V.

IX'

Lawrence Berkeley Laboratory

UNIVERSITY OF CALIFORNIA

Physics, Computer Science &

Mathematics Division

VORTEX OPERATORS IN GAUGE FIELD THEORIES

Joseph Polchinski
(Ph.D. thesis) mwmm&iom
MASTER
July 1980

Prepared for the U.S. Department of Energy under Contract W-7405-ENG-48

 VORTEX OPERATORS IS GAUGE FIELD THEORIES

Joseph Polchinski
Lawrence Berkeley Laboratory
University of California
Berkeley, California 94720
ABSTRACT

Ws study several related aspects of the 't Hooft vortex opera

tor* The first chapter reviews the current picture of the vacuum 01

quantum chromodynamics, the idea of dual field theories, and the idee

of the vortex operator*

The second chapter deals with the Abelian vortex operator v/rit-

ten in terms of elementary fields and with the calculation of its

Green's functions. The Dirac veto problem appears in a new guise*

We present a two dimensional "solvable model" of a Dirac string*

This leads us to a new solution of the veto problem; we discuss its

extension to four dimensions. We then show how the Green's functions

can be expressed rore neatly in terns of Wu and Yang's geometrical

r
idea of "sections". The renorealization of the Green's functions o.

two kinds of Abelian looplike operators, the Wilson loop and the vor­

tex operator, is studied* In each case the possible divergences are

easily determined with the aid of the operator product expansion, and

for both operators only an overall multiplicative renormalization is

needed* In the case of the vortex this involves a surprising cancel­

lation*

In the third chapter we discuss the dependence of the Green's

functions of the Wilson and 't Hooft operators on the nature of the

vacuum* We emphasize the cluster properties of the Green's functions

rather than the vacuum expectation values* We explain 't Hooft's

Fannie and John Hertz Foundation Fellow

,n,».mrT--.-- " -iTtS'JHUMITBf

 iii

Acknowledgements

I would like to thank my thesis advisor, Professor Stanley Man-

delstam, for such guidance and instruction- I am grateful to Profes­

sor Korkut Bardakci for many helpful and enjoyable conversations, and

to Bob Cahn for advice and encouragement* 1 would like to collec­

tively thank everyone in the LBL and U. C. Berkeley theory groups,

especially the members of the 1978-79 and 1979-80 study groups and

softtall teams, for fun times and for helping me to improve my under­

standing of physics*

I am grateful to the Fannie and John Hertz Foundation for five

years of generous support, and to ay parents for their continual sup­

port and encouragement* Finally, I would like to thank my fiancee

and sweetheart, Dorothy Chun, for love and for sharing with me the

various crises of graduate school*

This work was supported by the U. S. Department of Energy under

Contract No. W-7405-ENG-48.
 VORTEX OPERATORS IN GAUGE FIELD THEORIES

Table of Contents

Acknowledgements il

1. Introduction

A. The Vacuum of Quantum Chromodynamlcs 1

B. Vortex Operators 6

2. The Abelian Vortex Operator

A. The Vortex Operator In Terms of Elementary Fields 12

B. Green's Functions of Vortex Operators 18

C. Renormalization of Looplike Operators 31

3. Cluster Properties of Looplike Operators

A- Cluster Properties in a General Gauge Theory 44

B. Vortex Operators in an Abelian Higgs Phase 54

4. Dnquantized Vortex Operators

A- Continuum Theories with Fields

in the Fundamental Representation 65

B. A Lattice Model 87

5. Conclusions 100

References 102

Figure Captions 105

Figures 108

iv
 CHAPTER 1

INTRODUCTION

A. The Vacuum of Quantum Chromodvnamies

In the years aince the discovery of asymptotic freedom, confi­

dence that Quantum Chromodynamics might be the complete theory of the

strong interactions has grown steadily. At high energy, although

quantitative results have been limited by the size of the effective

coupling constant and the need to factor out incalculable long­

distance effects, the qualitative features of QCD appear to be

correct • A recent important success was the observation of three

jet events, attributed to gluon bremstrahlung, at PETRA. At low

energy, various theoretical approaches have shed light on the nature

2
of the QCD vacuum , and there is evidence to support Che long­
standing hope that quarks are confined in QCD.

There is now a standard picture of the QCD vacuum. It appears

that fluctuations involving the creation of quark-antiquark pairs can

be neglected as a first approximation, so that hadrons can be treated

as first-quantized quarks in interaction with a second quantized

quarkless SU(3) gauge theory. The justification for this is empiri-

3
cal : the success of the OZI rule and the fact, supported by hadron

spectroscopy and by measurement of parton distributions, that baryons

are predominantly qqq (not qqqqq or qqqqqqq) and that pions are

predominantly qq (not qqqq, etc.). The theoretical basis for this is

uncertain. It is a feature of SU(N) QCD in the large N limit, but

the belief that SI)(3) QCD resembles Su(N) QCD is in fact based on

comparison with phenomenology, not on a calculation of the 1/N

corrections.
1
 The appearance of the pure SU(3) vacuum depends on the inverse
6 + -
distance scale k at which it is probed • A process such as e e

annihilation to hadrons at c m . energy E probes the vacuum structure

at scale k-E, while measurement of thermodynamic quantities in a

gauge theory at temperature T probes the vacuum at scale k-T. At

short distance, k > s A » (AQPJ, is the srile parameter of QCD, approx­

0CD

imately 400 Mev), perturbation theory is accurate and the vacuum is

dominated by configurations near the configuration of least classical

action (A -0 in most gauges). On a somewhat longer scale, other

classical configurations (instantocs) give non-negligible effects ' ,

but the functional integral is still dominated by a small class of

configurations which are near classical solutions. On a still longer

scale, k<V\ , both the perturbative and lnstanton calculations have

0CD

uncontrolled divergences.

A different approach, the lattice approximation, gives a picture

of the gauge theory vacuum at extremely long distance. At this

scale, the vacuum looks like a coherent superposition of every possi-

7
ble configuration. In the Hamiltonian lattice formulation this is
signalled by the fact that the vacuum is an eigenstate of E , the

operator conjugate to the vector potential. In the Euclidean formu-

8 1
lation it is signalled by the fact that — , the coefficient of the
8
action in the functional integral, goes to zero, so that configura­
tions contribute purely according to entropy (volume in function
space), without regard to action. In both formulations, the Wilson
loop satisies the area 3 aw, implying a linear potential between
quarks. The important recent advance in this picture has been the
 3
g
Improved calculation of the lattice beta function, by both analytic

and Monte Carlo methods, indicating that Che only UV fixed point is

at zero coupling. This Implies that aa the lattice spacing is taken

to zero, the lattice approximation turns smoothly into the continuum

theory studied in perturbation theory. Qualitative features such as

quark confinement uould then persist in the continuum theory, and the

ratio of the scring tension to / W Q has also been calculated and

found to be of the correct order of magnitude.

The infrared divergences of the perturbative and instanton cal­

culations can then be understood as a consequence of expanding around

a qualitatively wrong vacuum, analogous to the infrared divergences

which appear when one expands a two dimensional slgma model around
11 12

the asymmetric vacuum , which is known by Coleman's theorem to be

incorrect. It would be desireable to have an approximation which

starts with the correct sort of vacuum, that is, one dominated by

entropy rather than by a few classical configurations in the

infrared. The lattice approximation has this property, and is very

useful, but the information it can give is limited by its crrdenass.

A different approach is suggested by the idea of duality. A wide

cla.>s of lattice theories, naaely Abelian (0(1) or 2 ) gauge and gen­

eralized gauge theories, in any number of dimensions, can be rewrit-

13
ten in terms of a dual gauge theory . The coupling constant of the
dual theory is roughly the Inverse of the original coupling constant.

Thus, if one theory is entropy dominated (large coupling) and there­

fore difficult to treat, the othar will have a small coupling and can

be expanded around the configurations of least action. This suggests

a search for a dual description of continuum non-Abelian gauge

 4

theories.

The idea of duality can also be reached from a different point

14
of view . Magnetic flux cannot penetrate a 11(1) Higgs vacuum (such

as a superconductor), but is squeezed into tubes, within which the

vacuum is normal (unbroken) and flux can exist. If a monopole and

an^loonopole were placed in a superconductor, the flux between them

would form a tube, with an energy proportional to its length, so that

a linear potential would bind the monopole and antimonopole. This is

exactly the behavior expected for the QCD vacuum, except that (color)

electric rather than magnetic sources are to be confined. The QCD

vacuum might then be expected to look like a colored superconductor,

but with electric and magnetic quantities interchanged, so that it

would be a coherent state of color-magnetic monopoles where the

superconductor is a coherent state of electric charges (Cooper

pairs). Duality in a gauge theory involves just this interchange of

electric and magnetic quantities. Further, in non-Abelian gauge

theories, monopoles or mocopole-like configurations occur

automatically; they need not be put in by hand as in the Abellan

Dirac theory .

't Hoofu and Mandelatam have suggested a kind of operator,

the vortex operator, in terms of which this duality might be

described- This thesis is concerned with various properties of these

operators in continuum field theories. In the remainder of this

chapter, we introduce the vortex operator and discuss some of its

general properties. In Chapter 2 the Abelian vortex curator is

written in terms of elementary fields. We emphasize a simple form in

which it is the exponential of a surface integral, and we explore in

detail the requirement that the operator be independent of the par­

ticular surface chosen* The Dirac veto problem, known from magnetic

monopole theory, is encountered, and we suggest a new solution to

this problem. Ke then discuss the renormallzation of Green's func­

tions of vortex operators, and a nice cancellation of divergences is

found- For comparison, the renormalization of another looplike

operator, the Abelian Wilson loop, is also studied* Chapter 3 deals

with the Green's functions of vortex operators at long distance,

their cluster properties* We describe the various possible phases of

a gauge theory in terms of the cluster properties of vortex and Wil-

son operators, and explain a result of 't Hooft* which restricts the

possible phases. We then show that in any Abelian theory without

magnetic aonopoles and with no physical massless particle, the vortex

operator must have a surface-like cluster property and so satisfy an

area law* Finally, we show how this cluster property emerges from

the graphical expansion in an Abelian Hlggs theory.

For one to be able to define a vortes operator, the fields in

the theory must satisfy a quantization condition GT, for the gauge

group must be non-trivial),' QCD without quarks satisfies this condi­

tion but the introduction of quarks violates it. In Chapter 4 we

examine the case where the quantization condition is not satisfied*

We show that any vortex-like operator will so longer have the impor­

tant property of being "local looplike"* Nevertheless, in an Abelian

theory we are able t;> find an operator with some reasonable proper­

ties: the energy that it creates is infrared finite, and its cluster

property distinguishes a perturbative Hlggs phase from a QED-like

phase* We speculate on the properties of this operator in non-

 6

Abelian and confining theories, and on the resolution of an apparent

paradox involving a result of 't Hooft. We then examine a lattice

model where the quantization condition is not satisfied. We find

that our extension of the vortex operator is a good order parameter

and correctly indicates that a Higgs/confining phase transition is

taking place, when the field which violates the quantization condi­

tion is weakly coupled. Some preliminary results indicate that our

operator is also a good order parameter when this field is strongly

coupled, in the sense that discontinuous behavior of the operator

implies a discontinuity in the phase diagram; it leads us to find a

phase transition that might not have been expected. Chapter 5 sum­

marizes our findings.

In summary, Chapter 2 deals with technical details related to

the Green's functions of vortex operators. Chapter 3 discusses known

relations between the behavior of the vertex operator and the nature

of the vacuum, but with the emphasis on cluster property rather than

vacuum expectation value. Chapter 4 then uses some of these elements

to investigate systems where vortex operators were not known to ba

useful, and some light is shed on the phase structure of these sys­

tems.

v
£• ortex Operators

The vortex operator, like the Wilson loop, is associated with a

closed curve in spacetime. Let us consider a general "local loop-

3
like" operator L associated with a closed curve C lying in the R

plane t-t , and let us consider its commutation relations with other

operators at time t . By definition, L commute.} with observable

 (that is, gauge invariant) local operators associated with a point x

not on C. This Implies that for a gauge dependent field J(x )

8
LJCx*) - tCx> L (1.1)

where }(x)° Is f after a gauge transformation g(x) and g(x) is a

gauge transformation associated with the operator L. For instance,

in the usual gauges (axial, covariant. Coulomb), the gauge transfor­

mation associated with the Wilson loop is simply 1 everywhere.

-* 3
It is important tnat g(x) is not defined in all of spa.-e, R ,
3 3
but only in R - C. R - C is multiply connected: curves are dis­
tinguished by their winding number around C. As a result there can
be a non trivial effect, a topology, associated with g ' x ) . As x
describes a path around C, g(x) describes a path through the gauge
group. If the gauge group is simply connected, there is nothing
interesting about this, but if it is multiply connected the path may
20
lie in a non-nrivlal element of 11, (G) • For different paths of the

same winding number, continuity requires that this element be the

same. Since paths of winding number other than one can be generated

as a product of paths of winding number one (traversed backwards for

negative winding number) the homotopy class associated with winding

number one determines that for any other path. The class associated

with L is gauge invariant: since any gauge transformation can be

continuously deformed into one which Is unity in an arbitrarily large

volume (Including the whole of L) so Che gauge transformed g(x) can

be continuously deformed into its original value; homotopic invari­

ants are therefore unchanged. Thus, there is a gauge invariant quan­

tum number, a homotopy class, associated with any looplike operator.

 Operators for which this class is not the trivial ope are called
21
vortex operators. The name comes frcu the Nielsen-Olesen vortex ,a

classical solution in a Higgs theory associated with twisted boundary

conditions for the Higgs field• In a completely broken Higgs theory

the set of classical vacua is identical to 6, so that there will be

20
solitons with quantum number in II, (G) . These are the Nielsen-
Olesen vortices, and when a vortex operator acts on the vacuum of the
Higgs theory it produces a state containing such a soIIton.

In an Abelian gauge theory, a vortex operator creates a loor of

magnetic flux just as the Wilson-operator creates a loop of electric

flux. In the dual (Abelian) gauge theories mentioned earlier, the

Wilson loops of one theory are mapped into vortex operators of the

other. In the long distance, large coupling, limit of non-Abelian

(as well as Abelian) gauge theories, the vacuum approaches an eigen-

state of a simple vortex operator. Thus, they are attractive opera­

tors to considei If one tries instead to consider duality in terms

of pointlike, monopole creation, operators, one finds that there is

no associated topological quantum number: n_(G) is trivial for any

Lie gauge group G. In this, there is an interesting analogy between

electric and magnetic quantities: the pointlike operators In non-

Abelian theories (gauge fields or monopoles) have no gauge or topo­

logical invariance, while the looplike operators (Wilson loops or

vortices) do.

The quantum number associated with a vortex can be characterized

in a different way. Consider the path described by g(x ) as x winds

once around C. This path in G defines in a natural way

(g og"g~ 5|) a path in the simply connected covering group G. By a

 ••20 "^
well-known connection between 17, (G) and the center of G , when x
returns to <ts starting position g(x) need not return to its origi­

nal value but is multiplied by an element Z of the center of G.

Because g(x ) is single valued, Z. in G is mapped into 1 in G by the

ususl homomorphism. Vortex operators can thus be considered to have

quantum numbers in Z(2)/Z(G), the quotient of the centers of the two

groups, which is isomorphic to 11, (G).

One can show from equation 1.1 that

V(C C }
L(C)W (C) - V (C')UC)(z ) ' '
s s sl (1.2)

where Z in representation s is z , times the identity, w(C,C') is

the winding number of C through C, and W ( C ) is the Wilson loop in

representation s associated with curve C* (assumed here to lie in the

t«t hyperplane):

W ( C ) - Tr P e p{ig^ ,dx A^( )^>

6 X c i x (1.3)

with y the generator a in representation s, Tr representing the

trace, and F representing the usual path ordering. Thus, although

L(C) commutes with every local gauge invariant not on C, if it is a

vortex operator it will not commute with certain Wilson loops linking

C. Equation 1.2 characterizes completely the topological character

of the vortex operator; it is only Z, (or « , for all s ) , not g(x)

that can be defined in a gauge invariant way. Note that equation 1.2

is entirely dual between L(C) and W(C'); one cannot say that one is a

topological operator and the other is not* The asymmetry arises only

when one introduces gauge dependent quantities. The one genuine

asymmetry is that the Hamiltonian is relatively simple in terms of

 10

the Wilson loops (or the related vector potentials), but does not

appear to have a simple form in terms of vortex operators; this asym­

metry may or may not be permanent•

For a field ^ In representation s., equation 1.1 becomes

L(C)*6f) - g ^ x V o O U C ) (1.4)

where g,(x) is g(x) in representation s.. In order that equation

1.4 be well-defined, it is necessary that z , be 1. This Is Indeed

V-
t'.ie case, as Z in G maps into 1 in G and s, is single valued In G

(by definition, the gauge group G is the smallest group with the Lie

algebra of G such that every field in the theory lies in a single

valued representation). If we add new fields to the theory which are

In representations s' not single valued in G (and thus, by defini­

tion, enlarge the gauge group to G') those vortices for which z ,. is

not 1 no longer make sense. In other words, T1AG') is smaller than

II, (G)« z . -1 for all fields •+ is the quantization condition

* S lid

referred to above. Note that It is equivalent to saying that vortex

operators oust commute with W for all fields •+ in the theory.

V
We can illustrate these ideas for the group SU(n)/Z , an SU(n)

gauge theory with all fields in the adjoint representation. IL f°r

this group is Z , so that the vortices carry an n-fold quantum

number, p(L). The representions alco carry an n-fold quantum number

q(s), the "quark number" of the representation: if s lies in the pro­

duct of m fundamental representations, m—qfs) .^ . Equation 1.2

becomes
 11

L(C)W (C) - W (C')L(C)exp<2irip(I.)q(«)w(C,C*)/n}

g s (1.5)

We see that the SU(n) gauge theory without quarks (whose gauge group

is really SU(n)/Z ) has vortex operators, while if quarks are added

the gauge group becomes SU(n), which is simply connected, and vorttx

operators can no longer be defined. More will be said about this in

Chapters 3 and 4.

The short distance properties of the vortex operator, near C,

are still unspecified; the vortex operators are a class, not a single

operator. Certain results, such as 't Hooft's concerning possible

phases, are independent of the short distance properties. Others, in

particular those related to dynamics and renormalizstion, depend on

the details near C. In this thesis, Chapters 3 and 4 are primarily

concerned with long distance properties, independent of the detailed

form of the vortex operator. Chapter 2 deals with some of each.

