions. The plaquette variables and action contributions
are invariant. under the "confused gauge transformation"

(52)

rather than simply under the usual gauge transformation

('3)

Here I.,. are quenched randomly chosen outer automorphisms of the gauge
group K .
 157

iv) The plaquette action So is taken to be a quenched random function of the

piaquette variable Uo . Gauge invariance requires So to be a function defined
over the space of conjugac:y classcs21 for the group K , i.e. So takes the same
value for two group elements Uo related by an inner automorphism. We
therefore expand So on characters for the various representation r of the
group K
(M)
•
where the coefficient.s Pr are quenched random variables satisfying

(55)

and f denotes the complex conjugate representation to r. The character for

representation r is
,.(Ue) = 1t{p.(Ue)} (56)
where p~(Uc) iA the representation matrix for Uo in the representation r . The
coefficients Pr are chosen , independently for each plaquette and for each pair
of conjugate representations, from a gaussian distribution . The distribution
is chosen to have zero average and a width which decreases with tbe dimen-
sion of the representation, in order to obtain a convergen~ and reasonably
smooth plaquette action So .
v) The link variables U(-) are allowed to take values on a group space and
the sroup structure varies randomly from link to link, as in Fig. 1.3. The
plaquette variables Uo are then constructed from common factor groupa for
the groups of the surrounding links. QuUr automorphism ambiguities then
naturally arise in identifying factor r;roups of different link uoupa that just
happen to be isomorphic. It is therefore essential to consider the confused
gauge transformations of modification iii) .
By introducing the above modifications of ordinary lattice gause theory, we
simulate the action for a gauge glass. We now consider the consequences of these
various modifications and how their glassy nature may cause a breakdown, at the
Planck scale, of many oCthe gauge ,;roups that are p r esen~ a priori in this gauge
glass.
Spontaneous symmetry breakdown occurs when the vacuum state is not invari-
ant under the global gauge symmetry. This is expected by analogy with the Higgs
mechanism, in which sauge bOlons acquire a nonzero mass when the vacuum state
is not globally gauge invariant . The gauge glass plaquett.e action oC Eq. (04) does
not contain an explicit Higgs field, but it is still possible to investigate the vacuum
state oCthe theory, in the classical approximation, by considering the energetically
preferred value oCeach plaquette variable Un separately. If the vacuum value of a
link variable U(--) , or a plaquette variable Un, does not commute with all the
elements oCthe gauge group K , the vacuum will not be invariant under slobal gauge
transCormations . Hence we require all the link variables, and consequently all the
plaqueUe variables , in the vacuum state to belong to the centre of the gauge group
158

FiS. 7.3. A laUice p~e model with an irftsuW lanke and ru.dom p:oupt aft the Yuioualinb.

K ; otherwise the gauge symmetry will spontaneously collapse. We review below

the conditions a gauge group K sbould satisfy in order to resist such a collapse:Z3 ,2e
J. A6.ftllce 0/ owter ..tom0'l'Ai6m,. The gauge gla.ss action , for a sau!e group
K having outer auLomorpbisms, is invariant under the confused transformation of
Eq. (52). We anume that the 'Vacuum values of the corresponding confused link
variable. U(G) take values in the centre of the group. These vacuum values will
not be invariant under Slaha! gauge transformations, A(z) = A, unless they obey

/.,.(A) = A (57)

for all the outer automorphisms I~ ., randomly chosen throughout the lattice. It
follows that the vacuum state is only symmetric under t he subgroup. which is len
invariant by all the outer automorphisms of K . So the gauge group K spontaneously
breaks down to a subgroup without outer &utomorphisms (or at least with very few).
The gauge particles corresponding to directions in the Lie algebra of K, which are
not left invariant by the outer automorphisms, acquire masses of the order of the
Planck mass. 13 Inner automorphisms can be transrormed into background gauge
fields. These background fi..elds may give masses to some of the gauge particles
via a Rigs-lib mechanism, which is essentially equivalent to the mechanisms to
be discussed below associated with frustration of the gauge glass and with other
vacuum fields in the gauge glass.
The confusion mechanism discussed here is, of course, essentially the problem
of identification inconsistencies for gauge glass degrees of freedom in another guise.
 159

Consequently it is possible kJ circumvent. t.he confusion mechanism and allow a gauge

group with a few outer automorphisms to survive, by including matter degrees of
freedom.
II. A Mil-trivial centn. If the vacuum state is 10 be invariant. under global
gauge transformations of II. poup K with a trivial centre, all the plaquette variables
Ua must take on the value of the unit. element in K. However the coefficients fJr
in tht! gauge glass plaquette action of Eq. (54) may have either sign . This means
that. the unit element is as likely to correspond to an energy maximum u to an
energy minimum. Thus minimising the pJaqueUe energy will, in about one out of
two cases, bring the plaquetLe variable away from the unit element and hence oul of
tbe centre, for a group with a trivial centre. This, in turn, means that the vacuum is
not invariant under the global part of the gauge group K , which is thereby broken .
So, in a gauge glass, groups with a trivial centre are highly susceptible to collapse
and only ,roups with a non-trivial centre survive.
III. A conru:d4!tl C4!njrt. A similar argument to the foregoing , applied to a set
of neighbouring plaquetLel!l instead of to a single plaquette, suggests that. gauge
groupa with a topologically disconnected centre tend to collapse. The plaquettes,
which t.ogether constitute the surface of a generalised three-cube of the amorphous
lattice, must satisfy the Bianchi identity. This identity arises from the fact that
each link is contained in two plaquettes of the cell and the plaquette variables Ua
are therefore not independent. If the Ua lie in the centre of the group K , i.e. the
Abelian subgroup which commuLel!l with all elements of the group, t.he product. of
the plaqueHe variables forming the surface of a cell must. equal the unit element :

IIUc =1. (58)

c ""
The Bianchi identity, Eq. (58) , is not automatically satisfied when each individ-
ual plaquett.e variable is chosen to minimise its energy. The energetically preferred
individual plaquette variables UOpm must belong to the centre to prevent collapse
of the gauge group K but, due to the random Corm oC the plaquette action, we
expect the product
F= IT UO pm (59)
c ""
to be a random centre element. We define F to be the Crustration of the cell 11' . If
F I- 1, the true vaccum values Ua .....: oC the plaquett.e variables cannot equal Ua pm:
there must be some compromise between the various plaquett.es forming the cell .
It is p088ible a compromise can be achieved by small adjustments of the plaque-
tt.e variables. without leaving the cent.re of the voup. if the centre is continuous.
However if the centre of the group is not also topologically connected, the frustra-
tion F for some cells will belong to a different connectedness component to that of
the unit element. It is then probably necessary to adjust aU the plaquett.e variables
forming such a cell, leading to deviations Ua prdUa .....: - I between Uapm and Ua Va<: '
These deviations correspond to "magnetic: fluxes" leaving the frustrated cell , which
160

add up to a path in group space leading (rom I to F . Such a path must leave
the centre if I and F helons; to different connectedness components. It follows that
lOme of the deviations Ua pnr!Io....., -1 are non-central. Since the Ua pre!' are assumed
to belong to t.he centre, this means that sorne of t.he vacuum plaquette variables
Ua .,.., are non-central and the gauge SlooP K collapses. In order to avoid 8uch a
gaug: symmetry breakdown, it is necusary that the gauge group K haa a connected
centre. However it. is not Inflici~t.2.
IV. Coftjwgac, cia .. ' pate wilh onl, centrul ,jng.lariliu. The plaquette ac-
tion Sa is defined over conj~&Cy d&Sl space, Lui", the same value for two group
e1ementa related by a similarity bansformation. The minimum of the plaquette en-
ergy therefore tend. to occur at. the extremities, or singularities, of conjugacy class
space. The vacuum plaquette variables are thus likely to take up these sin!Ular
values, which must then belons to the centre of the sroup in order to avoid the
spontaneous breakdown of the sauge symmetry.
The sinsularities are typieally eorners or edges of conjugacy class space. For
example the eonjusaey class spaces for the rank one groups 5U(2) and 50(3) are
line segment. and their sinz:ul&rities are the end points. The conjugacy classes for
5U(2) and 50(3) are characterized by the angle of rotation about some direction in
the three dimensions. The end points eonespond to angles of 0 and 211' for 5U(2)
and to angles of 0 and 11' for 50(3). For 5U(2) both end points belong to the centre,
correspondins to the group elements land -I; while for 50(3) only the end point
correspondin,; to the unit element I is a centre element . The non-eentral end point
sinsularity of 50(3) conjupcy class space represents the class of rotations by an
angle 11'. If a plaquette variable in the vacuum takes on a value in this singular
,.--class, the 50(3) gau,;e group spontaneously breaks down . 5U(2) is favoured to
survive .. a sau.&e group, since it has only central eornell.
Another type of sinsularity is a ~eaned "top hat" point, which lies in the inte-
rior of conjusacy class space. For example the &loup of rotations in four dimensions

SO(4) = SU(2)xSU(2)/«l,I), (-l,-1)) (60)

has a conjusacy clasa space with the geometry of a paper hat. The class space for
the covering group 5U(2»(5U(2) in the square formed by the cross product of two
line segment., representins the eonju,;acy class spaces of the two 5U(2) groups. The
conju,;aey class space of 50(") corresponds to the paper hat , formed by folding the
square along a diagonal and then gluing the two halves of the diagonal together:
the midpoint of the diaconal corresponds to the top of the hat. This singular point
at the top of the hat represents the class of .. x .. orthogonal matrices with two
eigenvalues equal to -1 and two eisenvalues equal to +1. It is non-central and , if a
plaquette takes on iLl value in the vacuum, the 50(") gauge group collapses.
The above four conditions favour the standard model group

SMG = S(U(2)xU(3» (61)

 161

as being particularly well suited to avoid spontaneoull collapse in the gauge glass
and to survive as a low energy symmetry group. The SMG is the one group with
the Lie algebra U(I)xSU(2)xSU(3) which has a non·trivial connected centre :

