A

SLAC-PUB-2372
December .979
CT/E)

ELEMENTS OF QUANTUM CHROMODCTAMICS

J . D. Bjorken

Stanford Linear Accelerator Center

Stanford University, Stanford, California 94305

the

Summer Institute on Particle Physios,

Stanford Linear Accelerator Center, California

July o-io, 1979

Work supported by the Department of Energy under contract number

DE-AC03-76SFG0515.
 MASTER- > -
TABLE OF CONTENTS

iter

\, Introduction 4

2. The solution to QCD 10

3. Quantization: Introducing the Femtounlverse 21

4. Instantons 44

5. Pure QCD: Gluonlum Phenomenology 53

6. Inclusion of Heavy Sourcce: Onluns Phenomenology •• &B

7. Inclusion of Light Fermlons *3

8. Ideas About Confinement 106

9. Alternatives to QCD 109

10. Conclusions 110

Appendix

A. Causa' Law, Constraint Equations, and Gauge Fixing

Ambiguities 112

B. Instantons In Cure QCD 116

. DISCLAIMER —

MSTOIUTOII M ' i : . ;
%
 ACXNWLEDGEMENT

Before beginning, 1 wish to acknowledge the Invaluable and generous

help of Marvin weinstein In Introducing me to much of the material

covered in these lectures. While there are many shortcomings of these

notes, they are not to be blamed on hlia{ they would have been much

poorer without his help and insights.

 - A -

1. INTRODUCTION

Quantum chromodynomics should need little introduction, since it

already permeates almost all descriptions of strong interaction phenomen

nowadays. We recount here very briefly the basic motivation, beginning

with some factual evidence:

1. Quarks of fractional charge and three colors seem to be require

as constituents of hadrons in order to understand the spectrum of hadron

and their resonances,

2. The ratio of electron-quark to neutrino-quark deep Inelastic

scattering argues strongly for fractional charge of the quarks.

3. In addition to the spectroscopic evidence, the observed width

of the decay n •+ 2y and the large cross section for e e •* hadrons ii

successfully understood provided there are 3 colors of quarks.

4. Tha color-symmetry should be exact (or very nearly so); other­

wise we would expect additional low-lying color non-singlet hadron state

states for which there ia no empirical evidence.

Based on this evidence, an analogy between color and charge is an

attractive one. Just a« the conserved charge is closely related to the

electromagnetic force, one may search for a strong force related in a

similar way to the conserved color quantum-numbers. The most imediate

answer to this, and the one most similar to quantum electrodynamics (QED

is quantum chromodynamics (QCD). The beginning of the analogy between

QED and QCD is exhibited in Table 1.

In that table the bold-faced quantities are 3 > 3 matrices acting

on column vectors of quark fields

 QCD

Conserved quantum 3 colors

charge
numbers

Symnetry group SUC 3)

Transformation _ ( (i/2>i«».
property of s o u r c e e ** e iQ6
^ e
( e l e c t r o n or quark) i,j-1,2,3

Current d e n s i t y A
•|q % i < i ; A-i,2,.

Quanta of f o r c e
m a s s l e a s photon e i g h t massless gluons
coupled to c u r r e n t s

6
Field-variables
(potentials)

Caige i n v a r i a n t a s u 3 3 u
s u b s t i t u t i o n (when -r^- * •—- - ieA*(x) -r~ - jr- - ieA (x)
X u
s c t i n g on quarks) K V P

q<«) - [ q C«)
2

[x) (l.D
The matrices J A are the 8 independent 3>(3 hennitian traeeless matrices

of Cell-M*nn, and are exhibited in Section 3.

The comparison in Table 1 could go on and on. That in fact will be

the case throughout these lectures. For now it suffices to say that QCD

in the weak coupling limit can be formulated in a way very similar to

QED, Wt may «ak whether, once having that formulation In hand, it pro­

duce! any useful results. The answer is yes; it appears Co provide a

consistent description of those hadron processes that depend on short

distances only, as well as provide some justification for the applicability

of parton-model ideas to hadron processes. The list of these includes

tha value of R(e e •*• hadrons) and the approximate scaling behavior of

deap-lnelastic lepton-nucleon structure functions. In addition the

deviations from scaling axhlblt the qualitative behavior expected from

the theory. Some claim the significance is precise and quantitative,

but ue shall not raise that question here.

An Important property of QCD that distinguishes it from QED Is

"asymptotic freedom." In QED, the effective charge at short distances

becomes larger as a consequence of vacuum polarization. If for QED we

write in •omentum space the force between static charges as

2 C1 2)
V q ) - «*"<ft
C -

Z 3ir
o(q ) a(m*) o*

Thus at sufficiently short distances (but In practice ridiculously short)*

QED becomes s strong-coupling theory. As we shall discuss in more

detail later, In QCD the vacuum-poi-arization has the opposite sign; the

corresponding equation is

1 I (33-ZN ) f „Z
(1.4)
2
• - V + log ^
«(q ) a(M ) 12 n H

where N, (-3 or 4) is the relevant number of quark flavors. Thus at

-14
sufficiently lar^e distances (large compared to ID en) QCD becomes

a strong-coupling theory. This is both good and bad news: good news
because something non-perturbatlve is needed at large distances to pro­

vide a mechanism for confinement of color: neither the gluons nor quarks

which appear in the field equations appear In the spectrum of hadrons.

What exists Is only a hypothesis that only color singlet combinations

of quarks and gluons can exist as isolated particles. The bad news is

just that we don't know how to calculate with — and even to formulate —

the theory on a large distance scale.

It Is therefore a serious question whether QCD Is a theory at all.

Let us compare the situation with QED. There one can follow textbook,
1
approaches to the subject — specifically two particular textbooks,

hereafter to be known as Book 1 and Book II. The Book I approach uses

the classical equations of motion and common sense to motivate rules for

Feynman-dlagrama. The emphaais is on leam-by-doing and Intuition, with

a relatively casual attitude toward a strict systematic logical develop­

ment, X think most of contemporary perturbative QCD does not get far

beyond this level. Beyond the relatively rock-solid predictions to which

we already alluded, there do not exist clear rules which divide phenomena

dependent on the non-perturbative, confining part of the theory from

phenomena dependent on the purely perturbative aspect.

In QED, the Book II approach is the strict logical one of canonical

quantization of a classical field theory. In the presence of interactions,

but in the weak-coupling regime, the LSZ formalism car be used to relate

asymptotic states of physical, isolated electrons and photons to the

field variables. However, the formalism relics rather heavily on these

asymptotic scattering states — In other words, on large distances. There­

fore this approach appears to bf closed in QCD. Actually as we shall

 - a -

see, the situation is not quite that bad. However, X think perturbative

QCD is much mora restricted than QED, for reasons that hopefully become

clearer as we proceed.

These lectures will concentrate for the most part on the Book II

approach. Tht.^ is not meant to imply disparagement of the Book X method;

2
Stan Brodsky will most ably cover that. Nor is it meant to downgrade
3
other approaches, in particular those based on path-integral formulations.

Indeed) use of path-integral techniques has thus far been the most success­

ful mode of attack on the difficult mathematical problems posed by QCD.

The rules for- diagrams are most efficiently derived in that way (especially

in correctly accounting for the subtleties associated ,.lth Faddeev-Popov

ghosts). Also the demonstration of confinement in strong-coupling lattice

QCD acid the studies of instantons are best done within the path-Integral

formalism. Our rationale for avoiding it here is simply to look at the

subject in a slightly different way [after all, some problems, such as

the hydrogen atom, are more difficult using path-Integrals]. as well as

to spare the less theoretically oriented reader the unavoidable preliminary

technology needed to set up the path-integral formalism. In any case,

the problem confronting everyone is difficult; all attacks should be brought

to bear.

We shall classify the subject-matter into three stages oE increasing

complexity. In the first we consider "pure" QCD in the absence of any

ferr Ions or other sources. The QED analogue is the (trivial) theory of

free non-interacting photons. For QCD this Is presumably the (nontrivial)

theory of interacting gluonium (color-singlet bound glue) states. The

second stage allows the introduction only of superheavy quarks in addition

 - 9 -

to the pure glue, with emphasis on the non-relativistic limit of the

quark motion. Tn QED this is not much more than the theory of the Coulomb

interaction — In other words, nothing but all of chemistry. In QCD,

this such should already allow study of basic questions of confinement;

e.g., the nature of the static potential between heavy quarks at large

distances. Only In the third stage will we introduce the light quarks

u, d, s. The presence of copious vacuum polarization and pair cre~-ion

•hould modify the structure of the theory in a major way, Nevertheless

It can be hoped that these modifications have less to do with the exist­

ence of confinement, and more to do with the nature of the spectrum of

confined hadrons.

These lectures will be organized as follows. In Section 2, after

caking some assumptions about the nature of QCD, we shall describe what

v« think the solution of the theory looks like, first for Stage I, and

than for Stages II and III. This will not even bo at Book I level — call

it Book Zero. For the most part wo will not question whether it is the

solution; that problem is too hard. But if it weren't, it isn't clear

we would ever be very interested in QCD as a theory to be applied to the

real world.

In Section 3, ue dldcuas canonical quantization of QCD with use

B
of a Banllconian formulation (in A- 0 gauge). We shall encounter formal

problems such as gauge-anblguities and topological classifications of the

states (B-vacuua, instantons, and all that) not met in QED. This will be

attempted at a descriptive, relatively painless level, details being left

as an appendix to .these lectures. Sections 5, 6, and 7 describe in more

detail the properties of Stages I, II, and III of QCD as wc sketched

above. In Sect Ion 8 we briefly sketch and compare various approaches to

the confinement problem. In Section 9 we very briefly discuss more

radical alternatives to QCD, such as the 3tring model or the Pati-Solao

program. Section 10 is devoted to conclusions and to an assessment of

the experimental situation. Two appendices are devoted to a more technical

exposition. We urge that it be studied in parallel with the main text.

2. THE SOLUTION TO QCD

In this section we shall be optimists and look through rose-colored

glasses -it what the solution to the theory should qualitatively look like

if all goes well. In doing this we make the following set of assumptions.

1. At distances small compared to 10" cm, the theory is Just that

of unconfined gluons and quarks interacting with each other via a coupling

„_„„,. of snail strength: *_

astrong „
strong
Mwt£^ 0.2.

2. Only color-singlet particle states exist in isolation and none

11
of them ate massless at any stage.

3. All confinement effects are "soft," i.e., characterized by a

momentum scale £ 1 G*V.

4. "Halve" parton-model ideas may be used to connect what goes on

at sHort distances to hadro.. structure.

How let us advance to Stage I and ask what the theory is like. At

short distances the only quanta are color-octet spln-1 gluons. At large

distances we must form composites of these quanta to form color singlets.

The details of how to do this depend upon questions of confinement.

Nevertheless we can make some educated guesses, based upon what happens

to up and down quarks. At short distances these quarks have color and
 - 11 -

5
negligible mass (a few MeV, according to current-algebra ideas ). ?ut

In a. hadron the quarks can be considered to have a nonvanishing mass

(-300 MeV) and to move non-relativistically. If this is also 50 for

6
gluons, then the S-wave states can be classified rather easily; their

wave functions must be gauge-invariant and color-singlet. The simplest

are two-^luon bound states. They are all SU(3) singlet and C-cven states.

C-odd states can be formed from 3 gluons (decay products of 0 and T

are a familiar example) as well as C-even.

Two color-electric fields might bind to J • 0 and 2 states, while

two color-magnetic fields might bind to another distinct 0 and 2 hyper-

fine multiplet. An electric gluon also might bind to a magnetic gluon;

In this case thu parity is reversed and the triplet is 0 , 1 , 2 . (The

0~ state may be regarded as suspect inasmuch as the operator which creates

it is Tr E « B, which is a total divergence and which plays an important

7
role in instanton phenomena and the U(l) problem; see Section 6 and

Appendix G.)

Just as for ordinary hadrons, these states should possess a composite

structure and have a rich spectrum of excited states. They should scatter

from each other like hadrons generally do. The big question is evidently

their typical mass. The only mass in the theory is the mass scale at

which perturbation theory breaks down; by hypothesis -1 GeV. Estimates

from the MIT bag model have given masscH of 1.0-1.5 GeV for the lowest

lying gluDnium states. But, to say the least, this is an uncertain

8
business. There is a school of thought (which I don't understand) that

puts gluonium masses considerably higher. We shall return to the

pheno nology of such objects in Section 4, but now choose to quickly

move an to Stage II.

In Stage II, ve introduce heavy color-triplet quarks such as charm

or bottom quarks. Actually for this idealization superheavy quarks Q

with mass >100 GeV would serve the purpose best, so that the size o£ their

bound-atate uave functions would clearly be small compared to the con­

finement radius. The properties of such superheavy quarkonlum would be

expectc co be well-calculated perturbatively. The "binding energy"

calculated from perturbation theory would be hydrogenic, -a M (-1 GeV),

1 - 1 5
and the size (a^M** ) - 1 0 cm.

Niv let us imagine scattering an electron from this heavy quarkenium

2 4 X
system at very large momentum transfer — say, Q - 10 Gev . The struck

quark will move semi-relativistically Cr - 2) away from its partner.

Because of its large inertia, nothing to do with the confinement mechanism

can immediately stop it. By hypothesis (specifically, assumption 3 made

In the beginning of this section) the rate of momentum transfer to gluon

degrees of freedom, potential energy, etc. cannot exceed a few GeV per
-22
fermi of travel. Thus after some time (say 10 sec., which is a long

time), the heavy quark and antiquark will be separated by a large distance

(say, >10 cm). However by hypothesis (assumption 2) the quarks are

not themselves free because they are color triplets, not color singlets.

Nor will any finite number of (octet) gluons locally dressing or screening
9
the struck quark do any good. Pair-creation of supnrheavy quarks might

be invoked, but this is a very unlikely process because of the large

rest mass of the quarks Q. Indeed if somehow a virtual QQ pair could be

created the final state would consist of two heavy quarkonia. But the
 - n-
total mass would have to be > 4H , and the total energy of the system

nead not be that larpe.

Wt are thus unable to avoid the conclusion that during the separation

of Q and Q some spoor of gluon degrees of freedom were left behind In the

space between them. The QQ system Is only able to be considered as one

complex extended (color-singlet) system and not two Isolated ones. What

region of space is likely to be affected? We may consider three possi­

bilities: (a) very big, (b) string like, and (c) minimal, as shown In

Figure 1. The very big system looks Inefficient (too much stored energy)

and also appears to depend upon degrees of freedom of long wavelength.

By hypothesis (assumption 2) there are no massless color-singlet modes,

and configurations of such large spatial extension (even though they are

not In isolation) make more serious the problem of absence of maesless

•odes. While this Is rather feeble hand waving, we are led to consider

more favorably case (b). Here a string of fixed thickness connects the
-
Q and Q . Evidently the diameter of the string should be related to the

confining scale and be between 10~ and 10~ cm. The energy per unit

length should be constant, and can be estimated from "old" hadron

spectroscopy or from the charmonlum potential. In either case an energy

par unit length of 1 GeV per ferml is a reasonable guess.

The minimal caso (c) also supposes a string configuration, but

allows increasing constriction as the string gets longer and longer.

But auch a constriction seems to violate our starting hypothesis. Given

a constriction of size small compared to the confining radius, momenta

larga compared to tha confinement scale will be Involved in the confine-

sent mechanism. This violates our assumption 3.

 I®
(o)

r©

-T©

Fig. 1. Three views of the color field becween widely separated

heavy quarks.
 - 15 -

Therefore we find solution <b) the preferred one and infer that (in

Stage II) « string composed of gluon degrees of freedom with characteristic

thickness of order the confinement scale and an energy per unit length

-1 GeV/f connects widely separated superheavy quarks. Furthermore this

string Is universal; it depends only on the color triality of the qusrk

source and not on specific color representations (e.g., j>, ljy, 2Aj etc.)

of the sources. To see this, first suppose that the superheavy quark

were a color octet. Then upon separation oi Q and Q there will be no

string at all: the color of the quark can be locally screened by an

octet gluon. In other words thare should be color-singlet states of Q

(or "5> bound to a gluon. Now consider a quark In a different representa­

tion of nonvanishing triality such as 6. A string connecting a 6 and 6

would a, priori be expected to have a different energy per unit length

than for a string connected 3^ and 2_ — probably larger, because there is

more color-flux, if the energy per unit length is larger, then it is

energetically favorable for the sources to be partially screened by

gluons to form an effective source which is 3. o r

A- F o r
example T is

contained in jS ® El. In a similar way, any color representation can be

o t
reduced to jL, 3.* 2. W multiplication by a suitable number of JTs.