 SHATTER 2

THE ABELIAN VORTEX OPERATOR

A. The Vortex Operator in Terms of Elementary Fields

We first review the ideas from Chapter 1 for a 0(1) theory. The

covering group of U(l) is R ; a general element of R is a real

number y. In the representation of charge e, y becomes exp(iye). A

general looplike operator L(C) is then associated, through equation

-> 1 -> -> ->

1.1, with a function y ( x ) , the R version of g ( x ) . Like g"(x), y(x )

need not be single-valued, but exp(iy(x )e} must be if fields of

charge e are present. It follows that when x winds once around C,

y(x ) must change by 2irp/e . , where e . is the unit of charge and p

is any integer. Vortex operators are thus characterized by an

integer p. Equation 1.2 is now

L(C)W ( C ) - S (C')L(C)exp{ipqw(C,C')/e } Bin (2.1)

where U is the Wilson loop

q

W ( C ) - exp{iqf ..dx A (x)}

( i i (2.2)

For those fields actually present, q is a multiple of e and the

phase factor in equation 2.1 is 1.

Differentiating equation 2.1 with respect to q and setting q-0,

and then using Stokes's theorem to relate the line integral of A to

->
the surface integral of the magnetic field B ,

> > EiE 3 > >

L(C)B (x ) - <B (x ) +| f dx '6' (x -x ')>L(C>
± i c i ( 2 > 3 )

min

Although equation 2.3 refers to a commutator directly on C, it fol-

12
 13

lows from equation 2.1 and is therefore true for any vortex operator

independent of its short distance details. This is the source of the

statement that vortex operators create a loop of magnetic flux; the

-» -»
dual equation, replacing L by U , B by E and 2np/e . by e is also
e min
true. There i s no corresponding local version of equation 1.2 for a
non-Abelian theory. Equation 1.2 might then be taken as a definition
16
of non-Abelian magnetic flux .

An operator satisfying equation 2.1 i s

3
V (C) - e x p { ^ - j y x £•£OO-KJj 0 0 ) ( 2 . 4 )

* min r

where 6,p, and z are cylindrical coordinates, 8 i s a unit vector in

the 6 direction, and j n i s the charge density. From the canonical

commutators (we take the A -0 gauge for convenience, but V(C) i s

gauge invariant and so * i l l be i t s commutators with gauge invariants

such as equation 2.1)

> 3
[ E ( x ) . A ( ^ ) ] - 1J
i J i (i^) (2.5a)

[Link])} - -e^it-fwCy) (2.5b)

one finds

_> -ipSe. _>

r
V ( C W x ) - exp[— ]*(x)V(C) (2.6a)
p p
min

V (C)A (x) - ( A ( x ) + i e x p { - i p 8 / e > » e x p { i p e / e

p i i Bln 1 i]iln »V (C)
p (2.6b)

> >
V (C)E (x ) - E (x )V (C)
j) 1 ± p (2.6c)

This i s the Abelian form of equation 1.1. The curve C i s here the
 14

z-axis; g(x) Is exp(-ipej. A vortex operator for any C and any g(x)

can be constructed in the same way*

The coordinate 6 must have a discontinuity of 2ir on a semi-

infinite surface S bounded by the z-axis (such as the half-plane

8-0«2n). The exponent in the definition of V (C) is therefore

P

discontinuous but the operator itself has no discontinuity on the

surface, as can be seen from Its commutators. These are completely

independent of where we choose to define the discontinuity of ©.

Using oS-^ + disc(8), V (C) can be rewritten
P »
3
V (C) - exp{^_[JVi x (j (x*)-«» E (x*))« + 2nX dn E s ]} ( 2 . )
7
p
min

-* v -* o

j„(x )-o.E (x ) does not vanish as an operator in the A «0 gauge, but

by Gauss's lac it vanishes in gauge invariant Green's functions* The

gauge invariant Green's functions are therefore the same for V (C) as

for
V '(C) - KpfSfiEj. d n E > { 2 > 8 )
r
min

If one evaluates the commutator of V '(C) with «(x )"^*(x )D.^(x) (D.

is the covariant derivative o -ie/A (x)), it does not appear to van­

ish on S; it must, however, because [V (C),e(x)]«0 on S and % is

P

gauge invariant* The problem is that V is too singular for canoni­

cal commutators to be correct; if one evaluates the Green's functions

of V '(C)0(x) using the methods of the next section, one finds tbat
P

they have no equal time discontinuity on S. V ' is a more convenient

form of the operator when one is discussing Green's functions*

One might wonder whether the commutators 2.6 are really correct

in field theory, even for V, or whether some anomaly will develop due
 15

to the discontinuity of ©• V and V' are poorly defined because the

discontinuity of 6 is sharp; if we define them by amearlng Che

discontinuity and taking the smearing to zero, does the limiting

oerator satisfy 2.6? A simple example shows this to be a valid

worry. Consider a theory with fermions, +, in two spacetime dimen­

sions (they may even be free fernions), and consider the operator

a(x) - e x p ^ d x ' j 2rri?(x')y *Cx')>

0 ( 2 # 9 )

where x is the spacetime point (X..X,) and the Integral runs along

the equal time path from (x_,x.) to ( X , C D ) .

Q From the commutator

2
r?(y)y f(y).*(z)jS(x -y ) - -S (y-z),*(y)
o 0 0 (2.10)

we find, by the same canonical manipulations that lead from 2.5b to

2.6a, that LQ(x).i'(z)]-0 at equal times, so that fl(x) would be a c-

22
number. On the other hand, from boson equivalence ,

aW - exp{j"dx' -2\jSic> ?(x'))

Xj

- exp<2\S?l0<x)>

- consfftxXl+yjWx) (2.11)

If the fermions are massive, the leading piece of ^(l+y,)^ is in fact

a c-number, but if they are massless, iKl+y,)^ has no c-number piece

and [a.'fl is not identically zero.

One may also see this in a different way by evaluating

G(x,y,z) - < £(x) 7(y) *(z) > (2.12)

for free fermions. G(x,y,z) is poorly defined because the support in

 16
T
the exponent in equation ?..9 is too singular. .f one smears the sup­

port (necessarily Into the time direction) over a small distance A

(call the resulting operator fl, and the Green's function G. ) , we can

calculate G. (x,y,z) directly. As A goes to zero, for free messive

fermions G. (x,y,z) approaches < V-(y) V'(z) > times a constant, while

for massless fermions it approaches

<?(x)(l+y )V-(x) *(y) •f(z) >

5

times a constant.

This is not a serious problem. By regulating £ In i slightly

more complicated way, one may obtain the desired limit. Consider the

operator

where

2
H (x) - exp{ifd z f (z)?(x+z),Kx+z)>
A A (2.13)

and f. (z.,z ) is a family of functions with support in |zj < A» If

f. is defined so that as A-»0,

2 2
J*d z f.(z)/|z| -» co (2.14a)

2
J"d z f (z)/|z| -» 0
A (2.14b)

t
then as A-»0, i (l+)' )^H approaches a c-number even for massless
5

fermions. If one then defines a regulated fl(x) as JCL(x)H. (x), its

limit will be a c-number. Equations 2.14 are correct for free fer­

mions or with a super-renormalizeable interaction; as one might

expect, with a Thirring interaction there would have to be different

 17

powers of r in the integrands-

It is Interesting to repeat some of the above analysis for the

Schuing&r model, without using bosonisation. Using Gauss's law, we

get

fl^(x) - e x p { - - ^ J ^ z S*(z)E(x+z)) (2.15)

where 5\ (z) is the delta function, smeared over a distance A in

spacetime, and E, the electric field, is a (pseudo-)scalar in two

dimensions* The analogy between this and equation 2*6 is clear*

We would like to compare

t < ) >
G(c.y) •T *f
c ff J ,p 5
(x)
V y (2-16)

with the same quantity with 1 in place of tL • H e r e

J 5 is the axial

current -/-Y^y rA and c is a spacetime curve circling the spacetime

point y at some distance large compared to A but small compared to

2
1/e (which has dimensions of length). Because of the latter stipu­
lation, we can neglect all but the leading term from perturbation
theory.

Using the anamolous divergence equation,

we can rewrite equation 2*16 as

„2
G(c,y) - ^ X r
d E
* << > « V x y >
* ( 2 , 1 8
>
From <E(x)E(y)>-6 (jc-y)-K)(e ) , it follows that

< *(x) j^(y> > - - iSt fi^( _y) + o(e) x (2.19)

G(c,y) is then simply -2, independent of A, whereas when & is

 is

replaced by 1 in the matr.U element, 2.18 is zero: the limit cf

41 as A goes to sero is not a c-number. 0. has much in common with a

Olrac string; equation 2*II shows that there is a lump of fliu: vsiiich

does not gi t'.way when A gees to zero. Our result here i? chat a mas-

nive fermlon does not (eel this flux in the limit, whereas a massless

fermiou does, but that a massiess fermion can be "shielded" from the

flux by the additional regulator 2.9. Massive and massless bosons

appear to behave like massive fermions: The Green's function analo­

gous to 2.12 goes to the free propagator in the limit.

Our results for the four dimensional case are not so complete.

We discuss them, and more of the analogy with the Dirac string, in

the next section.

JB. Green's Functions of Vortex Operators

In this section we will discuss the calculation of Green's func­

tions of the V'(C) form of the vortex operator in Euclidean space-

time. For simplicity we shall assume only one charged field, scalars

j»(x) with charge e, mass m, and four point interaction A; the gen­

eralization to more fields and to fermions is straightforward. We

can now consider a vortex operator associated with a general closed

curve in spacetime. By analogy to equation 2.8 we define

-exp^de^pCy)} (2.20)

where S is any 2-surface whose boundary is C and *...,«1. There is

no factor of 1 in the exponent because F »iE in Euclidean space-

time. To show that 2.20 is correct, we would like to check that it

 19

satisfies equation 2.1, and that its gauge invariant Green's func­

tions are Independent of the mirfjop S.

Consider now a general Green's function of V :

4
< V '(C.S) exp<rd xJ (x)A (x)+K(x)*(x)+L(x)tf*(x)> >
p p p
where J , K, and L are general external sources. The expansion of V'

involves terns with arbitrarily large powers of 1/e, but owing to the

exponential form of V , these sun up in a convenient way:

< >
V
p' ' ( C S )
3SL ' e X p
t l f kih'TrnhH- (2-22)
r
khran

when* < > indicates the connected Green's function. The connected
c
k+h
Green's function is of order at least e (e being the charge),

except for k+h-2, m-n-0, for which it starts at order e . The sun in
_2
2.22 therefore starts at order e • Further, the graphs contributing

at each order in e are readily classified: a graph with n internal

photon lines (internal means both ends connected to charged lines)

and n external lines (one end attached to a charged line and one
2n-tn e

attached to J ) is of order e , independent of the number of

vortex photon lines (defined as lines which run from S, the vortex

operator, to a charged line). The connected graph with no charged

—2
lines (one photon line with both ends attached to S) is of order e .

For example, the h-m-n-0 tern, which is independent of J , K,

and L, and is the only term which contributes to < V '(C,S) >, is
P
 20

2 2
2 ( 2 - 2 5 )
< V '(C,S) > - exp{ ^ H ( 0 ) A ( S )
2
P e-'

2
E d X d y ( x y ) 2 l n d VS ; 2
" " l^C |/c u " ~ " e t { A / > ) + 0(e )>

v S v S v S - 1 v S
vhere A(S) is the area of the surface S, & is < i ) D } , D is
H H H
S -ieA (x) and A ' (X) is given by the graph of figure la. The first

two terms are from the one photon graph. The third is the sum of all

graphs with one charged line connected to S by any number of photon

lines. The first term is S dependent and quadratically divergent.

We shall postpone further discussion of 2.Z3 until the end of this

section. We will consider until then "reduced" Green's functions,

with the h-m-n-0 term divided out.

To investigate the dependence on S of gauge invariant reduced

Green's functions, consider first the order e term of

< F (x) V '<C,S) > / < V '(C,S) >, which is equal to

fp P P
2
Vp><*> - $<*> < -">
vS
(Parentheses on subscripts indicate antisymmetrization). A was

defined above; 2.24 defines F as its curl.

HP
i (X) F S (X) x )
? -P ' P " fJW<p*«py < W S V >o
d o + IS <x y) ( 2 - 2 5 )
- - ^ J W V *«PVJ V F ^ P *^ "
i s
where < > indicates the free propagator and A_ the Euclidean

scalar propagator:

2
& A (x-y)--6*(x-y)
E (2.26)

Using the fact that a completely antisymmetric 5-r. ens or vanishes in

 21

four dimensions, Che quantity in square brackets can be rewritten as

After surface integration the last two terns vanifh, because S-S' has

no boundary, while the remaining term gives, using 2.26,

V y )
^>i^>--fJW< ^ P

We see that the Green's function is S-dependent, but only when x lies

directly on S or S'. However, when x lies on S, < F fx) V '(C,S) >

is not well defined due to problems of operator ordering- If we

resolve this ordering problem by taking a limit as x approaches S,

v
F is S independent everywhere. In effect this is taking the T-

product, whereas 2.24 defines the T*-product :

2 28
Tlfy,)] - TNFJJW] ^fjV^SV-y)
+ <' >
is independent of S.

Now let us go one order further, and consider

G(P.C.S) - (2.29)

< d*(x)exp{iej dx' A. (x')}jl(y) V '(C.S) > / < V '(C.S) >

p

r P " P
Q

where P is some path from y to x. To lowest (e > order, this gauge

invariant Green's function is given by all graphs of the form shown

in figure lb. These graphs sum up to give

vS
G(P,C,S) - -A <y,x)expUeXpdx' Ajf(x')> (2-30)
 22

v
Using the path integral rep* Mentation for the propagator ^ (y»x),
ve can write

S
G(P,C,S) - -exp<iej' dx* A^ {x')> (2.31)
r r
2
D S
T * *- « P { # t - \ -| y p + ioi Aj; (z)}
p

with Dz indicating the integral over all paths z(t) such that z(0)«x

and z(s)-y, and z is dz /dt. As long as A is smooth (we must tem­

porarily smear out the surface S) the path integral can be made
25
mathematically precise ; at this level we simply have an external

field problem and there are no short distance difficulties. The

right hand side of 2.31 depends on A v

only through the phase it ".tors

8 d l n ? > (2 32)
«p{lef ,dx* A]; (,-)> - « P < T W ) » p ' « y
c p a *
where C" is the closed curve formed by joining P with z(c) and R(C')

id any 2-surface whose boundary is C . To obtain equation 2.32 we

had to use equation 2.20, which defines the S-dependent T * [ F 7 ] . We

cannot try to simply define this S-dependence away by replacing

T F a s t h e
T*[F^g] with l ~ l » 8 result would in general depend on the

choice of the arbitrary surface R.

From 2.27 and 2.32, it follows that changing the surface on

which V ' is defined changes the phase associated with the path C by

a factor

ex [Link]* _ .do-^ dfl£ S (x-y) > 4 (2.33)

P s s R p

In four dimensions, a closed curve, C , links a closed 2-surface, S-

S', a definite, integral, number of times, H. (leaving out for now

 23

those particle paths which actually intersect S or S'). Any surface

R bounded by C will then intersect S-S' at N points (intersections

are defined In an oriented way, so that it is the net number of

intersections that is counted)- Setting up local coordinate systems

on S-S' and on E near such an intersection, one finds that the double

integral in 2.33 is exactly 2N-* The phase then changes by

exp{2llipN } under the change of surface, which is 1 if p is an

Integer as required by the earlier quantization condition*

The above argument is extremely familiar: it is just the argu­

ment that one can have magnetic monopoles in an Abellan gauge theory,

If the monopole charge is quantized so that the charged particles

26

don't "see" the monopole's Dirac string . In fact, just as the Wil­

son loop can be regarded as the world-line of a classical charged

particle, the vortex operator can be regarded as the world-line of a

classical monopole, with the surface S as the world-sheet of the

Dirac string* There is one problem with the Dirac string, and that

problem is also present here: what of paths which actually pass

through the surface S? If we smear the integral defining V ' (or

that for V ) and let the smearing vanish as a limit, can we neglect

paths passing through S because they are of "measure zero", or do

they contribute in a sufficiently singular way that their effect does

not go away?

This is exactly the question that was raised from a different

point of view at the end of the last section* In our two dimensional

model it was found, in effect, that when the support of our Dirac

string (there it was a "Dirac lump") goes to zero with the total flux

staying constant, under some conditions (massless fermions) the

 24

effect of paths through the lump survives in the limit, while under

others (massive fermions and bosons) the effect vanishes. Even in

the massless fermion case It was possible to obtain the desired limit

by adding additional regulators to the operator, which can be thought

of as correcting the action for paths that pass through the lump.

The four dimensional case is not so easily analyzed* For S an

infinite 2-plane (so that C is infinitely far off In some direction)

the charged field propagator is easily calculated with the flux

smeared, and one finds th.-.t it approaches the free propagator for

massive or massless fermions, or for bosons- In the presence of this

2-plane plus an additional smooth A -field, however, an argument

similar to that for the Schwinger model (based on the anomalous

divergence of the axial current) shows that the effect of the Dirac

string survives in the liait if the fermions are massless. An extra

regulator of the same form as H in 2.15 (essentially giving the fer­

mions a mass very near the string) corrects this particular problem.

We have not been able to show, however, that this is the only addi­

tional operator needed in general, or that no correction is ever

needed for massive fermions or bosons. It seems quite likely, how­

ever, that if we do need to correct the action for paths passing

through the string, this correction will always take the form of

local operators near the string. The whole point Is, perhaps, moot,

as we shall see that the Green's functions are uniquely determined

without knowing the detailed form of the extra regulator (just as in

the two dimensional case, we do not really need to know the form of

Hi once we know that Q. is a c-number, all of its Green's functions

are fixed up to an overall constant!)-

 25

Many other solutions for the Dirac veto problem have been given-
26
Brandt, Nerl, and Zvanzlger have rewritten a field theory of

charges and nonopoles as an integral over all numbers and configura­

tions of particle paths* They then define the action associated with

configurations having charge paths Intersecting monopole strings as

the limit of that for non-intersecting configurations- In a monopole

field theory, our solution would have the form of a smeared Dirac

string and an additional non-local charge oonopole interaction along

the string.

In the remainder of this thesis we assume that V "(C,S) can be

defined in such a way as to have the desired limit, so that G(P,C,S)

is independent of S- It then follows that there is a gauge transfor­

mation g(S,S';x), defined except when x lies on S or S', such that

VS VS
A '(y.x) " g(S,S';y)A g*<S,S';x) (2.34a)

S S
A* '(x) - A^ (x) + |g(S,S';x>& g*(S,S*;x) (2.34b)

Any gauge independent quantity constructed from A and A Is there­

fore S-independent as veil*

At least In the present case, where there is only a classical

monopole, we can evaluate quantities such as G(P,C,S) or A (x,y)

without using either an explicit form for V '(C,S) or the path

integral prescription. Equations 2.34 by themselves give sufficient

information to determine the Green's functions we want, using a

geometrical approach (in the sense of differential geometry) due to

27
Wu and Vang . Consider two non-intersecting surfaces, S and S ,

each having C as its boundary. Take spacetime with 2, a thin tube

 26

containing C, removed• This space can be covered by two overlapping

open regions, Qj and Q,, such that S, lies entirely in Q,, and S.
vS VS 2

lies entirely in Qj. A (x) is defined for x in Qj, and A (x) is

defined for x in Q„; in the overlap region these functions are

related by 2.34b. They are given, as before, by the simple one pho­

ton graph, and g(S,,S.;x) can be determined in terms of this graph.