S(U(2)xU(3)) = U(1)xSU(2)xSU(3)/D. (62)

where the discrete group divided out

D, = (hOI' E Z.) (63)

is generated from the central element

(64)
oft.he group U(1)xSU(2)xSU{3). Here ZN denotes the additive group of integers
modulo N and IN denotes the unit element of SUCH). There are no non-central
comera or top hat singularities in the conjugacy class space of the SMG. Also it
has only one outer automorphism, i.e. complex conjugation, under which the chira!
quark-lepton representations are anyway not. symmetric.
Let WI DOW consider more systematically the type of group favoured by the four
requirements I-IV. In general, the necessi'y for a continuous centre requires that
the gauge group have an Abelian invariant subgroup or U(I) factor . For the SU(N)
groups the extremities of conjuS&CY e1aS1 space all correspond to central elements
exp(i21fmjN), mE ZH . This is not true for other simple groups, which are therefore
expected to spontaneously break down in the sause SlaSl. So we are led to cORsider
groups with a Lie algebra consisting of a dired product of various SUCH) factors
and an abelian U(I) factor. Multiple occurrences of the same subalgebra do not
survive, due to the confusion mecl1anism of condition 1.
The centre of the coverins group for a Lie algebra, consisting of a direct prod·
uet of SUCH) factors and a U(l) factor, is not connected. In order to satisfy the
connected centre condition III , it is necesaary to divide out some discrete central sub--
group, which has a non-empty intersection with all the connectedness components
of the centre of the covering ~oup . Some elements in all connectedness components
of the centre of the covering group are thereby identified with the unit element. If
the N values for the various SU(N) factors are all mutually prime, it is p065ible to
generate the required discrete central subgroup from a single centre element, as the
discrete ~oup D3 is constructed, in Eq. (63), for the SMG from the centre element
of Eq. (64) . However if the various H values have common divisors, such a discrete
subgroup cannot be generated from a single centre element . An extra generator,
with zero component along the U(l) group, is then needed. This extra generator is
identified with the unit element , when the discrete subgroup is divided o ut of the
covering group. In this case, the resulting factor ~oup must have a non·central
singularity in conjugacy elMS space , analogous to the top hat singularity for the
SO(4) group of Eq. (60). Condition IV is therefore violated if any of the H values
have common divisors.
162

Since «aug!! groupi with a .ma11 number of dimensioDs are most likely to arise by
chance in a sauge glus , we expect the preferred N values to be as small as possible,
as well &II mutually prime. We therefore conclude that the gauge glass model , and
hence random dynamics, favours a gauge poup of the form:

c'P) = U(I)xSU(2)xSU(3)x5U(5)x ... x5U(P) / Dp (65)

where the SU(M) groups are taken for M running through all prime numbers greater
than or equal to 2 and less than or equal to P. The discrete group divided out

(66)

is generated from the centre element

hp =(exp(i2~iNp) , -/2,ex.p(i2l1"/3)13 ,
exp(i2l1"m5/5)la •... , exp(i2"mp/P)Ip} . (67)

Here
Np = 2 x 3 )( 5 x ... x P (68)

is !.he product of all prime numbers less than or equal to P and the parameters mn
for n > 3. are integers not divisible by n . For P > 3, there are several different ver·
sionl of C< P ) corresponding to the different values of the integers mi . ro7" , . I mI' .
The first three members of the series of G<P) groups are

c") = U(I) (69)

c") = U(2) (70)
c") = SMG = 5(U(2)xU(3» . (71)

The groups with P > 3 contain SU(N) lubalgebru with numerically larger reno~
maJisatkm group beta functions , in tbe WlUa! notation,21 than tbose of the SMG j
therefore they might be expected to confine at a relatively high enere scale . We
make a more careful estimate of this confinement enere scale in the next s ubsection.
Chiral fermion fields are required to avoid spontaneous collapse of tbe gauge
poup G<P), by confusion under charge conjugation, in the gauge glass. However, in
general, we expect the maUer fields to belong 1.0 as small representations as possible,
if the gauge synunetry is to appear accidentally in a field theory glass. States
belonging to matter fi eld representations of (;f.P) must obey hI'1/! = 1/!, leading to
the quantisation rule for weak hypercharge:

~ + ~ "duality" + ~ "triality"
+ ~5 "quintality" + .. . ,+"; "P . ality" = O(mod 1) . (72)
 16J

Here the "N-ality" is the rank index 27 of the representation I/J under the SU(N)
subgroup: it counts, modulo N , the number of Ii representations used to build up
the representation 1/1.
In the ease of the SMG , Eq. (12) expresses the usual electric charge quantisa-
tion rule. Eq. (1) of Chapter Ill, of the standard model. ~ we have already seen in
Chaptel 3.2, the experimentally observed patt.ern of SMG matter field representa-
tions is the smallest ORC, consistent with the ~uirementl of cancellation of gauge
and mixed anomalies and fermion ma.ss protection. This minimal fermion spectrum
is just repeated a number of times, corresponding to the number of quark-lepton
generations . We consider the number of generations, predicted by fitt ing the fine
structure constanta of the standard model using random dynamics, in the following
subsection.

1.!. 4. Numerical prtdictions from random d,nGmics

In addition to the many features of random dynamics which favour the survival
of the SMG as a low enerK)' synunetry group, one of the m06t suggestive results is
the prediction 2S1 of the number of quark and lepton generations: N,en = 3.
As discussed in the previous subsection practically any gauge group may be
found locally, at the fundamental scale Ip , in a field theory &lass. In general, how·
ever, we only expect gauge group.l of relatively small dimension to survive at larger
distance scales. A randomly selected compact connected group K, having a dimen-
sion of a given order of magnitude, will typically have a simply connect.ed covering
group i< equal to the direct product of a number of repeated factors of U(l) , SU(2) ,
SU(3) , sotS) ""

K =U(I) x U(I) x ... x U(I) x SU(2) x ... x SU(2)

(73)
x SU(3) x ... x SU(3) x SOtS) x ....

The randomly selected gauge group K will in general be obtained by dividing out
some discrete subgroup D of the centre from K:

K=K/D . (74)

On the one hand gauge groups K which have a charge conjugation like automor-
phism, but no. cbiral fermions to protect them, will collapse by confusion with their
own cbarge conjugate. On the other hand it is most likely that the surviving gauge
groups have as few matter fields as possible, in small chiral representations. The
chiral fermions must, however. satisfy the conditions for the cancellation of gauge
and mixed anomalies. In order to obtain a minimal set of chiral fermions, one is
therefore led to consider fermions belonging to non-trivial representations of several
of the subgroup factors in k :
1) The same chiral fermions can thereby protect several of the factors against charge
conjugation confusion, and 2) the division of k by the discrete subgroup D may
.64

require the amaUcat allowed representations of K to couple to more than one sause
lubgroup factor .
As an approximation, motivated by the random dynamica u pectation of sm&ll
wral fermion representations, we IbaU ipore the pouible existence of fe.rmioru:
involving higher dimensional gauge groups than U(l), 5U(2) and 5U(3) . Even the
I1I1a.Ilest representation. for, uY. the SU(P) groups, with P prime and greater than
3, have P = 5,7 _. particles and quickly become rather large compared to the
"
irreducible representations of the standard model. FUrthermore if tbe representa-
tions are non-trivial under both SU(P) and 5U(2) or 5U(3) , even more particles are
required, Le. 2P or 3P. As discussed in Chapter 3.2, the anomaly cancellation con-
ditions (or chiral fermion. can be fulfilled in a very particle economic way, using JUBt
the U(l) , 5U(2) and 5U(3) goups of the standard model: in fact one quark-lepton
generation constitutes the smallest anomaly free and mass protected representa-
tion of the standard model group. This principle of small chiral representations
then fa..-ours the s.ituatk:m in which any higher dimensional SU(P) goups have no
fermion representations at all. We will estimate later the confinement energy swe
for such higher dimensional SU(P) gauge groups, should they exist.
Consequently we may il1'\aline reducing the gauge group K to a direct product
of several factors of U(l). SU(2) and SU(3) . each combined to form the standard
model group
SMG = S(U(2) x U(3)) . (75)
Each SMG factor is provided with jUlt one generation of chiral fermions; the min-
imum number required to avoid collapse by confusion. Any left over uncombined
factors of U(l) , SU(2) or SU (3) are expected to collapse by one of the mechanisms
discussed in the previous subsection . We are thus led to consider the di rect product
of N~n copies of the standard model group:
KN... = (SMG). x (SMG), x ... X (SMG)N.. . (76)
This direct product goup KNra has outer automorphisms. corresponding to the
permutations of the isomorphic SMG factors . It therefore. in turn. collapses hy
confusion to the diagonal subgroup:
SMG,;.. = {(h. h •... • h)lhE SMG) (77)
which has no outer automorphisms unprotected by chiral fermions . It is the 'mass.-
less' gauge b0600S associated with this diagonal subgroup, which are to be identified
with the phenomenologically relevant gauge particles. The remai ning gauge b080ns
all acquire a masa of the order of the confusion breakdown energy scale.
The breakdown of the gauge group KNra to SMGdi.. is expected to occur just
below the Planck scale, where the three gauge couplings (gndi-c of the diagonal
subgroup are related to the gauge couplings (g?h , (gl), •. ..• (9nN••• of the parent
sroup factors (SMGh . (SMGh •... • (SMG)N••• :
1 1 1 1
(g;')di.. = -g;')(I + -g"')( + ... + (gi') N.u (78)
 165

This result is easily obtained by using the fact that (or an oscillation in the U(l),
SU(2) or SU(3) component of the diagonal subgroup field A:.I.~(%) ' all the paunt
fields A: 1(z) , A:,(z) •. .. , A: ..... _ (z) oscillate in the same way :

(79)

The Lagrangian density (or I Uch a diagonal subgroup oscillation becomes

(80)

which, using Eq. (79), may be rewritten:

I( Igig
I
C,; .. (.) = -. -')( + -')( + ... + (
2
I)
')
9 N...
,
(F,:.,.,(,)) . (8 1)

This is of the usual form

(82)

provided Eq, (78) is satisfied (or the U(1) . 5U(2) and 5U(3) gauge coupling con-
stants.
It follows from Eq. (78) that the gauge coupling constant. (gl)di .. (or the diagc>
nal subgroup are s maller than the gauge coupling constanta (91)1. (gl), •.. . , (g?)N•••
for the parent group. In particular, if the latter are all taken to be equal we would
have
(g;h
(9i )d iac = r;;.r- . (83)
V",..n
There is thus a relationship between the standard model gauge coupling con-
stants, identified with (9;)di.. , and the gauge coupling constants of the parent group
KN.O& at the confusion breakdown enerr;y scale PI;. The confusion scale Pc is pre-
sumably of t he same order of magnitude but less than the fundamental cut-oR',
which we associate with the Planck mass pp . The requirement t hat the parent
gauge group K N ... should not be in a confining phase, in t he small energy gap
between Pc and pp , is a crucial ingredient of the random dynamics prediction re-
lating Nrn to the phenomenologically measured gauge coupling constants. Usually
confinement is defined by the presence, at sufficiently large distances, of a confining
force which prevents the separation of the confi ned particles. In the followin& we
shall use the concept of confinement at a certain scale of energy or of length. If the
running gauge coupling constant, at a certain enerr;y scale p , is strong enough that
a confining force is already active at distances of order p_l , we shall say that there
ill confinement at the scale p . In fact we are really not &0 much concerned with
166

confining particleS, as with the correlation length being less than or equal to }.I-1.
So we shall use the word 'confinement' 1.0 describe such a strong eouplinr; pbase,
even when we ignore matLer fields.
If the parent direct product gauge group KN.. ~ is in a confining phase, there
are no correlations over distances as large as the confusion iensth scale J.I; 1 . It is
then hard to ima~ne that any confusion breakdown of KN... can make .uch a fidd
theory fundion &8 a Yan,-Milla theory, which could be identified with the standard
model at large distances . We therefore expect that. the parent ,auge poup coupling
constants (9ih.(gih , ... ,('vi)N.- must have upper limits (gi)crit :