Thus either there is no string or else it is coupled to an effective

source made of 3_ and 3_,

Now let us return to the original problem of dynamics of the QQ

system. Evidently, the first approximation is that they move seml-

classlcally in a confining linear potential. But haw does the system

decay? The two mechanisms available are radiation of gLuonium and, for

E > 4H-, dissociation into two quarkonium states. Each of these

 - 16 -

processes can be further subdivided Into soft and hard processes. We

discuss the hard processes first. The Q and Q are In a bound state with

very little damping and therefore pass by each other twice each period

of the Ration. At each passage they may, with finite but small probability,

undergo a hard collision such as

Q + Q-^Q + Q + g

q + q + q + q + q-t-q (2.1)

with g denoting a hard (pointlike) gluoR with high p . L These being pro­

cesses dependent on short distances only, they can in principle be calcu­

lated via perturbative QCD. The first lowers the QQ su^energy by a large

amount at the cost of production of a jet of gluoniura quanta. The second

leaves two lower-energy QQ quarkonium systems. Inasmuch as there is a

finite probability <1 per period of revolution for such hard processes,

the width for decay via them will be a finite fraction <1 of the level

spacing. (In fact, in 1 + 1 dimensional QCD, decay into quarkonium pairs

is the only mechanism available and specific calcula.ions support this

10
conclusion. )

Turning to "soft" processes, we can imagine QQ pair creation by the

string itself, provided 1^ is long enough to contain internal energy

greater than the rest mass of the pair; i.e.,

TL > 2M Q

where the string tension T is, again, the energy per unit length. The

probability of such a long string breaking into a smallish QQ system

would be expected a priori to be extremely small. The QED analogy is

electron-positron'pair creation in a constant electric field E. This

11 1 2
was calculated long ago by Hciuenberg and Euler; the answer i s
 Z
dp ~ „ -AAT(M /E)
V
dt " * (2.2)

Converting to our units, the probability per unit time rould be

£•' (2.3)

where A is the are* of the string, and we have made the identification

of B (ctnergy density) with r/A (string tension per unit area). Thus

t".. probability of soft breaking of the string is exponentially small

because of the large quark mass. Notice howwer that there Is every

reason to expect that when the light quarks are Introduced this mechanism

will be important.

Finally we cone to the moat Important mechanism — soft radiation

of gluonium. By this we mean radiation of gluonium In the direction of

motion of the q: it Is simply brensstrahlung. Because gluonium has a

non-p*rturbative extended structure its coupling constant to heavy

t

13
quarks la uncertain. But there IS not ouch reason to expect It to be

tttremaly small. It follows that as t'mj Q moves away from Q, the rate

of conversion of quark energy Into the gluonium radiation should be

comparable to the rate of energy delivered into the lengthening string.

Thus per period the energy lost to gluonlum radiation is a finite frac­

tion of the total excitation energy of the system. Ua conclude that the

widths of the excited states are broad compared to the level spacings.

The width* may even be a finite fraction of the mass of the system.

Bene* the discrete levels we maintained up to this paint dissolve Into

M. continuum. Actually the details are not too important, because there

will clearly be quite a change when we go to Stage ZII. Suffice It to

 - 18 -

say that in Stage II, heavy quark pair production is never of great

importance and thi: central mechanism of energy storage in excited

quatV.on.la is in the universal string, while the central mechanism of

energy dissipation is in soft gluonium radiation.

Before leaving Stage II, we should mention that the ground stnte

quarkonla QQ can annihilate into 2 or 3 hard gluons, via a process

calculable In perturbation theory. This, along with the hard-gluoa

bremsstrahlung in QQ collisions, provides us with an argument that

gluonium states must exist. After all, one might try to entertain the

idea that pure QCD is trivial; i.e., while pointlike gluons exist at

short distances, all they do is make strings but no finite-mass (as

opposed to infinite-mass) particles to populate an LSZ-type Hilbert-

space. However, by the above hard processes the quark sources produce

at short distances distinct gluon quanta carrying away energy and

directed momenta. Energy conservation demands this be materialized into

asymptotic states of finite energy. Given our assumption 3 of soft

confining forces, it is hard to come up with a scenario of the dynamics

vhlch avoids these gluon jets being composed of a collection of gluonium

states.

The existence of strings in Stage II also has Implications for

Stage I: there should exist in pure QCD mccastable closed strings of

large circumference — a kind of soliton. Such strings will shrink as

1
they emit gluonia, eventually merging into the quantum gluonium spectrum.

Finally we go to Stage III, where light quarks are Introduced. A

light quark q is one whose mass (as measured by short-distance probes) is

small compared to the confinement scale. Thus u, d, s, may be considered

 - 19 -

light, and charm Is borderline, but probably better considered heavy:

pair creation of cc in hadron processes is relatively rare. The new

features emergent with inclusion of light quarks are of course the spectrum

of ordinary hadrons, and the dissolution of the universal string. There

Is no longer any rationale for string production. The color of a heavy

quark Q can be screened locally by production of qq pairs — a process

which 1 B energetically cheap. Furthermore even were a string to be

•ade, it would rapidly break into many pieces through the Hcisenberg-

Euler pair-creation mechanism we already discussed.

We may ask what happens tD the gluonia. They can now decay into

ordinary mesons made of qq (cf.. Figure Z). There is an Important issue

of how nuch Zweig-rule suppression of these decays is operational h?re.

If there is enough for the width to be small compared to 100 MeV, the

gluonia should remain distinct members of the family of hadron resonances•

Although the issue seems to involve a certain amount of witchcraft,

consensus (but not unanimous) seems to be that in fact the states ought

to be narrow. We return to this question again in Section 5. Finally,

how light quarks effect confinement in collision processes — especially

—
deep-Inelastic processes has been discussed in detail extensively in
15
the past, and ve shall not cover this ground again here,

We have given very little attention to Stage III, which is, after

all, real life. Why should Scages 1 and II be relevant at all? The

reason lius in the belief (!) that the phenomenon of color confinement

is a universal one Examination of the perhaps more limited and tractable

issues embodied in StageB I and II might shed enough light on the nature

of confinement so that, even though Stage III may be more complicated to

 /'/////ft

Fig. 2. Mechanism [ox gluonlum decay into ordinary mosons.

mar .gft mathematically, che notion of color-confinement will no longer be

an Implausible or mysterious one.

It may he that confinement is not universal, e.g.. in Stages 1 and

II gluons and heavy quarks are not confined, and only after light quarks

arc introduced does confinement emerge. This vould entail a more specific

view of confinement dyna-nics, and might well be an encouragement to

search for mechanisms of confinement which do not depend upon an under­

lying non-Abelian gauge theory.

3. QUANTIZATION: INTRODUCING THE I. .TTOWJIVERSE

We now turn to the formulation of the theo?y. [those who are faint­

hearted and frightened by staple equations are invited to rejoin us below

Equation (3.43),] Up to a certain point, this is a parroting of the

canonical quantization of Q£D. Whenever one is in doubt he should retreat

to QED by replacing the triplet of quark fields

/ 'iW \

q(x) - I q (x)
2

\,«/ 3

by a single electron field ^(x) and by replacing the octet Df gluon

16
fields, described by a 3 * 3 tracelesc matrix

8
x J A
V > I E i Aj (3-2)
A=l

by the electromagnetic field. The ateps for constructing the theory are:

1. Start with the free Dirac equntion(s) for the 3 quacks

'(ly-m)q(x) - (iy^ J L - j < * )
n q - 0 (3.3)
 - 22 -

2. Make the gauRn Invariant substitution

which leads to

Uy-eA-nOqCx) = (l£-n)q(x) - 0 (3.5)

Hare e will be a strong coupling constant: a . (More about its normaliza­

tion later.)

3. Lock at local gauaa transformations, analogous to those In <jED

q<x) - S 0 0 q'(x) (3.6)

tjj(x) is a unitary } * 3 matrix that depends upon apace and time]

and find out the transformation lav that A. must satisfy in order to keep

the equation form-invariant. A few lines of algebra show that it is

l l
A - S A'S" -isiS_ „ S A - ~s + 1 / J S _ \ -l
s ( 3 7 )

Thin say look a little unfamiliar J In QED

leA<1
S - e- "' (3.8)

and evidently commutes with A because the "matrices" are 1 x 1 ; hauce in

QED

(In QCD there will be somewhat more emphasis on the actual gauge trans­

formations g(x) and leas on their generators A(x).)

4. Build the pauno fields F^,,

There are smooth ways nd clumsy ways of doing this. A smooth way

is to notice that in QED

 f
[V"v]lt<)0 = -ie F y v «(x) (3.10)

for any function * , and with the covariant derivative

u
iD _ i 3 _ CA" (3.11)

already defined in Equation (3.4). The same device works here, leading

to the definition

D 3 A
* " -r ..»2,,l • -3A+i rA,Al e (3.12)

The extra term quadratic in the A_*s coming from the non-coanuting 3 * 3

••trices causes all the extra grief in QCD, This will make the Maxwell

equations nonlinear: the gauge fields, which carry color, couple to

themselves as well as to natter fields*

The fields £ , unlike the A , do not both translate and rotate

under a gauge-transformation, they only rotate. For example from

Equation (3.12) it follows that

The easy way to check this ia to notice that the covariant derivative

also only rotates

1
s - s B; S "
v (3.W)

i . e . , for a l l functions * ,

(±\ - a A j *<x) - S(x) ( i 8 - eV)s-J-0c) «(x) y (3.15)

Then we can agsin use Equation (3.12) to establish Equation (3.13),

 5. Write down the "Maxwell Equations"

To do this, notice that in cleur analogy with QED the interaction

should be

v
i*l h - q-^% - I E TY
A
I „AU
a U

which serves to define the 3 * 3 matrix J , Under the gauge transforma­

17
tion. Equation (3.6), it Is not hard to show that J will rotate

J.. - ^Efq^iNu"
2 £-(•"„ 4 H;I - S S , !

Thus, to be gauge-Invariant, the Maxwell equations must behave in the

same way. This is again ensured by constructing them via use of the

covarlant divergence:

fc.r] - [ v v r ] - •? (3.18)

These equations along with the Dlrac equation can be derived from a
10
Lagranglan, but we bypass all that and take directly Equations (3.5),

(3.12) and (3.IB) to define the dynamics of the theory.

So far, ve have mainly seen similarities of QCD to QED. As we

already noted, the big difference between QED and i)CD is that there is

a term quadratic In A in F . Along with the teni In D proportional

to A , this means that the left-hand side of the Mixwell equations have

terms quadratic and cubic in A . This causes all the headaches in' QCD

(but, one hopes, also the seeds of the confinemen', phenomenon). The

physics of this is that the gluon fields themselves possess color and

therefore act as an additional source of color field.

 - 25 -

We now must replace the classical variables with operators (in

Helsenberg picture) and quantize. To implement this we will go to a
Huilltonlaa formalism. Before doing this, i t is useful to
6. Choose a gauge (for us, jJL • 0)
We do this with aft eye toward the next steps, namely identifying
canonical variables and carrying out a canonical quantization procedure.
Before uroceeding we remark that this i s not the only path to take at
this point. For small e we may just follow Book I, identify Feynman
rules, and start calculating. A more formal but very efficient way of
doing that is to get the rules via the path-integral formalism. For
covariant formalisms, one runs into technical complications — the so-
19
called Faddcev-Fopov g h o s t s . The canonical formalisms avoid these and
amount to choo ing one space-time component of tha potentials A to
vanish. There are three basic choices; either 1 » 0 (temporal or
canonical gauge), A_ • 0 (axial gauge), or A_ + A. - 0 (null-plane or
light-cone gauge). The l a t t e r two, which pick out a space axis, are
convenient for colliaear collision processes—namely, the parton-model
type of qCD applanation. We shall not be so interested in that kind of
thing at this point, but rather In large-distance questions associated
with confinement In pure QCD, and then in the properties of the theory
au e
with additional heavy quark sources added. Therefore, £n " ° S S 1&
convenient, especially since this keeps D. - 3/3t clean and helps to
Isolate the real dynamics from the phony dynamics of time-dependent
lauge-transformations.

To see that this choice is always possible, return to Equation (3.7)

and set Al • 0. This Implies we must find an j> such
 1
A . ' - i f

For gauge potentials which vanish at t * -*>, t h i s i s a Schrodinger

equation and can be solved formally
t
- i e / ^ C x , t ' ) dt' (3 .21)
S - Te **
with T the time-ordering symbol, (Remember A. Is a 3 x 3 matrix; i t
doesn't commute vlth itself at different times.) How
7, Write out the equations of motion I n A . * 0 gauge:
Dirac Equation
i | f - [a • (p-4>+ eojq (3.22)
"Ampere's law":
1
3E
1 i 1
~ - (Sx2) - e J = (?« Sj^-lee^jA^.B^-eJ (3-23)

Definition of B:

L 3 k
t " <% * b = (? " V ~ i« l j k [& , A ] (3.24)

Definition of E;

E - - 7- (3-25)

"Gauss' Law":
1 1
2 • E = 3 • § - iefA , E ] - eJp
 - 27 -

These look structurally like the QED Maxwell's equations, as written on

countless T-shirts. There are only three distinction (other than the

ubiquitous 3 x 3 matrices). Two arc the nonlinear terms in Ampere's

Law and in the definition of B, and the third is the presence of the

covarlant divergence p. In Gauss' law. Notice that Gauss' Law is an

equation of constraint not an equation of motion. This will be

t

Important in the subsequent interpretation of the theory.

With auch a similarity to QED it should be no surprise that the

above equations of motion can be derived from a Hamiltonian which is

essentially the same as In (JED. This leads us to the next step:

8. Construct a Hamiltonian and identify canonical coordinates:

The result Is an utterly unexceptional analogue of the QED

Haalltonlan (in this gauge) ;

3 2 2 3 +
H- (PxM' - Tr/d x [ E ( X ) + B ( X ) ] + ^ d x q ja • (p-eA)+6m|q

6
E + B t 3 2 7 )
- | Z / ^ [ ^ i w ] + / a W , « . ( ? - o X > + &mjq '

Just as in QED, the canonical coordinates are the A(x). (cf. Equation

(3.2)), and the momenta conjugate to A(x) are the E(x)^. Becruse the

magnetic field

has a tern quadratic In A, the Hamiltonian has terms Cubic and quart1c

in A. The Hamilton equations of motion are

 q
3 P ^ * a E i E
A at at

(This second equation requires a little care In the integrations by

parts; the covarlant gradient D con be Integrated by parts, just like V.)

Let us Ignore foe a while the Haoiltonian of the tensions. Let ui

furthermore temporarily Imagine replacing the Integral Id x by a sura

over a million coordinates x . (That should be fine-grained enough fur

a lot of purposes.) Then each A,(x) is a canonical coordinate q and

E
each ,(x) is its conjugate momentum p . The Hamt1«-ontan of this aome-

vhat mutilated version of pure QCD ia nothing more than a good old

Schrodlnger lUmlltonlan

6
2ft « 1 0
• - £ i*l+*[*i 1 2 4 M 1 0 6) 3
< -30)

of a particle in a 24-million dimensional space moving in a static

potential. The potential (the B term) is non-negative and at most quartic

In the coordinates. It is important not to forget this homely analoRV.

especially when the going gets rougher (as it regrettably will). Finally

9. Impose canonical commutation relations on the conjugate variables

A(x) and E(x), It Is time (in fact somewhat overdue) to make clear our

conventions of notation and normalization. The 3 * 3 matrices A(x). are

expanded in terms of 8 Independent canonical fields which are coefficient

of Gell-Matin's X-matricfcSi
 A/x) - £ -1 A*M »*
A-l

foiol /°° ' \ /o o o

100 i -ooo)
4 >, • o o -l
(000/ \ 1 00 / \o 1 0

/0 - i 0 \ / 0 0 -i
i,- i 00 » - 5 00 0 i -^r
8 0 1 0
\ o o o / \ i o o / \oo-z;

/l 00 \ /O 0 0
i , - I o -l o J, - oo i

2 (3 34
[iA-i ] - « A B c i
B C ' >
defining the structure-constants of the SU(3) algebra. There 19 an

analogous definition for E(x), and the canonical equal tiae confutation
20
relations axe taken to b e

[E*(J,t),A*(?,t>] - -U ±i
3
s " « <»- y) 0.35)

These give a consistent canonical formulation and lead with no difficulty

to Interpreting two out of our of Che rtaxvell-equatlons In terms of

the Hanlltoalan equations of notion

A E (3 36
%£ - -*[»• H - i i*- « H - Q ->
Finally, before leaving this introduction to formal quantisation of QCD,

we should add one more item to the list:

10. Check the consistency of the quantization procedure with Lore ttz

covariance

We shall not do thin here; the only substantive problem lies with

Lorentz-boosts. There as in QED, the generator should Include a gauge-

transformation of the fields designed to restore Aj. •> 0 gauge in the new

frame. The formal covarlance of temporal-gauge quantisation has baen

21
checked, although I have not found a direct check along the lines used

in Book II.

The astute reader familiar with Chapter 14 of Book II will not-ice

that we have not been very good parrft-i: there are some differences In

what is done there and what we do here. In BooH II, the longitudinal

degrees of frf'dom of the electromagnetic potential A(x) are eliminated

from thft beginning by a choice of gauge, leaving only the transverse

field to describe the two physical polarization states of the photon.

In what la done above, longitudinal photons (or gluons) arc quantized

*• well, tfhat does this mean? It appears we have overachieved and

introduced too many coordinates Into the dynamics. The key to under­

standing this lies in Gauss' law, which we hero Interpret as an equation

of constraint on allowed solutions of the Great Big Schrodinger Equation.

That la, not all solutions of

HT /q. q \ - C* {*.....,«
fi ,.\ (3-37)
V
e\ l' H , H
2 4 * 10 / \6 c
24x 10 ) 6

are deemed physically acceptable— only thosa for which

,lU . „ W.3B)
|j5 • |(x) - «£„MJ%
 - 31 -

are allowed physical states. Note that (in QCD) this is a set of 8 * 10
L
equations; one for eac color and coordinate x! the meaning of all this

la somewhat clearer when one identifies (cf. Appendix A for details) the

Gauss' law operator in the above equation as the generator of time-

Independent gauge-transformations. Equation (3.38) means that the Great

Big Wave-function * is required to be invariant under time-independent

gauge transformations. Hot all solutions of the Great Big Schrodlnger

Equation will have this property. If we think of these gauge-transformations

as "rotations" of the coordinates it means ¥ must be "S-wave" in 8 n 10

of the 24 * 10 coordinates, if it is to be physically acceptable. Notice

also that because B is invariant under time-independent gauge transforma­

22
tions, it commutes with the Gauss' Law operator, so these 8 x 10

coordinates are symmetry degrees of freedom.