Any other gauge and S-dependent quantity can similarly be represented

in as a pair of functions (or, for A, as four functions, since it has

two arguments), each defined only in a certain region but related in

the overlap region by the gauge transformation g(S,,S ;x). Gauge

dependent quantities are thus "sections": sets of functions each

defined only in an open region, but such that the regions cover all

of spacetime (minus C in this case) and such that the functions are

related in the overlap of two or more regions.

Integrals of gauge (and therefore S-) independent products of
4
sections over R - C can then be defined: in each region the

integrand is defined in terms of the function which exists in that

region; in overlap regions this is unambiguous because the integrand

V V
is S-independent. Wit.l an inner product based on this integral, D D

acting on sections can be made self-adjoint* It follows that A. as a

section exists and is unique. A. can be obtained from it as a func­

tion defined everywhere except on x. Power counting arguments, as

developed in Che next section, indicate that the limit as the cutoff

C is removed exists, at least order by order in p.

We may now consider higher order corrections to Green's func­

tions. All graphs with a given configuration of charged lines and

 27

internal and external photon lines are of the same order in e; sum­

ming over all numbers of vortex photon lines gives one graph of the

same configuration with no vortex photon lines but with "effective"

A
propagators and vertices. The pboton propagator and j*V vertex are

unchanged, and the charged propagator becomes - ^ (x,y). The f*<fk

* 2
vertex picks up an extra term from graphs where a j* j»A vertex is

connected to one vortex photon line and one other photon line; it

becomes -i (x). Equation 2.22 then becomes

>
<V ' ( C S ) V
"JKL ' < p ' ' (C S)
• " > ' hSLr
2
(2-35)
r r
hmn

< [jA J A ^ t j ^ x K*]»IjA I**)* > cvS

where the prime on the sum excludes the term h-k-1-0 and the sub­
script cvS indicates all connected graphs constructed out of the
effective vertices and propagators. Examples of higher order correc­
tions to < A (x) V '(C,S) > / < V '(C,S> > are shown in figure 2.
f P P

It is then evident that higher order corrections to gauge

invariant reduced Green's functions are S-independent: Under a change

of S, the phase factors from two propagators meeting at a ^*j*A ver­

tex or an external source cancel; the total change in the propagators

and vertex at a f*#A vertex vanishes. Actually, this is not

strictly true if we have defined the vortex operator by smearing S

and adding additional operators. The Feynman integrals in coordinate

space include points lying in S, for which 2.34 do not hold. As in

the case of the propagator, we would expect to be able to "repair"

the Feynman integrals with additional corrections to the definition

of V_'(C,S). Again, we need never know the form of these

 28

corrections: regarding the propagators and vertex functions as sec­

tions, the Feynman integral for each graph can be written as the

invariant integral discussed above, and the result is unique.

We can now demonstrate equation 2.3:

i T< F (X) V '(C,S) > - d

p (T-T*)< f (x) V '(C,S) >
p

- & (T-T*)< f (x) V '(C,S) >

U Up p O

-2&f dx'
c
4
S (x-x')< V '(C,S) > p (2.36)

The first equation follows because T*(F ) is defined as

fipotp„b orT*(A,)
<•
p and so its divergence vanishes identically. The

second equality (the subscript indicates the lowest order graph, fig­

ure la) follows because any higher order graphs for the T* Green's

function, such those of figure 2, are S-independent and therefore

continuous when x is at S: they do not contribute to (T-T*). The

final equality follows from equation 2,28, Since & F (x)-0 is true
P PP
as an operator equation, 2,36 represents a commutator and is in fact

the covariant version of 2.3. Equation 2.3 is an operator equation

and we have only considered one matrix element of it. The same argu­

ment can be readily applied to any gauge invariant matrix element.

Equation 2.1 is also true. If we consider < W ( C ) V '(C,S) >,

only the graph of figure 3a has a discontinuity whenever C crosses

P < 1
S: the Green's function jumps by a factor exp[ 1 . This is the

covariant form of 2.1. Figures 3b, 3c, etc. are S-independent and

therefore continuous. Equation 2.3 thus exponentiates to give 2.1.

Ordinarily this would not be true, because in general the commutators

 29

of B do not determine those of the Wilson loop (although they do in

naive canonical manipulations) because graphs such as 3c which depend

on the composite nature of W ( C ) have discontinuities; this 1 B not a

problem here. The location of the discontinuity of

< U ( C ) V '(C,S) > does depend on S, unlike the Green's functions of

1 P

local gauge invariants! It seems, however, that the phase of this

Green's function is not observable, but only the net change when C'

is moved and then returned to its original position- The surface

swept out by C is then closed, so that it links C and intersects S a

definite number of times. Tnis fact will be used in the next

chapter.

For later use ve would like to examine the S-dependence of

gauge-dependent quantities. From 2.34 and the effective Feynman

rules, it follows that

< A (x)...*(y)...d*(z)...V '(C,S') > / < V '(C,S') >

p p (2.37)

- < A (x)+-ig(S,S';x)a g*(S,S';x)...g(S,S'jy)d(y)...

r r

g*(S,S';z)(l*(z)...v '(C,S) > / < V '(C,S) >

p p

We shall also be interested in the singularities of gauge dependent

v
quantities near S. The singular behavior of A (x) is:

vS 2 2
A (x) r * n, n r / r
a o n (2.38a)
p e arfSpp lor 2 ^ p

where n, and n, are orthogonal unit vectors lying in S and r is the

vector from x to the nearest point on S. From 2.31b and 2.38a we

derive
 30

g(S,S';x) C exprip8(x)l (2.38b)

where S' Is any surface distant from S and x. 0(x) is defined by

taking a 2-plane normal to S and containing x: S will intersect this

plane in one point and 0(x) is defined as the angle around this

point, from x to an arbitrary fixed direction in the plane. It fol­

lows from 2.34a that for x near S, A (x,y) is exp[-ip0(x)] times a

g

non-singular function, and for y near S it is exp[ip6(y)J times a

non-singular function.

We have found that reduced gauge invariant Green's functions are

S independent; we return now to the factor that we divided out, equa­

tion 2.23. < V '(C,S) > is given by the graph with one photon start­

ing and ending on S, plus the sum of all vacuum bubbles constructed

out of the effective propagators and vertices. For example, the

determinant term in 2.23 is from the graph which is just one closed

loop of the effective propagator A • As discussed earlier, graphs

constructed from the effective propagators and vertices are all S

independent. The only S dependence is that which we have found

explicitly, the 6 (0) term. We can take this term to be an artifact

of the way we have defined V ' when two of the fields in the expan­

sion of the exponential are at the same point, and divide it out of

the definition: all Green's functions are then S-independent. It is

good that this term can be identified so unambiguously, so that

srtificial S dependence can be distinguished from a real, physical

dependence of the vacuum expectation value of the vortex operator on

the area of the minimal surface spanning C. Of the surviving terms

in 2.23, the first is exactly the same as the leading term in the
 31

expectation value of the Wilson loop, with the replacement of — for

e. The next Is of a form familiar from functional integrals. In

fact, < V '(C,S) > can be Interpreted either in the normal way as a

functional integral over continuous A and j* fields with V '(C,S)

P

inserted into the integrand, or as a functional integral with no

Insertion in the integrand but with the A and f fields fixed to have

the discontinuities 2.37 on S. Since we want to be careful about

divergence problems, we will stay with the first interpretation.

£ . Renormalization of Looplike liberators

In the preceding section, we neglected renormalization. We did

not specify whether quantities were bare or renormalized, we did not

include graphs with counterterms, and we did not consider the conver­

gence of the various graphs. These points are the subject of the

present section. We include first, as an illustration of some of the

ideas, a short section on the renormalization of the Wilson loop

operator.

1. Renormalization of Green's Functions of the Wilson Loop

The Wilson loop Is a composite operator involving products of

arbitrarily many elementary fields. The associated divergences, how­

ever, turn out to be easily analyzed, at least in the Abelian case:

all matrix elements can be made finite by one overall multiplication

28 29 30

of the operator . Gervais and heveu and Polyakov have shown by

the use of elegant methods that the aame is true of the non-Abelian

Wilson loop.

A general Green's function of the Wilson operator, < W (C) > T ,

defined by analogy to equation 2.21, is given by

 32

C 2 39
< «e< > >JKL " « P . S vThibu < ' >
khmn

The quantity eA (y) is invariant under renormalization, due to the

Ward identity: e A « e A , where subscripts r and o represent renor-

malized and bare quantities, respectively* When it is necessary to

take a particular renormallzation scheme, we will use BPHZ at zero

momentum (this is acceptable when the charged fields are all massive,

which, for simplicity, we assume)* If we then take the fields cou­

pled to J, K, and L to be the renormaiized ones, the connected

Green's functions in equation 2*39 are all renorraalized: they are

finite when not evaluated at the same spacecime point, and they are

integrable over spacetime regions that include coincident points* If

we take J (x), K(x), and L(x) to be smooth, the associated x

integrals all exist*

The only possible divergence comes when k is greater than 1, so

that there Is a multiple integral over C. This may diverge when two

or more integrands approach each other along the loop. To emphasize

the region in the multiple loop integral where j of the arguments

approach each other along the loop, we represent 2*39 as

(ie)^dxX dy c X o f <. y 2 r *dy j y

i
6( 5 ly^yjl - x) < A (y )...A (y ) ••• >
flf 1 y j

3 2 ( 2 4 0 >
u^fL** x " Xc^te^B-
c 'i« 2j» • •*'i
jyy
w
-
 33

J
S( S |w l - l)< A^CyjU.(y +itw )...A (y -htw.)
1 1 2 y 1 •••>,,

where the final ellipsis in the Green's functions indicates the

remaining operators, coupled to external sources or to distant points

on C To get the leading x-*0 behavior of the Green's function we

31
use the operator product expansion to write the product of j A's in

the Green's function as

8 <xw 1
o«--.y 2 ''V*

xv
+ S lo( ... y( '
fl 2 '" "• J c w )
j * u l A < y )+ h i
8 h e r
operators

This form places all of the x dependence In the coefficient functions

g.. We implicitly use a covarlant gauge so as to avoid direction

dependendent singularities, as found, for instance, in the axial

gauge•

At fixed x, the j points cannot be coincident' We cannot

assume, however, that the w-integrations converge, as there will be

regions in integration space where some subset of the points come

together. We will assume a coordinate-space version of Weinberg's

32
theorem , which we have not proved but which seems quite plausible:

that it suffices to consider just the x-integrations for each subset

of points, and if naive power counting indicates that every one of

these is convergent then the whole integral will be. In fact this is

not necessary: at least when C is an infinite straight line the

Green's functions can be written In momentum space. Weinberg's

theorem and BPHZ subtraction may then be applied rigorously to verify

the conclusions reached below. It Is then very plausible that for a

 34

smooth curve C the leading divergences are the same as for the

straight line. The coordinate-space argument is shorter and perhaps

more interesting'

Because A Is of dimension 1, at each order in perturbation

theory (we are renormalizing order by order) g n _ _ _ y(xw) is of order

x times logarithms and g _ (xw) is of order x J

times loga­

rithms; the higher coefficient functions are all smaller as x-*0.

The x-integration associated with the operator 1 is thee linearly

divergent. Performing the x integration with a distance cutoff A "

and the w integration leaving out those subregions where subsets of

points become coincident, the c-number piece of 2.40 becomes

, dy n <y ) + f i n l t e t e r m s - 2 4 1
{ *C/V>J c lc* « l >* "* * c < ' )
where R is 0(A), n (y ) is a unit vector tangent to C at y , and the

ellipsis in the matrix element is the same as in 2.40. If C were

straight, it would be clear why n must appear: there is no other

available vector. For C a smooth curve, the leading divergence is

the same as for a straight line, since points close together don't

see the curvature; therefore n (y ) appears. To (lit that another

way, curvature, the lowest dimensional measure of the actual shape of

C, is of dimension two; its coefficient must be two powers less

divergent than linear (that is, convergent). This argument fails if

C has a kink, as the curvature there is infinite. In general there

will be a composite divergence associated with a kink that is one

power weaker than the leading divergence, from endpoint effects in

the integration. There would therefore be an additional logarithmic

divergence for each kink. This was also found by Gervais and
 35

29
Neveu . The Integral In 2.41 Is simply the perimeter P(C), and so

the divergence corresponding to the c-number in the operator product

expansion (which arises from graphs with no external lines} can be

removed by multiplication of the Wilson loop by an overall factor

expI-R(A)P(C)].

The only other divergence is in the coefficient of A (x). This

is logarithmic and is given by graphs which have external lines but

are such that cutting a single photon line separates all of the

external lines from the Wilson loop. As one might expect, this

divergence is actually absent due to the Ward identity* In fact, in

31
our particular scheme (zero momentum BPHZ) ,

g (xw G C x , x 0 ) ( 2 - 4 2 )
ioc-..yu 2 "V " ...y « V' V o t /

where G is the j-photon Green's function with one leg (the one with

index u) truncated and set to zero momentum. G satisfies a Ward

identity

k G ( x w , x w ! k ) 0 ( 2 , 4 3 )
u *...y f J 2"" j "

From 2.43 and the fact that G is continuous at k "0 (for massive

charged particles) it follows that G vanishes at zero momentum and so

g. is identically zero and that there is no divergence coming from

connected graphs with external lines* In fact, g. can be shown to

vanish by gauge invariance in any renormalization scheme.

We find, then, that all matrix elements of the Abelian Wilson

loop can be made finite by one overall multiplication*

2. Renormalization of Green's Functions of the Vortex Operator

 36

The analysis of the vortex operator proceeds much like the

analysis of the Wilson loop, and the result is the same: an overall

multiplication makes all matrix elements finite. There is a poten­

tial logarithmic operator divergence, as for the Wilson loop, but its

coefficient again turns out to vanish. The vanishing is here more

intricate, involving cancellation of field renormalization diver­

gences against composite operator divergences. Our nicest result,

that T[F^_] is finite for an infinite straight vortex, can be

obtained in a few lines (equations 2.48-2.50). The rest of this sec­

tion is simply power counting to establish that this implies the

cutoff independence of all matrix elements of vortex operators for

any curve*

F j
We first must ask whether the combinations and -£• appearing

in 2.4 and 2.20 refer to the bare or the renormalized fields and

charges, since these combinations are not invariant under renormali­

zation. A canonical argument indicates that they must be bare-

Equations 2.5 hold only for the bare quantities; if the renormalized

quantities were to appear on tfca left hand side of 2*5, an extra fac­

tor of z~ would be needed on the right hand side. In order to have

a finite commutator with the Wilson loop, as in 2.1, or with charged

F i
0 t
fields, as in 2.6, it is then clear that we must have °'PP and P.
e e
o o

We can also reach this same conclusion from the Green's func­

tions. The discussion of the last section did not include graphs

with countercerms. This is correct if all fields and couplings,

including those in the definition of V ', are the bare ones (we must

have a cutoff at this point). Otherwise, graphs with counterterms

 37

enter and spoil the quantization condition* The exponent in the

definition of V ' must therefore be cutoff dependent* If we had

started by defining V ' with F and e , we would have found that

p/Z,. [Link] p, was an integer, so that the cutoff dependence would

merely be shifted into p* Either way, writing V ' in terms of cutoff

independent p, e, and F gives

rr

d f ( y ) } ( 2 4 4
V(C,S) - ~ P < ^ S < j . r o ; j i ' >

with an explicitly cutoff dependent exponent. Note also that this

implies that the total coefficient (all counterterms summed) of the

one photon graph for < e F

r
V '(C,S) > or for < e F V '(C,S) >
° o opp p r rpp p
is np, with no factors of Z_. This in turn implies that equation

2*36, which comes entirely from the one photon graph, is correct with

F and e either both bare or both renormalized; explicitly,

f ( x ) V ( C S ) > E d 5 t 5 A ( x _ x ) < V C S
V * rpp p' ' -T ?C y ' p'< ' > >(2.36')

We now investigate the cutoff dependence of the Green's func­

tions of the vortex operator. Starting with the expression 2*22, we

can analyze the divergences of the Green's functions very much as we

did for the Wilson loop* There are two differences* One is that V,

unlike W, contains a manifest cutoff dependence from field renormali-

zation, as discussed above* The second difference is in the analysis

of the composite divergences, where the vortex involves an integral

over a two- rather than one-dimensional surface, and the field being

integrated is of dimension two rather than one* This second differ­

ence is small: because the exponent is dimensionless for both opera-

 38

tors, in each case there is only a small nuober of divergences. The

analog of (2.40) when j points cone together is

fC S (2 45>
< V ' > >JKL " <?>'£,* ^ " V i V ' 2 - * ^ -
6(J 2 |w | -1)< ^a(y,)F (y -hw >...F (y -h» ) • • • >
x y8 1 2 ff 1 ;J c

The small x expansion for the operator product is now

h 1
n ~ a VS ..„(*"»»'•••""J' (2.46)
Oorp,/o, . . . ,up 2 J

+ h y }
i«p.» ff ;«ri/»V—"V-V i
+ b ( x w , ( , c , )
2«p,y6,... ,pp z w,) '* i* i + •• •

The expansion includes only gauge invariant operators, because the

-21 J
operator product is gauge invariant, h. is of order x and h, and

h_ are of order x ^ , so that the x-integration for the coefficient

of 1 is quadratically divergent, and those for the coefficients of

F and <t*f are logarithmically divergent.

We have neglected the extra operators in V (C,S) that we con-

P
eluded were needed to insure S-independeoce* In the two dimensional

case, equations 2.14 indicate that the effective dimension of

2

d z f. <z) in units of mass is greater than '2 but less than -1. The

total dimension of the exponent of H is then negative; it is a "soft"

operator. We expect that this property will hold in general. The

inclusion of these additional operators, then, will give additional

contributions to the coefficient functions in 2*41, but will not lead

to any stronger singularities.

 39

The divergence proportional to 1 implies a common infinite fac­

tor of the form exp[R'(A)A(S>) In every matrix element of V ', with

P
2
A(S) the area of the surface S, and with R' 0(A )• Just such a

divergence was found from the one photon graph in the last section,

where it was also shown that such a term, being explicitly S depen­

dent, could not arise from any higher order graph. It can therefore

be unambiguously divided out. The operator 1 in the expansion also

gives rise to a- linearly divergent term proportional to the perimeter

FCC), from edge effects in the surface integrals. This was also

found from the one photon graph; for this divergence we would expect

that higher order graphs will also contribute. At any rate, It can

be divided out by a factor of the form exp[-R"(A)P(C>], with R "

0(A).