(9i)1 :5 (9i)"';t. (gih < (g;)crh •...• (gi)N... :5 (9,)ent . (84)

The critical values (9i)crit are determined from the requirement that. the direct prod-
uct gauge group KN l u must not confine at energies above ~he confusion scale Pc,
which is close to the Planck mass: pc .... 10111 GeV. Otherwise all the gauge bosons
would be replaced by g1ueballs with m&S8e8 of order 10111 GeV and not even the di-
agonal subgroup could survive as a low energy gaug: group . These critical couplings
are calculated in the mean field approximation (MFA) of lattice gauge theory, where
they correspond to a phillie transition between confined and Coulomb-like phlllles.
The MFA seems an appropriate way La charactert!ie ~he physical behaviour of the
fields at the I/Juice .tc/J/e. It is, namely, a very local approximation !.bat does not
take into account correlations over dis~aneea much larxer than the lattice spacing;
in this sense the MFA totally iXnores long waveleD.l;th ef£ecta. We thus obtain the
upper bounds
(gi (Pc»di.. (,; )<ri.
::; JNpn (8.)

for the standard model gauge coupling constants at the confusion scale Pc.
There are even reasons t.o believe that the inequality, Eq. (85), should be satu-
rated. The field theory glass random action, Eq. (44), should be imagined lUI chosen
from a probability distribution, leading dominantly to action contributions Sr(q,(i»
which are roughly constant, for continuity reaaons, over small regions in the space
of ,(i) configurations. The definition of the probability measure for the functions
Sr(,(i)) thus requires a metric to be introduced on the ,(i) configuration space,
so that. the typical chosen function is approximately constant over a distance of
order unity in this metric . In random dynamics the chances of a gauge symmetry
appearing accidentally are the better, the smaller are the orbita, as measured in this
metric, of any st81ting confil!;uraton ,(i) under the approximate gauge transforma-
tions. Whenever an approximate Yanl!;-Mills theory is present in a field theory ,1&88,
it should be pouible, at least crudely, to identify some co-ordinate:s of the field the-
ory glass configuration space with the link variables of a lattice gauge theory. The
approximate gaul!;e symmetry is m08t likely to occur by chance, when the variation
of a link variable over the gauge &roup volume corresponds to a variation over a
relatively small volume in configuration spate. In this case one also expecLs that the
 167

variation of a plaquetLe variable, conatruct.ed from the link variables, over ita range
will eonespond to & relatively small variation in configuration apace. This, in turn,
means a slow variation of tbe randomly selected action contributions S .. (f(i» as a
fundion of t.he plaquette variables. Since the effective gauge field action is propor·
ir
tional to I we conclude that gauge symmetries are most likely to appear by chance
for strong sauge couplings g', As discussed above, 8uch a gauge symmetry cannot.
survi.ve at. long wavelengths, if the coupling is atrong enough to cause confinement in
the small ener«)" gap between the fundamental Kale pp and the confusion scale Pc.
So the most likely coupling reIIultin& from the field theory glass, which is still able
to contribute to low eoer!)' physica, muat be the largest p088ible consistent with
the existence of a Coulomb-like phase in the energy sap. This means uluration of
our inequality, Eq. (85) .
We have argued above that small orbits of the gauge group are favoured in the
field theory glass configuration space, by giving a better chance for the accidental
appearance of the gauge symmetry. This, in tum , suggests that the gauge group
it.self should be small in some physical sense, in order to favour such accidental
symmetry. Our condwion that strong gauge group couplings are most likely to
appear in a field theory glass actually corresponds to the gauge group being small,
in terms of a group metric normalised by the inverse squares of the gauge cou-
pling constants. The gauge coupling constants then act &8 units of distance in the
gauge group manifold; there are different such units along the different invariant
subgroups. Normalising in this way, the Yang-Mills Lagrangian density .cYM(J:) is
naturally expressed in terms of a mdric telllOr 9•• defined by

(86)
Here then
(87)

for a and b denoting basil vectors in the ith invariant lubalgebra.

Assuming that the inequality of Eq. (85) is sMurated, it is possible to express
the standard model gauge coupling constants (9i(Pc))diac. for Pc _ 10 111 GeV, in
Lerms of the MFA critical va!ues3(l :

(,nait =('~(l))crit =1.98 , (88)

('~)cnl = (g~(2))coi l = 1.14 , (89)
(9~)ail = (g~(3))coit = 0.82 . (90)

In first approximation we here ignore the difference bet~ U(N) and SU (N)
critical couplings. The predicted values of gi(P,,)diaa may then be extrapolated,
by means of the renormalisation group equations {paper 16], from Pc down to the
expe.rimentalscale p-l00 GeV. Inc.ludingjust the standard model fields for N,,,n
generations in the renormalisation group extrapolation , rather good agreement is
168

obtained with the aperiment.ally measured A:&uge coupling constants for Nee" ~ 3
to 4.
The concept of an MFA critical coupling (9t)crit for the U(l) group is ambigu-
ous by a factor of 6, depending on the choice of the weak hypercbarge quantum
lllqll&ntlim ' The standard model putieles which are neutral under t.be non-Abelian
groups SU(2) and SU(3), such as the right-handed electron, have even integer weak
bypercharge: hquant\lm =1, corresponding La "he critical value (gl)~t given in
Eq. (88). However generally the s\andard model particles have weak hypercbarge
values, see Table 3.1, which are multiples of one-third: il/qIlMtum = 1,correspond-
ing to a six times larger value for (91)mt - It is not a priori obvioua whether to use
1
!Jlqllantllm = 1 or as the U(l) quantum. A speculative random dynamics analysis
of the critical couplings for SMG = S(U(2) x U(3» lattice sause theory resolves
the ambisuity in favour of lYquMtum = 1 and, hence, the smaller value for (91 )erit
siven in Eq. (88).
In this way fiLs are obtained of Nrn to the experimental values of the U(l),
SU(2) and SU (3) gauge coupling conatanLs:

NI'fft = 3.4 from U(l) (91)

NI«o. = 3.5 from SU(2) (92)
N... = 3.1 from 5U(3) . (93)

An earlier version u of this model similarly p~icted Nlm ~ 3 quite accurately,

prior to the LEP measurement of the number of light neutrino lpecie.II .3 1
I( the random dynamics argument for the IJ&turation of the inequality Eq. (85)
is not accepted, our p~ictionl for N lm become upper limitll:

NI'fft :f 3.5 from SU(2) (94)

NI'fft < 3.1 from SU(3) . (95)

Almost any further particles, which couple to the standard model SU(2) or SU(3)
gauge fields , populating the 'desert' between p ~ 100 GeV and p~ ::::: lOll GeV would
effectively violate these inequalities, due to their modification of the renormalisation
sroup beta functions . In particular supersymmetric particles much below the 10 111
GeV scale are not allowed by random dynamics. Also technicolour models arc
ruled out, unless there arc very few tec.hni-particles. Random dynamics cuentially
predicLa the existence of the 'desert' . The crucial inpedient in these predictions
is tbe aaaumcd embedding of the standud model gause group, as the diagonal
subsroup, in a dired product of N,m copies of the SMG = S(U(2)xU(3» group at
the confusion scale pc - 10 1• GeV .
We will now briefly consider the possible physic&! existence of the higher dimen·
sional SU(5) , SU(7), ... ,SU(P) gause goupa, which appear as factors in the random
dynamics favoured series of groups cC P ) of Eq. (65) . Let us first estimate at what
ener~ scales they might confine in the 'desert'. For tbis purpose, we assume that
 169

the above confusion scheme generalises to a direct product gauge poup of N&en
facton of C<P). which break. down to its diagonal subgroup at the confinement
scale Pc ~ lOll GeV. It turns out that the MFA critical coupling constants ( UN)cril.
for SU(N) gauge poups, are liven approximately by

N(g~ )<:nl ~ 2.5 . (96)

Assumin, critical couplings for the parent direct product fact.or groups at the eon-
fusion scale leads, by saturation of Eq. (85) , to the result

'( » _ N(g),)<ri. _ ~
(N IN pc -
..'
(97)
N
...
- N .

The renormalisation p-oup beta funct ion for the coupling parameter (Ng'k)-l is
N -independent:
d ( I ) 11
dlnp. NgJ..(p) = P1/N'~ = 24 ...
2
(98)

ignoring the contribution of any matter fields . Thus, in the absence of matter fields,
the SUCH) gauge couplinp would all diverge at approximately the same confinement
energy scale PCOfIr:

_
Pr:onr - pcexp
(_ 24.' .
11
NI")
2.5 (99)

= 6 x 10- 12 ",c (100)

7
~6 X 10 GeV . (101)

The inclusion of fermion matter fields coupled to the SU(N) gauge fie1ds would
lower the confining energy scale, as happens in QeD. The contribution of, say, an
l:!.. representation of fermion fields to the beta function becomes smaller, relative
to the gauge fie1d contributions, as N increases; so the higher dimensional SU(N)
groups should confine at higher energy scales, as intuitively expected.
According to the quantisation rule Eq. (72) , the matter fields for the SU(P)
groups, with P prime, should at least carry some weak hypercharge. We must there-
fore consider whether such matter fields, populatinr; the 'desert', are consistent with
the above random dynamics results for H len . In order to construct an anomaly free
and mass protected fermion representation, using ~ and~· irreducible representa-
tions, one is led to include at least one representation with I! I > 1. The contribution

U(I) gauge coupling constant, is at least equivalent to that of *

of such a P -dimensional representation with III > I to the beta function , for the
ordinary quark-
lepton generations. It happens that approximately half of the N,en-dependence,
of the above random dynamics predicted standard model gauge couplings, is due