Uhat should one do about, this? One way, appropriate to QED, is to

Identify the (In that case) 10 symmetry coordinates and their conjugate

nomenta and explicitly excise them from the Ramiitunian and therefore from

the formalism. Because of the linearity of QED this is easy tj do. one

breaks up the field into longitudinal and transverse pieces

t-^ + tj.
E - E + Ej,
L (3.39)

and writes
3 2
H - / d x \\s-lM + E*(x) + (5 - fy ]

3 t
+ / d x e ( x ) j 3 - ( p - e ^ - e ^ ) +• Smj e<x) (3.40)
Gauss' Law allows one Co eliminate E.

5 3 J (l (3 41)
V*> ' ^ V * > " -= J" " 4 ^ 7 | o " -
in favor of the Instantaneous Coulomb-interaction

Then the only vestige of longitudinal degrees rf freedom left is the

a • L term in the Dirac Hamiltonian. A phase transformation on the

Dlrac field easily removes this as well.

In QCD it is much more difficult to eliminate the "trivial" symmetry

variables, because of the nonlinear terms in H which couple together

23
longitudinal and transverse modes of the fields. Gribov has demonstrated

specific difficulties which block a generalization of Che QED Coulomb-

gauge quantization procedure. 7>,CS2 are sketched in Appendix A; suffice

it to say that things go along reasonably in parallel with QED, until

ecus T potentials) appeared in the formalism. The corresponding

operator emergent in QCD is 1/V * 0. But Gribov showed that there exist

gauge potentials A(x) for which homogenous sol-itions exist

7 - D(A) *(x) » 0 (3.43)

so that v • D is not invertible. This problem his deep ramifications and

is difficult if not impossible to elude by alternative choices of gauge-

fixing. At present there exists no. fully satisfa.tory formulation of

ll
QCD In terms of only physical degrees of frecdom. ' *

Given this situation, an llternativc is just to leave in the extra

degrees of freedom. After all, the quantization of the theory is

 - 33 -

Satisfactory and the problem is only one of frustration in not properly

coping with a gigantic residual symmetry. And it may be no worse to

deal with 24 * 10 coordinates then 16 * 10 . The problem is that explicit

lnvariance under Cine-Independent gauge trans format ions must be maintained

in every subsequent step in the development of the. theory. Otherwise the

unphysical solutions of the Great Big Schrodingcr Equation will mix with

the physical ones, and there Is no assurance that this does not create

some kind of nonsense In the resultant physics.

We emphasize that this difficulty only occurs for sufficiently large

gauge-fields, much larger (roughly by a factor a ) than commonly found

in the weak-coupling limit. Thus weak-coupling, short-distance applica­

tions should hopefully not be affected by the gauge-fixing ambiguity.

In particular it will not be seen in any finite order of perturbation

theory.

Let us now surviarize where things stand. We have seen that it is

possible to formulate (in A_ " 0 gauge) a straightforward quantization af

QCD in a way analogous to QED. Nevertheless there arose some problems,

which we here enumerate:

a. In the presence of strong fields the gauge cannot be completely

fixed in a satisfactory manner: ur.physical degrees of freedom are

quantized and a careful selection of the physical subset of solutions

to the Great Big Schrodinger Equation must be made, namely those con­

sistent with Gauss' Law.

b. Even assuming weak coupling, the basic quanta in the formalism

ate colored quarks and gluons, which do not (or at least should not)

appear as asyaptotic states in real life. Thus there is no obviously

 - 34 -

satisfactory way of Introducing and constructing an S-matrix as done

25
In QED» without first resolving the confinement question. It seems

necessary to solve QCD in order to formulate it.

c. Asymptotic freedom implies that although at short distances the

coupling constant may be chosen (consistently) to be small, it becomes

large at large distances. But then the nonlinearities of the theory

become pernicious.

In addition to the above three problems, which are no doubt inter­

connected, there is a fouvch which is to be discussed in Section 4 (in

order to spare the reader too much of a dose of formalism all at one

blow):

d. The inatanton phenomenon complicates the structure even of the

QCD vacuum. Effects of small instantons can be controlled, but what

happens at larger distances is not in good control.

All the above complication is minimized If we could somehow restrict

everything about the fonrulation to ahort-diatances. It ia possible to

do this: we enn Just quantize the theory in a sufficiently tinv box.

While this obviously leaves something to be desired, It does provide

solutions to the above p" '''-its: perturbation-theory is universally

applicable; the coupling-constant can be taken as always email, and the

dangerous large instantons don't fit in the box. Quarks and gluons are

the physical quanta and are unconfined, provided the confinement radius

ia large compared to the size o.c the box. Gauge-fixing ambiguities do

not arise because the field strengths relevant to typical scattering

processes are too small.

 But is this really satisfactory? Have we destroyed QCR in order

to save it? After all, we do not fit Into such a box, and what is a

quantum theory if it does not include us as observers? I think this

aspect can in principle be eddressed in a reasonably satisfactory way.

let us imagine that there exists a world out there built out of super­

heavy charged fermions which we call femtofeniions and which interact

with each other by exchange of a different octet of femtogluons carrying

femtocolor. The mass-scale of this world Is taken to be 10 times

larger than our own; hence by dimensional analysis the distance-scale

is «10~ that of our own (hence the femto-pref ix). The femtaquarks

will bind into femtonucleons and femtomesons. tf we give the femtoquarks

electric charges ±Z/3, ±1/3 in the usual way, and introduce a femtoelectron
1A
of mass 5 x 10 MeV, we can make femto-nuclei, femto-atoms, and so on,

up to — you guessed it — femtophysicists. Femtophysicists will build

accelerators, magnets (with field-integral /B 6.1 - 10 times ours), and

the like, and do experiments on not only the constituents of femtomatter,

but also the constituents of ordinary matter, now? Figure 3 shows a

way; build an accelerator (the one shown, a rennrkahle technical accom­

plishment of the femtophysicists, is -".03 A in length) which can create

a photon beam: ordinary quarks and leptons can be pair-produced and then

separated from feratomatter beams. Thus femtophysicists in principle

could scatter ordinary quarks and gluons from each other and measure the

relevant S-matrix elements. They could even communicate the results to

us via electromagnetic radiation. (However femtophysicists would have

to be patient souls: one bit of information transmitted per ten femtodays

corresponds to ~- one bit per nanosecond to us.)

 v r
TAKB17 lUflOINQS -

'^1/F*%-
( \ W AL'INC ftOAD

2-80
3662 A 28

FIR. 3. Femto-SLAC.
 - 37 -

Could actually such a fcmtouniverse exist? If one were nearby we

would certainly know it. Just the black-body radiation from a feoto-

earth would consist of It) photons per second, eacl of energy -20 TeV.

Be sure to avoid the fcmto-sun. However, none of thi3 is realistic: we

have ignored gravity. Gravity is not a scale invariant interaction; the

-9
gravitational force between a ferotoproton and femtoelectron is ~10 the

electromagnetic force. The gravitational force between "macroscopic

pieces of fcmtomattcr will be crushingly large*

But we digress. While our femtounlverse is not- completely realistic,

it should indicate that even with a tiny quantization-volume, we can

think in reasonably physical terms about scattering processes, now

Involving quarks and gluons ~ and not gluonia, quarkonia, or hadrons.

There is one major problem of principle: a typical beam pipe at femto-

SLAC has a diameter ^10 cm. The beans will be even smaller; hence

they have an unavoidably lares <p, >, large compared to the confinement-

scale of a .ew hundred KeV. Thus quarks and gluons Ln the bean will

have large momentum uncertainty — we can consider them as virtual, or

off-shell. There will also be infrared-divergence problems in calculation

of S-matrix elements similar to those encountered in the QED of masaless

26
electrons. In that case Lee and Nauonberg showed that the conventionally

'"fined S-matrix depends upon the quantization volume. Another way of

Saying this might be that the outcome of physical scattering experiments

depends oon the details of the preparation of initial beams aa well as

the properties of the final-state detection apparatus. Kinoshlta

as well as Lee and Nauenberg, * have given formal recipes on how to

2 1

Avoid these infrared problems by summation over groups of initial as

 - 38 -

well as final states. I am not sure these problems are fully under control

in QCD, Rut here we shall assume they arc. Specifically, we assume that

in the fcmtounlversc

a. Asymptotic fields and states of "physical" colored quarks and

ZS
gluons exist in the way described by LSZ.

b. These fields can be formally related to the fields appearing

in the equations of motion by a U-matrix (c£.. Book II, Chapters 16-17)

which can be calculated perturbatively by a diagrammatic expansion.

While wa have not explicit? done all this, I an confident that the

net result of such a program would be the rules for diagrams obtained by

Book. I or path-integral methods.

What good is all this? Perhaps we can provide a satisfactory theory

for femtophyslciscs, but does that help us directly? While I am aot sure

of the ansi er, I think it is yes. First of all, it should be the case

that any process which in principle can be measured by fcmtophyslcigts

as well as by us can be calculated perturbatively. Let ua look at some

good examples:

a. (e a~ •* hadrons). To lowest order in a, this one is fine.

Femtophysiciuts should be able to make femto PEP. R(e e -*• hadrons)

appears to be a very clean QCD calculation.

b. O(YY "*• hadrons). For this one we should restrict our attention

to virtual photons. There may still be a problem because of presence

o£ large longitudinal distances at high energies in this process (cf.,

Figure 4 ) . The distance-scale is iz ~ E /m - where tn - (>>1 GeV ) is Che

Y qq qq
mass of the virtual fcrmton pair created by the Y-ray.
 - 39 -

Fig, 4. Large longitudinal distances in Y - Y scattering.

 -40-

c. (e + q -• e + q + gluons). For large x, this process depends on

distances smaller than the size of the box. For small x, large longitu­

dinal distances are again. Important, and infrared effects associated with

the Initial beams may need to be taken into account.

27 20
d. (q + q -*• u + u + gluons). The Drell-Yan process ' appears to

be applicable — at l»ast for the total cross section (where one measures

orly this attenuation of the initial quark beams). The p. distribution of

dileptons appears also to be an acceptable process, but only at large

<v
Processes Involving hadrona must of course take Into account the

relationship (if any) between the quark and gluon beams of the femto-

physlclst and the equivalent quark and gluon beams of the parton model.

For sufficiently coarse-grained measurements Involving hadrons it is

plausible that the two kinds of incident beams are equivalent. In the

iase of the deep-inelastic structure functions relevant to electro-

production and neutrino interactions, one can go further and derive the
2

Q -depsnden-i of the moments of the structure functions without introduc­

ing the parton model. One need only rely jn the Wilson operator-product
29
expansion. However the basis of that expansion is ir. fact an analysis

to arlbtrary order of perturbation theory- He have only been able to

control perturbation theory by restricting ourselves to life in the

femtouniverae. But in the femtouniverse the Q -dependence (at very large

2
Q ) of electroproductlon moments does not depend in any way on the con-
posftion of the quark beam. Thus it is extremely reasonable to generalize
 - 41 -

this behavior to the proton aa well. Maybe it is even a rigorous state­

ment, but 1 don't know enough about rigor to be able to tell.

A somewhat stronger version of this Idea is widely used in applying

perturbative QCD to physical processes. As we already indicated, most

perturbative calculations in QCD will run into difficulty with infrared

o
divergences associated with coHir ar emission of gluons or quarks. This

is the kind of problem dealt with by Kinoshita and by Lee and

26
Hiueoberg for massless QED, and involve macroscopic distances (where

confinement effects will creep in). However one can define quantities

which are free of infrared singularities and which are calculable In

QCD perturbation theory. A prototypical example Is the angular distribu­

tion and angular correlations of quark and gluon energy produced in

30
* e annihilation. It la an observable In the femtounlverse which is

Independent of details of apparatus or beams. It is tempting (and

probably correct) to assert that this prediction is also true for the

hedrons measured by ordinary physicists as well. If our assumption that

the confinement mechanism is "soft" is true, this result will follow.

But there Is an additional assumption involved, namely that between the

acale of the femtouniversc and the scale of ordinary macroscopic measure­

ments, the jet of quarks and gluons is not significantly defocuscd or

•elf-focused. Nevertheless, this assumption is very reasonable.

GeneralJ-.ing, we may postulate that whenever one finds and calculates

pertutbatively a QCD process which is infrared finite, the answer is

correct. In other words, if the calculation is not obviously wrong,

it's right, Bui It is possible that this postulate, while not obviously

wrong, may not be right.

 - 42 •

So far we have used the femtouniverse as a guide and criterion for

the applicability of QCD perturbation theory ideas. In addition, the

femtouniverse may be of use In attacking non-perturbative questions as

well. Let vs accept that we can understand in principle everything about

-14
1
the theory in a box of dimension T" cm. Now stack together a million

such boxes, with the fields coupled at the boundaries in the vay appro­

priate for the full theory. Then the theory in the resulting cube of

dimension 10 cm should already provide a description of confinement:

free gluons and quarks should begin to disappear as asymptotic states,

and gluonla, ouarkonia, and ordinary hadrons should appear: they easily

fit Into a box of that size. The problem is to identify and properly

couple together the crucial degrees of freedom present in the small boxes.

The spirit of such an approach is very close to lattice-calculations:

we return to that apj-jach in Section 8.

A clue as to ho-/ one might proceed comes from closer study of the

structure of the theory in the snail box. This has been examined in some
31
detail, although some points are not yet fully understood. Everything

appears to be almost the same as in QED. Of course, it Is important to

use Fourier series, not Fourier integrals. With use of periodic boundary

conditions, the gluon state of lowest non-vaniahiag momentum has nomentunt

J 1
k - 2u/^r' » 1 GeV. As in QED, longitudinal gluons of it I 0 can be

eliminated by use of Coulomb-gauge. The only node of the potential

A.(x) which seems to behave in a way different from QED is the constant

k • 0 mode, which we label as the vacuum mode. The Hamiltoniar. for this
f
vacuum mode, ignoring couplings to degrees of freedom with ' - ^ 0. ".a.n

be written
 + f
"l-fi.3 | - ^ ) S '«•.'«»' WA,
A-i,...,8 (3.«)

A, i 1 -A . V 1 [ A , , , ik*x , . "I
+
AVOc) - — QV Z - ——— a.(k,e) e + h.c.
1 L J
/i * wo /vT5K
A
/5

with IT, and Q, canonieally conjugate pairs; i . e .

- -i-=j (3.46)
35?

The problem of finding the eigenvalues and eigenfunctloas of the

Hamiltonian i s equivalent to solving the Schrodinger equation of a single
non-ralativistic particle moving in a (not very symmetric) quartic
potential in a 24-dimensional space. I t turns out that the "particle" i s
confined, i . e . , the wave functions are nonnalizable. The Hamilton!in has
an energy-spectrum with many low-lying levels. Properties of the solution
are as follows:
1. In low-lying states

1
A1/3 0.«)
(« >
and the dispersion around the mean value appears to be small.

2. In general the level spacings of excited states are in propor­

tion to (e /V) . This is to be compared with the lowest-Lying gluon

modes, whose level spacings are-2IT

Thus the energy-spacings arc
 - 4i -

small. By the uncertainty relation fit iE > 1, this means the period of

the motion is large; the motion is slow*

3. The louest-lying set of levels appears to be a rotational band,

i.e., the polarization-vector of the vacuum field tumbles in space. The

energy should bo given by

The effect of these vacuum rotor-modes on the physics in the femto-

unlverse Is obscure but probably minimal. This is because of tht dense

level spacing of the rotor-modes. Very little energy Is needed to excite

rotor states, and their motions should not interfere with the dynamics of

the gluons and quarks occurring on a much higher energy scale. [For

instance, we do not worry much about excitation of photon states during

a hard collision of a high energy hsdron with a piece of matter.] However,

these vacuum modes may, upon coupling of a million femtouniverscs

together, contribute to the long-wavelength aspects of the physics at the

larger distance scale. Hence it is possible that they may be of relevance

to an understanding of the confinement issue.

U. DiSTAKTOMS

There is yet another co plication present in <JCD not encountered in

QED. The states of the theory may be classified according to the topo­

logical structure of the gauge potentials. In our quantitation procedure

— especially in the feratounivcrse - we considered only weak gauge-fields

A — and only pu*-e gauge transformations which can be reached from the

Identity transformation continuously, i.e., by a product of infinitesimal

transformations. However, there exist pure gauge potentials A(x)

(potentials for which E * B • 0) which cannot be reached continuously

from A • 0 without pulling the gauge potential in from the boundary of

the quantization volume. An example of t h i s is a spherical "gauge bubble"

(Figure 5) given by

1 1
4(») - i u S u " with ...-'i'*'..- !'^"' (4.1)

a a A w l t f t £
where x * C ^ . T - . T - ) = (*,,K>*-0* ^ increasing from zero at

i • 0 to a multiple of i at r • », If f(r) * 0, than £ - 1 and if

n
f(r) • iiT, tj * (-l) . Thus £ is nontrivial only in the region where

3f/3r is nonvinishlng. This region can be taken as a shell of arbitrary

radius R. Continuous gauge transformation can change the shape of f(r),

nodlfy R, and distort the bubble. But they cannot untangle the topological

tviatB which cone from coupling internal-symmetry generators T to space

coordinates r-
32
It turns o u t that the amount of topological twist In the gauge-

bubble can be characterized by a "topological quantum number":

N - - - ~ T r / d x S y ^(x) A^x) A ^ x )
3
k (4.2)

Thia formula Ifl valid if and only if A. is "pure gauge." The quantity N

1 B invariant under continuous time-independent gauge transformations

which vanish at the boundary of the quantization volume. Furthermore

for A't which, vanish on the boundary, N can take only integer values.