TOP logarithmic divergence from the h and h_ can be removed by

the addition of three counterterms to the exponent of V *(C,S) (as in

P

the case of the Wilson loop, this conclusion can be verified

rigorously by use of Weinberg's theorem and BPHZ subtraction when C

is a straight line):

c 2 47a)
i -JV^p^y) <-

c 3 - J^-ldo^ |«-*<y)*(y) (2.47c)

I do- | appears in c. for the same reason that dy, n (y,) (which is
orp J lor cr l
just |dy,|) appears in 2.41: there is no other Euclidean invariant

form for the leading divergence*

 40

The counterterm c, is forbidden by by CF invariance. do- „F _

2 ' oc[J orp
is CP even; as a result, so is any logarithmic composite divergence

(because a smooth surface is CP invariant to the extent that its cur­

vature can be neglected), while c, is CF odd. c, is forbidden by the

S-independence of gauge invariant Green's functions. The easiest way

to see this is to consider a surface S which doubles back on itself.

It is important that because of S-lndependence, there is no new com­

posite divergence associated with this doubling back. The doubled

surface cancels out in the definition of V '(C,S), because do- 0 is

oriented; it must therefore cancel out in any divergences. It does

cancel in c, and c„, but not in c,, as |d<r ,1 is not oriented. We

1 2 3 orj5
conclude that all of the composite divergence can be removed by a

term of the form t(A)'C,'

There is still the second source of cutoff dependence in the

Green's functions of V '(C,S), the factor of Z- in 2.44. J

We see that
P
this is of exactly the same form as the counterterm 2.47a; an

appropriate choice of f(A) can remove this divergence as well, leav­

ing every Green's function of the vortex operator finite.

Since there is only one unknown function of the cutoff, we can

determine it by calculating one single Green's function of one par­

ticular vortex operator. A convenient choice is

T< F (x) V '(Z.Y) > / < V '(Z.Y) > (2.48)

rpp P P
a n d Y i B t n e 0 .
where Z i s the l i n e y "yi"y2"°
Q half-plane y "y."°t y , *
0

By Euclidean invariance, this must be given by

2 2 2
a ( x>c )) * „,x
, x , _ ++ b(x?)
b ( x )x x . , 6u ,. ,, (2.49)
z' upor3 ZQC z' z(ji p)3
 41

where x Is the vector from Z to x which Is perpendicular to Z.

ZfJ
Under parity, V '(Z,Y) goes into V '(Z,Y'), where Y' h half-
i B t e

P P
< 0
plane y "yj"0> y ,
o (note carefully that p doesn't change sign). The

gauge invariant Green's functions of these two operators axe equal-

Under x--»x , x.-*-x , the Green's function 2.48 then has natural
0

parity: space-space components are invariant and space-time com­

ponents change sign. Only the first term is 2.49 has natural parity;
2
it must be that b(x )«0. Equation 2.36' then requires that

/ 2.
a (
p , 2,-3/2
V * " IT'V (2.50)
r

Remarkably, the Green's function 2.48 is completely determined, and

it is cutoff independent without counterterms: f(A) " 0. The factor

of Z, in 2.44 provides just the cutoff dependence to cancel that in

33
the composite divergence

In terms of graphs, the cancellation we have found is this: The

order e contribution to 2.48 is figure 2a. Expanding the effective

vertex and propagator in terms of the usual ones gives all graphs

with one charged loop and no internal photons, such as those of fig­

ure 4. The graphs of figure 4a have the usual divergence; ordinarily

this would be cancelled by a counterterm from the order e~ graph,

figure la. Here, this counterterm is absent owing to the factor of

Z, in 2.44* However, the graphs of figure 4b are also divergent when

the integration over the momenta of the photons attached to V ' is

included. What we have found above is that the divergences of 4b

exactly cancel those of 4a. This cancellation has an interesting

feature: the different graphs are proportional to different powers of

p, depending on how many photon lines attach to V '. When we let p

 42

vary, the full cancellation occurs only for i n r — a l p, as only then

can we say that the Green's function is independent of S and that

2.49 is the only allowed form.

The fact that the matrix elements of the vortex operator turn

out to be automatically finite is rather important- If ve had had to

add a counterterm 2.47a to the exponent of V ', its commutation rela­

tions 2.1 and 2.3 would then have become cutoff dependent* We hope

to find that there are relations between the commutation relations of

V ' (that is, the fact that it is a vortex operator) and its Green's

functions; this would seem unlikely if cutoff independent commutators

had been incompatible with cutoff independent Green's functions.

This is remeniscent of the situation with Noether currents, genera­

tors of exact symmetries: there also ve wish to ascribe physical sig­

nificance to commutators, and there also the commutators and the

Green's functions are simultaneously finite.

There Is one weakness in the above analysis. What we have

really shown with the operator product analysis is that the counter-

term c. suffices to remove all divergences from Green's functions of

V when they are expanded order by order in p (thus, in 2.44 we have

isolated all graphs of order p ) . It would be preferable if we could

first sum to all orders of p, getting effective propagators and ver­

tices as before, and then analyze the divergences directly from the

short distance properties of these effective propagators and ver­

tices. We do not expect that our conclusions would change; however,

we shall see in the next chapter that «n expansion of the Green's

functions in powers of p can sometimes lead to erroneous conclusions,

so it would be good to have an analysis of the divergences which did

 *3
not rely on such an expansion* Note that the flniteness of 2.48 was

independent of the expansion in p. It would also be interesting to

see what effect hard P- or CP- violating interactions would have on

the analysis*
 CHAPTER 3

CLUSTER PROPERTIES OF LOOPLIKE OPERATORS

A. Cluster Properties in a. General Gauge Theory

In this chapter we discuss some of the long distance features of

the Green's functions of operators associated with closed curves-

The present section is quite general, and applies to an Abelian or

non-Abelian theory, though we shall be particularly concerned with

the non-Abelian case. In the next seccion we return to an Abelian

theory and extend the results of the previous chapter to the case

where the symmetry is spontaneously broken.

The spectrum of a given gauge field theory might have one of

many different forms. It might be "QED-like", where the physical

particles resemble closely those in the Lagrangian: nassless vector

particles and unconfirmed charged (that is, non-singlet) particles.

It might be confining, with all physical particles gauge singlets.

It might be Higgs-like, with the gauge particle massive and the gauge

symmetry broken either by a fundamental or a dynamical Higgs field.

It might be intermediate between these possibilities, or it may have

some other form altogether. Which form, or phase, is realized will

depend on the details of the dynamics.

It is desireable to have some order parameter, some precise dis­

tinction between the phases. For Instance, the QED-like phase is

distinguished by the presence of a massless vector particle. The

confining phase of a non-Abelian theory Is not so readily defined.

14
As pointed out by Handelstam , there is no global non-Abelian

charge: the current is gauge covarlant, so its integral over space

44
 45

has little meaning• It is therefore imprecise to say that a physical

particle is a gauge singlet. Nor is the identification of the non-

Abelian Biggs phase obvious. The expectation value of the Hlggs

field is not gauge invariant, and there is no satisfactory gauge to

use. The Coulomb and covarlant gauges do not exist in a non-Abelian

34 14

theory , and the axial gauge is too singular • The confining and

Hlggs phases can be defined precisely by the Wilson and 't Hooft

criteria, relating to the vacuum expectation values of large Wilson

and vortex loops. We shall return to these after we show how the

phases can be characterized by the cluster properties of the Wilson

and vortex operators.

Consider the Euclidean Green's function

G(x,C) - < «(x) W (C) >

g c

- < e(x) w (c) > - < e(x) > < w ( o > s (3-D
8
with 8)(x) some local gauge invariant operator such as F flFye. In

the phases we have mentioned above, this function will behave In one

of three ways when C is very large:

Short distance clustering: G(x,C) falls off exponentially with

d(x,C), the distance between x and C.

Surface clustering: G(x,C) is nonvanishing near the minimal sur­

face, S spanning C, and falls off exponentially away from that sur­

face.

Long distance clustering: G(x,C) falls off as a power of d(x,C).

These alternatives may be better understood by considering a

 typical state of the system in a 3-surface cutting perpendicularly

through C. In this 3-surface, one sees a source/antisource pair, s

and s, where C intersects the surface. If the clustering is short

range, there are only short range, Yukawa, fields around the sources,

and vacuum elsewhere* If the clustering is surface-like, there is a

tube of non-vacuum joining s and s, whose energy per unit length

gives rise to a linear potential between the external sources. If

the clustering is long range, s and ~B have Coulomb-like fields with a

power law fall-off. Long range clustering is only possible if there

are massless particles. The other two types of clustering each have

a characteristic scale (the range of the Yukawa field or the thick­

ness of the tube) which are determined by the mass m, of the lightest

particle; for quarkless QCD this is presumably of the general order

of A Q C D . AS U L P ( C ) is taken to zero, either by shrinking C or by

letting m. go to zero, the first two cluster properties turn continu­

ously into the third.

Short distance clustering is analogous to that for pointlike

fields in a massive theory, where the general connected two-point

Green's function falls off exponentially with distance. Long dis­

tance clustering is analogous to that for pointlike fields in a mass-

less theory. Surface clustering is a new feature; it seems to arise

when there is a flux which can neither spread nor be shielded. One

might also image more general cluster properties, of course- These

three, however, seem to cover all those which have arisen in various

gauge theories and models.

We can also consider the cluster properties of Green's functions

of the vortex operator V (C) (p designating the homotopy class); the

 47

same three possibilities seem to arise. The cluster property appears

to be largely independent of the operator • ( * ) • For instance, if a

tube of non-vacuum runs between s and s, we would expect most local

operators to have within the tube an expectation value different from

that which they have in vacuum. It may, however, depend on the

representation s of the Wilson loop, or the homotopy class p of the

vortex operator.

A general phase, then, may be characterized by which of the

three cluster properties is realized for each representation and for

each homotopy class. A confining theory is one in which the Wilson

loop, at least in some'representations, has surface clustering, as

this implies linear confinement of external charges in those

representations; this is the Wilson criterion . The 't Hooft cri­

terion defines a (completely broken) Higgs theory as one in which

some of the vortex operators have surface clustering, as this implies

that magnetic flux is forming into tubes.

This classification is closely related to the usual classifica­

tion of phases in terms of the vacuum expectation values of the Wil­

son and vortex operators. For a very large curve C, the vacuum

expectation value of a general looplike operator X(C) will be dom­

inated by expf-S ,J» where

4
S c l - J*d x < L(x) X(C) > c / < X(C) > (3.2)

and L(x) is the Lagrangian density. The connected Green's function

in 3.2 is a special case of 3.1. For short distance clustering the

integrand will be nonzero only for x near C, so that the whole

integral is proportional to P(C): < X(C) > follows a perimeter law.

 48

For surface clustering, the integrand vlll be nonzero only for x near

the minimal surface, so that the whole integral is proportional to

A(S„). In phases without massless particles, there is a one-to-one

correspondence between an area law and surface clustering, and

between a perimeter law and short range clustering*

For long range clustering, the integrand is proportional to

d(x,C) ; the behavior of the integral depends on the particular value

of k. In QED k is -It; the integral therefore is dominated by small

values of d(x,C) ard is proportional to P. If k were -3, the

integral over d(x,C) would diverge logarithmically until it was cut

off at roughly the linear scale R of the curve C; it is proportional

to R-ln(R). We use R rather than P because P implies a certain shape

dependence, which aeed not hold here. If k were -2, the integral
2
would be proportional to R , as it is for surface clustering. In

this case there also be a linear potential between external sources,

but without the formation of a flux tube. There would be, however, a

strongly interacting massless particle, which is not observed.

Furthermore, it is not clear that it is possible to find a consistent

physical picture in which k is -2.

Restricting attention for now to phases without massless parti­

cles, the Uilson loop operator for each representation and the vortex

operator for each homotopy class will have either short range or sur­

face clustering. The cluster properties for the Wilson loop in dif­

ferent representa:ions are related. If there are fields of represen­

tation s in the theory, it will always be energetically favorable for

s and s external sources, when they are far apart, to pull an ss pair

out of the vacuum and shield themselves rather than form a flux tube.
 49

W will then always have short range clustering. Similarly, one can

see that If for any representation r, W has short range clustering,

this will also be true of any other representation r' which is con­

tained in the direct product of one r and any number of s representa­

tions. The cluster property is therefore a function only on the quo­

tient, {r>/{s}, where {r> is the set of all representations of the

gauge group, and <s} is the set of those representations carried by

fields or products of fields. Incidentally, for any compact gauge

group, the above quotient is isomorphic to II, (G), so the set of dis­

tinct types of Wilson loops is isomorphic to the set of distinct vor­

tex operators (though there is no unique natural mapping between the

two sets). One further restriction on the clustering is that it must

be the same for a representation and its conjugate (or for a homotopy

element and its inverse), as the looplike opcvators for these differ

only in the direction of their "flux", so that they can be turned

into one another by a spacetime rotation.

't Hooft has shown that relations can be obtained between the

cluster properties of Wilson loops and vortex operators In phases

18
without massless particles. One result is that ii: a Wilson loop

and vortex operator do not commute, they cannot both have short range

clustering. To Bee this, let C and C be large loops, C lying in the

3-plane t»0 and C* lying in the 3-plane t«-R (we single out one

Euclidean direction and call it t ) , where R>>mT . Let them have the

spatial orientation of figure 5a, linked but separated by the large

distance R. Consider < V (C) W ( C ) >, where p and s are such that
P s
the operators do not commute, when C is translated as follows:

first, forward in time to t-Rj then, in a spacelike direction to the

 50

spatial configuration of figure 5b (unlinked, and separated by R ) ;

then, back to t—R; finally, back to the position of figure 5a. The

Green's function must now have returned to its original value. C and

C are always separated by at least R. If both operators have short.

range clustering, there can be no correlation between them. In other

words, C is essentially being moved through vacuum, so that

< V (C) W (C') > is constant by translational invariance. This is

p s
true except when C crosses S; :he Green's function then jumps by a

phase. C sweeps out a closed surface, which links C a definite

number of times. It therefore crosses S this same number of times,

independent of the choice of S. Taking the special case where S is

the minimal surface spanning C, we see that C crosses this only

once, on its first leg to t-R.

The only change in the Green's function during this round trip

is one change of phase, when C crosses S, corresponding to the non-

commutation of the two operators. This is inconsistent with the

Green's function being single-valued. It is therefore impossible for

both operators to have short range clustering. If either one has

surface clustering, the Green's function changes not only when C

crosses S, but also when one loop moves through the flux tube of the

other. If there are massless particles and one or both loops have

long range clustering, the Green's function can change even when the

two loops are well separated. In either of the latter two cases,

there is no contradiction with the single-valuedness of the Green's

function. We have tried, above, to emphasize the distinction between

the arbitrariness of phase of the Green's function associated with a

change of S, and the fact that it is well defined, and single valued.
 51

once S is fixed•

35
More recently, using a more powerful framework , 't Hooft has

been able to place even stronger restrictions on the cluster proper­

ties in phases without massless particles. We shall refer to some of

his results later.

We now illustrate the above ideas. In a free Abelian theory,

one readily finds that all Wilson and vortex operators satisfy a per'

imeter law. For the vortex operator only the second term of 2.23

survives; an almost identical expression holds for the Wilson loop.

External charges and monopoles have Coulombic fields, and connected

Green's functions all satisfy long range clustering, with various

powers of d(x,C). For the special case of •(x)-L(x), It—4, as men­

tioned above. For QED with massive charged fields, the infrared pro­

perties are essentially those of the free theory, and the above con­

clusions still hold.

In a pure SU(3) gauge theory (QCO with quark loops suppressed),

there are three different kinds of representations, distinguished by

"triality", the SU(3) version of the quark number discussed before

equation 1.5. The representations of zero triality can all be formed

as a product of adjoint representations; since there are always

adjoint fields present (gluons) these can never have surface cluster­

ing. Assuming on phenomenological grounds the absence of massless

glueballs, the Wilson loops for the representations of zero triality

will always have short range clustering. The representations of tri-

alitles one and two will always have the same clustering, as they are

conjugate to each other. The Wilson criterion, defining a confining

 theory, is chat this should be surface-like for quarkless QCD, so

that quarks are bound by a linear potential until the flux tube

breaks by creation of quark pairs. In this, the standard, picture, a

linear potential can never form between gluons. Physical gluons must

be assumed to be absent because they are shielded in some way: a high

energy gluon resulting from an interaction would pull more gluons out

of the vacuum and form a singlet glueball. As for the vortex opera­

tors, there are two homotopy classes, other than the trivial one,

corresponding to the elements exp[+2ni/3] of the center of SU(3). 't

35
Hooft has shown that these will have short distance clustering , if

the triality one and two Wilson loops have surface clustering.

When the contribution of quark loops is added, it becomes possi­

ble to shield any Wilson loop, so that all Wilson loops will satisfy

a perimeter law. At the sane time, the effective gauge group becomes

SU(3) rather than SU(3)/Z . This is simply connected and does not

permit vortex operators to be defined in a simple way. Thus, insofar

as the Wilson and 't Hoofc criteria are the only distinctions between

phases, it appears that an SU(3) (or SU(n)) gauge theory wich fields

in the fundamental representation can exist only in a single phase,

regardless of the dynamics. This phase may resemble a confining or a

Higgs theory extremely closely, but it would be possible to change

the theory in a continuous way from one into the other. This possi­

bility has been much ciscussed, e.g. in references 36, 37 and 38. We

shall have more to say about it in the next chapter.

Lastly, we consider a perturbative Eiggs theory, one in which

the coupling is quite small so that we may use a covariant gauge

34
without worrying about the Gribov ambiguity • This enables us to
 53

discuss gauge dependent quantities such at the vacuum expectation

value of the Higgs field. Ve shall take the gauge group G to be com­

pletely broken. M, the set of classical vacua (ainina of the Higgs

potential), is then identical to G and there are no massless parti­

cles. Let us consider

# ( x ) - < fix)
vS
V_'(C,S) > / < V '(C,S) >
p (3.3)

where j»(x) is the Higgs field, C is a large curve and S is taken, for

convenience, to lie far away from the minimal surface of C. In a

three surface which is perpendicular to C and cuts it at two points

we have figure 6. Near the small loop 1, which is far from C and

from Its minimal surface, gauge invariant connected Green's functions

will vanish and f (x) will be position dependent but its values will

lie in the set H.