SMG factors in Eq. (76) being N,en . Thus *

to the renormaiisatKm r;roup extrapoJatKm; the other half is due to the number of
extra generations contributing to
the beta function would lower our prediction Eq. (91), for the number of ordinary
170

quark-lepton generations, by roughly ~:::=:~ . For, say, SU(5) this would lower tbe
U( I) prediction from N,ft!. = 3.4 to N ltn :::::: 2.7, which is just consistent with th ree
quark·lepton generations.
However , if the confusion mechanism is to work properly, the same matter field
representations should he associated with each of the N~ facLors of SU(P) in the
parent dired product factor group . Then, (or three generations, the correction
to the fitted value, Eq . (91) , for N&efI would be increased by a factor of three to
3:: so the prediction , Eq. (91). would be lowered to N~ ::=: 1.3 or less , in strong
disagreement with three quark-lepton generat.ions. It. would he possible to avoid "his
disagreement, if the weak hypercharge quantum could he taken to be !I/q.. 1
antum :;:
instead of the preferred value ht·q.. ..,I .. m :;: I. We)'l fermions coupling as both a p .
dimensional representation of 5U(P), and as a doublet of 5U(2) or a triplet of
5U(3) , would lower the predictions of Eqs. (92) and (93) for Nitn by approximately
=. This could be tolerated for one representation , but not for the N,~n = 3 copies
required to make confusion work.
Thus, if matter fields are assigned to the higher dimensional SU(P) groups,
random dynamice does not really fit with three quark. lepton generations ; at least, if
tYq.....tUtft = I is used . We conclude that there are no such matter fields populating

the 'desert '. The higher dimensional 5U(P) gauge fields - should they exist -
would then be hard to detect: SU(P) glueballs might o nly function as dark matter
in the Universe.
Another nUlllerical reault from ralldom dynamica, but with oilly order of mAl;·
nitude accuracy, is a computer simulation of the quark.lepton mau spectrum and
weak mixing angles, using approximately conserved random 'horizontal' U(l) chiraJ
charges (Paper IIJ . Good quaJitative agreement is obtained with the experimental
masses and mixing angles, but with evidence for some unifying interaction beyond
the standard model:
a) The similarity between the mass spectra for the three charged fermion families ,
Le . the Q = I quarks, Q = -1 quarks and Q = -I leptons , indicates a need
for correlations between right..-handed quarks and between quarks and leptons.
These correlations are most easily introduced by postulating the existence of
righlrhanded currents and leptoquark currents. Without such correlations be--
tween the horizontal quantum numbers on rigbt·handed quarks, our computer
simulation predicts too large fluctuations in the logarithm of a fermion mass in·
side a given generation. In other words, the masses of the electron, up quark and
down quark are much more similar than expected from just the standard model,
when they are treated as the lightest members of three independent fermion fam-
ilies. This is also true of the second (e, 8, p) generation. The predicted spread
in the logarithm of fermion mass is smallest for the heaviest or third generation:
phenomenologically the third generation has the largest spread and , for a top
quark mass mE R$ 100 - 200 GeV , is in agreement with the model prediction
without using any right..-handed or leptoquark symmetry. This would suggest
that righlrhanded cunents and leptoquark currents are needed for just the two
 111

lighter generations.
b) The smallness of the off-diagonal elements of the quark weak coupling matrir'2
also .uUests t.he existence of correlationl between the right-handed quark com-
ponenta. The large weak interaction transition matrix e1emenLs between quarks
in the same generation, u _ d, C _ I and ~ _ b, and smaIl inter-genuation
transition matrix clementa would only occur accidentally. if not enforced by a
correlation between the random horuontal quantum numbers inside a genera-
tion.
There au two (ree parameter. in the simulation , which can be fitted to give the
overall mass scale, set by the W -mass, and the strength of the horizontal symmetry
breaking interaction respectively. The reaulta of the computer simulation aze not
very sensitive to the number of types of horizontal cbiral charge introduced , nor to
the possible existence of further generations.
After elimination of the two fitted para.nw;ters, our main prediction for the
fermion maases (paper 111 becomes:

(102)

where m. denotes the ith generation mass within a family of quarks or leptons of a
given electric charge. It agrees rather well with experiment, but is of course not a
very strong prediction. If right-handed currents are introduced , the quark mixing
matrix elements are naturally reproduced, witbin the large statistical fluctuations
of our model,32 but with a somewhat low average value for the Cabibbo angle
which mixes d and 6 quarks. The pbase angle, in the standa.rd pa.rtide data group
parameterisation33 of the weak mixinr; matrix of [Paper 141, is of order unity in our
model. The correct order of m&«nitude is therefore obtained for the CP violating
parameter c in kaon decay.
Another general conclusion from our statistical analysis is that grand unified
models, such as SU(5) or SO(10), whicb put some Iefl.·banded fermions in the same
representation as their antiparticles, tend to give one or more cases of families with
a pair of almost mass degenerate particles.34 The lack , phenomenologically, of such
inter·generation approximate mass degeneracies disfavours grand unified models like
SU(5) 0' SO(\O).
Grand unified models based on simple gauge groups like SU(5) or SO( IO} are, in
any case, not expected. to occur in random dynamics, according to the general field
theory glass scenario discussed in the previous subsection. However gauge fields
coupling La the rigbt-.handed components of quarks and leptons, e.g. an SU(2)R
gauge group, are also l'Io'expected to survive the confusion breakdown mechanism of
random dynamiQ. Thus tbere is an apparent conflid between random dynamics and
the suU;ested existence of right-handed currenLs from our mass matrix simulation.
It is assumed in our mass matrix simulation model that the chiral horizontal
symmetry group commut.es with the standard model gauge p-oup. However, random
In

dynamics BUgests the existence of an intermediate &&u&e sroup

K, = (SMG), x (SMG), x (SMG), ( 103)

at the confusion Bcale; with one direct product factor (SMG); for each quark-lepton
generation. With 8ueh a direct product unified gaut;e &roup, it il possible that the
separate sause quantum numbers of (SMG}t, (SMGh and (SMGh could play the
role of horizontal quantum numbers and be responsible for the different mass scales
of the three r;enerations.
Let us now Bununarize tbe results obtained from random dynamics :
1) The standard model t;aup;e group SMG = S(U(2)x U(3» is 81.ron,ly favoured.
2) The number of generatioos, Npn = 3, ilsuccessfully related to the phenomeoo...
loKicai values of the gauge eouplinS constants fOf tbe three invariant subgroups
of SMG . This result essentially requires II. 'desert' between presently available
energies and the Planck scale.
3) The successful but rather crude prediction, Eq. (102), for the ratio of IOf!;arithms
of quark or lepton mass ratioe. Also tbe weak mixinf!; matrix is statistically
predicted , in qualitative ageemen\ with aU phenomenology: but the luuested
existence of an SU(2)R f!;auge group would violate the confusion mechanism of
random dynamics.
It is of course rather danf!;erous, in general , to conclude backwards from a suc-
cessful prediction that the underlying theory is correct. This is especially true of
random dynamics predictionl, b~ause they should follow from almost all models:
so if they turn out to agree with experiment, they may not be suUettive of any
particular theory. On the other hand if a random dynamics prediction fails, one
should really Jearn something.
It would indeed be moet unreliable to conclude anything about the truth of
random dynamics from the successful result 3) above. The computer simulation of
quark and lepton masses uses a type of mass matrix, which could crudely arise in
many models that would seem realistic from more conventional points of view.
The results 1) and 2) seem more promising evidence in favour of random dy-
namics, but they are far from conclusive. At first sight it may seem that one uaea
many unconventional assumptionl built into the field theory glus, such as dis-
cretized Ipac~time and many gauge groups of the same type at the fundamental
level. However, M the end , it is the confusion breakdown mechanism which is the
important ingredient in deriving both 1) and 2) . So the main supported element
of random dynamics, if anytbing at all, may be ' confusion' and the lack of outer
automorphisms in the gauf!;e ~oup of nature. AA previously emphasized , tbe fact
that so many of the symmetries found in the laws of nature can be derived is a
sttone argument in favour of the random dynamics model.
A priori the main evidence .,;ainst random dynamics is provided by the well-
known fine-tuning problema: 1) the hierarchy problem of why the W:i: and ZO
masses are so small compared to the Planck mass, 2) the vanishing coemologi-
cal constant problem and 3) the problem of CP symmetry in Itrong interactions.
 173

Baby universe theory may resolve some of these problems, but this subject is still
controversial. Fine-tuning problems are of course especially worrying for random
dynamics , which is baaed on the principle of no fine-tun ing taken to its extreme.

7.3. Classification oC Symmetry Derivations

Bow c::&n it happen that symmetries appear in physical models without having
been put in? We attempt to answer this question here, by classifying the symmetry
derivations collected in this book .
We first consider classifying our examples of symmetry appearance, according
to whether or not the synunetry is really present in the language of the original
model, in which it is to be derived . For example the quark and lepton flavour -
strangeness, charm, muon number etc . - conservation laws are true symmetries
which are really present in the strong and electromagnetic. interactions (paper 8).
We note here that the qUMk flavour conservation laws appear as symmetries of
the standard model in the low energy limit, where the weak interactions can be
neglected; nonetheless the symmetry operations really leave the model invariant in
this limit. However, in the case of the macroscopic scaling symmetry discussed in
Chapter II, the scaling transformations cannot really be performed , in an unambigu-
ous way, on the molecules of the underlying model of molecular physics. Similarly
the formal derivation of gauge symmetry {paper 26J is an example in which the
symmetry is not present in the original model.
For convenience we shall refer to the original model, in which the symmetry,
is to be derived, as the "fundamental" model or theory. We shall then divide the
symmetry derivations into two main groupa:
I. Symmetry really present in the fundamental theory.
II. Symmetry not really present in the fundamental theory.
It. must be admitted t.hat in some of our examples it is a slight. exaggeration
to talk about a synunetry 'derivation' , when there is already a larger symmetry
manifestly built into the fundamental model considered. A symmetry derived in this
way will be called a ' transformed symmetry' . For example, if Poincare invariance
is considered La arise from general relativity &8 a special case of diffeomorphism
symmetry, see Chapter 6.2.1 , it is c1eMly a rather mildly transformed symmetry. A
more genuine transformed symmetry is the case of the Kaluza-Klein model (Paper
20J, in which diffeomorphism symmetry turns into gauge symmetry. In cases of
transformed symmetry, the symmetry is really there in the fundamental model and
belongs to group Ii we subclassify transformed. symmetries as category 1.0.
For some symmetries of group 1, it could simply be that there are so many
assumptions, principles or restrictaons, built. into tbe fundamental model under
consideration, that it is not. possible for tbe structure in different parts ofthe model
to behave in a different manner. When the restrictions imposed on a model are so
strong &8 to imply some symmetry, we say there is a 'symmetry by restriction' and
subclassify it as category 1.1. This would be analogous to what happens in, say,
group theory: Many groups have automorphisms, i.e. symmetry maps compatible
174