He have neither motivated this construction nor demonstrated these

raiulto; that ia business relegated to Appendix B. Here it suffices to

 ieA = UVLT

U = - l outside

Aj<0

(o)

lef(r)
, J = e -ioT.ff(r)

(b)

Fig. 5. A gauge-bubble of pure gauge-potential (E - B • O^,

 • 47 •

know that there is a way of characterizing and classifying the gauge-

bubbles. If wo start (perturbatively) from any weak-field configuration

connected smoothly to A - 0, the topological quantum number M is zero.

Thus it would seem that wc night be able to restrict our attention lo

those N - 0 configurations, since the only way to reach others is by

bringing the gauge field in from the boundary of the quantisation volume,

and bouiidary-effects (for sufficiently large universes!) shouldn't matter.

However, this is not the case; there exists a dynamical communication

through quantum-mechanical tunneling which 1G • volume effect. To see

how this comes about, it is necessary to again think of QCD in terms of

the Great Big Schrodinger Equation with A(x) the coordinates and £(x)

the momenta. (It is important to recall here Equation (3.30} and the

accompanying admonition-) A gauge bubble with M j* 0 is a configuration

1 /*2 3
with classically zero energy since V(A) » y #B d x = 0 for a "pure gauge."
If we go directly toward A « 0 by scaling down A, i . e . , by letting
A(x) •*• f(t)A(x), 0 £ f < 1, the magnetic field no longer vanishes because
the intermediate configurations are not "pure gauge." (This can be
checked by direct calci'-nion: B. - ief(l - f)f"A ,A/le .) Hisice there
is n potential barrier between configurations with N 4 0 and N « 0; the
potential energy for all paths in A-space which smoothly coi aect A >- 0
33
with the gauge bubble is non-vanishing.
Nt-w if we start with a (physical) solution V (A) of the Great Big
Schrodinger Equation restricted to A/s with N = 0, i t would appear to be
possible for the wave function to leak through th . barrier. This turns
out to be the cane: thi
figurations nad onward.
 - 48 -

Schematically, the situation looks like Figure 6; the potential la (in a

sense) periodic. The important tunnoling couplings will evidently be

nearest-neighbor with JANJ - 1. It should not be surprising to anyone

slightly familiar vith solid-state physics that the energy eigenfunctions

will be BloCh-waves

This is known as the 0-basis. Physical observables will depend upon e,

but oust depend on it only to the extent that the tunneling amplitudes

a-e non-vanishing. Therefore, it is important to estimate the tunneling.

For jmall e, one must tunnel through a thick barrier (cf.. Equation (4.2)

which implies [a • A • LJ £ 1, with L the size of the region in space in

which A(x) is non-vanishing), and a scmiclassleal estimate Is appropriate,

The rule for obtaining this amplitude can be abstracted from experience

with nonrelacivistic quantum-mechanics, and some details of how to get

it are in Appendix B. The recipe is as follows. To find the amplitude:

1. Change t to it (i.e., go to Euclidean metric).

2. find i classical solution of the Euclidean equations of motion

(thn QCD version of Maxwell equations) which as t -+ ~= reduces to A = 0

and as t -*• -Ho reduces to a gauge bubble.

3. Calculate the action S • / L dt of this classical solution.

s
Then the tuniicling amplitude is * e~ .

The required classical solution was found by Bela/In, PolyakoV,

31
Schwartz anJ Tyupkln. * There are, in fact, L class of such solutions

characterized by
 3 2
fv(A)=/d xTrg

J J«3*J

Fig. 6. Effective potential energy in field space as a functioi

of winding-number N.
 a. Size- The fields F are concentrated in a characteristic

4-dimcnsional volume of size X, where \ can be chosen at will. For

p - /|t| + r » A the fields F 2

fall off as p~ , are proportional to

plicated angular dependence.

b. Orientation in internal space: the T'S in the gauge bubble

(and fields F ) can be chosen from the 8 ^-matrices in many waya.

c. Location in space-time. The center of the configuration is

arbitrary. These classical solutions are the instantens. The gauge-

bubbles are not instantons. Iastanfcons arc in some sense a dynamical

agent responsible for their creation.

Because the field strengths F, from diiaensional arguments, are

~l/cX Inside the instantoa, and because the action S of QCD is

JTr/"d xF4
F uy
- X f d^V-T C*.4)

the tunneling amplitude is small. The correct numerical calculation

gives

e
Tunneling amplitude - e ' - e~ *'2 a3
(4.5)

The effect is nonanalytic in a and could not be seen in en ordinary

perturbation expansion.

Because an instanton can be located anywhere in space-time, we can

have gauge-bubbles continuously created and destroyed all ever the place

f.ad all the time. This is what makes their effects relavant i the

coupling at the N-vacuua is a volume, not a surface flli for example,

n o t only does t h e energy <H> of a 9-vaeuum depend upon 9 but a l s o t h e

energy-density. The formula i s 35,35

®7?
The predominant ex^-nentlal of the tunneling amplitude is embellished by
T
various factors, h e factor cos9 reflects the periodicity of the 0-vacuua;

i.e.i the conjugate operator N h3s integer eigenvalues (c£., Equation

(4.3)). The factor (8n /e ) arisen because of the large number of

degrees oE freedom possible for an instanton (scale-size, group orienta­

tion, location) and is a measure of the "entropy" of this gas of instantons

in space-time- The weight factor \~ for the integration, over scale sizes

X is determined by dimensional analysis. Finally, the value of e to be

used should match the scale of the ins tan ton — although already here the

rigor of the analysis begins to unwind somewhat.

If we restrict ourselves to the fentouniverse, it is easy to sec that

the most important instantons are those which fit snugly inside the box.

This is because the running-coupling constant Equation (1.4) is

3
_fe 2- * _ * 2z^i l o g ,v
e*tt) « CM a„<M*) (4.7)

with M a very larpe mass scale ( » 1 GeV) where we normalize a (« 1 ) .

The coupling constant a increases as the scale A increases, and although

the Growth of a is only logarithmic, its effect is large because of itc

presence in the exponential. Thus the tunneling factrr

*V - „2"A.«'>
S
2
, "-Tt
M (4.9)
depends like the 8th power of the scale-size and dominates over the

d> \~ factor. In a femtouniverse the vacuum energy is (disregarding

numerical factors <, 10) from Equation (4.6) is then found to be

. •)» ~~\"
1l
J J
~'"t t * tit 1
vV3
- cos9
,|j] p,^,^]»"V -rt )

unset of instanton-related affects is consequently extremely abrupt; in

the perturbativc regime an increase of distance scale by a factor 2

increases the instanton effects by a factor -500. It ig also regrettable

but true that as the tunneling amplitude gets large, the semi-classical
37
approximation used to compute it breaks down. Thus it is not ouch of

an overstatement to say that whenever an instanton calculation can be

believed, the result is physically unimportant. Cor example, there have

been estimates of the QCD modification to the colliding-beam R coming

38
from couplings of the quark pairs to instantons. They find a contri­

bution which behaves as s~ , with the crucial value of Js occurring at

about 1 GeV. A similar situation also exists for deep inelastic

scattering.

While the instanton does not upset in any way perturbatlve QCD, it

does play an important role in two other areas. One has Co do with CP

violation in the strong interactions and another has to do with chlral

symmetry breaking and the resolution of the "U(D problem" — the problem

of the origin of the large mass of the TI and/or n' mesons. We shall

return to these questions in Section 7, There is also the possibility

that lnstantons play a significant role In the understanding of the con­

finement question. This will be considered in Section 8.

5. PURE QCP: GLUOMIUM PHENOMENOLOGY

We have argued In Section 2 that pure QCD (i.e.,Stage I) Is the

theory of interactions of massive, color-singlet (and SU(3) flavor-

singlet) bound states of gluons, and that it is unlikely that these

states disappear when light quarks arc introduced. Instead they decay

(with not extraordinarily large width) into open channels of ordinary

mesons. It is an important test of QCD to find decisive evidence for or

against the existence of these states. In order to do that, some

theoretical guidance is clearly of use. It Is surprising to me, given

the degree cf"hubris extant that QCD is the correct theory of strong
39 l,0 1 1
interactions, how little discussion , > * there has been of this

problem. We approach the question in three stages: (a) the gluonium

spectrum, (b) decay schemes, and (c) production mechanisms.

a. Spectrum: Properties of bound states of two gluoas—in particular

4 1
their mass—can be motivated in several ways. First of all, * we can expect

that local operators built of products of the gluon fields and having

appropriate quantum numbers uill create gluonlua states. Just as local

products of quark fields (e.g., q.Yc°.jJ 4<Y *Ii» etc,) create meson states

when applied to the vacuum. The simplest such operators are

8
A A
x
pw afi
2C *uu^
A-l ap *•*** F w n i c n
should create gluonia of various spins and
parities from Che vacuum. Consider the spin-zero operator F (x)
T r F x F W
" u yC ) (x) for simplicity and look at Its two-point-function:

+ 1 1 X 2 2
D (x) - ^ x e ' * <0|F (x)F (0)lO> (5.1)

(An object very similar to this measures o(v + v -*• hadrone) through single-
gravlton annihilation).
 _ 54 -

At short distances (laxge q ) , QCD perturbation theory allows a

computation:

+ 2 4 2 2
D (q ) » (const) q eCq ) q + - (5.2)

At some larger distance scale the perturbation expansion breaks dawn,

and the spectrum cuts off among a presumably discrete collection of

gluonium states (Figure 7). Me have two choices: either the gluonium

mass is "reasonable" (1-2 GeV, say) or else the gluonlum mass is large

and QCD perturbation theory breaks down at a surprisingly large momentum

scale (» 2 GeV). We shall concentrate our attention here on the first

possibility. This option is somewhat reinforced by calculations by Jaffe

2
and Johnson in the context of the HIT bag-model." In the bag model, the

vacuum in the region occupied by quarks and gluons (the bag) is modified

and has, by hypothesis, a different energy density. The kinetic energy

of the quarks or gluons in the bag provides a pressure which balances the

pressure created by the (true) vacuum at the vails of the bag. Jaffe and

Johnson found masses "-1-1.5 GeV, although nowadays they have less confidence
143
In that estimate. Possible s-wave configurations of two and three

gluona axe exhibited in Table 5-1, along with local operators which create

them. The <p ,. (J,H) are the spin wave-functions which couple the fields

into • state of definite J and M. This classification might be called

1 0
the "naive glui-.lum model." * It Is at least as naive as the naive

nonrelativlstic quark spectroscopy, naive parton model, naive Drell-Yan

formula, etc.

There is nothing very definite that one can say regarding the masses-

and especially level spaclngs—of these candidate states (not all of which
 Perturbation
Calculation

~a2
Gluonium .Par ton
Dominance Model

Fig, 7. Schematic ot "luonlum spectral function, normalized to

i t s high energy behavior.
 - 56 -

TABLE 5-1

p c
J Field Operator

£ EJE*»«(J,H)

•H EWU,!!)

1J

2^ BJB*« (J.H)

<£ «&&*%*
1
3 ^"S*™
2** ^H^w-w
•t Qft^w*
•HI $°& \^ m

i
2
3 wv».»
£ w^\*
3 -
•tf-^W-H)
need exist at narrow resonances). The question of hyperflnc spllttMgs

is for example problematical. The splittings might be considerably larger

than the typical few hundred M Q V for hadrons, because of the larger con-

etltuent spin.
 Another approach to glv.oniua properties is to try to connect them

up to properties of the Pomeranchuk trajectory,'*'' Inasmuch as two-gluon,

t-channel exchange lias vacuum quantum numbers and a "naive" intercept at

J • 1. Again, the recurrences of that trajectory would tend to lie In

the 1-2 GcV mass range. The spin" i-.iu parities moat naturally suggested

•re 1~~, 2**, 3~~, etc.

A somewhat related attack is to assume that OZI-vlolating processes

•re mediated by gluonia. The systematica of these processes (e.g.,

t-^hadrons, i|»* -t-^nir) might then provide some insight into mas Bee or
115
couplings of the gluonia. This has been studied rather extensively.

Nevertheless, while some progress Is made in correlating the data, no

clear picture of the gluonlum spectra emerges.

b. Decay modes: The decay modes of gluonia are of course sensitive

to their mass. The masses are In turn sensitive to, among other things,

the hyperfinc splittings within a multiplet, which we emphasized might be

larger than for systems built from quarks, because of the larger spin of

the gluon. QCD tends to put antiparallel spin alignment (in color singlet

icates) lowest in mass (e.g., m < n ; BL, < n , etc.) so ue might expect

the lowest J In a multiplet to lie lowest. We shall, however, remain un­

prejudiced for awhile, and classify decay, of gluonia into two-body

channels of the s-wave and p-wave qq meson states. This may be a credible

guide for gluonlua states with mass < 1.5-2 GeV. The final-state mesons

we consider arc the S Q pseudoacalar (P) and S^ vector (V) nonetfl, along
3
vith the (C-odd) P x (A*) and (C-even) P Q . -(S, A, T) nonets. The only

two-body channels we allow are those which contain no more than one P-wave

•anon. The two mesons in the final state muat be coupled to an SU(3)
singlet. Table 5-2 gives a listing of the meson-states available which

are consistent with SU(3) symmetry. (We assume no singlet-octet nixing

1 1 3 3

for S Q and p. mesons, and ideal mixing for S and P mesons.) The
B
resultant J of the final states are alno recorded for £• Q and some

a • 1 configurations. These two tables can be used together to search

out likely decay modes of the gluonium candidates we have identified.

The reader is urged to do this for himself and in chat way reach the

proper state of enlightenment. This writer cannot claim to have reached

that state himself, and offers here only a motley collection of

observations;

1. Phase-shift analysis in irir and KK channels should tie a good

place to look for narrow 0 and 2 low-mass gluonium channels. It was

w
once argued that the S* is a gluonium state, and the Pnqq state lies

higher—perhaps to be identified with the very broad c "resonance" at

-1300 MeV.

2. The 0 E >5 state has the same quantum numbers as the n' and may

be mixed with it. (The decay n* •*• fr, however, is well-accounted for ill

terms of Just the qq component.) This gluonlun channel may be especially

1 6
tricky * because this channel has the ssme quantum numbers as the divergence

of that axial current which la nonvanishlng because of Instanton effects

(cf, Sct-tion VII). However, other than n'» the only other low-mass open

channel Is IT6.

3. Many of the states have the rather ^distinguished 3TT, 4n, 5ti

channels open to them; those offer relatively little hope for a novel

search method. Among the (low-spin) gluonia we can put In t M s category

1 + + +
are the l*"", 2 ~ , 2**, 0 " , 1~", l"**. 1 " and 2 " .
 c
Final Mesons / c s-vave) J P C
(p-wave) Components

PP d"" - iriT KK, nn, n'n'

f

Ptf +
i " 0~, 1 , 2 wp, KK*, Cn + n, trf-f)
+
PA' l"~ o - . 1 - , 2+" +
KB, KQ,
+
JS o' 1** ITS, KK, (n+n*. S*+e?)

PA 1- 0**, 1 ,
W
i» irA r KQ, <n+n', D)

PT 2-+ 1*+. 2+*, 3** TIAJ, KK , Cn+n', f+f)

W «**. 2 + f
etc. Pfl, K K , HXJ, $$

VA* a*. l"\ I* etc. PB, KB,?

vs i~
1
o - . I-*", a " 4
p$, K K , (u>+$, S +c?)

VA o~, 1~. 2~ etc. pA^ K Q, (LH-$, D?)

VT 1~, 2~", 3 —
etc. p A ^ K*K**, uf, *f*

Notation:

P: \ «-» TI, K, n, n'

A': \ (i-h B, Q, 1, 1

V; \ (i~> P. K*, « , •

5: 3
P 0 <o~> t , K. S*. e

A: \ u"> Al, Q, •:, D

T: \ (2**) A2, K , f, £

4. The L states arc candidates for resonance production in e e

storage rings. However, the width into e e~ may be very small, and in

any event hard o calculate. Furthermore, such states are rather far

down the list (3 bound gluons) and may be fairly massive (or for that
 - 60 -

matter nonexistent). The activity at DCI and ADONE should evidently be

watched with this in mind.

5. Because the gluonia are SU(3) singlets, decays into mesons con­

taining strange quarks may not be suppressed as much as is customary in

"old" physics.

6, The tabulation ue have made allows decays for all gluoalum

candidates except for the last of the list—the 3 state of three magnetic

gluons. Decay into p-wave A- or A- channels are open in that case.

Before moving onto questions of production-mechanisms, Wd reiterate

that this catalogue is not a list of predicted gluonium states. Some,

but not necessarily all of the channels listed should support discrete

gluonium states. Were QCD to be a true theory, it could provide success­

ful predictions of the spectrum of gluonia comparable to its successful

postdictiona of the spectrum of hadrons.

c. Production cechanisms: Half the momentum of an energetic proton

is not in quark-momentum and is presumably carried by gluon degrees of

freedom. It follows that there ought to be a sizable production cross

section for gluonia in hadron-hadron collision*. Why hasn't it been found?

A possible answer is that there are so many different states of comparable

mass and so many decay modes far each tha*. for any given channel aE is

very small. Another possible answer is that there hasn't been a sufficiently

vigorous search.

Beyond the self-evident searches in hadron collisions and in photo-

production, other methods with more specificity would be especially

useful. The physics of onia is an especially attractive possibility.