If we have an Abelian theory, we know from 2.37 that f (x) will

be a-expIipO] as we traverse the infinitesimal loop 1 and ft goes from

0 to 2ft. Here a is the magnitude of the vacuum expectation value of

?. M is the set of complex numbers of modulus a. On the loop 1,

v
j» (x) is seen to describe an element of the homotopy group II, (M)

identical to the element of II. (G) associated with the vortex opera­

tor. This idea generalizes to a non-Abelian theory: there too there

will be a £auge transformation g(S,S';x) governing the S-dependence

of gauge dependent quantities, which will give rise to a nontrivial

element of the honotopy group when x circles S or S*; this causes

j* (x) trace out the corresponding element of II, W

V
when x travels on

loop 1.
 We may now imagine enlarging the loop, sliding it off S, taking

it to the position of loop 2 and shrinking it to a point, without

ever getting close to c or c. Because the fields are singular only

at S, ft (x) must be essentially constant on loop 2: it maps out an

element of the trivial homotopy class* By definition there is no way

to continuously deform an element of one homotopy class into an ele­

ment of another while staying within M. It follows that, somewhere

between 1 and 2, <f (x) took values outside of H; this must be the

case at least within a tube between c and c This is precisely the

argument by which one shows that when the Higgs field at spatial

infinity in two space dimensions maps out a non-trivial element of

the homotopy group, there must be a "lump", a Nielsen-Olesen vor-

20 21 —
tex ' , somewhere in space. The tube between c and c 1 B a
vS

Nielsen-Olesen vortex. Where f (x) does not lie in M, gauge invari­

ant connected Green's functions such as that for the Higgs potential

will be non-vanishing. It follows that the vortex operator in a com­

pletely broken small coupling Higgs phase has surface clustering.

This is the source of the 't Hooft criterion. One may also check,

order by order, that the Green's functions of the Wilson loop are

short range, because all fields are massive.

J). Vortex Operators in an Abelian Higgs Phase

In the last section we saw that the vortex operator in a pertur-

bative Higgs phase has a surface-like cluster property. For an

Abelian theory we can obtain a stronger result: the vortex operator

can never have short range clustering, so that in any Abelian theory

without massless particles it will have a surface clustering and obey

an area law. From 2.36' and Gauss's law.

 55

T< f V {C S) V (C S) (3 4
JV'Jp pp<*> p ' » * - " 7 ^ ' ' * P '>
where C Is a large curve and B Is a 2-sphere linking C (we might Ima­

gine figure 5a as showing the 3-surface containing C, with C as the

equator of B; the rest of B extends Into the t-direction). All

fields and charges in this section are taken to be renormalized• The

integral over B is independent of the radius of B. On the other

hand, since the Green's function in 3.4 is gauge invariant, short

range clustering would require it to fall exponentially when the

radius of B is greater than SL ; this is Inconsistent with equation

3.4.

What we have shown is really quite simple: magnetic flux can

never be shielded. If the absence of massless fields then makes it

impossible to have a Coulomb field, magnetic flux can only form into

tubes. One could see this directly from the operator equation

5'B(x)»P (x), where p (x) is a purely external magnetic charge

density; this equation is one component of 2.36'. Given that mag­

netic flux is confined, one can extend 't Hooft's more recent
35
results to the Abelian case to show that all Wilson loops must obey

a perimeter law. It follows that an Abelian theory (without magnetic

monopole fields) never confines. This result was anticipated by Man-

14
delstam on the basis of the existence of the Abelian Coulomb gauge.

One might try to argue in a different way that the Green's func­

tions of the vortex operator had to be short ranged in a phase

without massless particles. Take a large curve C, with the surface S

far away from the minimal surface, as it was In figure 6. Consider

 56

< «(x) V '(C.S) > / < V *(C,S) >

P c p

< l y)li (3 5)
- J u •<*> ffsK W
1 ? *» -
where again 0(x) is any gauge invariant operator and where x is a

point near the minimal surface of C but far from C itself. Because

theie are no oassless particles, every Green'a function in 3.5 falls

exponentially for x distant from S. In particular they are vanish-

ingly small when x is on the minimal surface, so that surface clus­

tering is Impossible. Further, since < 9(x) V '(C,S) > Is exponen-

tialy small for x far from S, and also independent of S, it is

exponentially small except for x near C. This is in direct disagree­

ment with what was shown above. The problem must lie in the expan­

sion 3.5: while this expansion is formally correct, the long distance

behavior of the sum is not the saue as that of the individual terms.

This is the source of our statement, at the end of the section on

renormalization, that the expansion in powers of p is not to be

trusted.

In the remainder of this section, we shall extend to the case of

a spontaneously broken Abelian theory the rules for calculating

Green's functions Involving V '(C,S) in terms of effective propaga­

tors and vertices. We shall be particularly interested in seeing the

surface-like cluster property emerge. We shall also be interested in

the following question: In the presence of a vortex operator we

expect the Hlggs field to have a certain "twist", or singularity; yet

we generally, in a Higgs theory, expand around a constant, non-

singular Higgs field. How are these compatible? Ue shall see that
 57

the twisted Uiggs field arises in a natural way.

First, we will sketch the derivation of the Feynman rules in a

Higgs phase. Consider the functional integral

Z(J) - JDf exp{ - " j M i j ^ - P W + J ^ > (3.6)

Here, for convenience, we have Joined all the fields into a single

field ifr. > where the index i includes quantum numbers, spacetime

indices, and spacetime position. M,, is the quadratic part of the

4
action, P(f) is the interaction Lagrangian, polynomial in I ., and J,

is an external source, whose functional derivatives are the Green's

functions of •+. Writing +-p.+a , where a. is the value of +. about

which we wish to expand the functional integral, we may write 3.6 as

ZCJ) - Pp e* < - ^P i j P i P j - l l y . ^ - ^ y - t - j " ?(p)

a p ( a a J J a 3 7 )
" i »i P> " T i /'ij<P> " ' • ' " i P i " i i> < -

- ex < P V l - M a £- - P(£) -
lj i V i l (^) - • • • }

1 } ( 3 - 8 )
exp< • f ( M " > V J ij ;

Here, we have Taylor expanded the interaction F, with the subscripts

indicating derivatives with respect to p . The ellipses indicate the

rest of the Taylor expansion, which, of course, has a finite number

of terms. Overall constants in Z(J) are thrown out because we are

always Interested in Green's functions with the vacuum-to-vacuum

amplitude divided out.

From 3.8 we may read off the Feynman rules, and compare them

with the usual rules (obtained by taking 3.8 with a - 0 ) . The

 58

propagator is the 6ame, (M~ ) . There are new vertices, from the

Taylor expansion of the interaction, which in graphs can be

represented as interaction vertices with some number of external

lines replaced with a "tadpoles" representing multiplication by a..

Some of these new vertices are quadratic, and could also be summed

into the propagator, rather than being treated as interactions.

There is also a new one-point interaction M , . a , .

ij i

To illustrate these rules ve consider

K Z{3)
u
i " *i * "
Z ( J )
~ 1
Jj ( 3 , 9 )

This equation is represented in figure 7. The first term is from the

action of the functional derivative on the term J.a.. In all other

i i
terms the derivative acts on the quadratic in the last line of 3.8,

giving rise to a propagator and a connected graph. The second term

of 3.9 is from the graph where this propagator ends on a one-point

vertex M a . . In all other graphs this propagator ends on a vertex

v from P (which may have one or more tadpoles attached). Ve iden­

v
tify the one particle irreducible (1P1) part containing v, as v,

plus all vertices which cannot be separated from v. by cutting a sin­

gle line (v way be the only vertex in V), plus all lines connecting

two vertices of /"• Note that all vertices of / are from P; none are

of the form H.,a, as these can be separated from the rest of the

graph by cutting a single line. Y Is then independent of a.,, since

by definition no tadpoles are included in it. F*" is defined as

the sum of all such Y with n external lines besides the one which is
 59

connected to the original propagator. Consider now the external

lines of P"s these may be a tadpole or a propagator, and the propaga­

tor may end on a M. .a vertex or a vertex from P (which then defines

a new 1PI part). These are the same three choices we originally had,

and sum up to give u again; summing over all numbers of external

lines, and checking on the combinatorics, we get the last term of

3.9.

The first two terms of 3.9 sum to zero; only the last term,

vhich is independent of a , contributes. Iterating figure 7 graphi-

39
cally, we see that u. is given by the sum of all "tadpole trees"

Throughout this section, "tree" refers to a term In the iteration of

an equation such as figure 7, not to a graph without loops. Equation

3.9 can be rewritten

, - 0 (3.10)
i^™ ' -u v k-o k ! , i
V •J "3
k 1
2
k
i i 1
where

r +K (3.11a)
^i" h n
n n
r - r . n+2 (3.11b)

are the 1PI Green's functions as usually defined (by the Legendre
40
transformation) and

0 0
l„ k
rev) - - i i n - r v . . . , v ---v
J J J
0.12)
k-0 ' l k l -"k
Is the quantum effective potential. Equation 3.10 indicates that u

is an extremum of the effective potential; in fact it must be an

absolute minimum or one will eventually run into pathologies in the

 60

Green's functions indicative of an unstable vacuum

It is not surprising that u is formally independent of a ; for­

mally, the expansion represents the whole integral, regardless of

what value of the field we expand around. The sum of tadpole trees

is, however, ambiguous when there is more than one minimum, as in the

case of broken symmetry. If we try to sum the trees iteratively by

gathering the last two terms of 3.9 into one term, which we treat as

being small, and taking the first term, a , as a first appoximation

to u , the result now depends on the value of a. chosen. The value

of a. determines which extremum of the effective potential the sum

converges to; for some values of a. it diverges. The most efficient

way to sum the trees is to calculate the effective potential and

39
locate its absolute minima. Taylor has emphasized the role of the

effective potential in summing the tadpole trees. This is a rather

trivial example of the Idea that the convergence of a perturbation

series depends on having the right sort of vacuum as a first approxi­

mation; one might hope, however that Che effective potential would

also be a powerful tool in the less trivial case where a composite

42
operator is acquiring a vacuum expectation value •

We can now write down a general form for a connected k-point

Green's function (k>l) In a Uiggs phase:

< y, . . .^ > . ?S - 1
i -- G i ... u ---u (3.13)
*1 *k c
n-0 n!
V--Vl" n h ,J j
n
where G is given by the sum of all connected Feynman graphs such that

there is no way to cut a single line and be left with two pieces, one

of which is connected to none of the external points i ,...,i, . G is

 61

1FI in the indices j, but not in the indices i. To see 3.13, note

that G is essentially the general Feymnan graph with all tadpoles and

trees amputated, and the u. in 3.13 then represent the sum of all

possible attachments of tadpoles and trees.

In the case at hand, an Abelian Higgs theory, ^ stands for the

fields A (x) and ?(x). u is expected to be zero for A and some

nonzero value u (which can be chosen real) for (< and f . That is, we

draw graphs where either end of a charged line may end on a tadpole,

whose value is u. The presence of G would complicate the graphical

analysis for the vortex operator, so instead of 3.13 we use the

equivalent

< 1* • • •V' > - i (all connected graphs, with trees) (3.14)

1
1 \ c

The meaning of 3.14 is made clearer by comparing it with 3.13. The

tadpoles u. have each been replaced by the tree sum to which they are

equivalent; there is therefore no longer a 1FI requirement on the

graphs. We may see that 3.14 is correct at k«l, as it is there sim­

ply the sum of trees, which is u .

Equation 3-14 holds in a symmetric as well as in a Higgs phase;

the only difference is that in a symmetric phase the fields as nor­

mally defined have vanishing expectation value, so the trees sum to

zero and we can neglect them. To evaluate <V '(C,S)>, equation 2.22,
P

in a Higgs phase, fix the configuration of charged lines, and inter­

nal and external photon lines, summing over all numbers of vortex

photons. As before, fc3 get a graph of the same configuration, but

with no vortex photons and with all propagators and vertices replaced

by the effective ones. Summing over all configurations, we simply

 62

recover equation 2.35 (which includes graphs with trees).

The Higgs phase is distinguished, not by the formal expression

for the Green's functions, but by the nature of the tree sum. We

emphasize this by rewriting the Green's function in 2-35 in a form

analogous to 3.13, that is, with the trees summed. We find, again in

shorthand notation,

< 1 > c v s C3,15)

V"\ "
1? 1 -rvS vS vS
5 TT G. - V. .J. l - V, Ju.l • • • u.
J
r

where G is partially 1PI as before, but is now constructed from the

effective vertices and propagators, and u. is the sum of trees with

vS
effective propagators and vertices, u satisfies the iterative

equation shown in figure 8. It therefore is the extremum of the

"effective potential in the presence of the vortex"

i 2 ij i j k m l k! i!--.^ i x i k

s
Here v represents A , 0, and j!*, I" *- the sum of IPI graphs with
1 is

the effective propagators and vertices, and the first term is expli­

citly

One may verify that under a change of surface

vS VS
r '(A„,?,?*) - T ( A ,g(S,S')*?,g(S,S')i>*) (3.18)
r r

From 3.18, it follows that

 63

S
A* '(x) - A " ( X ) (3.19a)

tS tS
* ' ( y ) - g(S,S';y)d (y) (3.19b)

where the superscript tS indicates the extremum of the effective

potential, that is, the sum of tree graphs constructed with the

effective propagators and vertices for the operator V '(C,S). The

tS
expectation value of <f is just f • For the expectation value of A ,
there are the graphs where the external photon ends on a vertex—
tS
these sum up to A — and there is the graph where it ends on S—this

is simply A • It follows from 2.34b and 3.19a that the expectation

value of A , which is A +A , satisfies 2.37, as does the expecta-

F P P
tion value of (••

The Green's functions are then given by graphs constructed from

the effective propagators and vertices, with tadpoles corresponding

to the tree sums for A and <f in the presence of the vortex, rather

than their free values. The tree sum for $ has the expected singu­

larity on S, as we can see from 3.19b. Th-'.s is equally true in Che

Higgs and symmetric phase; the difference, again, is in the nature of

the extremum of the effective potential. If we take a vortex opera­

tor with C and S bounded, distant from C and S the propagators and

vertices are essentially the sane as in vacuum, and so will be the

vS
effective potential. The tadpole u will therefore have its free

value. For the symmetric theory this is 0 for all fields; in the
iof
Hlggs theory it is a constant ue for ^ and zero for A . Arguing

as in the last section, in the Biggs phas- there must be at least a

surface (of some thickness) spanning C on which the modulus of the

expectation value of the Higgs field is not u (it must in fact go

 64

through zero). In the symmetric phase, however, f (x) may have its

vacuum value of zero everywhere (except very close to C ) , as 3.19 is

linear in jt and so is trivially satisfied by zero. The surface clus­

ter property is seen to emerge from the sum of all trees, even though

each term in the sum can be shown as before to have short range clus­

tering.

I
 CHAPTER 4

UNQUANTIZED VORTEX OPERATORS

A. Continuum Theories with Fields in the Fundamental Representation

In the foregoing chapters, a certain quantization condition has

played a central role, this is, that if there are fields present in

a representation s, the vortex operator oust commute with the Wilson

loop for the representation s. We consider here the case of a system

which does not have vortex operators, but which closely resembles a

different system which does.

At least two such systems are of interest- The first is an

ordinary superconductor. The Cooper pairs, which form the Biggs

field, have charge 2e. This field will support flux tubes quantized

in units of •s—, and tubes with a single unit of flux are in fact
A3
observed experimentally • We would expect to be able to define an

operator which creates such a flux tube, and whose surface-like clus­

ter property would be a signature of the stability of the tube and

the existence of the Higgs phase. On the other hand, electrons and

holes, with charge +e, are also present, and as a result we seem to

be able to define the vortex operator only for even numbers of flux

units. Since these "sea" electrons and holes do not appear to radi­

cally alter the properties of the single-unit flux tubes, the quanti­

zation condition does not seem to reflect the physics of the system'

The second system is QCD, where, as discussed in section 3A, it

appears that the introduction of quarks has removed all distinction

between phases. In an SU(3J gauge theory without fields in the fun­

damental representation, we would expect to be able to distinguish a

65
 66

confining from a Higgs phase by the stability of electric or magnetic

flux tubes. It is clear that the introduction of quarks makes it

impossible to observe an electric flux tube, since they can shield an

electric source. It is not at all clear, though, why they should

make it Impossible to observe a magnetic flux tube, and we might

still expect to be able to define an operator whose cluster property

would enable us to say definitely that QCD is or is not In a Higgs

phase.

In this chapter, we first show that without the quantization

condition, a vortex operator cannot be "local looplike", in the sense

of commuting with every gauge invariant operator at a spacelike

separation from the given closed curve. By analogy with QCD, we call

representations which violate the quantization condition fundamental,

and those which respect it adjoint. (This designation depends on

which vortex we are considering, of course: electrons are fundamental

with respect to odd-unit Abrikosov vortices but adjoint with respect

to even-unit vortices.) We show that we can define a vortex-like

operator which has a surface-like cluster property in a phase with

adjoint Higgs field and fundamental non-Higgs field. We argue that

in a confining phase with uhe particles in the fundamental represen­

tation weakly coupled the operator would not have surface-like clus­

tering, so it does serve to characterize the Higgs phase. This

implies that there can be a Higgs-confining phase transition when

particles in the fundamental representation are present, provided the

Higgs field is in the adjoint representation. We point out a possi­

ble paradox involving 't Hooft's restriction on the possible cluster

properties, and suggest its resolution. We then discuss the

 67

extension of these ideas to a non-Abelian theory* In a second sec­

tion we give a short discussion of a lattice gauge theory with two

matter fields, one fundamental and one adjoint, which provides an

excellent model for our problem. We show that when the fundamental

field is weakly coupled, there are the expected phase transitions,

and that our modified vortex operator is a good order parameter.

When the fundamental field is strongly coupled, it is not clear

whether our operator should still be a good order parameter, as it is

quite nonlocal. We investigate this question and find some indica­

tion that it is a good order parameter; we then discuss why this

might be so, and -Uscuss areas for future investigation.