with the group composition law . The &XtoffiS of sroup theory are so 'restrictive'
that, very often, they lead to different parts of the group becoming so similar that
it. is possible La find aut.omorphisms, which map different such structurally identical
pieces into each other.
For example, in QeD there a re several quark flavours in colour triplet represen-
tations, whose interactions are restricted by the following conditions: (a) Poincare
invariante, (b) colour gauge invariance, (e) renormalisability, (d) cancellation of
gauge anomalies and (e) smallness of representationa. These restrictions are 80
strong that, when the quark IIUUI8e8 can be neglected, a symmetry results under
unitary transformations independently on the left-handed and on the tight-handed
components oribe quark fields : there is a U(Nr)L x U(Nr)R chiral symmetry, but
with an axial U( l ) anomaly [paper 10} .
In addilion to having restrictions, it might be that the vacuum, or some other
important structure, adjusts itself in such a way that some symmetry results. \Ve
then say that we have a 'symmetry by adjustment', which we subclassify as category
1.2. An example of this phenomenon is the Peccei.Quinn mechanism [paper 9i,
leadin!!; to CP invariance for tbe stron!!; inLeractions. In this case it is really two
Kalar fields with non·zero vacuum expectation values, which adjust themselves to
minimise the vacuum energy density and whose relative phase then cancels the
effects of the 5U(3) topological 8 term.
It is only acceptable to consider symmctries that are not really present in the
fundamental theory &8 examples of symmetry derivation, if they exist as symmetries
for a new language describing the physics in terms of degrees of freedom relevant to
the limit of conditions under eonsideration. Such a new language is ofl.en introduced
when stepping down from one level to the next on the quantum staircase (Fig. 1.1 ).
A symmetry may then happen to be present in the language appropriate to the latter
but not in that of the former level. In the case of macroscopic scaling symmetry,
for example, we pass from the language of molecular dynamics to the language
of macroscopic physics. For group II symmetry derivations, we therefore h ave a
relation R corresponding to translation between the two languages. This relation
R between the 'fundamental language' of the original 'model ' and the 'effective
language' of the new field of physics will, in general , not be one to one .
If many sLat.ements in the effective language correspond to the same statement
in the fundamental language, it may be possible to identify a symmetry of the for·
mer simply by transforming the effective language, without changing anyth ing in
the fundamental language. We shall call an example of this type a ' formal sym.
metry derivation' and subclassiry it as category 11.1. Many of the gauge symmetry
derivations in this book are of this cat.egory.
On the other hand , if many st&tement.s in the fundamental language correspond
to a sin&ie statement in the effective language, it means that much of the content
in the fundamental theory is ignored in the effective language. Such a procedure of
ignoring structure opens up the p088ibility of finding more symmetries; these 'IiYm.
metries by ignoring structure' are subdassified as category 11.2. Scaling symmetry
 I7S

in macroscopic physics is of this type, since one haa to ignore a larse amoun~ of
detailed information , about the positions etc. of the molecules, in order that such a
sealing symmetry may he implemented . A huge number of molecular configuratioJUI
are described, approximately of couree, by a given density and velocity field for the
8uid in the effective macroscopic language.
The above suggested classification scheme is not entirely free of ambiguity : it
turns out to be a matter of degree whether or not a symmetry can be considered as
really puaent in the fundamental model. It namely depends on bow much structure
is considered to be part of the fundamental model and is required to be respected
by the symmetry. A class of examples of such doubtful classifications is what we
could call the 'hidden symmetries', Taken literally, hidden symmetries should mean
thOR that are really present in the fundamental model hut are not immediately
recognised. So we should expect them to be classified under category 1.1; symmetry
by restriction. (We here use the term 'bidden symmetries' in a narrower sense than
in Chapter 4.2 and 4.3 where it was taken to include local syoundries, such as
SUeS) symmetry in N = 8 supergravity; local gauge symmetries are more properly
classified above as formalsymmetriea:.)
A typical example of a hidden syounetry i..!I the 0(4) symmetry of the hydrogen
atom or of a planet--sun system [paper 4]. For an exact non-relativistic Coulomb
force system, there is in one sense an exact 0 (4) symmetry; the Hamiltonian is
invariant under the four dimensional rotation group 0(4) . However it is really only
a true symmetry if certain structure is abstracted (rom the lIystem. There is, for
instance, D O underlying isometry of the geometrical space of positions of the electron,
or the planet, corresponding to the 0(4) transformations. An 0(4) transformation
ads on the full position and momentum phase space, in general mixing up position
with momentum, and is not compatible with the spatial metric, i.e ., it maps a
pair of point.. into a pair with a different spatial separation . This means that the
position space metric is part of the structure that must be abstracted from the
Cundamental model, iC the 0(4) transformations are to be considered a symmetry
group of the model. If this suucture - the metric - is taken to be an integral part
of the Cundamental model, we must reclassify the bidden 0(4) symmetry derivation
as 11.2: symmetry by ignoring structure. The 0(4) transformations only constitute
a synunetry of the effective model obtained by ignoring some structure, including
the position space and iLl metric .
The 50(32) and Es )( Es symmetries of the heterotic strin!: [Paper 22] are also
hidden symmetries, which can quite analogously be classified as category 1.1 or
category 11.2. In this cue it is necessary to ignore the structural difference between
a soliton and a phonon mode of excitation of the strinr;. Similarly the hidden non-
compact Er(+7) symmetry of the N = 8 superr;ravity model (Paper 18] may be
classified under 1.1 or 11.2, dependinr; on the point of view .
So we shall classify the symmetry derivations, in Tables 7.1- 7.5, under the fol-
lowing five headings:
1.0. Transformed symmetries
176

1.1. Symmetries by restriction

1.2. Symmetries by adjustment
Il.l. Formal symmetry derivations
11.2. Symmetries by ignoring structure
The columns of the tables will be used to indicate (1) the role played by any limit
in the symmetry derivation, and (2) the sug:esting principle motivating the idea
that the symmetry might exist.
We consider first the possibility that a) no limit is needed in order to derive
the symmetry. For instance the hidden symmetries discussed above, such as the
0 (4) symmetry of the hydrogen atom, the SU(N) syrnmetry of the N · dimerusional
harmonic oscillator ete., are valid exactly inside the idealised models considered:
the hydrogen atom is taken to be purely non. relativistic , etc . It is not necessary to
restrict the region of study to, lay, low energy inside the models considered and the
hidden symmetries are placed in column a) : no limit taken.
The other p068ibility b) is, of course, that a limit is significant either by ba)
simply causing the validity of the symmetry in that limit, or b,B) suuesting a new
point of view or language appropriate to the limit. The two cases ba) and b,B)
apply to group I and group II symmetry derivations respectively, i.e. according to
whether or not the symmetry is truly present in the fundamental model. Typical
symmetry derivations of class ba) are those for which the symmetry appears at low
energy, due to the suppression of symmetry breaking terms towards the infrared
by the renormaliaation group equations . The best working renormalisation group
symmetry derivation is that of Lorentz invafiance in the infrared, for non-covariant
gauge theories with non-covariant chiral fermions [Papers 24, 25). MacrOl!lcopic
scaling symmetry, as discussed above , is an exsmple in which the significance of
the limit is rather to influence the choice of lan!Uage and, thereby, to make the
symmetry manifest itself, as far as the relevant degrees of freedom are concerned.
When we here talk about 'limits ', we mainly have in mind a corner in the
space of possible physical conditions; typically where the partides have low enough
energy and momentum so that, for example, they can be produced under available
experimental conditions. However , in a few of the symmetry derivatioIlB presented
in this book, it is crucial that some parameters of the fundamental model tend to
a special value, say zero. Such a limit cannot be reacbed simply by restricting the
experimental conditions; one must rather hope that nature has chosen , for some
as yet to be underst.ood reason, to select parameter values clOlie to the limit. An
example of a parameter, relevant to understanding the origin of symmetries, is the
ratio of a quark mass m t to the QeD mass scale parameter AQco . It is, for instance,
the smallness of this ratio for some quark flavours that is responsible for the chila!
symmetry of the strong interactions [Paper 1OJ.
We therefore distinguish between 'experimenta l' limits and 'parameter' limits in
t he tables, by introducing separate columns labelled bae:) and bop) , for symmetries
which are truly present in &ome working limit of the physical conditions, e.g. the low
energy limit, and some parameter limit respectively. We have found no examples
 177

of a parameter limit in the group II symmetry derivations and we, therefore, have
& single column labelled b,6) in Tables 1.4 and 7.5.

The firat. column of Tables 7.1- 7.5 is used Lo pve a 8uggestin~ principle, where
appropriate, from which the idea of the symmetry group might be &aid to originate.
For inst&nu, many of the fundamental models considered contain the geometry of
Minkowski space-time, or of just three-dimensional space, as an ingredient in their
structure. 'True symmetries of such a model must transform, although perhaps
trivially, this geometrical structure in an automorphic way. This leads to the idea
that symmetries of the geometry are especially promising candidate symmetries
for the full model. Derived synunetries which can be interpreted in this way will
be ca]Jed geometrically lugested symmetries. We have found sev~al symmetry
proposal mechanisms useful in classifyinS symmetry derivationa and we list them
below:
Geometry
Fermion number
Analytic continuation
Repetition or permutation
Stability sub&roup
Approximate symmetry