In the <iCD framework, it Is manifest that heavy oniua only couples to
1 7
ordinary hadrons via gluons. * for example in the process e e -* * + X

or e e~ •+ T + X (Figure 8a), the system X could be rich in gluonium states.

A related Idea uses dlffractive electroproduction (Figure 8b)

u + nucleus •+• u + ¥ + gluoniua + nucleus (5.3)

Naither of these processes Is blessed with a big cross section. Probably

the most attractive option is to resonantly produce V (or perhaps T) and

look at the radiative decay "^

* -t- y + X

T- T + X (5.4)

The principal decay mechanism for ¥ or T is supposed to be 3-gluon

annihilation (Figure 9a) irlth subsequent materialization into hadrons.

The radiative decay should therefore proceed via a virtual -ygg channel

(Figure 9b), ideal for formation of gluonium as the state X. Ferturbative

QCD (not to he trusted much * at the f mass-scale) predicts *

1 9 1 8

$^f£ ~ «* (5.5)
To good approximation the gluons (or v-rays) are uniformly distributed In

Che Dalitz triangle; thi3 immediately leads to an energy spectrum of the

T peaked (Figure 10a) at the high end (higher order radiative corrections

round off the upper end; we don't worry about that here) • Thus the mass

spectrum of the recoiling system X—which hopefully contains the gluonia—

is peaked (Figure 10b) at the low energy end. In this region we would

expect the gluonium resonances to dominate and turn the specrrum into
 gluonium
V

@-*-
Pomeron • giuonium

( b) M«A»

Fig. 8. Mechanisms for giuonium production: (a) e e -* ijig» and

Cb) uN - n«gN,
 • * "

-„
1 (o)

Fig. 9. Gluon (a) and gluonlum (b> p r o d u c t i o n in i> ( o r v) d e c a y s .

 After Radiation
Correction -y

U)

Fig. 10. Spectra In (a) y-ray energy and (b), (c), recoiling mass
for the process <t> •* Y + X .
 - {>•> -

something like Figure 10c. There la not yet much known about the jiclusive
•Ingle v spectrum from "-decays. The decays

f •*• Y H

yn'

y£

have been measured. If they dominate the mass spectrum n fi 2 Ge\T and

arc duel to the perturbative (parton) estimate they should account for

"•4a% of the decays V -*• y + X. The measured exclusive channels actually

account for '•1/27. of the "P-decays, Implying by this bookkeeping a branching-

ratio ~1Z for all the radiative decays.

A study of these exclusive channels from a QCD viewpoint has been

made by Kramner who looks in detail at the angular correlations In the yf

50
channel. Assuming mediation by two transverse gluona (Figure 11) coupled
3 —
to a P qq system, the helicity state of the J - 2 TTTT system (it is
2

51
predominately ±2) is predicted and compared with the data on the Tin

angular correlations. The agreement with QCD is good and appears to be

52
nontrlvial.

If these exclusive channels are really mediated by two gluons, then

there should be other analogous channels which are reasonably well

predicted by the quark-model. In particular, a brief look at Table 5-2

lets one accumulate the following list.

V * tt'Y AlY

HY Elf

6y A2y

S*Y f'Y (5.6)

 - 66 -

Fig. 11. Mechanism for the decay chain ty •*• yt •* yms.

But an even more Interesting question is what lies beyond. If nirrov

gluonlum states dominate In the region from H = 1.4 GeV to :i - 2 GeV, they

should provide -30X of all radiative decay mode#. The Y-ray energies arc

1 GeV, and probably badly buried in contamination from ir decays. A 27.

Y~iay energy resolution corresponds to a resolution in gluonium mass of

order 30 MeV. It may be unrealistic to try to resolve any gluonlum

lines by measurement of the recoil y-rays alone—even using Crystal Ball—

and reduction of background by looking for exclusive gluonium decay

channels may be needed. Here one might try for some of those involving

neutral decays, e.g. n- But it will be difficult. A scenario appropriate

for the Crystal Ball might be

¥ •*• y + gluonlum 1/21

Ln + n 12
LYY LYY 38% * 38X - 142 (5.7)

lecays, even with i

branching ratio assumed for the i\n decay-channel.

If the mass-scale for gluonia i s "surprisingly" large, say ~3 GeV,

then i t is necessary to u t i l i z e the T decays. For the T, the QCD pre­

8
diction i s "

r(T •» T + X)
•3% (5.8)
r(T •+ a l l )

with again the same spectral shape for the recoil-mass distribution. Fo'

a 3 GeV gluonium system recoiling against the Y- the y-ray will have 902

of the endpoint energy, namely E - 4.3 GeV. A 1% Y-ray energy-resolution

now yields a 300 HeV mass-resolution. Again final-state information on

 - 68 -

the gluonium decay-products probably has to be invoked. Because of the

large cuss, this will, be even more difficult than in the previous case.

Finally, if clean gluon-Jets are isolated in T-decays, decays of even

more massive onia, or in hard processes such as « e •*• q + q + g, it is

plausible chat the "leading particle" In such <x Jet will often be

gluonium. Thus in such kinematic regions it Is especially appropriate

to search for the decay channels enumerated in Table 5-2.

6. mCLUSIOH OF HEAVY SOURCES; 0N1UH PHENOMENOLOGY

In Stage II, we modify the pure QCD of gluonium by introduction of

heavy-quark sources. la theory (cf. Section 2) a universal linear confin­

ing potential is expected between such heavy sources, provided they have

non-vanishing color-triality. But beyond this theoretical issue, Stage II

is of Interest as a description of the phenomenology of charmonium,

bottomonium, toponium, or any other even, heavier oaia which might eventually

be observed.

We start with the idealized situation of very heavy quarks, Q, whose

binding can be described accurately in terms of the short distance r

potential. The static interaction energy of Q and Q in a color-singlet

state can be read off from the expression for the quark-current source of

the color field. Equation (3.16), inasmuch as the subsequent quantization

procedure followed accurately (to lowest order in a ) that of QED:

and that the 3 * 3 matrix Z* (IV) is color-invariant, hence a multiple
i
A
of the unit matrix. Thus, using the normalization in Equation (3.33)

X J J 3
A-l A-l ^

Just about everything about this system is analogous fa QED. The binding-

energy of superheavy onlum of mass M ~ 2H. is thus

and the level-spacingft are hydrDgenic. Thus the 2S end IS levels are

split by an amount

The Bohr-radius of the onium is

~ 3 1 « 3 ....
9 S

Thus pure Coulomblc "size" of bottomonium (taking a > 0.2) is £0.3f, not

especially small.

The value of o_, just as in QED, depends upon momentum transfer as a

consequence of vacuum-polarization effects. The dependence on distance-

scale has already been cited in Equation (1.4). ond it is appropriate to

here discuas this contribution in more detail. A diagrammatic analysis

53
can be carried out; in Coulomb-gauge the vertex and self-energy inser­

tions (Figure 12) are unimportant. All that happens is a cancellation

of divergent renormalization constants Z. and Z. just as in QED. The

 c- <T~ C
(o) (b) (c)

Fig. 12. Vertex and eelf-energy Insertions for the Coulomb energy.
corrections C"ae from the three diagrams In Figure 13. Only the first

(Figure 13a) earning from fermion-loops, exists in QED. It tends to

Increase the charge at short distances because of vacuum polarlzabillty.

The calculation Is just like QED vacuum polarization except for the group

factors. Because the polarization loop is diagonal and Independent of

the color label A • 1, 2,... ,8, we may choose A • 3, which measures the

color-lsospin of the quarks. Since the coupling of A (x) to quark sources

Is

T cT <6,6)
f"S"f 3" 3
the modification from QED is just to make the replacement

2 2 2
• k — c.o E
lors
3-*[®
*-
*(-*) *o|-K
•*
«•»
of quarks

I s QED, v e have

2
ta> 2
4-%[l-. (q )] (6.8)
1 1

_ | .2

4
Thus i n QCD we get from t h i s c o n t r i b u t i o n

'"-£"«*. 4
4

o r s £ t e r renormaliz-ition and summation of t h e g e o a e t r l c series

Z Z 6 2
a(q ) a(» ) " q
 _:0' .•_G-
3" q-T q"

n-n (e) I*«AI*

13. Vacuum polarization corrections to the Coulomb interaction:

(a) quark pairs, (b) transverse gLuon pairs, and (c) correc­
tion to instantaneous interaction via coupling to transverse
gluons.
The second contribution (Figure 13b) from pair production of transverse
gluons by the Coulomb field is physically similar, and In fact has the
same sijpi as the first contribution . The coupling of gluons to the
Coulomb-field is through the convection-current, with polarization vectors
c and e' dotted together (Figure 14a)
jt^ttertag _ ( E + E .) * . p (6.12)

For the crossed process, there is a minus-sign (Figure 14b)

1
J** * ~ (E - E-) t • t<

~ljJ.t-P (6.13)
showing that there is an explicit |<j| factor emerging in the vacuum

polarization.

Since the polarization vectors can be safely dotted together and do

not enter the guts of the calculation, the structure of the vac:

polarisation amplitude is the same as for a pair of spin-zero gluons,

which in turn is 1/4 of that for spin-1/2 quanta (given the same charge

of the source) . (Recall that R , - 1/4 for a pair of charged splnless

e e"
point particles.) Hence there are only countlog-factors to consider.

From this source we therefore get

•— spin zero vs. spin 1/2

2
1 °„ «
' • (gluons
£^)- • i • ^ 10, AT (6.14)

polarization Gum. 1 L-spin J1/2 result

l—spin
 V'"
ij-Tt (a) (b) 1**3*11

Fig. 14. Scattering (a) and pair-production (b) of czanivetse gluons

by a Coulaab field.
The gluon i s o s p i n sum i s

ttnd upon a g a i n summing che geometric s c r i e s , wc g e t

f6 16)
-TVTVG^)
o(q ) a{m ) ^ '
1 0 8
^
q
-
:
again the Vrong" sign £o_ asymptotic freedom.

The third contribution (Figure 13c) does not exist In QED and comes

from the coupling of the Coulomb field of the quarks to the vacuum fluctua­

tion of the (transverse) gauge quanta. It has the important sign-change

leading to asymptotic freedom, i.e. the decrease of the effective coupling

st short distances. It is in fact much larger than the other two and

doninates everything. The physics of the third term is not 100Z trans­
51
parent * But is roughly as follows: the Coulomb field created by the

quarks itself carries color and therefore interacts with the vacuum

fluctuations of the (transverse) color fields. This jitters the Coulomb

field and rounds off the short-distance singularity. Whatever the validity

of this hand-waving, the calculation follows directly from the formula

2 -»••*•
for Gauss' Law," expanded out to order e . One gets (ignoring A • ,E
t e r l n , w h i c h
• A T ' ~T is already accounted for) from Equation (3.26)

7 • 1

Upon i n v e r t i n g th'.s for EL

Nov we can put in plane waves for A and, in the last tern, contract theo

together to obtain a vacuum-polarization contribution. After routine

Fourier transformation (Figure 13c) the structure of the term is

% 2 3 2 2 W .19)
o [° £l ej 2|k|( ) 2n \t -1J| 9 ° J
The Clebsch-Gordan i s t h i s t i n e twice as big, because E_ and E. are

not identical degrees of freedom

3AB 2
Z-r Cf ) - 2CD + 4<|) - 3 (6.20)

A,B

Using E * k - 0, and going to the zero-frequency limit, we get a

correction
A
J»
4'iia * 3 •-= q • I =—=• • ~ ~ r " ~ — l o g —=•
3 3 2 2
» Jq i s A ISI ' =

In computing tho "electrostatic" energy

2
jP" 2 C $ • | / r x J„ i J + . . . 0 (6.22)

we get another such tern from the other E, , and a third identical 'lerm

from contracting the 0(e) contributions from each of the E, factors. Thus

the entire contribution from this source to tho effect "e charge is three

times what we have written down. While not at all obvious, the geometric
55
sum does survive these comblnatorLC complications and the net change in

coupling-constant from this last source is

e — log —z (6.23)
 We now can add together the three contributions. Writing in momentum-

space, the interaction energy as

I W2)
V(,)-. 1^-5-

we have

(a) _ ^ _ Y M 2
° 8
< M 2
> 2 M 2

111 q

2
... a (M ) 2
t b >
c - - ^v—log-2=5- (6.16)

3 2
(c) , °«<" > • »2 ..,.,
c' = + - log 5_- (6.23)

where i;-. c the sure over fermion flavors goes only over those fermions

whose masses ere small compared to the momentum scale of interest. Sum­

ming up the three contributions leads to the aforementioned formula for

the running coupling constant a(q ) :

2
(33 - 2N ) J.
1 1 t f f t l 3\, M • rr—f- log SL (J.24)
2
afH )

We see "asymptotic freedom" is controlled by the modifications to "Gauss'

Law" due to the vacuum fluctuations of the transverse fields in inter­

action with the Coulomb field of the sources.

Notice the definition of the coupling constant a (ir) is such that

at |q| = VT, Che higher-order corrections vanish. Here there is some

56
difference with the traditional QED renormalization program, which

defines the physical charge in terms of the force at large distances

(as q •*• 0). In QCD, weak counting is at short distances only, and there

is no unique choice of mass-scale at which to normalize. This leads to

arbitrariness in definition of the coupling constant; this freedom can

lead to confusion in the interpretation of the theory. One must choose a

normalization at some very short-distance scale M (a good choice is

M ~ I O where perturbation-theory is manifestly good. At this seal '< vm

may for example choose the definition such that the vacuum-polarization

corrections to the static potential vanish. Then, by convention, the

coupling-constant at this mass scale is defined to be

a.CM ) -= 2
^ ~ ^ (6.25)
(33 - 2X Q?»
f l o g >£
2
A

2
The dlmensionless number a (M ) is traded In for the parameter A which is

an energy-scale. The convenience cf this definition rests in ti.i fact that,

for all values of q for which the perturbative calculation, Equation (6.24)

is accurate (and only for such values) one hafl

a
a,(q i *~ a (q ) " •£ (6,26)
2
133 - ZH.(q )]lof! 3-
£ Z
A

Thus the expression is form-invariant; \ does not explicitly depend upon

the choice of the scale at which the coupling constant is defined- it

least to this order of approximation.

What is the significance of the parameter A? It is evidently a rouRh

measure of the point at which perturbation-theory breaks down. However,

when q — A , there irf no Justification left for believing the form of

Equation (6,26), Also, while the parameter A is claimed to be measured

reasonably accurately (let us say, between 300 and 700 MeV), this does

not mean that we have accurate knowledge of the coupling-constant. For

2 *
example. Figure IS shows the spread in o (q ) allowed by the uncertainties

in knowledge of A, We may conclude that it makes little sense to even

2 2 2

consider perturbation-theory below Q — 1 Gev , and that Q in excess of

5 to 10 Gev are needed to bring a down to a small enough value to trust

quantitative results.

Actually the slippery nature of coupling-constant renoroalization,

57
when combined with the proven renormalizability of QCD, becomes a

powerful tool for summing higher orders of perturbation-theory. This is

the "renoroalization-group" method. The application to the behavior of

the QCD running-coupling constant is just like that of QED. Without any

explicit diagrammatic calculation, the Q -dependence resulting from sua-

maticn of leading and next-to-leading logarithms can be determined. In

particular one justifies that the leading approximation is tho summation

of the simple vacuum-polarization nibbles into a geometric scriea, a

result which is not so obvious in QCD as in QED. A further result of the

rmormallzat ion-group calculations is that if one starts with a small

coupling constant a CM ) defined at a short distance scale H and tries to

2

accurately extend it back to large distances, the convergence is rather

slow. The formula for a becomes, after summation of next-to-lead.'ng

58
log. :

,..<«v(*.4)

, , 03-2N,) „2 3(153-19N) « (q')

(M
> " 12, l0
SS h 8 - -T
l0 §

"* q 2n<33-2N.r «(lO

0.8 v- i 1 1 r

0.6

0.4 A=700 MeV -

0.2

Fig. 15. Q dependence oE the tunning coupling constant for

A - 500*200 MeV.
with a the "standard form," Equation (6.26), for the running coupling
8
constant. As q gets small, the correction becomes Important whan the

1* never large, despite that fact that

,lo log_ _
Bu 3 ( 6 > 2 8 )

"a"'CO log q

giving a very slow rate of convergence.

Let us now return to the question of onlun bound states. The ques­

tion of higher-order corrections to (and even existence Of} the potential

has been extensively studied. For the most part, the development runs In

parallel with the treatment of QED bound states. There is one somewhat
59
n«w feature which appears in high orders, and that is chat the

Interaction-energy has terms a £n a . These come about from diagrams

(

as In Figure 16 where the intemedl**:e (transverse) gluou is very soft,

and produces an infrared divergex.ee. The infrared divergence is cut off

by the energy-splitting of the intermediate (color-octet!) QQ configuration

from the color-singlet QQ bound state. But beyond this complication there

la no signal of confinement other than the behavior of the running

coupling-constant.

How does all this compare with the properties of the ¥ and T systems ?