Consider first the operator given by 2.4 or 2.8 when p is not an

integer. The commutation relation 2.6a is then discontinuous on the

surface S where the angle 8 is discontinuous. It then follows that

the gauge invariant operator

> > >

• U . ^ l P ) - **(x )expUeJ- dx' A <x '))*('y )
p 1 i (4-D

will not commute with V (C,S) or V '(C.S) if P crosses S. This is

P P
-> ->
true even if x and y are very close together but on opposite sides
-> -*
of S, so that #(x ,y ;P) is essentially local. The local operator
* -* -* ->
f (K)D(»(X) can be obtained from B(x,y;P) as a limit when y -*x ; it
fails to commute with V or V ' when x is on S. We also note that a
P P
meson-like bound state of f and 0* at position z is created by a

superposition of operators 4.1 with x and y near z. For z near S,

the V and V ' will commute with some of these operators (those which
P P
do not cross S) but not with others, so that it multiplies only cer­

tain components of the meson wavefunction by a phase, leaving

 68

something which no longer looks like the same kind of meson- He con­

clude that when the quantization condition is satisfied, the opera­

tors 2.4 or 2.8 are surface-like rather than looplike; their Green's

functions would be expected to cluster near the surface S, and they

would obey an area law, regardless of the nature of the vacuum-

He might wonder whether by being cleverer we could find a local

looplike operator which still satisfied equation 2.1, or, if not,

just how nonlocal the operator must be. We can make a general argu­

ment that if there are fields present in som2 representation s, and

if an operator L multiplies the Wilson loop for this representation

by a nontrivial phase z (the essential feature of a vortex opera­

tor), then the commutators of L with certain gauge invariants can

fall off no faster than 1/p, where p is the distance from the vortex:

Note first that

( )W ( t!P ) , f>1
' V i ^ V i l ^ - ^ s . j ^ • "..k » .,ia "' ' « .. < l

- iir At)V„ ,.(t,?;t+t')4 ,(t) S^it-f) (4.2)

S»X o 1 J.
4 S pi

Here <f is a field of representation s and it is its canonical momen­

tum; the operators in 4.2 are then a canonical momentum at one point

joined by parallel transport along a path to a field at another

point. Equation 4.2 says that the commutator Joins two such opera­

tors into one whose path Is the sum of the first two. From this it

follows that a large Wilson loop of radius p can be written as the

multiple commutator of 2np/d short segments of length d (to join the

 69

last two endpoints we actually need to take the difference of this

multiple commutator with the same multiple commutator with all the

orders of n and f reversed). We can use the Jacobi identity to

write the commutator of our vortex-like operator L with this large

Wilson loop as the sum of 2irp/d terms each of which involves the com­

mutator of L with a single short segment* Since the total commutator

is of order 1 (that is, a phase, independent of p) the commutator of

L with the short segment can be no smaller than 0(d/p) as d becomes

small or p becomes large- The local operator IT .D <p . with D the

covariant derivative can be obtained from the segment operators as a

limit; its commutator can fall off no faster than 1/p.

The above argument relies heavily on canonical commutators,

which are always suspect, but we believe that the principle is

correct. Certainly it is correct for the lattice gauge theories we

will consider later, in the language of the lattice, when the quant­

ization condition is not satisfied the operators 2.4 and 2.8 intro­

duce a "frustration" on the surface S; uhis frustration can only be

44

spread out, never eliminated . The resulting operator is then asso­

ciated with a 3-surface, not a curve or 2-surface (it is not possible

to spread it out still further, into the fourth direction). Here we

will generally take this surface to be a fixed time-plane, but when

we discuss Green's functions we may take any closed 3-surface con­

taining C.

We now consider an Abelian theory again, with a single charged

scalar field <f of charge e and mass m, and with the gauge symmetry

unbroken. We look for an operator V (C) satisfying 2.1 (with e . «e)

p min
but with p non-integral. The spin of the charged field makes no
 70

difference—a fermion works out the sane way. Me remain, for a

while, in a single tine slice. For convenience, we will now take C

to be the t-axis, though everything we will do generalizes to any

curve« We would like to construct an operator which has its frustra­

tion spread out, and see if it is useful as an order parameter- He

will look for an operator which creates a minimum amount of energy;

that is, we will choose V to minimize

P
+
E » <v| H |v> - <0|V H V |0> (4.3)
v p p

where H is the Hamiltonian, |0> is the vacuum, and |v>«V |0>. This

energy will be of interest both as a guide in constructing V and in

its own right. If the 1/p falloff in the commutators implies a 1/p
2 ->

or 1/p falloff in the energy density <v|H(x)|v>. the total energy

per unit length will be infrared divergent. This would imply infin­

ite energy for an isolated Abrikosov vortex, and a long range

interaction between pairs of vortices.

The minimum of £ can be determined without actually construct­
ing V . We first smear out 2.1 near the z-axis, to remove the infin­
ite energy associated with an infinitely thin flux tube. The condi­
tion on V will be
P
w c v ex i a dx x f( 2)) (4 A)
v « ( 0 - < ') q p P<- ^ fc' i i P *
where p is again a cylindrical coordinate and

*<p > - ~ J
2
p>» (4.5a)
P
2) 1
f (
P ' ~Z P < a
(4.5b)
 71

The integral in 4.4 gives the winding number of C around the z-axis
when C" does not pass within distance a of that axis; this reproduces
2.1. Taking derivatives of 4-4 with respect to e, we find the commu­
tators of V with products of magnetic field operators:
P

VBp (^•••J _(x„> i ( 4 . 6 )

[B (t.)-m. (X\)J...IB (t )-m (£))v

1 1 m m
with

Note that M i s nonzero only for p<a. 4.6 and 4.7 a r e correct with B

and e e i t h e r both bare or both renormalized; we s h a l l take a l l quan­

t i t i e s to be renormalized u n t i l further n o t i c e . Equation 4.6 i s

equivalent to 4 . 4 ; i t implies the following c o n s t r a i n t on the s t a t e

|v>-.

<v| B o f j J . - . B . it )
1 a |v> ( 4 % 8 )

1 ID

- <0\ IB. (x.)-Hi, (£ V - • • [B 6? )4M, 6? ) |0>

1 , 1 1 . 1 2. m 1 m
1 1 m m

For m«l and 2 r e s p e c t i v e l y .

>
<v| B (x ) |v> - K (t)
1 ± (4.9a)

<v| B (x* )B (x* ) |v> - H^ZjM

1 1 2 (? )
2 <4.9b)

>
+ <0| B (x )B (x* ) |0>
i 1 j 2
 72

If 4.9a were the only constraint on the state |v> (together with

<v|^(x)|v>-0, which follows becauae the symmetry is unbroken), the

minimum energy would be given immediately by the effective poten-

,*0,41
tial

E v - U(? )M (4.10)

d i ( , , )
S.^rJ*
n»l V *n H.i 1 I l —Vl n ^

r
i , . . . i <« !• — •»,!>
1 n

where A M i s any vector potential such that 5»A ( x ) « M (x ) and

f*° , (x, x ) (4.11)

l n

a
- J*Ldt,...dt r , , ( x . , 0 ; x , , t _ ; . . . ; x ,t )
*-^D 2 n 1....1 1 2 2 n n
l n

In 4.11 the P are the 1PI Green's functions of the vector poten­

tial, and we have written out the time coordinate explicitly.

Satisfying 4.8 for all m requires a generalization of the effec-

42
tive potential . Equation 4.10, however, Is correct to the first
-2 0
two orders in e, e and e . To see this, consider adding a term

3 >
J*d x J (?)A (x ) i i

to the Hamiltonlan, with J. an external source chosen so that

<B(x)> .H(x) J (4.12)

where the subscript J Indicates an expectation value In the ground

state of the new Hamilconian. This ground state Is the state of

 73

minimum energy subject to the constraint 4*12 (or 4-9a) and its
41
energy is given by 4.10 • Such expectation values can be evaluated

in terms of Feynman graphs; the graphs for the matrix element 4.12

are shown in figure 9- The graphs for the matrix element

-> -> -> ->
<B(x)B(y)> are shown in figure 10. We see that 4.9b is satisfied
to order e
-2 and to e0, but fails at order e2 (note that ->
A „ and J.
M 1
are of order e~ >. To produce a state satisfying 4.9b, we would have

to add to the Uamiltonian a term of the form

3 3 >
Jd x d y K (?,'y*)A (x )A (^)
ij i j

2
with the bilinear source K, of order e ; this would only affect the
2
energy at order e . The same is true for the higher constraints.

Since the charged field contributes first at order e , we will

only study the energy to this order, and 4.10 is satisfactory.

Z
3 3 ->2 -> *? 1 - 4 4
E --jV*M (x) - * £j*\
v
2
— *\
6 ( oop(x x + e2
< W i <*i>-Vi " n":!.i n i'-- n> °( >
1 n I n

The order e term is from graphs with one charged loop propagating in

the external field A „ . The cutoff dependence from the Z, in the

zero loop term cancels the divergence from the 1 loop [Link] with n«2;

the effective potential is well known to be ultraviolet finite. The

four-dimensional Integrals in 4.13 can be taken either in Minkowski

or Euclidean space; this is just a contour rotation. It is con­

venient to take them in Euclidean space; the 1P1 Green's function

then falls off at least as exp(-m|x.-x | ) • We can write

 74

1 n X n
fr-'fn o^ PPl'F '--Fn
2

where

and x is the mean of x,,...,x . Because 8>1, the arguments of P on

the right-hand side are no closer than those of P on the left; P then

also falls at least as fast as exp(-m|x -x.I )• Inserting 4.14 in

r
i j max
4.13 and integrating by parts gives

\ - $&** H (x ) - 1 J±r J*\...d*

2
Vhjk
4
<' > 16

11*2

6 ( t ? ( ? ( + ( e 2 )
l > V l > V i , ^ 2 > — - V l D * n > ^ ; i ,^. . . i n< V " V °

(There is no surface problem with tne integration by parts, as can be

seen by considering the limit of a large but finite curve C ) .

M,(x ,) is nonzero only for x , within a distance a of the curve C,

and r falls off rapidly at large distance. Therefore, the integrand

is small when any of zne x. is far from the origin: there is no

infrared divergence in 4-16. In fact, the energy density in the

state can be shown to fall off exponentially. Similarly it may be

checked that the interaction between distant vortices due to the

electron sea is exponentially small (to this order in e ) .

Equation 4.13 is a perturbation series in p, which we have

learned to distrust. This particular series, however, can be shown

to converge for any p. The reason is essentially dimensional: each

 75

extra power of p in 4.13 comes with a;i extra propagator, 0(m ) , and

an extra power of A , which turns out to be 0(a ) , a being the

N
smearing radius. As long as we take a > -J~, the series converges;

since we are only interested in the infrared divergence of the

energy, we may take a to be as large as we like.

We now try to find a simple operator which satisfies 4.4 and

which creates the minimum energy state. Our first try is

3
V - exp{iJYl x ^ O O - t o ? ) } (4.17)

This is also the form of the vortex operator when there are i.o

charged fields at all. If we need to consider a specific fcrm for

A , we expect the rotationally symmetric form p8pf(p )/e tn have the

-»
lowest energy. With this choice of A„, the commutators of V, have
n l,p

the minimal 1/p falloff discussed earlier; with other c' ->ices they

fall off more slowly, at least in some directions. To calculate the

energy 4.3, we use

3
Vt B v. ^ •= K - i [ ; d x A ( ? ) - f ( ? ) , H J
v (4.\8)
l,p l,p * M

1 3 _ > - > - * -> 3 - » - » - » .>

-•jljd x A ( x ) - E ( x ) , tjd y A ( y ) - E ( y ) , H ]]+•••
M M

We will evaluate 4.18 using canonical commutators, so we must take

all quantities to be bare. The second term on the right-hand side of

4.18 can be obtained directly from the equations of motion; it is

_ 3 - » - » - » -» -»-»-*-»
J*dx B (x ) -M (x ) - j (x ) -A (x ) M (4.19)

with j the current. The third term of 4.18, with the double commu­

tator, can now be evaluated; it is

 76

3 2
Jd x -| if U) + e ^ O O ^ O O ^ U ) (A.20)

It is important that because of gauge invariance, the Hamiltonian is

not fully normal ordered (though it is permitted to add a constant to

H so that the energy of the vacuum vanishes), and so the product f*p

in 4.20 is not normal ordered* The triple and higher commutators

vanish, so the energy E of the state V. |0> is the vacuum expecta­

tion value of H (zero by definition) plus the vacuum expectation

value of 4.19 (zero by rotational invariance) plus the vacuum expec­

tation value of 4-20, which is

2
Ej - Jd x \ K lt)
3
+t*(t)<0\?*(x)t(t)\0> (4.21)

The vacuum expectation value of (»*(* is quadratically divergent,

so V, appears to be a poor guess (with charged fermions, the same

quadrati- divergence arises from a Schwinger term). The problem is

that it creates a coherent state; coherent states in interacting

20
theories generally have divergent energies . In particular, V.
l»P

creates a photon field, but the vacuum fluctuations of the charged

field are not correlated with the photons. To correct this we try

3 3
V, - V, exp{ijd x d y F U , ^ ) * * U )0(y) (4.22)

+ G(x ,y )j>(x )tt(y ) + G*(x ,y )?*(x )n*(y ) + H(x ,y )n*(x )ir(y )>

1»P

where n(x ) is again the canonical momentum and F, G, and H are unk­

nown functions. The energy of the state V, |0> is found to be

*,p
 77
3 2 2
E - J*d x <j£ <£)
2 + «AIH (l?)+H of)|A>> + 0 ( e )
0 1 (4.23)
0 > a n d
where |A> * A I

H (l?) - :it*6Orr0O +tt*(x)>tf(2)i

0 (4.24a)

K^t) - ieA^t/oo'oV?) - nt)tt*($-i] (4.24b)

2->2 -> * -> -»

H- and H are quadratic in the fields and are both of order e .

2
Interaction terms contribute at order e and have been dropped. U

is normal ordered, essentially because the energy of the vacuum is

defined to be zero, but H is not normal ordered.

The functions F, G, and H can always be chosen so that IA> is

1 a
the ground state of H + ' • we require a' A " A j» where the a
Q

annihilate eigenstates of H (plane wave states) and the a', annihi­

late eigenstates of H +H . The particular form of A depends on how

we pair the eigenstates of the two Hamiltonians (perturbation theory

in H gives a natural pairing). We will not determine an explicit

form forA> but will simply work out the ground state energy. We do

this using stationary state perturbation theory in H,; the complete

J.
expression for the energy i s 45

E _ < 0 1 H ( 1 + - 1 1 1 0 0 1 1 1 1 5 1 | 0 > 4 2 5
A l F^H " ** ! " ( - >
A o
wherp |0> is the free field vacuum. Equation 4.25 can be expanded

perturbatively in Hj. The first term is Just the vacuum expectation

value of Hj, which Is the quadratic divergence found earlier. There

is a second quadratic divergence from second order perturbation

 78

theory in the first tern of H.. These two quadratic divergences can­

cel exactly (we use Pauli-Villars regulators to tell us how to add

divergent terms) leaving a logarithmic divergence. This logarithmic

->2
divergence cancels against the divergence from the M term, which is

present because the e which appears in the denominator of M Is bare

in this calculation.

These cancellations may be verified by direct calculation. A

simpler way to see them, however, is to note that the first order

term is exactly the contribution of the graph 11a for the effective

potential, while the divergent second order term can be cast in the

form

- Jd x d y <0| A (y)-j (y) ^ - A ( x ) - j (x ) |0>

M M ( 4 . 6)
2

- - J"dx d y J^dy <0| A ^ ) - ? (t)

3 3
0 exp(-H y ) t^t)-t(t)
() 0 |0>

- -|Td x d y 6(x ) T<0| t^(t)-fiy)

4 4
0 t (Z)'t(x)
H \0>z

where rne subscript 1 indicates the interaction representation and j

is the free field current. Equation 4.26 is just the contribution of

the graph of figure lib to the effective potential* The cancellation

of the quadratic divergences of graphs 11a and lib is well known;

further, the M term in 4.21 is exactly the zero loot term of 4.13,

so the logarithmic divergences cancel as well. In fact, the whole

expansion of 4.25 can be cast in the form of the one loop effective

potential: V. creates the state of minimum energy, to this order.

The idea of V, can be extended to higher orders in e, but the form

of the operator becomes more complicated in each order.

 79

With additional charged fields present, the one loop contribu­

tion to the effective potential is just a sum of contributions from

each charged field* In particular, when there is a Higgs field (of

charge ne/p, n an integer, so that it respects the quantization) the

energy of a flux tube is just the energy of a flux tube without the

field of charge e, plus the one loop term found above. The operator

which creates the minimum energy state, to this order, is simply a

product of terms: the term V , times a term of the form of A for

l,p "

each field not respecting the quantization, times a term of the form

exp{—^-Jti x Sj (x)} for each field respecting the quantization.

Q As

before there is a surface S from the discontinuity of 9. (Does ^ go

over to this latter form as we vary the charge of a field from

unquantized to quantized? It depends on how we choose to pair the

states when we define A., but they can be so chosen) We should note

that when one of the fields which does not respect the quantization

acquires an expectation value, the minimum energy becomes infrared

divergent (the energy density goes as 1/p) and V no longer has any
1 z,p

nice properties. This is in accord with expectations: we would not

expect to have flux tubes then, even classically.

The commutator of V. with the Hamiltonian density can be

checked to fall off as 1/p, whereas the energy density falls off much

faster:
<0|V H ( ) V |0> ( A 2 7 )
2,p ^ 2,p - "P^T) '

The point is that the operator V, has been carefully tailored (by

the minimum energy construction) so that the state V |0> very

*tP
closely resembles the v( -uum state when we move away from the curve
 80

C. We might then expect that when we go to four dimensions and look

at the connected Green'B functions of V. , there would be no anamo-

2»P
lous clustering near the 3-plane t-0 (on which V. is defined) but
*»P

far from the curve C. This is in fact the case. For instance, we

have checked that in a QED-like phase, the connected Green's function

2
< 8(x,y;P) V, (C) > (see 4.1) falls off as [|P|/d(x,C)] in all

directions when the length of P, |P|, is snail compared to d(x,C).

This is the same behavior as when the quantization condition is

satisfied, and holds even though the commutator of 8 with V„ falls

2,p

off only as |P|/d(x,C). Similarly, in a Higgs phase, the same

Green's function falls exponentially when x is not near the minimal

surface. It can also bs seen that when there is a Higgs field which

respects the quantization and another field which does not, the vor­

tex operator still has a surface-like cluster property: The Higgs

field still has the singularity 2.37, so the uiscussion following

equation 3.3 still holds, and there must be a surface spanning C on

which the expectation value of the Uiggs field is zero and the expec­

tation value of the Higgs potential is greater than its vacuum expec­

tation value. Cluster properties and vacuum expectation values of

V in perturbative phases are then the same as for the vortex
*iP
operator in theories with only adjoint fields.