The first four of the above symmetry sugestin& principles are used to subclassify
the restriction symmetries of Table 7.2. Some of these symmetry derivation exam-
ples were difficult for us to subclassify and. for such 'left-over ' cases, we len the fint
column blank. Tbe princip~ of' Analytic continuation ' is, in practice, applied 1.0 the
Lorentz group, in order 1.0 derive the CPT theorem and 1.0 motivate the related idea
of particle-antiparticle or mar&e conju,ation symmetry. The other two restriction
symmetry subclassifications - 'Repetition or permutation' and 'Fermion number'
- represent rather ,eneral ideas. In the case of ' Repetition or permutation' in-
spired symmetries, some objects in the model - e.g. particles - are restricted so
strongly that they have identical behaviour in a.lmost all respects; thus they be-
come related to each other by the permutation symmetry operation.. In quantum
mechanics, even superpositions of such identically behaving objects are considered;
so the permutation symmetry is extended to a symmetry under the transformations
of a unitary group. Similarly the idea of various types of 'Fermion number' sug-
gests the existence or generalised cbar&e conservation laws and synunetry under the
a8B0Ciated phase tranarormatioru.
The subclassification of the formal symmetry derivations in Table 7.4 required
us to introduce the last two symmetry proposal mechanisms listed above. Several of
the formal symmetry examples are sugested by considerin& a field takin, values in a
homogeneous space. This homogeneous space is identifiable with the space of coseta
GI H, where G is the group of transformations of this space and H is a subgroup in
G: the subgroup H is the 'Stability subgroup ' for a point in the homogeneous space.
The stability sub&roup H becomes the gauge symmetry group in the examples
considered, which we therdore identiry as 'Stability subgroup' suggested . In the
178

formal r;auge symmetry derivations of (paper 26] and of the field theory glass, an
approximate gauge symmetry of the fundamental model is shown to manifest it.seIC
as an exact symmetry physically, by the inverse Higgs mechanism. The gauge
synundry group derived is of course sugested by the 'Approximate symmetry ' of
the fundamental model.
The tranformed symmetries of Table 7.1 are already naturally suggested by the
original symmetry ~oup of the fundamentaJ model considered , whic.h simply gets
transformed into the derived symmetry. So we have crossed out the first column
of Table 7.1 as redundant. However it must be admitted that sometimes , as in
tbe case of general relativity, the symmetry group of the fundamental model is so
huge that the choice of subgroup to be transformed requires (udher motivation.
The Poine&re subgroup is luuested as the physicaJly relevant symmetry, since it is
the only surviving unbroken symmetry in the usual flat space-time vacuum. Our
IIG-called derivation of Poincare invariance from leneraJ relativity could therefore
be said to be 'sugested by the Oat universe'.
We now present Tables 7.1- 7.5 which constitute a classification of all of the
symmetry derivations discussed in this book, according to the above principles.
The number in front of each entry refers to Table 7.6, which is a complete list of the
symmetry derivations given in the order in which they occur in our commentary.
Table 1.6 also contains cro&&-reference& to the classifi cations given in Tables 1.1-
1.5 and indicates when a symmetry derivation does not work or only works rather
inaccurately.

LO Tranalonned .ynvneuies

Suue-tinc -) bo.) I bop)

principle No Jimil taken Symmet.ry tn.d,. JII"'M'IlI in
workin, limit pvamel.u limil

Karu-.Kkin TrIlDll.tion.aJ.
"
sause .ymmetry
32
invuiance of &1_

G&\lP aymmetry
"
f1un atrinp

DiII'eomorphi.an
"
aymmetry from atnn,;.

30 PoUu:ar.! iD~

en- ,mer.u reLati..it,.

 179

Table 7.2. C1auificatioD of .yrometl')' deriftlionl: Symmelne. by ralriction.

co 5' ........... b, _Iridioo.

.) bo, bo,)
5061""'"
priKipie No liM" l&lea 5,,,, ..etQ' tral, p~.1 i.
_'Wo,Ii",,' po. • .mctu Ii"';!

Cooob ... 01 i. nriUOl! , Paril, of ...... -..pk:

"
f ...... ocaIi .... duUc:i1)'

c....mctry U
f, o,",
Lo, •• '" i • ...na.c.c
."'' 'SO
• Spi_ deca.pb,
i •• 1ooIaioc ,.,....,.

" Bjo.loc. ",coli.,

Lo ••• " iu.nUlOt

,f""0'"... ••_ ..
p
oIi .........

, QUlk "¥OIl",;, qeD , Q • .,k ....... '" i.

... dQEO SUad ud loIodeI

U 8&11_ •• mIter B U Bu,.... ....loft.

jud.eIi., weak ....111&17) (iIod.di", ......... _ 017)

Lcpl.ollb"".111

fUlllio ........
" Lep",. . .""'".
£ • •£ • .£.
(... ddi., wu.I< ...,-...1,)
"
£ • .L•• £ .
(u.d.di", _ ....... _at,)

Lepl.cla . . _ .
"
dlf• ...,_ L . -L • .L. -L .

~L{......

"
S.udud Modd

11-£ ftom
"
SU(5) CUT

A...l,Uc .,.,.!.i...u.... •"omC!r.af,<

QCD
.....,;.,........
QED
&ad
C
• a..,," co.j .... tia. C
...... SUadudloiodd

3 C""

Ikpdi ....
"f._
otriq:
SO(3~)ud 1:•• £.
1M ....rolic " c.uoll,.mmct.y
SU( .... ~)L. 5U (.... rll!.

,.... ......tio. n._.co..._

"
llooo "'" .."."al
"" ...... (GIN )

, 0( ' )-., .........., of C ...~ " __ '" In...

.yd.",. . .10... "................1Ooa V""p
• SU(.}-.,,,,_.,
lHIi .... uio.al kunooaic
of

oociJl.Io,

" If,t . ' 1"''''''''''''<1

frono I'd '.P"II,a";11
180

Table 7.3. Claaaificatioll ohynunetry deriYMJoN: Symmetriel by adjustment .

.., 5'._.... b,. odj.. ~1

S-umlaa
priJlcipk 1'1.
.)
Ii,.it t . . .
....
5' __ "1 lnIb ~I ,.
... ,)

_kia, Ii";! pt.f . . .der limit

Aaol,.tM: COIIIl...tiooo 10" II CP, T, P

r- f'Q-aloded
...I a - ' " 01 u- Sludud MocId

u' _ " CP, T, P fr_ "'117

..i _ ,.,..... deoI qeD " Ct>, T, pm- b.alIr
..i __ o. ...... aI
$lqdud. Mo6d

11.1 F_ol .,...IIMU]' .maw..

.) ")
S""uli., priacipic No h lit take. Li "'t ,"U.. t..lu.&."C"
p ....., ..,allld.rJ

\ /~"Cpo-I"(>•.....,.
' our .,-_, ,.
X 22 H- I·"~ ,, __ t., ;&
S •.b;;IiI, .. bv-p
~ CIH_"'d
/ n SU(. h....c 1l'_"'"'7 i.
1'1 _' ' .PUI.a';"
Appnru ..... k .,...mdrJ

"'" "'" '1--..,.

/'
/ l4 For..l.Ilppcuaaor of '"qI'

Table 7.5. Cluaifie&tioll olayrlll'Dl:Uy derivatiorUl: lpaorin, structure.

11.2 5'. -.... lor ••ori. . 11...,,-,,,

.) ")
Sa"",tl", pri.aple No li';tlaRa Uai' _u_
'e
l ...... cj"; ", . , . - .,

""",,,0 > MIICIOKopic .cali.,

Rcpetil.i.o•• ad pc •• • 1.&IOoo. 50(n) ...d B • • S. !n>oa
"
IIIoc H\.uoW t4riac
, 0( 4)-o' . - " J of .1<1",&" uo.

•wSU(a}-o,.met.,
-ic
01
ooribl«
oHIiaeasio.ol

" 6-,("1""' . - " from N.,

•• per.....; \)'
 lSI

Table 7.6. Lilt or Iynvnetry derivation •.

CIo&ptel' U. 5' ........... r""" ....-.d ..uyiotic plo,oi.co

. "- pio: oeaIiq fro • .oIen1.. plo.,...,. 11.2•...,
• ........1<1
, Parit, i .uri"... d. lloe elutio: ptopfttM:o fII. ..,~
._U1
'.!.I.u
•• 0(4}-&1 • • U," ~ _ _ &ad 1M k)',hoc....1OIa
£pIpei' 4] I.I .~
• 11.2••

•• SUI ..j-.,.._'1' for 'M oHIi........ ol ..... _ic ...n. ....

(Papel' $) 1.1 .•
• 11.1...

..,\&1
•• Sep.t. •• t.: ............""" <II .pi_ uod 0 • _,_Iu - . 1 ...
ia .000_.dUioioUc ..'-'c plo.;pia (Spi. dec ... pI;",) ,...........,
1.I ..... c

• Wic_ SU(4r-,.-ur al • .doe.. pIo,.;..

[f""", I] ntlou i._. . .
, SU(I)-o,.",et., til Itroa& il""octJ,oo,.
{PIpet 7J ntH< i._...."
ChpW' III. 5,.. _ . .... fro. , ... S...... dud Model

l .2. Su oq .... 4 8ecu_~.."" ilt.t<Ktiou

•• n........ 1''''_' ' .... q .......... 4 lepto.. 1,-, qeD &ad QED 1.1 ••
[papu .] ru..... ""'~,
• a. ...., • ..,.j.pliooo c r..,. qeD u.d QED
r",. SWId. .d WoOd
1.1 ••
1.1......
{PI",' 21 ...JrUc COOIli. . .liooo

". PuilJ' P""'" Puai-Q.... ac.....kd 51....!...!. Wodd

(P",en ' " IJ 1.l.I.u
,_"1
H. 1'1.. ~ T ...4 Ct' f , - I'«ooi.Q.i.. ",,~ded SIAad.,d 1.1.10..
....... [Paper'l
uaJ,llc COOI'l ..............
,.........,
". o.;• .ol.,.. _., SU( N,h.IliSU(N,h!. 1.1.b.,
{Po.,.,.. Ill] R.po<''' '& &ad """' . ..........
182

T.ble 7.6. Com 'd.

J .S. r .u StaMud. Model

u. Bu _ ...M. ClHMfYalioa B l l .bu ('ad . ....,.. ~)
[P.~ III 1.1... (ud . .. _al,)
Ic, ... ~ •• _'""

". ~ la_, co...... . . . t • •L• .£. 1.1.1••• (iad. _o.b)

1.1... (<DId. .....wr)
.., .. ioto . . .1>n

U. Ltpt<MI h_. "..........

I"" .-'.di ... _ o.t ... _ oJ, )
L . _ L. ,L . _ L. L1 .•
"'.111;"" .....be<
". Bal}'CMl ••• be . . . a.. Irptooo . . .M. 8-L
(.-.n ,adoodi ....... a.k . _.., )
LI••
I< •..u- .....be<
". F1&_. COOlon ....... lot ..,.,,01 ........La (elM ............... )
[Paper U J
1.-.1>.0,
..,petitiD.

OL B;o.kc. oahC U .b...

,""",''''
". Co.""'.....
-"
;.y~ f,.,. ""oJi., 1.1 ••

a..pkt ..... Ik_d l,Ioe S L~ud Wodd

• • 1. c ....... ;~u.;., •

". 8 ... _ ....be . ..i ... 1qo1.oa ....Iter B-J.. f ..... S U(~) · G U T 1. 1••
............ ",be.

n.
• . 2.
U(I )-&" ..,Q __ u,
ia CP"- ...........
Hidde. IDaJ . , . _. , ...t ", ...ic,.] , .... booooo. i. _-li_

lPapc:. i1)
11.1.""
....... -.lei •

...b;ti\J' .. ""... p

". H-,u,< '1",mct. , fro. lueral G/N-oilm • ...o.kl 11.1.""