Koit of the spectroscopic facta of those systems rest on a nonrelativlstic

nodal of quark motion and are not very specific to QCD. Given the
60
assumption of nonrelatlvistic motion, Quigg, Rosner, and Thacker have

used the level-spaclngs of the ¥ and T states, along with the wave-functions
 o 5

::iq
it-Tt 3*»2M1

Fig. 16. H-diagran contribution to the Coulomb energy.

reconstruct the form of the potential. They find (Figure 17) a rather

good linear potential at moderate distances, with some tendencies toward

a 1/r singularity at short distances. The slope of the linear potential

2
(string-tension?) is 1 GeV/f in agreement with vhst is inferred from the
51
string model of ordinary hadrons. John Richardson has turned things

around and taken the QCD form, Equation Id.2b) and made a very simple

ad hoc change jhich turns the potential at large distances into a linear

potential. He writes, In momentum space

1 6 t A 2
V(q) - §• 12= g- * • - T (6.29)
2
<33-2H.)o. lo U 4> <33-2N )q
8 + f
x
AT
Aq behavior as q -*• 0 corresponds to a linearly rising potential. In
taro* of a string model, the coefficient is related to the string-tension
T as follows:

2 2 2
T - \ q o(q ) I == 3 3 ^ fi * 0.25 ± 0.15 CeV ~ 1 CeV/fermi (6.30)
2
9. -0
Whether or not this has any fundamental basis, this potential does well
In describing the level-Qtructure of charmonium. Other calculations for
62
even heavier onia have been made, as shown in Fifiure 18. One sees
that the pure QCO Coulomb potential is not applicable until the ontum
mass exceeds - 100 GcV.
Does QCD have much to say about fine and hyperfino-structure? Here
the situation is not very clear for several reasons. First of a l l , the
detailed origin of the linear component of the potential is theoretically
unclear; hence it is not reasonable to expect spin-dependent refinements
Fig. 17. Typlcnl charmonium potential as reconstructed by Thackcr,
Qulgg, and Rosncr (we choose n - 1.2 GcV, and E - 3.8 GcV).
 I l l 1 1
1 i

-^_3D
^4St^-——

—-£2~"~~ ^^__J^^
--is -
2P -
ID -
. 2S
IP -

-
, IS, , i 1
16 64
n (GeV)
Q

Fig. 18. Oniuo Levels as a function of mass. From Kef. 50.

to It to bo under control. Perhaps the most reliable (and also ocst
successful) QCD application is the pattern of hyperflno splittings. These
axe typically short-dlstunce In nature and operative in a-states, so chat
one may hope that a perturbative estimate is at least qualitatively
adequate. The most interesting result from QCD Is that the signs of the
splitting come out correctly both for qq and qqq systems. Thus for qq
systems, the relative signs between charge and spin couplings are opposite
to (JED, because of the £, • \~ factor, and S, lies above S_. this
pattern is universal among the quarks; p, u, d l i e above ir, n> n and D
*
l i e s above D, Thus Y had better l i e above n , although the magnitude of
the splitting is not certain.
For qq systems, the agreement persists. The QCD prediction ifl that
a (1236) Ilea above the nucleoli. This can be seen fairly easily; in going
from a S. o e* to e S_ e e In QED, the sign of the spin-dependtnt force
changes sign-, hence is attractive, However, In QCD, there i s s t i l l the
V • _X, color factor which will be negative In 3 and positive in 6. Thus
the energy between two quarks goes as

Then In the &, where each quark pair Is In a spin-triplet atate vith
( o • Oj) - +1
t

4 + k X + (6 32)
<"IVlnl > " -<\ • »2 2 • 3 S • h> ' * 1 4 -
end there is n c repulsion which is optimally strong. Thus H must lie
below &.
The situation regarding fine-otrucruri splittings is confused. A
Breit-type hamlltonlan built In analogy to QED predicts quite a large
 - 87 -

amount of spin-orbit splitting among ordinary hadron multiplies. However,

there Is relatively little observed. The problem is most acute In the

P-wave 70, L • 1 baryon states, where there is remarkably little spin-

orbit coupling found. There i£> some success in organizing the P. states

of the charmonium system using semiphcnomenologlcal potentials, but the

65
picture is not a completely simple one.

There axe many detailed studies of the level splittings from both a
63
QCD and directly phenomenological point of view. The above discussion

only scratches the surface. From the present point of view, many of the

analyses are not fully relevant because they incorporate the old evidence

for the n at 2.8 GeV as seen by DASP. A rather thorough study of baryon
G1
splittings has been made by Karl and Isgur, * who account for a remarkable

amount of data on 70, L * 1 and even on _56_, L = 2 from a semiphenomeno-

logical QCD starting-point.

Ua next turn to transitions between onium levels. In Stage II, many

of the low-lying levels are stable with respect to strong interactions,

because the only open channels contain (probably relatively massive)

gluonlum statas. Nevertheless, we may induce transitions (e.g., in the

fcmtounlverse) by slowly varying external gluon fields. Under this

circumstance, a multipolc-expansion Is suggestive. This has received

qultD a bit of study." Effective local color-singlct operators, e.g.,

ai written in Table 5-1 (gluonium field operators), will have calculable

couplings, Including transition-couplings, to thu onia. This is still a

long way from estimating hadronic widths for onium transitions. For

that one needs to know the spectrum of gluania and their decay widths

Into ordinary hadrons. However, some information on relative widths can

be gleaned by analysis of angular-momentum barrier factors present in

the various transition channels. For example the p x (3510) state is

forbidden to decay into two gluons, and operators higher order in a need

to be invoked to induce the decay. It follous that iti hadronic width

would be expected to be smaller than for the ?. . x-»tates. This seems

to be the case experimentally.

The annihilation channels, where the Q-value is large compared to the

confinement scale, offer a circumstance where perturbative QCD should be

applicable (onium can happily annihilate even when, confined in the

femtouniverae; thus its decay-vidth should be calculable perturbatively).

The width for T •*• ggg can be stolen from the 1949 CO QED calculation by
67
Ore and Powell for the 3-photon annihilation of positronium.

r<T - ggg) - -^ (~\ * (WFC) • r ( 3

S l pOBitroniom •«• m > • y~) C&,33)

The factor (WFC) is a wave-function correcCion originating fron the fact

that the onium wave function is not purely Coulomhic. The numerical

factor comes from the color wave function.

formula. Me may contemplate three choices for q :

2 2
a. q • m_: This is the time scale over which the annihilation
process takas place.
2 2
b. q • 1/9 oil. This corresponds to a distance scale equal to the
Compton wavelength of each gluon.
2 2 2
c. q - a m_. This corresponds roughly to the size of the onium hound
state (the Initial state should fit into the femtounlverse).
This gives a spread In possible <] values fion'-3 CelT to—100 Gelr, and
a corresponding uncertainty in a (from Figure 15 ) of a factor <4. Thus
the theoretical width for the T is (conservatively) uncertain to a factor
£60. That i s , any experimental value within that range could be rationalized
at being in agreement: with QCD.
Given this situation, a reasonable way to proc-ed i s to f i t the value
of a to the observed T width of —100 keV. In this way one arrives at a
value
a (T + 3g) ~ 0.2
s (6.30

This implies, from Figure 15, a Q value ~10 Gev , which i s quite
acceptable- Ue can also infer froa sltfple uimcnsional considerations that
2
the Q -value for ¥-decay should be about 10 tines smaller. Hence, were

perturbatlve QCD ideas to be acceptable for the V * ggg decay, we should

have

r<v - 3g> _ ret -.- 3g> . y v ,, .„

P(f * up) - 4r<T * up) L C ojj

Tha leptonic branching r a t i o for the T i s s t i l l poorly known. But putting

in numbers for these widths, one finds

•' I si 1.0 ± 0.2 («.36)

«W
Even qualitatively this is not what is expected. From Figure 15, there

should be a decrease in a of 502 to a factor 2 it. going from the * system

to the T system.
 - 90 -

The easiest conclusion to draw from ti:i.i Is that the mass of the ¥

Is much too low to allow quantitative applications of perturbative QCD.

This should be no surprise inasmuch as each gluon carries an average of "'I

••9
GeV of momentum. In addition, recently calculated radiative corrections

turn out to he remarkably large, even for the T system.

A perhaps better test of percutbativo. QCD is to look for the 3-jet

final state in T-decay, and also the radiative process

T * ggy (6.37)

The branching ratio for this process is estimated to be—3X. These ques­
2
tions are discussed In detailfayStan Bradsky.

7. INCLUSION OF LIGHT FEKMIQHS

Thus far we have not faced directly the real-life situation present

when the light fennlons are Included in the theory. He have already

indicated the basic changes which occur. The most important Is the

appearance of the ordinary hadrona. Qualitatively there Is not much of

an additional conceptual problem. The problem Is a quantitative one:

can we understand the classification and the mass-spectrum—especially

of the excited resonant states—and their couplings to each other? This

question is a very big and difficult one to handle theoretically, and

will not be directly attacked here. It would seem necessary to have a

rather firm control of the mechanism of confinement before one had firm

control of such questions.

Along with the Introduction of the light quarks goes the disappearar.ee
11
of the string, which can break due to (Heiseaberg-^ler) pair creation.

Actually, the lifetime of a piece of strlr^ Is somewhat uncertain. Xf

 - 91 -

Msons of large J can be considered as quarks rotating about each other

and connected by a piece of string, then their lifetimes (T as 10-100 MeV)

give some measure of the string lifetime. The relatively large widths of

chmnuniu-j states which lie above DD threshold are another indicator. But

in any case the linear potential becomes complex (absorptive) as r

Increases, and ve should not expect that the concept continues to make

much sense when, say, Re V(r) >> 1 CeV. This is because Im V(r) grows

with r (linearly?, quadratically7) is well as Re V(r).

Introduction of light quarks will also modify the properties of the

gluonia discussed in Section 4. Gluonia may mix with the ordinary mesons,

and will also decay into meson channels. However, as already mentioned,

these are not expected to be large effects. Eloquent argumentation fot

68
this has been given by Wit t e n on Che basis of the 1/N expansion, where

S i» the H of SO(H). One hopes H - 3 may be "large"; as H •+ » one can

argue that gluonia decouple from quarks (as veil as from each otherI).

While probably most theorists expect gluonia to survive as distinguishable

atatcs even in the presence of mixing with ordinary mesons, a group from
54
ITEP has challenged this view. Their argument is based on the QCD sum-

rules for the charmonium system and a generalization to the gluonla. The

QCD sum rules are interesting in their own right. Weighted integrals of

the colllding-beam cross section are related to matrix elements of various

local operators via the Wilson operator-product expansion. The structure

i»>

ds _, , m suras of vacuum matrix

/
n 8
N elements of local operators. (7.1)
 - 92 -

The ri£ht-hand side Is calculable In terms of short-distance properties of

QCD—the more so the smaller the value of N. With some mathematical

trickery they can sum things up and also write

. -s/H - .
f another sum of vacuum matrix
/0 as e K\S) -e l e m e n t a operators.
o f l Q C a l <7.2>

When they consider separately the R, associated with currents of u, d and

a and c' -)*-" M appropriately, the sun is saturated by p, u, and $. The

result is a successful calculation of masses and widths of these resonances

in terms of short-distance parameters of QCD!! They then turn 70

to

processes such as discussed in the beginning of Chapter V, created by local

2 2 -*• •*
c»loT-slnglet gauge-Invariant operators such as Tr(£ - £ ) or Tr E * B,

The corresponding sum rules are constructed analogously to Equation (7.2).

The right-hand side can be evaluated from the ^formation already obtained

from the previous sum rules. They find that the le^t-hand side can be

saturated by known mesonic states such -s E or n' And there is no need

for additional disrinct gluonium states. This does not of course prove

such states do not exist; however, it does undermine the notion that

Inclusion of light fermions does not significantly modify the properties

of the rluoaia, since without the fermions present there would be no

alternative but to saturate the QCD sum rules with gluonium states.

In the context of QCD, some of the most interesting aspects of the

presence of light quarks have to do with questions of symmetry--£ca?n the

approximate chiral flavor symmetry SU(3) ® SU(3) of current algebra to

the discrete symmetries of C, P and T. The issues involved are quite

 • 93 -

subtle, and ve shall concentrate our attention in this section an them.

The first Issues have to do with chiral symmetry. Introduction of the

fermions u , d, s,... adds to the QED Hamiltonian a term

".trong " > W QO. +

S *# ' <•* " % > «
4 +
^ ' j 7
< - ' 3

where the mass-matrix is usually presumed to originate from the Higgs-

meehanlsm of the electroweak interaction. Because the Higgs sector is

not expected to respect internal symmetries, it follows that m.. need not

be diagonal. In fact the mass matrix may even contain y,. Nevertheless,

because of the presumed diagonal and flavor-independent nature of the

kinetic-energy and i • A terms, one can redefine the fields Y. in such a

way that the mass term l n H . is diagonal and f, free. This is an

strong 5
important feature: it implies (or would seem to imply) that, despite P,

C, and/or T violating effects in the Higgs sector (or whatever) responsible

for quark mass generation, such symmetry violation will not find Its way

Into the strong interactions. This desirable result will be tempered some­

what In what follows: the Tr £ * JJ surface-term does induce CP violating

effects in strong Interactions via instantons. Vie shall come tD this

later on.

In the absence of the mass terms (a reasonable approximation for u,

d, a ) , the Hamiltonian is invariant under independent rotations (In flavor

space) of the left- and right-handed fermlon fields

U
V * * i J *Jt<x>

(7.4)
 u U
leading to a chiral (3) ® ^ R symmetry. This is broken only by the

"small" uaas terms, Eighteen vector and axial currents can be c"n«rmeted

which in the limit m , n , m •* 0 are formally conserved by the equations

of motion- However, one of them, the 9th flavor-singlet axial current,

is not conserved because the short-distance ultraviolet divergences of

the theory do not allow it. This phenomenon is the triangle-anomaly: the

divergent graphs of Figure 19 do not allow the shifts of origin in

momentum space required to obtain the formal vanishing of the divergence

71
of this axial current. Defining

Jy5 "
J
l5 2-d
l-u,d,...
\ Vp^i (7
' 5)

the result is

W5 s f _, ± •* . terms vanishing

where n f Is the relevant number of flavors (3 for u, d, s ) . The Tr JI • j*

factor should already be familiar from Section 4 on instantons and from

Appendix B. There we found that the term Tr £ • B, itself is a total

divergence

.2
r Tr E • B - -jj^ (7.7)

where the current

- ^ - TrfA 3 AD - ~ A A.A 1 (7.8)

an

is gauge-dependent. The operator N - / K-d x measured the topological

quantum number uued to classify the QCD vacuua. It might appear that the
Pig. 19. The triangle anomaly.
 - 96 -

U(l) ehiral symmetry could be salvaged by considering the summed current

& =• J + K which is conserved. The conserved charge would be in this

u u u —
case

ft/* 2 S
= "tot - «5 <• "£ • -^„ «Li " V
fl
+
'"f" "•»>
where N counts the number of gauge-bubbles present in tha QCD vacuum.

However, even in the absence o£ the light quarks, we have learned that we

cannot characterize the vacuum by the quantum number N. Because o£ gauge

invariance and vacuum tunnelling via iustantons, it is its conjugate

variable 6 that labels the vaeuua. So also it will be for the chiral

charge Q„; the variable (J conjugate to Q , which is a phase, is used to

g

characterize the chiral structure of the vacuum. This is the same as the

situation for spontaneous chital symmetry breakdown in the more conven­

tional context: the vacuum is not an eigenstate of the chiral charge.

We have not yet motivated why this chiral symmetry-breaking muse

occur. This has to do with the instanton phenomenon. The different

N-vacuua become coupled because of the existence of a non-vanishing

quantum tunnelling amplitude (the Instanton). With fermions present, we

must reanalyze the tunnelling process and keep track of how the quark-

states are affected by, say, the creation of an intitanton-induced gauge-

72
bubble. Consider a femtouniverse where the quark coupling to trans­

verse gluons nay be considered perturbatively (i.e., neglected). If we

are in an N = 0 sector the quark states are simple plane waves of definite

chirality (with the approximation m , =• 0) and quantized momenta.

 - 97 -

la there anything different a'.out the solutions in the presence of a

gauge-bubble? The answer is no: only a color-dependent gauge-phase

need be multiplied to the quark wavc-functions. The energy-eigenvalues

(which arc gauge invariant) remain unchanged at i • ,,~ , ± ~~TTf > e t c

-

[Hote: for convenience we choose to quantize ..-ith anciperiodic boundary

conditions V G ) * -?<- -r) to avoid cigenstates of zero energy,J However,

We nay ask what goes on as the tunnelling occurs, We shall analyze this

problem in first-quantization. That is, we treat each fermion state

Individually and only introduce the filled Dir3C-sea of negative-energy

elgenstates at a later point. Because the tunnelling is semi-classical

the coordinates of a given quark must be deformed along with the gauge-

field coordinates. In the intermediate configurations between K = 0 and

W - 1 there arc no pure gauges; the quark finds itself in real color-

electric and color-magnetic fields. Thus the levels will shift—and if

73 7
the tunnelling rate i3 slow, they will shift adiabsticslly: ' '' only

degenerate fermion states mix. However* as the tunnelling becomes complete,

one returns again to a pure gauge configuration—and therefore the same

*«t of energy levels as in the beginning. But the important and crucial

feature of thin process is that the matching of final levels to Initial

levels is nontrjvial. Some of the initial levels move upward; others

downward and end up, at N •> 1, in different states (c.f. Figure 20). What

does this mean? It means that, when one second-quantizes and—according

to Dirac hole theory—fills all negative-energy states, the tunnelling

1
phenomenon takes one from the N = 0 vacuum of filled fermion state - to

an N • 1 state which i3 not necessarily the N * 1 vacuum. In fact, if

any of the ferralon energy levels does cross zero, then the H -1 state
Fig. 20. Schematic picture of shifts in ferrtion energies as a
function of winding-number N.
 • <)9

which is reached will not be the naive N = 1 vacuum state. This is

actually what happens. In tunnelling from N = 0 to N = 1, one negative-

helicity state (for a given flavor) of originally positive energy dives

into the negative energy-sea, while one positive-helicity state emerges

75
from the negative-energy aca and joins with the positive energy states.