In order to have the correct commutation relation, we must have

3
V. - exp{iZ r d x X (^)."E (x*)} (4.28)
i,p J * r»n r

where A „ is cutoff independent: its curl is 4.7 with the denomina-

r,n
tor containing e • The Green's functions of V are then ultra-
r i ,p
violet divergent; to renormalize them requires adding counterterms to
 81

the exponent in all of space, not Just on the curve C. The operator

A provides just these counterterms: improving the ultraviolet

behavior of tt.3 energy has also improved the ultraviolet behavior of

the Green's functions. To show this, ve note that V„ can be writ-

2»P

V„ •= exp{ij"d x A ( x ) . E ( x ) + 0(p )} (4.29)

2,p » r,M r M
phys p h y s

where

* p h y g <?) -Z £ <?) + ? w )
3 r (4-30)

~f(t) » iej^du exp(iH .u)^*(lf)^(i )-^(lf)^*(^ )]exp(-iH u)

£
> ,
0 (4.31)

Performing the u integration, one sees the connection between equa­

tion 4.31 and the wavefunction correction in stationary state pertur­

bation theory . The higher order corrections in the exponent of

4.29 can be neglected because they are less singular (A„ has dimen-
->
sions of mass). T is bilinear in the charged field (the u integral

smears out the two charged fields) and so has the correct form 4.22.

The matrix elements of A will have a composite divergence from the

graph of figure 12a if F(x,y), G(x,y), or H(x ,y ) is too singular

-> ->
as y-»x ; this divergence has the form of figure 12b, which is the
->
same as the cutoff dependence in the first term 'if E , . We have
phys
calculated the ultraviolet behavior of the graph of figure 12a, and

find that its cutoff dependence exactly cancels that from the Z,.
->
E therefore has finite Green's functions as well as the finite
phys
commutator
 82

fE B ( )J 63( > <4 32)

phys,i<*>' r,j * "^ 4 *"^ '
x

The "physical field" £ , thus has certain desireable characteris-

phys
tics of the bare field (finite commutators), and certain desireable

characteristics of the renormalized field (finite Green's functions),

at the cost of being nonlocal- The Green's functions of our particu-

->
lar E . are divergent in the next order in e; it seems likely,

though, that a suitable operator can be constructed order by order in

e.

Equation 4.27, the exponential falloff of the energy density, is

not so much a property of the operator V. as it is a combined pro­

perty of this operator and the state |0>; as we have noted, V. has
*»P

been tailored to leave the vacuum almost invariant at large distance-

As we change parameters in the theory (such as charged field masses),

tne necessary V, will change. As an "order parameter" it might be

more appropriate to have one operator which we can apply to any

vacuum. This also is possible: we can simply take

V 3 ( ? - expdj^x t ( x ) . E
M p h y s O?)} (4.33)

where E . is chosen to have finite Green's functions and the right

_> ->
commutator, and to be of finite range. T (x ) as given by 4.31 does

not have finite range, but if we simply cut it off beyond some dis­

tance, it will be of finite range and still have finite commutators

and Green's functions. In a QED-like phase, the connected Green's

2
function < 8(x,y;P) V, (C) > falls off as [|P|/d(x,C)] , except
J
»P
within 0(m ) of the 3-plane t-0, where it only falls off as

|P|/d(x,C). Its Green's functions differ from those of V, , and

*»P
i

83

from those of the quantized vortex operator, by an additional piece

which is large only near the 3-plane on which the operator Is defined

and which falls off only as fast as the commutator. This Is what we

might expect for an operator not chosen to "fit onto" the vacuum

state just right. The same thing happens to V, in a Higgs phase:

->»P

the connected Green's function is still exponentially small away from

the 3-surface of the operator and away from the minimal 2-surface of

C (which need not lie in the 3-surface) but it now has a piece

0(|P|/d(x,C)) near the 3-surface. TK arguments requiring the Higgs

field to have a surface of zeroes and a large energy density on the

surface still hold. The connected Green's function of the Lagrangiar.

or Hamiltonian density with V is the same as In the case of the

quantized vortex operator, plus an additional piece near the 3-

surface which is of order d(x,C) ; the extra pitce is not of order

-1 -»
d(x,C) because <E . (x)L(y)> vanishes by C-invariance. The
pnys
integral of this over d(x,C) diverges logarithmically and is cut off

at the size of C, roughly P; the integral around the perimeter then

gives a total of 0([Link](P)), and the expectation value is the

exponential of this.

Sincu we do net have a perturbative model of confinement we must

be more heuristic vhen discussing a confining phase. Imagine a con­

fining theory, such as an Abelian theory with a magnetic monopole

Higgs field, to which we add a charged field which violates the

quantization for some of the vortex operators. Suppose that the

additional field affects the c/namics only weakly, as quarks are

believed to in QCD, so that we may expand in the number of loops of

the new field. Assume also that there is no absolutely massless

 84

plon, so that we nay talcs the loops to be localized In spacerime.

The new vortex operator, V, , is the old one times

exp{ij*d x A (x)'T(x))>
M Consider the vacuum expectation expectation

value of V, . Loops near C can only contribute 0(F)• Loops far

3,P
from the 3-surface (and therefore far from C) are unaffected by the

presence of the vortex operator (since the operator about which we

are expanding has short range clustering) and do not contribute to

the expectation value. Loops near the three surface but far from C

give contributions of the form

> (4 34)
^ V - A . V i A ' - ' V i <»„>< v^i>—*t ^ c -
1 n 1 n

This vanishes by C-invariance when n-1. When n-2 it is logarithmi­

cally divergent at large p, as A „ is 0(l/p), and for larger n there

i> no divergence. This is exactly as in the other phases, and leads

co the same P-ln(P) in the exponent. In the QED and confining phases

this piece dominates the usual perimeter term, but in a Higgs phase

the area term still dominates. In the same way, one may see that the

cluster property is [Link] short range plus a new piece near the t-0

plane, falling off slowly with p.

We appear to have a general rule for the cluster property when

fields in the fundamental representation are added: to the original

cluster property is added a piece which is non-zero only near the 3-

plane of the operator and which falls off rather slowly, like the

commutators. The cluster properties for the three kinds of phase are

still distinct. Note again that we may choose, for ease in identify­

ing the cluster property, to spread out the "frustration" on a 3-

surface other than t-0, so that we may take one which does not
1
i

85

contain the minimal surface. We are led to conjecture that in sys­

tems such as a superconductor or QCD, with the particles in the fun­

damental representation having only a small effect on the dynamics,

we can distinguish a Higgs phase from a confining phase based on the

expectation value or cluster property of our modified vortex opera­

tor.

There is a possible paradox here. As usually stated, the result

of 't Hooft is that either the vortex operator or the Wilson loop

must satisfy an area law. We have conjectured that in a QCD-like

theory, our modified vortex operator follows a P-ln(P) law, while the

Wilson loop in a confining theory with particles in the fundamental

representation satisfies a perimeter law because of screening.

Expresspj in terms of cluster properties, there is no paradox, 't

Hooft's result was that both loops could not have short range clus­

tering. The vortex operator here does not have short range cluster­

ing; spacetime near the t-0 surface does not look like vacuum. When

the Wilson loop in 't Hcoft's argument moves through this region, it

can give back the phase th3t it gets from the canonical commutator.

However, in QED-like and i-'iggs phases we were able to find an

operator which did not have this extra clustering. If this were als>

possible in a confining phase, there might then be a confict with 't

Hooft's result; the proof is too heuristic for us to be sure whether

there is. It may be that the vortex operators considered in this

chapter evade the proof automatically because of their nonlocal com­

mutators. We are inclined to believe that this is not the case, how­

ever. The proof considered only the Green's function <W(C')V(C)>,

with C lying entirely before or entirely after C. This Green's

 86

function depends only on the states <0|V(C) and V(C)|0>, not on any

other properties of V(C). Short range clustering for V(C) then

implies that these states are locally vacuum away from C; this should

be enough for the proof to go through.

We cannot be sure, then, whether short range clustering is in

fact excluded for any vortex operator when there are fields in the

fundamental representation, but we conjecture that it is* (In our

earlier construction of a vortex operator with short range clustering

(V ) , we had to treat the dynamics perturbatively; it is not at all

clear Lnat we could define such an operator for a confining vacuum.)

This leads to the related conjecture that it is not possible to

satisfy equation 4.8 (or its non-Abelian analogue) with a state of

finite energy in a confining theory. This would provide an addi­

tional means of distinguishing a confining theory.

Ws should mention that all of this chapter can be extended to a

non-Abelian theory. In that case, there are special problems with

defining a gauge invariant operator. In a weak coupling phase,

though, we can simply take all fields to be in the Coulomb gauge. We

can then define the vortex operator exactly as in the Abelian case,

except that we must add the "gluon color density" to the charge den­

sity in the exponent of V, and we must consider all of the fields and

currents in the exponent to lie in some arbitrary U(l) subgroup of

the gauge group. The perturbative analysis in this chapter then

changes only by the addition of a few color matrices, and the physi­

cal arguments and conclusions are all the same.

 87

Ji. A Lattice Model

We have recently begun to consider an Abelian lattice model in

which the above ideas apply and in which they can be tested in a more

precise way. Although our study of this model is far from complete,

we present some of the results. They support many of the conclusions

reached above, while opening up Interesting new questions about the

46
nature of order parameters and phase transitions .

The model contains an Abelian gauge field U defined on links,

a singly charged matter field f defined on sites, and a doubly

charged matter field $. defined on sites. The subscripts a, b, ...

refer to sites, so a link is labelled by two sites, its endpoints (in

a directed way), and a plaquette is labelled by its four corners.

The action is

S
" J P I 1.L 2,L>
S + (S + S
(4-35)

where the sums run over all plaquettes and all links respectively.

The gauge field action is

S p - -KRe(U ) p (4.36a)

U p is the product of link variables around the plaquette,

U .0. U ,11, for the plaquette abed. The matter field actions are
ab be cd da

S Re( U , )
. 1,L " -Pl
k 4 abS lb (*-36b)
ab

S R e ( U
2,L „ " - P 2 4 aW (*-36c)
ab

The variables U . , f , and p are complex numbers of modulus 1.

* *
U, , j», , and j>, , respectively, are their complex conjugates, not
 88

independent fields- The action is gauge invariant; the gauge

transformation is

U U ( 4 3 7 a )
ab - *a ab*b "

S'l.a'* V l . , <*-37b)

H,&-* «a*2,a (/
'- 37c)

where g is an arbitrary field of modulus 1.

With no matter fields (B -8,-0) this is a pure Abelian gauge

theory. On the lattice, however, this theory is not free and in fact

it is known to have two phases ' • When K is small (corresponding

to large coupling in a continuum theory) it is confining; when K

increases beyond a certain critical value K it changes to a Coulom-

bic theory, with a massless photon.

A vortex operator for a curve C can be defined as follows:

choose a surface S whose boundary is C, and identify all plaquettes

intersecting S but orthogonal to it. For instance, consider C to be

an infinite line in the 3-directioi. plus another infinite line run­

ning in the (-3)-dircetion and displaced in the 2-direction from the

first; S can be taken to be the flat surface (in the 2,3 direction)

lying between the two lines. Fart of this is drawn in figure 13a; we

have suppressed the 3-direction entirely, so that all the links and

plaquettes shown are perpendicular io the 3-direction, and the full

spacetime is obtained by stacking the figure on top of itself (in the

3-direction) an infinite number of times. The plaquettes we have

identified above are the 0,1 plaquettes drawn. The vortex operator
 89

is defined by replacing the action S with

iot
S - -KRe(e D )
p (4.38)

49
on the indicated plaquettes • It appears that this is a surface­

like operator, but by changing variables

iCf
U -» U e " (4.39)

on certain links, we can change the apparent surface S (see figure

L3b). Only the curve C is invariant. The expectation vaLue of the

vortex operator satisfies a perimeter law for both phases of this

49
pure gauge theory, for all or • Note that because there are no

charges as yet, there is a continuum of voitex operators 0<at<2ir.

The Abelian theory with either Bj or fi, non-zero, out not both,
36 37 38

was studied by Fradkin and Shenker and others ' . Even though

the matter fields are constrained to have unit magnitude, when 3 is

small they act like linear fields, of zero vacuum expectation value.

When S becomes large, though, the matter action is more and more

strongly constrained to stay near its minimum, and the fields become

Higgs-like. When there is a doubly charged matter field, the vortex

operator is only useful for or-n. It Is only for this value that

4-39 leaves the matter action invariant; for other values the vortex

operator is surface-like rather than looplike, and always has a

surface-like cluster property. When the matter field is singly

charged, there is no nontrivial value of or for which 4.39 leaves the

matter action invariant, and we can define no vortex operator. He

might then expect that in the doubly charged case we can make an

absolute distinction between Higgs and non-Uiggs phases on the basis

 90

of the cluster property of the or-n vortex operator, while in the

singly charged case this distinction nay be purely quantitative,

unless we find another order parameter. In fact, Fradkin and Shenker

showed that the doubly charged system has three phases, separated by

phase transitions, while the singly charged case has only two, with

the vacuum energy proven to be analytic between the regimes that

resemble Higgs and confining phases• The phase diagrams for the dou­

bly and singly charged systems are shown in figure 14.

The above work, relating to theories with only a single matter

field does not bear directly on what we have done* Restricting

attention to the ot-n vortex^ the doubly charged field is what we

have identified as sdjoint, and the singly charged field is what we

have called fundamental. Our analysis of the previous section only

applies to theories where the fundamental field is weakly coupled, so

that 3. must be small. When B, is small, we do not expect to see a

Higgs phase if there are no other matter fields. When p. is large,

we do not expect to see vortices even at the classical level, and the

analysis of the previous section does not apply. To test our ideas,

we must consider the gauge theory with both matter fields, with B

small.

The phase diagram for this theory is a cube, the three direc­

tions representing the couplings K, B , and B_. The face of the

cube on which £."0 is the doubly ciiarged theory (see figure 14

again); this face has a confining, a Coulombic, and a Higgs phase.

The arguments of the last section suggest chat if we consider a slice

of the cube at finite but small 3., we should still have three

phases: the transitions on the p -0 face shown in figure 14 should

 91

be walls that extend into the cube, not lines etched on the surface.

The small B region can be analyzed easily with the ideas of Fradkin

and Shenker, based on ideas of Hegner • If we expand the part of

the functional integral involving the singly charged field, we get a

sum over products of S for various sets of links. Each product

1,L
*
will involve various powers of 9. and 0. at each site- If the

powers of j>* and j» are not equal at each site, the functional

integral over $. gives zero. If they are equal on each site, the

integral gives 1, leaving Just the link variables from S, . The

condition from the (*, integration forces the remainii ^ link variables
4
to be formed into closed curves. The lowest term is of order 3.,

coming from the product S.l,ab. S, . S.

l,bc i,da, • The integral over -0,I
l,cd ,S,
leaves lust link product U . U. U U. : this term just changes the
J r J
ab be ca da
1 4
effective coefficient of the plaquette action from K to K+-gS .
Higher orders in 3. give traces of link variables around longer

curves. Very long curves are suppressed by large powers of 3 .

Fradkin and Shenker argue that these interactions cannot destabilizo

the phase transitions whet: B, is small. These arguments are based

more on lore fhan on rigor, but phase diagrams derived from them have

been verified nicely by numerical work • The [Link] is that for

small 8, the effect is just a shift in the effective coupling; as we

take our slices in the cube (at small fixed 3.) we see the same

structure as on the face, shifted to the left by the change in the

coupling. The small 3, expansion on the lattice closely resembles

our continuum arguments about the expansion in number of loops of the

fundamental field, though on the lattice there is at le<?st the hope

of more rigor.
 92

What is our modified vortex operator on the lattice? The defin­

ition 4.38 will no longer do. 4.38 replaces K with -K on the links

orthogonal to S (for or"n); we can change S by changing variables of

1 4
integration. The integration over the singly charged field adds -gji.

to the coefficient on every plaquette (the matter action doesn't know

14 14

about S ) . The coefficient is then K+-gB? on most links but -K+-gpT

on S. These differ in magnitude and not just sign, so no change of

integration variables can now shift S. The vortex operator is sur-

facelike (as we also argued earlier). Nor can we remedy this simply

by changing the sign of the matter action on certain links: that just

moves the problem elsewhere.

We can only spread the problem out: pick out some 3-surface X of

links, in which S lies. Such a surface is illustrated in figure 15.

On these links, change the charge 1 action to

S 5
'l,L K " - ^ ' " ^ t a V l b <4.40)
ab

where 9 is illustrated in figure 15, and the discontinuity of the

exponential is defined to lie on the surface S. When we now change

variables on some set of links, changing the apparent surface S in

the gauge field action, the apparent discontinuity in the matter

action moves as well. X may be any 3-surface containing C; S may be

any 2-aurface bounded by C and lying in X. No change of variables

will change the surface X (it is distinguished by the fact that the

product of the arguments of the "Re" in the matter actions on the

four links of a plaquette does not equal the argument of the "Re" for

the gauge field action on the plaquette); the operator depends on X

but not on S.
 93

When we integrate over the charge 1 field, the effective gauge

action that we obtain is unchanged on plaquettes not lying in X' On

plaquettes such as A, shown in the figure, the effective action

becomes, to leading order in p.,

2
- ReKK4~^exp{i y->)U ] p '4.41)

+ ) t e ( I1 C0 :i ( + J s i n § l l B (
- « iPi V - aPi - ^f "" V 8i ? f V
On plaquettes such as B which lie in S, the effective action is

exactly the negative of this: the K term has the opposite sign by

definition, while the induced terms have the opposite sign because o
16/ 2
the discontinuity of e .

The expectation value of our modified vortex operator in the

full theory is then equal to the expectation value of the operator

exp< I e > p

P*X

times the usual vortex operator, in the effective gauge theory after

Integration of the charge 1 field. •„ stands for the second and

third terns on the right hand side of 4.41. At small 6., the expec­

tation value of the exponential is the exponential of the expectation

value. The expectation value of In(U ) is zero by C imparlance (com­

p

plex conjugation here); the expectation value of Ke(U ) is some con­

stant. G-e' is of order a/p, so the integration over p is logarithm­

ically divergent as in the continuum case. The expectation value of

the modified vortex operator is then cxp(-[Link](F)) times the expecta­

tion value of the normal vortex operator for the effective gauge

theory. The total expectation value therefore changes from

 94

exp(-F-ln(F)) behavior to ezp(-A) behavior on just those curves where

the effective gauge theory (and the full theory as well) change from

confining to Higgs-like. The cluster properties are also the same as

expected from the continuum case. It appears that we have found a

good order parameter for the small B. region. We have not yet stu­

died the conjecture about finite energy states which arose from the

paradox with *t Hooft's result. This is probably studied most

readily in a Hamiltonian lattice theory.

As we travel into the cube which represents theory space, do the

phase walls we have found ever just end, so that the different phases

can be connected by a continuous path through the middle of the cube,

or are some of the phase regions completely walled off? Our modified

vortex operator has different behavicr in the Higgs and confining

regions; on any path connecting them its behavior must change discon-

tinuously. If we have a "good" order parameter, this will also imply

a nonanalyticlty in the vacuum energy and Green's functions. What

makes a good order parmeter is not well understood; one can find many

examples of an operator whose behavior is discontinuous even though

the physics Is not: the phase transition is "in the operator", not in

the vacuum. As a simple example, consider a scalar theory whose

2 2
Lagrangian density is L,(x)+m 0 and which has a phase transition at
2 2
• -E . As an "order parameter" take

2
exp{kfd*x * (x)>

The expectation value of this operator will not be discontinuous at

2 2 2 2
m -m , where the vacuum is discontinuous, but at m -m +k. This is a
c c
rather trivial example, but there are much more subtle ones; these
 95

"bad" order parameters disturb the vacuum rather than diagnosing it.