...bi~Ll .. bs .... p
1l0Un ., ..-..... i. N • ••• paC. u'"
4.1.
U. SU(I}-,".,c " - _' 1 1m. N .. ......., ••• ;11 1I. \.b,IIuM•.",1o _ .
[Papc. 18\ o'abitit, •• Itct-p

". ~+f) &Iobal SJ _ _ t1, .......

N _ .... ,.......;1, 1. 1.• or
[P.,... II J I.l.•
4.4. K&I .... K.... I""",

". C " Ie ., .. _ . , io Kal .... KIci. _odd

[P.,... 20/
1.0••

4.' . S,,;.p
H. SO(12) ......, .£, " __ 1}' f..,.. ~o.eooti<. oui., 1.1.- or 11.2••
[p.,... nl repe<i_

". C •• le 'J _ _ " f._ ouiop 1.6.•

". Oi'_p~;""

olti.p
o,. .. md.lJ 01 ulc •• &I,cacfOl «1.li.;ll r._ 1.0••
 183

T.ble 7.6. Cont'd .

1.0••

"
n . "haul&ti",...1 in .......... i. tu 10'"1 .a-...la&llo limit i • •
pa..lilr& • ..,.....

u.
[Weia'-al

[P~< 21]

1.I .b.. Doc..04. _.k,Hot,

lot ..,,,,ptoticalb' IHe

C"'pl.cr V1L

QCD;
(,0_ bahJ aai_ CIlooaded S'....d ... d Mod" lU... o<.-l
.....l,l'lic COIIU...lio-. ud
&_etry

Some of the placements of the symmetry derivations in the tables deserve furiher
explanation:

Ta61e 7.1: Tran${onned symmdriu

As mentioned above, some very special diffeomorphism symmetriel of general
relativit.y are transformed into Poincare invariance. Similarly special diffeomor-
phism symmetries, 85S0ciated with i) the n extr a compact space dimensions in
Kaluza-Klein theory and ii) the tw~di mensional space-time of the string world
sheet, are transformed into gauge symmetries. The translational symmetry of a
glassy theory at large distances is inherited from the large scale translational in-
variance present statistically in the assumed glass-like structure of the vacuum. This
derivation of translational invariance comes close to transformin8 a symmetry into
itself and it is a moot point whether it should really be counted as a symmetry
derivation.
184

Tdle 7.£: Srm mdriu 6, reJtridi"n

All the 'experimental' limits of column boe) correspond to the low energy or long
wavelen~b Jimi ~ , except for the case of Bjorken scaling in QeD. Sjorken scaling
symmetry appears in the opposite limit, when the energy is large compared to the
mass Kale parameter AqCD and the relevant quark masae8.
The quark flavour symmetries and also charge conjugation symmetry have been
classified under the ' no limit' column a) . as wen as under column bae). These
double placements are made, since the symmetries may be derived either from
the standard model, or from a model composed from just QeD and QED. If the
symmetriefl are coJUlidered to be derived from the standard model, it is necessary t.o
take a low energy limit in order to avoid weak interaction efreda. However, if the
'"fundamental" model is taken to consist of QeD and QED no such limit is n eeded.
Baryon number and lepton flavour conservation lawl are also presented twice in
the table, according to whether or no~ the effects of the weak interaction anomaly
are taken into account. The baryon and lepton number conservation laws are vi~
lated by sphaleron· like transitionsM between the topologica.lly distinct vacua of the
standard electroweak theory. In order to eliminate these baryon and lepton number
violating processes, it is necessary to take the limit of low temperature2 relative to
the electroweak phase transition temperature, T < 300 GeV; it may also be neces--
sary to take the limit of low energy3G relaLive to the electroweak SU(2) sphaleron
energy, E < 10 TeV. However, the differences between lepton flavour numbers,
L. - L,. etc., and between baryon number and lepton number, B - L, are anomaly
free and exactly conserved in the standard model; no limit is required.
We felt a lit tle doubtful about how to classify the symmetries resulting, in the
infrared limit, from the use of the renormalisation group equations. It is not suffi·
cient just to make use of the general restrictions on a renormalisable theory, in order
to ident ify these symmetries. It is also ne<:essary to calculate the ,8·functions for
the various coupling const&nts, in order to see how the latter develop towards the
infrared . There seems to be no II priori principle, as to when such a calculation will
result in the running coupling constants tending to values which respect some sy~
metry. In fact , it seems that most of the discussed cases of renormalisation group
symmetry derivation only work under somewhat spe<:ial conditions. The derivation
of Lorentz invariance is the most robust example. It is apparently rather aceidental
wbether or not a symmetry is derivable from the renormalisation group.
In order to shed more light on these renormalisation group derived symmetries,
the following brief discu.ssion of the ph)'llical mechanism underlying the renormali-
sation group may be helpfuL The renormalis.ation group effects are quantum effects
due to loop diagrams. This means that they reRect the effects of partieles being
virtually split into the 'bare' particles of the fundamental fields . More and more of
these splittings are taken into account, as we move towards tbe infrared limit. This
means that , in the far infrared , the observed particles are composed of very many
'bare' constituents, as illustrated in Fig. 7.4.
As an example of the effects of these many constituents, we may consider the
 1&5

p.-
p.'
w-
q z z
e- e' p.'
T- q T' T' p.- q
p.t e- y
e' w· Ii. T- 1'-
W-
e'
q
b b
Dressed Dressed
electron moon

Fi,. 7.4 . IDu..tr.uon ~ the compo&.iu. uructun or the dreNed. eie<:tn;lp....t the dn:a.cd rrI\lOn; thi •
• trudwe can be revul.ed by hip efleru probes..

derivation of Lorentz invariance in non-covariant electrodynamics [pape r 24]. The

different metric tensors for the phot.on , and for the len-handed and ri&hlrhanded
component.! of the electron, must be ILveraged over all their respective constituent
partic.ies. Consequently, in the infrared limit, 11.11 the particles have the same -
namely the aVl'!rage metric tensor.
In a s imilar way we CAn intuitively underlltand why any gauge symmetry vic>
lating terms, introduced into the pure Yang-Mills particle sector of a non-asympto-
tically Cree theory, tend to zero in the infrared limit [Paper 29J . The presence of
several fermion multiplets in a non-asymptotically Cree theory implies that the Yang
Mills particlee become more and more compo.se:d out or rermions in the inrrared.
The YanS· Mills selr.energy contributions, in the inrrared, then come dominantly
from loop diaUams consisting or two fermion lines. Due to the global non-Abelian
symmetry or the rermion kinetic energy and interaction terms in the Lagrangian
density, these rermionic contributions to the renormalised YanS-Mills kinetic en-
ergy term are automatically locally r;auge invariant. So, &8 these contributioIl8
become more and more dominant in the inrrared limit, the selr·interactions or the
Yang-Mills particiee tend to their gauge symmetric rorm.
Conformal invariance has been listed as r;eometrically inspired . This may be
considered a slight exaggeration, si nce the conrormal t ransrormations do not rorm
an isometry group or Minkowsld space-time, but rather rescale the metric by a
space-time dependent conrormal ractor Q( z) :

( 104)

However conrormal transrorma1.ions do act on space-time and preserve the important

Iightcone structure, mapping timelike vecwrs into timelike vecwl'll, null vectors into
null vectors and spacelike vectors into space like vectors . So we feel justified in listing
geometry as suggestive or conrormal symmetry.
186

There are just two entries in column bap) corresponding to symmetry derivatiOILI
relying on a parameter limit: (i) chiral symmetry, and (ii) flavour conservation for
neutral currents (the GIM mechanism) . Both examples rely on the relevant quark
masse. being small , compared to (i) AQCD and (ii) the W and Z gauge boson masses
respectively.

Table 1.3. Srmmetriu 6y adjustment

There are just the discrete symmetries P, CP and T in this class ; CP and T
are of course essentially equivalent according to the Bafely derived CPT theorem.
They are entered three times in the table: In the Peccei-Quinn two Higgs doublet
extension of the 6tandard model, the vacuum adjusLs itself to cancel out the effects
of the QeD topological 9 term; the low energy limit must be taken to avoid P and
CP violation effects from the weak interactions. In baby universe theory the effec-
tive coupling constants become dynamical variables and, more controversially, may
naturally adjust themselves to have very sharply peaked probability dis~ribu~ionsi
in the cue of the t.opological 8 parameter of QCD . it is predicted to peak at a P
and CP symmetric value. P and CP symmetries are then derived, if QCD is taken
to be the fundamental model, without the need for any limit . However if baby
universe theory is applied to the full standard model, it is again necessary to take
the low energy limit in order to avoid P and CP violating weak interaction effects.

Tobie 7.4. Formal Symmet,., Derivation

This t.able comprises the various derivationa of local gauge syrrunetry for field
theory models which , in their original formulation , a priori have no such symmetry.
In all the cases listed, the local gauge symmetry is introduced by a rather trivial
extension of the notation used: the synunetry is initially purely one of formalism.
It is of course p08Sible to introduce practically any symmetry by a careful choice of
formalism; however usually such a symmetry would be of no physical significance.
In order for a formally introduced synunetry to be considered a true symmetry
derivation , there must exist some reason why precisely the formalism chosen is
relevant in some physical conditions. The~ has to be a physical limit. under which
the new language or formalism is enforced on us, before the formal syrrunetry can
be accepted as being a physical symmetry; we have t.herefore cr08Sed out. column
a) , which is reserved for symmetry derivat.ions where no limit. is involved. The local
gauge symmetric formalism is needed in the low energy limit of modeJs where the
inverse Higgs mechanism operates.
The formal gauge symmetry under consideration would be broken by the non-
zero vacuum expedstion value of a scalar fi eld , analogoas to that of the Higgs
model , were it not. for quantum flu ctuations. Radiative conections can renormalise
t.he parameters; in part.icular t.he wo uld-be neliative mass squared Higr;s scalar fieJd
may be renormalised to have zero vacuum expectation value and turn out. t.o describe
an ordinary positive mass squared spin zero particle. The inverse Higgs mechanism
then operates and the quantum fluctuations are 80 large as to wash out all physical
 187

gauge symmetry breaking effects. The massive vector bOllOns, which lose their man
by this mechanism, may either already be present as elementary particles or arise
as composite particles in the fundamental model. In the first case vector boeons
are present as elementary particles and tbe fundamental model p08lleS8e8 a manifest
approximate gauge symmetry; in the second case they are absent but the model
contains, for example, fundamental scalar fields.
The successful operation of the inverse HiUS mechanism and the consequent
existence of an unbroken gauge symmetry is & maHer of dynamics. The parameters
of the fundamental model must lie in the appropriate range. [n the C P"- t model
the coupling constant f has to be greater tban some critical value37 lei in d = 2
space--time dimensions /e = O. For the general GIH sigma model, the inverse Higgs
mechanism is not expected to work if G is a non-compact group, as is tbe case
for E7(+7) in tbe N = 8 luperuavity model. In general, for the compact group
models considered, tbere is a whole phase where zero mass gauge particles exist : it
is not necessary to take particular limiting values of the coupling condants, but juat
values lying within lOme extended region of parameter space. We therefore do not
classify the gauge symmetry as arisin, in a parameter limit, but rather as needed
to describe the degrees of freedom surviving in the low energy limit .
At low energy, the massless gauge particles can be produced but the massive
scalar particles of the would-be Higgs field H cannot. The formalism shou ld reflect
this physical property and the H field should be removed from the effective field
theory in the low ener!), limit. This requirement that the would-be Bigs field
H should disappear from the formalism in the low energy limit is not satisfied in
the fundamental models considered: the gauge is fixed in the original formalism by
requiring the phase of the H field to vanish , arg H = 0, which clearly makes no
sense once the H fields is left. out. In this way the low energy limit drives us to use
a new formalism , which corresponds to choosing another gauge and is thus strongly
suggestive of a gauge invariant formalism.