Thus, the net effect is that the Instanton-induced transition from the

N • 0 vacuum to the N = 1 vacuum la suppressed; Instead the transition

simultaneously creates a pair of each flavor of "maasless" fennion, since

the levels of all flavors are shifted together. In our case, that means

the transition is accompanied by creation of three quarks and three

antiquarks

(vac) _ t+
N Q ( v a c ) ^ + im dd ss (7.10)

*
Renarkably enough, this process respects the conservation law Implied by

Equations <7.6) and (7.7):

A E< N
R i - N
L 1 )\" 2n f AN (7.11)

72
When applied to the weak-interaction gauge theory, this phenomenon
is even more spectacular. In a femtouniverse with dimension small com­
pared to mJJ , one expect3 to have an essentially unbroken SU(2) non- L

Abelian gauge theory. Thus the instanton creates one left-handed

fermion from each weak doublet, consistent with charge-conservation. For
three generations this means:
 - 100 -

This cannot be an accident; the triangle anomaly must sense at the hijft

momenta the imbalance of fermion levels Induced by gauge-bubbles and

nonvanishing values of /d x E_ • .B. One may recall that even in QED the

definition of tho electrical-current operator requires a careful sym-

metrizatioa between positive and negative energies. The monotonic shift

in levels of given chirality induced by an E * .B term will be felt not

only in terms of levels which cross zero energy but also (in the presence

of a high-momentum cutoff) of an Induced asymnctry at the highest momenta

75
and consequent multilation of the structure of the current operator.

What now are the stationary states? The tunnelling effective

Hamiltonian analogous to Equation (4,6) now has an extra factor to account

for the quark pair-creation we have found. Schematically

a C X )
^ Cuu) (dd) (¥s)e ~ s (2l) ... (7.12)

When H' is evaluated in a 6-vacuum, the H-dependcnt factor evidently

becomes a phase

For example, this means there exist |6B[ » |AL| - 3 virtual transitions

(vac) -*—»-(vac)
H N+1 + v + n + u" + (ccs) + T " + (ttb)
g

Thus a pure 3rd generation baryon could decay into a pure second-generation
antibaryon, an antineutron, and 3 antileptons. Regrettably (or perhaps
fortunately) the amplitude for this process is extremely small, of
order

- 2n , 2„ , -60
n
 - 101 -

For diagonal matrix-elements it is necessary to contract the fermion-

fields uii •*• <uu>, etc.; and such contractions vanish If the cusses of the

quarks are zero. Hence all instanton-related effects will now be

proportional to the product n m ,n , where these are the current-algebra

masses (a reasonable estimate is m — it HeV, m — 7 MoV, m — 150 HeV).

Furthermore the phase transformations on the quark fields necessary to

put the nass-matrix^tflnto standard form / ^q. Im Jq will also leave a

± a. x L
phase $ in the expression (Equation (7.13)) for the tunnelling energy In
a 0-vacuum. This phase is easily to be found to be
$ - arg det.tf (7.14)

and hence the effective angle relevant to observable effects, such as in

Equation (7.13) will be
-
T - S - * - 8 - arg d e t ^ / (7.15)

Thus even if strong-interactions have for some reason a boundary condition

8 " 0, after Inclusion of masts-generation (via Hlgga-electroveak mechanisms)
the effective angle Q cannot be expected to remain zero.
Thus far we hove found that the U_. ® U-_ symmetry has been broken by
instanton effects to an SU(3). & SW(3) ® U(l)„ symmetry. What about
spontaneous chiral symmetry breaking? The flavor-octet of axial currents
J q ir Y <l i s n o t la ei w l t h
U " ' ii 5*i P S^ instanton couplings or triangle
anomalies, and one may attack the problem In more conventional ways. Even
in QED, it has been argued that chlral symmetry can be spontaneously
76
broken, i.e., lepton mass can be generated seif-conslstently by the
 - 102 -

mechanism illustrated In Figure 21, The same mechanism with gluon replac­

ing photon might apply in QCD. A different line of argument has been
7
advanced by Callan, Dashen, and Gross," who exploit the Instanton

tunnelling and propose a mechanism as shown in Figure 22. The effective

instanton-induced six-fermion interaction in Equation (7,13) is used to

trigger the spontaneous symmetry breakdown. But this is difficult to

make quantitative. In addition, recent explicit calculations in strong-

79
coupling lattice gauge theories are also supportive that spontaneous

symmetry breaking of SU(3) chiral symmetry will occur in QCD. In all

cases much more needs to be dona before one can be fully convinced that '

QCD does Imply the spontaneously broken chiral SU(3) symmetry which

underlies the successful current-algebra and FCA.G phenomenology of

ordinary hadrons,

Finally, It is necessary to inquire into the implications of a

nonvanlshing value of 0 for the questions of C? violation. Evidently a

t«na in the Lagrangian

f - —2-r Tr E • B (7.16)
Z
2«7f ~ ~

which serves to fix the value of vacuum-8 is P-odd and C-even; hence

T-odd. While formally a total divergence, we have had ample evidence that

such a term produces nontrivial effects. The greatest threat lies in the
-24
experimental limit of 10 e-cm on the neutron electron dipole moment,
75
Implying a value of |s| of <, 10 . It is_ necessary that 3 be very small.

There are various points of view possible:

1. 9 is an arbitrary , irameter in the most general rcnormalizsble QCD

Lagrangian and, since it tn subject to divergent renonnalizations from

higher order loops, must be considered not Implausibly small.

 - 103 -

«L ">B <L «3

Fig. 2 1 . P o s s i b l e mechanism f o r spontaneous c h i r a l symmetry b r e a k i n g .

 T#r

Pig. 22. Possible Instancon mechanism for spontaneous chiral

symmetry breaking.
 2, f^r strong interactions one simply Imposes the condition 9 • 0

by hand as a symmetry criterion: T-invariance of strong '.nteractions is

demanded from the start, this does not solve the problem completely;

wtak interactions and Higga-eoupllngs can reintroduce a non-vanishing 6

by radiative effects. Furthermore, these effects can again be divergent,

so one can argue they are not small- On the other han it is not neces­

sarily so that the Infinity gives a large value of 3. Tor example, in

the popular SU{5) grand-unified model, the "infinite" CP violating

effects only leak into the strong interactions in U t h order. The

6 0
nominal order of magnitude i s

f log -

This size is safely "small" (— log ™ <. 1 for any reasonable value of •>).
(7.17)

3. V* saw that in the limit of vanishing quark mass, the vacuum

tunnelling (and therefore CP violating effect) vas suppressed, and that

In the presence of quark mass the tunnelling amplitude Is multiplied by

a factor proportional to m m.m or better

/J-i + if.jat. (7 . 18)

V U d s/ u d

Thus, vere the "hare" current-algebra mass of at least one quark to be zero,
91
v» could svold the problem. The best candidate 1 B the up-quark, whose

mass is estimated, from current algebra considerations, to be -4 MiV.

However, zero mass seems to go against successful current-algebra and

SU(3) calculations.
 - IG6 -

4. The mechanism of qua..,-mass generation can be modified. Peccoi

92
and Quiim found that by allowing an additional U(l) syrame: ry in the

Higgs-sectDr, which was then spontaneously broken, that the CP violating

phase could be sloughed off Into that U(l) phase of the Higgs sector.
0 t
However, Weinberg andtfilczek" *then showed that there should be an

almost massless 0 Goldstone-boson (the axion) with mass £, 10-100 kcV.

The couplings of this object to matter are sufficiently strong that it

05
should probably have been seen.

Among these options, probably the least unpalatable is option (2):

6 is put to zero by hand in the strong QCD theory, with the weak-

Interaction and Higgs contributions required to be a small perturbation.

We have In this section not provided much of any idea how these

effects are calculated quantitatively. The most appropriate and powerful

86
technique utilizes the Feynman path-integral, but even that method is

technically quite difficult. Considerable uncertainty remains in the

magnitude of these CP violating and other instanton-induced effects.

8. IDEAS ABOUT CONFINEMENT

A great deal of effort has gone into trying to understand the con­

finement problem, and considerable insight has been attained. Neverthe­

less, there is no general agreement that the question is understood. Here

we shall only mention in the most superficial way some of the approaches:

a. Lattice 0.CD; One approximates the continuous system with a single

cubic lattice, with quarks living on sites and gauge field living on
97
links. The basic element is a line integral

if I • Jx
 - 107 -

connecting neighboring sites A and B, and the action Is a sum of contri­

butions of the W s at a tine taken around an elementary face (plaquette)

of the lattice. This theory is simplest in the strong coupling limit.

87
There Wilson showed that them exists a linear potential between heavy

sources—even for lattice QED, The field chooses the shortest path between
2
sources, and any fluctuation costs extra powers of g in energy.

The central problems are to connect the strong-coupling limit to a

weak coupling theory at short distances, and to demonstrate a distinction

in that limit between the abelian QED and nonabclian QCD theories.

Ideally one should be able to compute the string-tension In terms of the

perturbation-theory parameter A which controls the value of the running

coupling constant a at short distances.

Important progress has been recently reported. Kogut, Pearson, and

38
Shlgemitsu calculate the strong-coupling expansion to several orders
89
and use Pade approximation to extrapolate toward weak coupling. Creutz,

using a technique originated by Wilson, reduces the problem to an equivalent

statistical-mechanical calculation of a free-energy, which he does via

Monte-Carla techniques on a computer. Both groups, as well as Wilson,

find evidence for an abrupt transition from weak to strong-coupling at a

critical distance-scale, and at a relatively small value of a — 0.1,

90 91
b. MIT Bag: The HIT bag model was originally formulated at a

level more phenomenologlcal than QCD. It views the vacuum as a complicated

medium, and a hadron a "hole" in the vacuum which is simpler, at least

as seen by quarks and gluons. There Is an energy cost in creating such a

hole; also the vacuum pressure on the hole is compensated by the pressure

of the quarks and/or gluons in the interior, leading to a stable hadron.

 - 108 -

This picture has enjoyed reasonable phenomenologicul success. The QCD

interest lies in making a connection of this picture to the QCD Hamiltonian.

The ideas, not necessarily exclusive, include:

92
1) Princeton program. Callun, Dashen, and Gross argue that the

exterior vacuum (viewed 4-dimensionally) is a dense plasma of instanton

events. However they argue that in the presence of color electric fields

(i.e., In bag interiors), instanton effects are suppressed and one has a

relatively dilute gas. Their calculation o£ relative properties of these

two phases rests heavily upon the use a£ analogies to statistical

mechanics and on taking account of couplings betwacn instantons.

2) Analogy with Heissner Effect; Electric-Magnetic Duality. A

monopole in a superconductor undergoes confinement in the way envisaged

for QCD: a quantized vortex line connects a monopole-antimonopole pair,

93
providing a linear potential. To exploit this analogy in QCD, one

needs to reverse the role of electric and magnetic field; the QCD string

contains electric flux, not magnetic. There does exist same E»-»B duality
914
in QCD. But this program is evidently a difficult one and at present

is not complete.

c. Schwin^er-Dyson Equations: Another attack uses the Schwinger-

95
Dysoa equations, which sum up the Feynman-diagram expansion. The goal

is typically to find a nonperturbative solution consistent with a q

behavior in the gluon propagator, signalling a confining potential.

Problems include how to truncate the infinite set of coupled nonlinear

equations for the Green's functions, how to enforce gauge-invarlance, and

how to interpret the result (the Green's function of the gluon Is itself
 - 109 -

gauge-dependent). On top of this is the straight technical problem, even

after brutal truncation, of solving complicated sets of nonlinear integral

equations.

d. qGD Strtoga* This approach treats a piece of bare string,

possibly closed, as the basic degree of freedom instead of point quanta.

B
- •+
i f A • ds
Thus the field degrees of freedom are: :e :. The problem is

to establish equations of motion and a consistent quantum-mechanical

96
formalism, and then determine properties of dressed strings.

All these, and others not mentioned, are the subject matter of

courageous and difficult research. There is optimism that the problem

can be understood; perhaps one of these approaches will provide some

answers.

9. ALTEBHATTVES TO QCD

It is becoming hard nowadays to find serious work on strong interac­

tions which does not start with QCD. This is less a consequence of over­

whelming evidence in favor of QCD as it is a consequence of a lack of

serious contenders. In my mind the two strongest contenders are .the string-

nodal and the Pati-Salam scheme. In the case of the string model, one

simply asserts (without any backup from field theory, etc.) that quarks

•re tied together by strings of unspecified structure and basic origin.

It has a close relationship with dual models and the topological expansion

of the S-matrlx. A major difference with QCD lies in the absence of

the gluon degrees of freedom. The recent PETHA data does not encourage

this point of view.

The Pati-Salara scheme, based on Han-Nambu integer-charged quarks, has

97
undergone a considerable degree of evolution. The apparent discrepancy
 - no -

with the fractional charge measured in electroproduction can be avoided

if the theory is gauged. However, it appears difficult if not impossible

to avoid a low mass (<£ GeV) boson U which mixes with the photon and
98
which has s large lcptonic width. This seems to be ruled out on

experimental grounds. Also, Che T lepton and b quark do not fit very

comfortably into the scheme.

tfhy the difficulty in finding alternatives? It is simply the short

but restrictive list of reasons listed in the introduction, plus the

problem of incorporation of the parton-oodel picture at short distances

(solved in QCD by the asymptotic freedom property).

QCD does pass the test on these issues. It possesses as well the

distinguished pedigree of being a local gauge theory liko QED, which

puts the color degree of freedom to work dynamically. It makes it no

surprise that it is so widely accepted. Nevertheless QCD does need much

better experimental support for it to be truly confirmed. For me, the

most reliable tests are those which can be imagined to be carried out in

a femtouniverae. These include (1) measurement of the c e~ total cross

section to an accuracy sufficient to see the radiative correction and

(2) measurement of a via e e"-» qqg, with the final partons In a highly
a

non-colllnear final state (i.e. — 120° away from their neighbors). One

will then want to check the gluon spin by measurement of angular correla­

tions in this or other processes.

10. CONCLUSIONS

QCD is a theory of strong Interactions which has a starting point as

fundamcnt.il as QED, While there are major mathematical difficulties in

even a successful formulation of QCD as a quantum f.'eld theory, these

 - Ill -

difficulties are ameliorated by quantizing the theory in a tiny box Cfemto-

univerae) with a small coupling constant. There do exist processes which

we observe and which in principle fit into the femtouniverse. These

processes can be calculated perturbatively; in the femtouniverse QCD con­

verges as well as QED.

However, as the box size grows there appear at least six crises:

1. The Gribov gauge-fixing ambiguity impedes the elimination of

unphysical degrees of freedom from the canonical (temporal-gauge)formalism.

2. The running coupling constant becomes strong at large distances,

and perturbation theory goes out of control.

3. Large instanton effects occur, complicating the question of

topological structure of the QCD vacuum.

4. The vacuum rotor-nodes mix with transverse-gluon modes and may

have significant physical effects.

5. Gluonlum and/or color-singlet hndrons should appear as asymptotic

states and quarks and gluons should disappear. How (and at what distance

scale) docs this happen?

6. With light quarks pTcsent, spontaneous breakdown of chiral sym­

metry should occur. Again at what distance scale (and how) does this

happen.

But while there remain many unanswered question;, about the large-

distance aspects of QCD, the small-distance behavior appears to be

comprehensible. If the theory is correct, the problem of understanding

strong-interact ion dynamics can to a great extent be decoupled from the

problem of Interrelating the strong, weak and electromagnetic force at

extremely short distances. That alone would be a great step forward In our

understanding of elementary particles.

 One of the Maxwell Equations

| D - |(x,t) - eJ0<x,t»<r - 0 (A.l)

is an equation of constraint. The existence of this constraint 13 con­

nected with the residual gaugc-invariance of the theory: the condition

A-"0 docs not completely fix the gauge. Tlrae-lndepcndent gauge trans­

formations

X - sX'S -1
+ ^CVS)S~ l
(A.2)

with

35

a -° ^-3)
can be still carried out. The Causs-Lau operator (5 • E - eJ°) is actu­

ally the infinitesimal generator of these gauge transformations. To see

this write

Six) - 1 - ieM(x) (A.4)

with SA(x) a 3x3 matrix. Then by definition

^i " ? 6' " I W W t ?< >1

_ X +
'<5A(x) (A.5)

Now in the big HUbcrt-spaee of quantized fields we write the unitary

transformation if

•if - 1 + i Tr /d y |D • E(y) - *±Jy)]

3
5A.(y) (A.6)

Then using tin* canonical commutators, we fl nJ

l
/TA(x)//~ - A(x) + ?A(x) (A.?)

with 5A(x) given above in Equation ( A . 5 ) .

 - in -

Likewise the unitary transformation of the quarks is also generated:

-1
<?/q<x)V/ - (1 + te5A(x))<i<x) (A.8)

The Uamiltonian has a very large symmetry under residual gauge trans­

formations. The idea of gauge-fixing is to find all the extra trivial

coordinates and remove them from the formalism. In QED this is easy to

do,-as was illustrated in Chapter 3.