Bad order parameters tend to be very ncnlocal, as was our simple

example. Since our modified vortex operator is more nonlocal than

the usual vortex operator, we must be suspicious of it.

We have started to look at the behavior of our operator at large

p . On the (} »0 face, shown in figure 14, the behavior of the

theory is well known; there is a region of analyticlty connecting

p.-0 with B •co. When B, becomes very large, the functional integral

for the expectation value of the vortex operator is dominated by the

configurations which minimize the charge 1 action, 4.40. One such

configuration is given by j» «1 everywhere, V "1 for links not on X,

-18/2
and U -e for links on X (^ is decoupled). This gives 4*40 its

minimum value, -2, everywhere. The gauge field action, 4.38 on S and
2
4.36a elsewhere, falls off as 1/p on X for this configuration. The

action of this configuration is not large on any plaquette; the total

action is again 0(P>ln(P)). Our modified vortex operator docs not

have surface-like behavior anywhere near 8 ^ o , p,"0; its behavior

appears to be continuous In the region in which analytlcity has be^n

proven.

The acid test for our operator comes on the &*<x> face.
2 This

face is best understood in the f »1 gauge, where the charge 2 action,

4.36c, constrains U to be 1 or -1 on every link. We then have a Z,

gauge theory coupled to a D(l) matter field (the charge 1 field),

which Is still a fundamental representation in our terminology.

Because the matter field is fundamental, we might expect thia face to

have a phase diagram like the 3 -0 face, with a phase In the small

B,, large K corner having free 2 . excitations (this would still be a

 96

Biggs phase from the point of view of U(l)) and one continuous phase

elsewhere. This is shown in figure 16. Near the left side of the

lower edge of the diagram, we can integrate out the charge 1 field

and we have simply a Z_ gauge theory at small K. This is a confining

49
theory ; the vortex operator satisfies a P-ln(P) law. Near the

j3,^o face, where we must minimize the charge 1 action, we can no

longer find a configuration which makes the matter action small

everywhere; our earlier configuration required U to take values other

than -1 and 1. Any configuration will have at least a surface of

plaquettes (bounded by C) on which the action is large. Our operator

must therefore make a discontinuous change to surface-like clustering

somewhere between edge 1 and edge Z.

This leads us to a closer examination of the 3-"oo face. When

the gauge field and the matter field are the "same size", U(l) and

U(l) or Z. and Z., the phase diagram of figure 16 is known to be

correct, but here the matter field has more structure than the gauge

field. Me therefore look at the K«0 edge, where only the action S
1,1.

survives (the whole effect of the charge 2 field is the constraint

U«+l). Summing over U on each link, we find that the functional

integrand (leaving out the vortex operator now) is a product of a

factor of

exp{-I > - exp<-p Re(?* ^ )> + e x p ^ R e ^ * ^ ) )

ab 1 a lb (4.42)

for each link. This factor is invariant under a, -»-a, : there is a

r r
la la*

surviving Z. gauge invariance. The action is therefore a function on

U(l)/Z , which is just a different U(l); we write it in terms of a

new field 0 defined by

 97

exp{ifl > - ^
a (4.43)

The action is now a sum of terms for each link

© ••©
a
I a b - - ln{2co»h(p cos- ^))
l (4.44a)

We shall see that this is a close approximation to the action of the

xy-model 52,

J 4 4Ab)
ab " " P V ^ W <'
AC small p.,

2
I a b - const - | p j C o s ( - ^ ) + 0(p*)

- const' - | ^ c o s ( » - e ) + 0(8*)
a b (4.45)

1 2
This is an xy model at j3 "Tpi-

At large j3 , the action grows quickly away from its minimum,

2 —1
and (8 - e ) is constrained to be 0(8, ) . The argument of the

hyperbolic cosine is then very large, so it is essentially an

exponential. He then get

2 + 0 ( < e 4 ) ( 4 4 6 )
- const + I p ^ V V Pl a"V '

Performing the same expansion for the xy model at large 8 , we find

that J is the same as 4.46 when 6 "B,/4.

At small 8,, the K-0, 8_"OD edge is an xy model at small B ,

which is known to be in a disordered phase; at large ft., it is an xy

S2
model at large 3 , which is known to be in an ordered phase . We
 98

conclude that there is a phase transition on this edge. We have not

yet studied the stability of this transition at finite K; we have no

reason to believe, however, that is does not join with the other

phase boundary, giving the phase diagraa of figure 17. We note that

the phase we have labelled "total Higgs" is walled off on the K*co

face (seen in figure 17) and or the p,-ro face (shown in figure 16).

The K-0 face is still terra incognita to us. There are known to be

phase transtions on exactly two of its edges, but we have not yet

analyzed the stability of these transitions away from the edge.

Our operator seems to be a good vortex operator for the whole

cube, or at least as much of it as we have been able to study. Is

there a physical basis for this? He believe that strong evidence for

it is provided by the nature of the discontinuity of the operator.

If the phase transition were "in the operator" we might the action

density to change discontiuuously within the support of the operator.

In our case, if we choose X to lie away from the minimal surface, we

see a discontinuous change from clustering (action density) near the

3-surface X to clustering near the minimal 2-surface* This seems as

if it should reflect a true change in the nature of the vacuum.

What we have found indicates that the "free" and "confining"

phases, with P-ln(P) behavior for our operator, should be totally

walled off from the "Higgs" (also known as "free Z " ) and "total

Higgs" phases, with area law behavior. It aay be difficult to show

this analytically, as the center of the cube is amenable to few

approximations; numerical studies should provide the answer. What we

can do, at the least, is to examine the stability of our new transi­

tion against small nonzero K and large non-infinite B,.

 99

Everything we have done has been directed at four dimensions,

but it can all be extended to three or to more than four; it vould be

interesting to study the phase structures of these theories as well.

Ve note that our result iapliea that the corresponding operator in

QCD should also be able to diagnose phase changes; it should enable

us to say that QCD is not in a Hlggs phase, and that any pith through

parameter space which bring it to a Higgs phase oust pass through a

phase transition. A careful study of this problem might reveal some­

thing of the nature of phase transitions and the meaning of order

parameters.
 CHAPTER 5

CONCLUSIONS

In this thesis, we dealt first with some technical aspects of

the vortex operator. Our resolution of the Dirac string problem, and

our two dimensional "Dirac lump", nay be more novel than they are

useful, at least for the present problem where the methods of Wu and

Yang can be used. For a field theory of electrons and aonopoles, It

may be helpful to use ideas akin to ours, as the monopole is no

longer classical and the method of writing a monopole field theory as

a sum over monopole paths is rather formal* He have also shown that

the divergence? of looplike operators can be analyzed in a straight­

forward way by the use of the operator product expansion* We believe

that any attempt tc treat QCD as a theory of looplike operators (Wil­

son loops or vortex operators) must include a careful treatment of

short-distance questions, along these lines*

In the next chapter we discussed the relation between the

Green's functions oi the vortex operator and the structure of the

vacuum. We have placed emphasis on the cluster properties of these

Green's functions as providing more detailed information than the

vacuum expectation value. We then showed two correct ways (diver­

gence equation and tree sum) to find the cluster property in an

Abelian Higgs phase, and one wrong way (expansion in p ) . Our discus­

sion of the graphical expansion in a Higgs phase emphasized the fact

that the Higgs phase differs not in the form of the expansion but in

the nature of the tree sum; the extension of this should be borne in

mind when dealing with spontaneous breakdown through a composite

100
 101

operator.

In the fourth chapter we showed that the idea of the vortex

operator can sometimes be extended to gauge groups with fields in the

fundamental representation, groups which are slaply connected. Con­

tinuum arguments indicated that this is true vhe.i the field In the

fundamental representation enters weakly into the dynamics. Study of

a lsttice model confirmed this, and also suggested that our operator

is a good order parameter at all values of the couplings. More study

of the physics here is required, but it aetjuB possible that the

analogous operator would distinguish the QCD vacuus from a Biggs

phase with quarks in a qualitative rather than merely quantitative

way.
 102

REFERENCES

1) One of nany recent reviews is T. F. Walsh, DESY preprint 80/48

(1980)

2) S. Mandelstam, lecture at "Common Trends in Particle and Con­

densed Matter Physics" Les Houches, 1980; ENS preprint LPTEK3
80/5

3) A recent review of the phenonenology is in E. Witten, Nucl.

Phys. B160 (1979) 57

4) 1 would like to thank Stuart Raby for a helpful discussion.

5) T. Applequist and R. Shankar, Phys. Rev. Sid (1978) 2952. 1

would like to thank Curt Flory for helpful discussions.

6) I. Affleck, Phys. Lett. 92B (1980) 149

7) J. Kogut and L. Susskind, Phys. Rev. Dll (1975) 395

8) K. Wilson, Phys. Rev. D10 (1974) 2445

9) J. Kogut, R. Pearson, and J. Shigemitsu, Fhys. Rev. Lett. 43

(1979) 484

10) M. Creutz, Phys. Rev. Lett. 43 (1979) 553

11) S. Elitzur, I.A.S. preprint (1979)

12) S. Coleman, Comm. Math. Phys. 31 (1973) 259

13) R. Savit, Rev. Mod. Phys. 52 (1980) 453

14) S. Mandelstam, Phys. Rev. D19 (1979) 2391

15) G. 't Hooft, Nucl. Phys. B79 (1974) 276

A. M. Polyakov, JETP Lett. 20 (1974) 194

16) T. T. Wu and C. N. Yang, in Properties of Matter under Unusual

Conditions, ed. H. Mark and S. Fernbach (Intersciece, New York,
1969) p. 349

17) A good review is P. Goddard and D. I. Olive, Rep. Prog. Phys. 41

(1978) 1357

18) G. 't Hooft, Nucl. Phys. B138 (1978) 1

 103

19) There is no reference 19.

20) For a discussion of homotopy groups (TL) and their relation to

monopolss and vortices, see S. Colemau, in New Phenomena in Sub-
nuclear Physics, ed- A. Zichichi (Plenum, New York, 1977), p.
297, or reference 17.

21) H. B. Nielsen and P. Olesen, Nucl. Phys. B61 (1973) 45

22) S. Coleman, Phys. Rev. Oil (1975) 2088

23) For a discussion of T and T* products, see K. Nishijima, Fields

and Particles (Benjamin, New York, 1969) section 5-5.

24) R. Feynman and A. Hibbs, Quantum Mechanics and Path Integrals

(McGraw-Hill, New York, 1965)

25) See, for instance, the review article by I. Gelfand and A.

Yaglom, J. Math. Fhys. 1 (1960) 48

26) R. Brandt, F. Neri, and D. Zwanisiger, Phys. Rev. D19 (1979) 1153

27) T. T. Wu and C. N. Yang, Fhys. Rev. D12 (1975) 3845

28) I would like to thank Professor K. Bardacki for helpful discus­

sions. The work of this section was done befcre the appearance
of references 29 and 30.

29) J. Gervais and A. Neveu, Nucl. Phys. B163 (1980) 189

30) A. M. Polyakov, Nucl. Phys. B164 (1980) 171

31) R. Zimmerman, in Lectures on Elementary Particles and Quantum

Field Theory, e<J. S. Deser, M. Grisaru, and B. Pendleton
(M.I.T., Cambridge 1970), vol. 1

32) S. Weinberg, Phys. Rev. 118 (1960) 838

33) Equation 2.50 was found independently by S. Coleman (U. C.

Berkeley seminar) in the context of the non-renormalization of
the product e g where g is the charge of an external magnetic
monopole. In this context the operator product arguments are a
demonstration that one can in fact aake all Green's functions
evaluated in the prescence of an external monopole finite by
renormalization of the monopole's charge.
 104

34) V. N. Grlbov, Leningrad preprint and SLAC-Trana 176 (1977)

35) G. 't Hooft, Nucl. Phys. B153 (1979) 141

36) E. Fradkin and S. Shenker, Phys. Rev. D19 (1979) 3682

37) T. Banks and E. Rabinovici, Nucl. Phys. B160 (1979) 349

38) L. Sussklnd, Phys. Rev. D20 (1979) 2610

39) J. C Taylor Gauge Theories of Weak Interactions (Cambridge,

Cambridge, 1976)

40) E. S. Abers and B. W. Lee, Phys. Rep. 9C (1973) 1

41) S. Coleman, in Properties of the Fundamental Interactions, ed. A.

Zichichi (Editrice Composltori, Bologna, 1973) p. 605

42) J. M. Cornwall, R. Jackiw, and E. Tomboulis, Phys. Rev. D10

(1974; 2428

43) D. Saint-James, G. Sarma, and E. J. Thomas, Type II Superconduc­

tivity (Pergamon Press, 1969)

44) I would like to thank Marvin Weinstein and Helen Quinn for a
helpful discussion.

45) G. Baym, Lectures on Quantum Mechanics (Benjamin, New York,

1969)

46) I would like to thank Steve Shenker for very helpful discus­
sions.

47) K. Osterwalder and E. Seller, Ann. Phys. (H. Y.) 110 (1978) 440

4S) A. Guth, Phys. Rev. D21 (1980) 2291

49) A. Ukawa, P. Windey, and A. Guth, Phys. Rev. D21 (1980) 1013

50) F. Wegner, J. Math. Phys. 12 (1971) 2259

51) M. Creutz, Phyr Rev. D21 (1980) 1006

52) R. Griffiths, in Phase Transitions and Critical Phenemena, ed.

C. Domb and M. Green (Academic, New York, 1972), vol. 1

53) In fact, the exact expression for V, is 4.29 with only the
*iP
first term in the exponent, but with the whole exponential u-
ordered. I would like to thank Professor S. Mandelstam for
pointing this out.
 105

FIGURE CAPTIONS

a. The graph for A (x). The double line represents the

surface S.

b. A typical graph for G(P,C,S). The curved line is the path F

of the line integral; the straight segments are scalar propaga­
tors •

a. The order e graph for < A (x) V *(C,S) > / < V '(C,S) >.
The heavily circled v's arc effective vertices, the lightly cir­
cled v's indicate effective propagators*
3
b. An order e graph for the same matrix element*
a. The discontinuous graph for < V '(C,S) W ( C ) >. The single
heavy line represents C*.

b. Another graph for the same Green's function*

c. Another graph, connected to C by three photons.

Graphs in the expansion of figure 2a.

a. The two graphs with field renormalization divergences*

b* Two of the graphs which have composite divergences*

a. The curves C and C , linked and separated by a large

distance E.

b* C and C* unlinkad and separated by K*

The Green's function 3*3, considered in a 3-surface. C and TJ

are the intersections of C with the 3-surface; S is the inter­
section of S with the 3-surface. The function 3*3 maps loop 1
into a nontrivial path in H; it maps loop 2 into a trivial path.

The graphical equation for u., represented by the u-tadpole; a

is the a-tadpole* The first two terms cancel*
vS
The graphical expansion for u. , represented by the v-tadpole*
->
Graphs for <B (x)>.

a* The order e~ graph.

 106

b. The order e series.

10. Graphs for <B(x,)B(x-)>

-2

a. The order e graph.

b. The order e series.

c An order e graph not satisfying A.9b.

2 ->
11. Order p graphs for the effective potential V(A >. H

12. a. A graph for a matrix element of A. with a potential composite

divergence.
b. The form of the divergence of figure 12a. The X represents
a cutoff dependent function.
13. a. The lattice vortex operator: the plaquette product Is multi­
plied by a phase on the indicated plaquettes. The whole picture
lies in a 3-plane perpendicular to the 3-direction. C is an
infinite line in the 3-direction plus an infinite line in the
(-3)-directlon; these lines intersect the pictured 3-plane only
at the points C and Z- The surface S intersects the 3-plane in
a line, shown running through the middle of the plaquettes. The
pictured plaquettes lie in the 0,1-direction; C and V are
separated in the 2-direction.

b. Changing variables on the nine heavily drawn links, we

change the apparent surface S.

14. The phase cube for our lattice theory; the little dreibein shows
the directions in which the various couplings increase. The
upper face, with 3,"0, is the doubly charged theory; B_
increases as we go upward on this face. The lower face, B?™"*
is the singly charged theory; p, increases as we go downward.
The phase diagr&as for these two theories are found in reference
36 or 51; the shaded region is where analyticity has been pro­
ven. The face itao is seen obliquely.

15. Part of the place of links X on which we spread out the frustra­
tion in the singly charged matter field. The plaquettes of S
are drawn in. The arcs indicate the angles C and 6'; their
difference is 0(l/p). The angles are defined to be
 107

discontinuous on S; the action induced on B i* therefore the

negative of that induced on A.
f a c e
16. The cube of figure 14, tilted forward to ahow the p,"oo -
with one possible phase diagram. Our operator has P-ln(P)
behavior near the left aide of the lower edge, 1, and area law
behavior near the upper edge, 2.

17. The X indicates the phase transition we have found on the

K-0, p -a>
2 edge; the daahed line is its expected continuation.

18. The cube tilted forward again, showing the f>,*D face. The K-0
face, not yet fully analyzed, la also exposed; a plausible phase
transition is shown.
 108

°)

b)

Figure 1
 109

b)

Figure 2
 110

a)

b)

c)

Figure 5
 Ill

Figure ft
 112

Figure 5
 113

2
O

Figure 6
 Ill*

M
«? <? +

oo
+ I*
n=o

Figure 7
 115

CO

* -I*
n=o

Figure 8
 116

a)

b)
Figure 9
 117

•J

J
a)

T J
+
X2 J
J
b)

j*s~ J

X
C) 2
Figure 10
 118

a)

V
^M M

b)

Figure 11
 119

a)

b)

Figure 12
 120

a)

b)
Figure 13
a

«*A*<
 122

--»

Figure 15
 123

/?,= fi = «>
2

/ / V /
n\ %
A=A=o

Figure 16
 121*

£,=A>=»

TOTAL X %
e
it HIGGS N
«? k X
- ^ . 7s
i
o II
o O \ FREE 8
IJ
^ ° 1
1
zz 2
fsj

^ 1 •
111
UJ o
ii

c\ ^ \HIGGS \
/ / V
<%\ . ^ \ F R E E '^
fir^°

Figure 17
 125

A
7U
K
A*

A=» jQtrO

*t CONFINING
8
II
o
(I II
TOTAL HIGGS
ii

y ^ T O T A L HIGGS /&
foNRNTNG J F f t E E Z , ^
s 0 =£
A A> ° ^

Figure 18