Table 7.5. S,mmelriu ., ignoring .tlnchlrt

Macroscopic scaling symmetry arises in the limit of a macroseopically large num-
ber of molecules, when the details of how the individual molecules distribute them-
selves locally is not important. A description in terms of macroscopic hydrodynamic
variables is then more appropriate than in terms of the more fundamental molecular
variables. In this new effective language of hydrodynamics, the detailed molecular
configurations are ignored and the scaling can be an exact symmetry.
The other entries in the table are what we call 'hidden symmetries', which
have also been included in Table 7.2. The classification chaaen depends on how
much structure is included in the fundamental model under consideration . For
instance, in the heterotic string model, the objects that a re transformed around by
the E, x E, or 50(32) symmetry operations are not all of the same kind . We may
consider the transformations acting on the massless excitations of the string. These
quasi particles are all connected with the 16 extra dimensions in the model and are
188

of two types: 1) phonon. and 2) solitons. The quasipartides form a basis for the
adjoint reresentation of the symmetry group, associated with the root diagram used
in the Cartan classification of the Lie algebra. Those basis vectors corresponding
to zero weight are physically the phonon., while those corresponding to the non~
zero weight.s are the soliton•. The twisting of the .trin,; in the 16-<1.imeosional toru8
space, whenever IUch II. soliton is present, is given by the weight of the corresponding
basis vector. The E. x E. transformations do not constitute aD exact symmetry,
if the structural difference between II. phonon and II. soliton is considered to be an
integral part of the model; the symmetry must then he classified M derived by
ignoring structure.

The many entries in the above discussed tables emphasize again that the origin
of II. large number of the symmetries of the laws of nature can , to a greater or lesser
extent, be explained. In a way, this main conclusion of our book is in conflict with
Einst.ein's well-known quotation: "Subtle is the Lord, but malicious lie is not". By
this statement, EinsLein meant that although nature might have a sophisticated
structure, she would not mislead us. However we would claim that some famous
physicists bave been misled about the fundamental physical laws by the experimen-
tally observed symmetries. For example, Heisenberg enshrined isospin symmetry as
one of the fundamental principles underlying his unified field theory of elementary
particles. l As we have already emphasized, isospin symmetry is now underst.ood as
an 'accidental' consequence of QeD and the smallness or the up and down quark
mas8e8. The moral of our book is that symmetries are not so holy.

a.efereaca
1. w. HeiH:oba-s, Ilttroe.clin t. 11e UltiJie~ Fiel~ n •• Y)' oJ £I ...ultl ..., P.rtid ., (Wiley,
1966).
2. V. K........m, V. RubakoY and M . Shapwhni....... P1, •. Lell. 1S5B, 36 (19M).
3 . S. Wein~, in Sccn~ Wo ...bh, OIl G,.It~ U.iji •• liu. (Birkb&UKT. 1981). p. 291.
4. J . C. Pati. in S.ctlt~ Wor.h"', OIl Gr... ~ Ultiji c.li ... (Birkhi....,r. 1981). p . 195.
S. J . C. Pati and A. Salam, PI" •. Re-. DI0 , 215 (1914); R. N. Mohap&tra and J . C. Pali. Pit, • .
Re.. Dll , 566, 2558 (1915) .
6. H . 8 . Nielac:n, in G••,c TIt.o"''' oJ lito Ei,lat;.., eck. R . R.aitio and J . Lind/on , (Sprir\&er-
Vcriac. 1983), p . 288; H. B. Nie!.en. D. L. Bermet and N. BNfte, in Recut D •• e/o,mCltt, i ..
O ... t .." Fief~ neo'l" , eck. J . Ambj."-, , B . DurhuUI and J . L. Pelenen, (E ...... icr SQence
PubUahera, 1985), p . 263.
7. D. Weift&wte:n. prj.ate oomnlunkatOon.
8. J . A. Wheder, ~Law withou~ La.~ , in Q .... hm T.h~Y)' alt.! Meanrtm elt l, edI.. J . A. Wheeler
-.ad W. H. Zw-ek, (PriDceton UniVCTlli~y PRH, 1983) , p . 182; J . A . Wheeler, MBeyond the black
hole~, in SOrlie Slnllt,cltu. ilt lite P....'orliolt, ed. H. Woolf, (Addi.on· Wesley, 1980) , p . 341 ;
R. P. Feynmaa. Tla. Claanl<ler .J Pi,.iu/ 1..111 (MIT P ....): G. F. Chew, in P .... erli •• pJ
lie F.,,~."Cltlall,,'eracti... , eel. E . ZK:hichi, (Edilrice Comp..ltori , 1973). p . 3; C . H. Woo.
~Mlsalon ImpwalbJe? A Look at Put Setbadu in ~he Search for Elementary Mat~r and
ror Univenal Synund.ri...~ , Zenlrum mr lnt.crdiAiplinin= FondtUJl&, Uni~nity of Bidelekl.
Preprint.
9. S. W. Hawkin,. Pi,•. Re.. DS7, 904 (1988): S. W. Hawkin, and R. Laflamme, Pit, •. Lel/ .
20eB 39 (1988); S. Coleman. Nul. P4, •. BS07, 867 (HI88); S. B. Giddinp and A. Stromin,a- ,
N ••1. Pit, •. BS07, 854 (1988).
10. J . 8. H&rt.leand S. W . H.wkir\& , P4, •. Re•. DU, 2960 (1983).
 189

III
l~
Pl~
S. c......... H .. l Pl, •. BU • • 043 (, ...), J. P .....lI. N..l •. B." '. ,<1 ('''~).
8 . Cnnatelnand M. B. WIoH, Pl,•. Ldl. 212B, 407 (1988); 8 . CnNltem and C. T . H.n, Pi" .
Left. 220B, no (1989).
13:. II. B. NieIHD and M. N..nom.iya, Pl,.. lin. LdL 112, 429 (1989); J. PnUiU, S. P. 1'ri¥ledi and
M. B. WiH, Pl,.. WIL 22,B, 26 (1989) .
J . PokhinUi, Pi" . LeU. 219a, 2$1 ( 1989).
W. FiKhla'and L. Suukind, Pl, •. Leu. 217B, 48 ( 1989); S . Coleman and K . Lee, Pl,•. Lelt.
2218, 242 (1989).
1 J . A . Wheeler, in Billdl, Ru, co "I~' : 19'7 Lcchru i. Mellemali" . .. '" PA";,, , eda. C.
DeWitt and J . A. Wheeler, (Benjamin, 1968). p . '242; 8 . S. !leWin , P.. , •. Rn . 160, 1113
(1967).
1 r T . &.nb, N.cL P",•. 8249, 332 (1935).
IS, S. Ch..dha, C. Litwin WId H. B. Ntet.en, Ilnpublishecl.
uL H. 8 . Nid.ea, Ado Pl,';cl P,t..i" 820 , 421 (1989).
'lIi. S. Wdnber&, Pl,•. Rn. Leu• • 2 , .sa (1089)~ J . J . s..z'i""c:o- d d . , Pl,•. Re... Lei!, .3, 1031
I ('98').
2 ~. S . C hadha and H. 8 . NidRn. "Natu.ralneu 01 Wqol Equ.a.lioD and 3 + 1 Dimnw.ionA1ily". in
A Report _ the ReKan:b Aetivitte. at the Niea Bohr In.titute and Nordita, 1984 , p . 117.
n,. V . K . F\a and H . B . NicJ.cn, N.d. Pl, •. 82311 , 167 (1984) .
. H. B. Nict.cn and N. Brcne, in P ...u: .•f IAe XVIIII.'n•• ,i.",1 5,,,,, . ...,,,, (Aadcmie de..
Wi...,nacbaClen der OOR. Benin-leuthm , 19&) .
. R. Gilmo~, Lie G,01I", Lit AI,dn, •• ~ S."'e Qf Ilei, A"lic. ,i•• " (Wile)', 1974).
25. H. B. NidlUl and N. Brene. Pl, •. Lell. 223B, 399 (1989) .
26. H. B. NielKn and N. Brene, Nul. PA" . 8224, 396 (1983) .
,t
21. L. 0 ' Ralfeattai&h. Gr01l, S,nd ...., .f G•• Tluriu (Carnhridle Universit), Preu).
~. K . HMMn, Nidi Bohr JNltiwte Pnprint NB I-HE-88-60 ( 19M).
211. O. L. 8enDeu, H. B. NielKn and I. Pied:, PA, •. LelL 'l0'8 , 27li (1988).
30. J. M. DrovJJe and J . B. Zuber, PA, •. R" . 102, 96(1983).
31. J . O),dak, in Pr.c. •J lA, IStA I.,. C. -.1. • • Hi,A Eat,,, PA,iu, p . 3.
31. A . eo.we, C. D . FroWn a...:I. H . 8 . NidMn, PA, •. Leu. UnD , 347 (ISM).
33. Particle Data Group, Pl, •. Lell. 2UD (1990) .
34. C. O. f'rou;al.t and H. 8. Nieben, Pl, •. Lell. 101lD, 487 (19111).
M. F . R. Klinkhamcr and N. S. Manton, PA, •. Rn. 030 , 2212 (19&4) .
' . V. A. Rubakov , in PrK . •, Ih ISlA /.,. Co flj. f . Hi,l Eat,,, Pl,.iu, p. 309; L. B. Ohan,
i"~., p . 319.
3 J. Va. A~r'eva and S. I. Aukov, N.d. Pl, •. BI1I2, 298 (1980): T . KUIO, H. Tuao and S.
Uehara, p...,. Th.r. Pl" . 5.,,1. IS , 122 ( 1 98~) .
REPRINTED PAPERS