How we ask what the corresponding procedure is in QCD. Evidently

ve would like to copy CjED as much as possible. He may do this as before

by Introducing transverse and longitudinal parts of £ and J and eliminat­

ing A. and E. from the theory. He begin with the electromagnetic poten­

tial and assume there exists a time-independent gauge-transformation

which renders A purely transverse

$ • I* - 0 or A' - J' (A.9)

This ia not self-evident; the equation for the gauge-transformation S is

- 1
V ' (S'^t-S ?^) - 0 (A. 10)
leA
which is not at all transparent. <Note that for QED, with S - e , the

above equation is just

2
V A - ? * A" (A.11)

which allows the solution

*<») - /> y w p ^
3
r ' *<>•>) (A l2
- >
If this ifl carried out, then & Ls removed from the llacnlltonian, inasuuch

«• H is invariant under such gauge transformations. To remove E , define

 - IU -

and eliminate £ using the Gauss' law constraint

D • E - D • V£ + D • E T

- V - D£ - ie[j\, - | ] - T eJg (A.U)

It is easiest at this point to expand all 3 * 3 matrices in terns of X-

matrices. Then $ can be obtained formally via

3
A
» (X) - fd y K ( , A)[ef
AD x yl
BCD
^(y)-E^(y) + J ° ]
C CA.15)

with the kernel K "defined" by

(3 • 8 > K(x.y;A)
M BC

! 8 0 3
B V K(«,y;A) AC - of* ? " • VK^ - 5 U-y> (A.16)

In s h o r t h a n d n o t a t i o n

K - ' ; — (A.17)
7 • D

With $ defined above, ve finally end up with the Hamiltonian

3 2 2 + 1
H - _/d x|Tr(E^ + B + <v£) ) + * ( a • <P" - e ^ ) + Bra)*} (A. 18)

This somewhat cumbersome form directly generalizes the procedure In QED.

tt Is quite satisfactory for weak-coupling applications, but unsatisfac­

tory fot strong coupling. This La because it is known that there exist

homogeneous solutlon» for the Green's function K. Roughly speaking, the

A • * term ie a potential which supports "bound states" provided A is

big enough. If the region where A is large has a size L this means A £

1/eL In order to compensate the kinetic-energy term.

Another way of seeing what this means is to look at the gauge-

condition

5 •% - 0 (A.19)
We assumed this could be done in a unique way. However, existence of

honogeneoua solutions of V • D£ - 0 implies that the above condition,

Equacion (A.19), is not unique: namely there exist gauge-equivalent

fields which each satisfy V * A. - 0. That la, when A. is chosen such

that these solutions are close to each other ue have

V - A - $ - ( & + ?£)-0 (A.20)

However, from Equation (A.6), (D'E-eJ.) generates gauge transformations

A-ft 3
d y (E • E(y)SA.(y),A(*)]

- £5A (A.21)

Hence

? - JJA - 0 (A. 22)

Hon-trlvial solutions of Equation (A. 22) thus imply thac fixing ]£

to be transverse does not mean that the gauge has been fixed.
 - 116 -

APPENDIX B

Even the vacuum of QCD turns out to have a quite complicated struc­

ture for essentially topological reasons. We shall enter this subject here

from what is apparently quite an oblique way. It is not the historical

path, but we hope it is more physically comprehensible." We start by

adding a terra to the Lagrangian density of the theory with the form

2 8
32. "™ - -

where t h e dual F of F i s defined a s

uv u\>

~uv 2 uvae^

That is, Che dual of E is B, and vice versa. This interaction term is

evidently parity-violating but C-conserving; hence by the TCP theorem

also time-reversal non-Invariant. We might ask why in the world one

would gratuitously throw such a coupling into the strong interactions.

One answer is why not: the term is renonnalizable ^ad there is a school

of thinking that says that one should write- down tin raost general retior-

malizafcle theory whenever possible. But even if otv restricts the

strong interactions by hand to be CP-lnvari.nit by setting 3 = 0, CP vio­

la.ion elsewhere in the theory can eventually -eak bj^k into the strong

interaction ana induce such J torn. So it is reasonable to expect such

a coupling at some level, even if it is q'.ite sail 1. However this is

not the only reason that such a term is of interest- The 9FF Interaction
 117 •

ia quite peculiar, because it turns out that the Lagrangian density y "

la a total divergence. This Implies that the Lagranginn L can be written

as a tine derivative of a function of A. To see this easily we go to

A--0 gauge and write

j . . S^ IT ft -ti\
J
4. ~

where ve have assumed that A vanishes as | x J-+" so that integration by

parts may be carried out. How when the Lagrangian contains a total time

derivative, it ia a simple matter to calculate its effect. Ve review

tbii for a system with a finite number of degrees of freedom. Let

dF(c. )

L L +
<vV • o<vV -it—

The equations of motion are unchanged:

old new

The Hamiltonian formalism changes a little. The new momentum is

However the Haallt,onian remains fona-invarlant (in order to preserve the

equations of notionl):
 old new

-".(-if--)
and

(B 6
[pi-'tj-fpi"" • « ] - - " « -'
Thus the effect of the extra term F can bo taken into accounb by incor­

porating an extra phase into the wave function. That is, if

then upon defining

1F(q
f(q,t) - e " V <q,t)
0 CB.8)

one finds

HCp.qWq.t) - i " ^ . ' ^ (B.9)

inasmuch as e i F ( l , )
pe* 1 F
^ - p - ~ . This looka like it has a trivial
3q
effect on the theory. The spectrum of elgenfunctions and eigenvalues of

H ara simply determined in terms of those of H_, In particular the

energy-eigenvalues are the sane. For QCD, the transformation on the

great big vavefunctlon for the field-theory is given by

_ i N < A ) e
? (A) - e
e v (A)
0 (B.10)

vlth
 119 •

Why 1« this of any interest at all? For pure time-independent gauges

(E_-JJ" 0) the operator N takes the form

B 1 2
i-^'/^'ijkWk
24» 2 < - >

with

I - «-% <B
- U)

It is here that topological considerations enter. If we make an infini­

tesimal, time-independent gauge transformation on the above; A which van­

ishes at « we find

ijGO + [*£(*). ^ ( x ) ] + V «4(x)± (B.14)

and

Tr d3x E + 3c
™' ^ 7 / I w [*^~&] «* 4 : (***A)(
- 0 (B.15)

Thus H Is Invariant under infinitesimal ^auge transformations which

approach the identity transformation as x -» °». Thus it is also invariant

under finite gauge transformations U of the same type which can be

reached from the identity continuously, Xt might seeo possible there­

fore to restrict ourselves to potentials A which can be reached continu­

ously from the trivial potential A = 0; in this case (in the absence of

gauge fields £ and B), N would vanish and the wave function would not

depend upon 6. What follows will demonstrate that things aren't that

simple and that (1) there exist other pure-gauge field-configurations A*

which cannot be reached from the identity by gauge-transformations (at

least those which,approach unity at x * « ) , and which are characterized

by integrr eigenvalues n of the N-operator, and (2) these n statu are

dynamically coupled together (by the Instanton phenomenon), and Chat

(3) the coupling is Important enough that physical observables depend

upon 6, despite the 1'act that its presence in the Lagrangian did not

affect the equations of notion of the theory.

To do this, ve simply exhibit a prototype example of a gauge func­

tion which produces a non-vanishing value of H. We write, as in the

text

U - C exp \~^- f (r)J (B.16)

with C some constant matrix and with the Pauli-matrices T • (X.,X ,X-,) 2

0 1 0\ /0 -i 0 \ /l 0 G\

( l f l
loo) . - ( i oo
2 V °"
0 0 0/ \0 0 0/ \0 0 0/ <B.17)
defined as the first 3 of Gell-Mann's A-matrices. The topological games

come from the curious coupling of internal-symmetry matrices r to coordi­

nates. In order that U •*• 1, as r •*• , we must have, as r •* •

<B

f(r) •*• irn and C - C-l) n

(B.18)

with n an integer. In order that U be nonsingular a3 r -*• 0, ve must also

n
have £(r) •+• 0 as r •+ 0; hence l£ * (-) as r •+• 0. The integer n will in

fact provide the characterization of the n-vacuua. Because N is invari­

ant under continuous gauge transformations (provided they are trivial at

the boundary) we may choose a Binple form for f(r), in particular push

it out near to the boundary of the region, which we take as a large

sphere of radius R. Notice that the action of U is to rotate a quark in

the internal space by nn as one goes out to •» from the origin. The axis
of rotation in the internal apace is dependent on the direction In real

•pace that one travels. To calculate N we take a snail solid angle

s m a 1 1 a n d
All near the north pole (x» - R; x.,x 2 ) calculate the components

of X with

and.J£ given by Equation (B.16). Elementary estimates give

iT
-i r i 1
ieA B U Cl — - sin f(r) + terns odd in x

-i TIT 2 I

t e a , l T (B 20
* 2 3H - >
The t e n s odd i n X and y w i l l vanish upon Averaging over a r e g i o n c e n ­

t e r e d •yrametrically about x * y - 0 . I n s e r t i o n i n t o Equation (B.12) for N

yields

2
AN ~ ~ r • 6 • 2 i /dxdy f ^ | sin f (r)f ' ( r j
/
24TT •/ -0 R
x,y sirall

2 /
(B.21)
?- •* -o

What have we accomplished? For one thing we have finally uncovered the
2
reason for the factor 3ZTT in Equation (B.l). But we emphasize that it

is not yet physica: we have merely found a topological classification

o£ gauge-potentials A in term of a single Integer n. If one restricts

oneself to potentials which vanish at spatial infinity, then there ia no

gauge-transformation obtainable front the Identity by a product of

 - 122 -

Infinitesimal transformations which takes a potential characterized by

one n-vilue to a potential characterized by a different n-value.

Enter the instanton. While the n-vacuua are only coupled by gauge

transformations nonvanlshing at the surface at », there can be dynamical

coupling af the n-vacuua. For simplicity consider traveling In A-space

(remember the A(x)'s are coordinates In the great big Hllbert-spaco; the

E(x)'s are momenta) from the origin A « 0 In a straight line to an A/x)

In the topological sector n « 1. The intermediate A's cannot be pure

gauge; hence by definition there will be B-field in between. Thus the

potential energy

2 22
V(A) - T r / d \ B (x) (B* >

Increases as one proceeds away from n - 0 and then must decrease again as

one reaches the n « 1 pure-gauge configuration (c.f. Figure 6). Because

the great big Hilbert spsce is so multidimensional, there are many paths

one may travel in going from A. - 0 to the n - 1 gauge configuration;

nevertheless they all share the feature that there is a potential barrier

(unless one travels via the coordinates on the boundary), However evert If

one cannot go around, one can still go through the barrier by quantum

mechanical tunnelling. Suppose that e is small. Wo saw that foe a glob of

potential with n > 0 and spatial extent ~ L , we must have dlmensionally

(c.f. Equation (B.12))

oW> 1 (B.23)
or

The intermediate B-field would then be expected to be

B~~V (b,25)
 V(A) - -r- CB.26)

The height of the barrier is ~e~ ; hence for small e we have a large

thick barrier and a. small tunnelling probability. For such a situation

(and only for such a situation) there exists an easy, albeit rather

crude, estimate for the tunnelling amplitude: it is the scmicLasslcal

approximation. To get the answer, again retreat to a system vith a

finite number of degrees of freedom.

H
* - | L Pi + V(q ...q ) lf n (B.27)
' i-i

with V(0) - 0 and VCQ., ...,&.) « 0, For a big chick barrier we write in
the classically forbidden region
S < q )
HO ~ e " (B.26)

The tunnelling amplitude is then-e"' ^Q) - ( )]^ m i e a s l s l a c g e

_2
(coefficient e ) , the unwritten factors which normalize the wave func­
tion axe relatively inconsequential. There does exist a rather sophisti­
100
cated technology for their calculation (the old-fashioned way involves
101
what is known as the Van Vleck determinant ) but we shall not go into
this aspect of the problem here. It will be enough to obtain S(Q). The
equation for S in semiclassical approximation (3V/34 small compared to
rate of falloff of i|0 is

*?&)
i Z ( £ - ) * »<q,> * 0 (B.29)
Other than a change In relative sign, this is just the Hamilton-Jacob!

equation for the phase point q in the potential v(q). The sign-charge

(which originates in the e~ instead of the WKB e ) allows us to Inter­

pret Equation (B,29) as the real classical motion of Lhe phase-point q

starting at rest at Che origin and ending at the point Q (again At rest)

in an attractive potential V'£q)=-V(q). The quantity S(Q) is, from

claasica.1 theory, the action J associated with such a classical notion

starting at 0 and ending at Q. To see this, simply retreat from the

Hamiltonian formalism, writing

Hence the action J is given by

J - y L dt - J <*t £ ^ ^ " - S(Q> - SCO) (B.31)

inasmuch as the notion begins at q - 0 and ends at q^'Q^*

One last point before reverting to the real continuum QCD problem:

instead of changing the sign of th»i potential, we could equally well

change t to it, since the equations of motion

v-*f «•«
undergo the desired sign change.
Ljt ua now recapitulate. We get a tunnelling amplitude by the fol­
lowing procedure
 1) replace t by It;

2) solve the classical equations for this system, requiring

E " B " 0 a t t + " (and for us jS(x) and £(x) •+ Q as x •+ =

as well);

3) calculate the classical action J for this solution. The

desired tunnelling amplitude is then ~ e ~ .

We see that we need a classical solution of th" Euclidean QCD equations

with finite action in order to couple together all the n-vacuua. This

solution was found by Bela"ln et al. " To find It, it 1: best to abandon
3

A_"0 gauge and exploit the full 4-dlmen5ional symmetry of Euclidean QCD.

n r
He search first for the form of the potentials A as p • /t + x -*• «.

They Mist be pure gauge, fall jff as p~ and involve the Paull-matrices

T. A choice which satisfies this-t

and
+ looks
it • rright is
CB.33)
/7T7
This has a singularity at the origin but, as we shall soon see, otherwise

has the right properties to induce a transition between states of differ­

ing n. Then a choice for Che solution Is

L
ieA - £(o)S~ S U (B.34)

with

f(p> - 1 p » -

ftp) » 0 p - 0 (B.35)
Direct, tedious calculation shows that
„2
B 36
f(p) - V T <- >
does the job. The potential A is indeed no longer singular at x - 0,

and the QCD equations of notion are satisfied,.

He now can put together all the pieces. In doing this let us again

recapitulate what we have done. He introduced a term into the Lagrangian

2
1
density ST of the form -*=—=• TrF T? , This term is a total divergence
wu

p V
16ir ~
63 K ; hence the Lagrangian L w . i modified by a total time derivative
e-jT- /K d x. which simply put a coordinate-dependent phase on the wave

function in the great big Hilbert apace. The phase-operator 9N

in a representation in which the potentials are diagonal, measures a

topological property of the potentials, characterized by an integer n.

Tunnelling between states of different n is possible if there exist

Classical solutions of the equations of motion with finite action, and

which (in A - 0 gauge) take one from a pure gauge configuration of given

n at t--°> to a pure gauge configuration of a. different n at t " + " . We

apparently found such a solution, albeit in a different gauge. To show

that the solution indeed inducea a tunnelling transition between states

of different n, we introduce in the instanton gauge potential at large

distances, Equation (B.34), the polar coordinates shown in Figure 23.

t • p cos iji

r - p ain * (S.37)

He may then write

1—
U - -e r" (B.38)

Hence at * • 0 (t-t-+ = , r-*-0) U - -1,

while at <l> • * (t-*-", r-t-0) U - + 1 , How make a continuous gauge trans­

formation to a new U such that

Fig. 23. Effec:ts of a gauge transformation in the asymptotic gauge
.wi.>i'"*.als surrounding an^instanton: (a) "Covajiant gauge'
) V

and (b) a gauge in which A = 0 in the past and A is an

expanding gauge-bubble in the future.
 e
JZ - " ~ ^ «#> (B.39)

with f(0) = 0 and f (TT) = TT unchanged, but where f(^i) makes the flip from

0 to TT at small angles (i.e., in the future). Thus in the post U = l and

A 3 0 , while in Che future U takes the same farm as we had for the proto­
a
type gauge configuration with n I. It follows that the instanton

induces a tunnelling transition between topological sectors with |fin| =• 1,

To calculate the tunnelling amplitude associated with a single

insranton, we need calculate only Che Euclidean action J of the instanton.

It is

* 2Tr/
rr/d*x | • § (B.40)

by the Schwartz Inequality. We use thia inequality because we recall

that E • B is our total divergence, already evaluated from topological

considerations. From Equations (B.2) and (0.11)

2 I t / A f . | - ^ / d t f S - &U [«->-«-->] - *£ (B.41)

J
e c e

Hence the tunnelling amplitude is S e" " . It is in fact the case that

the instanton solution is self-dual, i.e., E« B, and the inequality is an

equality.

Wii now can crudely estimate the effect on the wave function of the

system. The situation is analogous to the problem of a particle in a

periodic potential
The elgenftinctions arc not packets localized near A « 0 ( n ° 0 ) , but instead

the Bloca-waves

¥ (A) (B.42)
•*£•

This i3 in fact Che structure we already constructed from use of the

gauge-invariant surface term. This much follows essentially from a
requirement of gauge-invarlance and not o£ existence of instantons. What
the ins tan ton does i s to provide a coupling between n-vacuua which i s
not a surface-effect, but a volume effect. That i s , the effective
Bamiltonian for tunnelling between adjacent n-vacuua will, for dimen­
sional reasons, be

B M
H ' ^ V / ^ B fin.a-Z <B.«)

The volume factor V occurs because the Instanton can be located anywhere
in space ( i . e . , the choice of which set of coordinates A,(x) do the
tunnelling i s open and must be Bummed over. Likewise the instanton size
\ is arbitrary and must be summed aver). Since H' has dimension of
inverse length, dimensional analysis gives the weight factor. There is
an additional factor e~ which requires study of the Van Vleck determinant
0 0 , 1 0 1
which normalizes the wave function at the classical turning p o i n t . "
The integral over X is infrared-divergent; hence in a finite and very
s> 11 volume, the most important Instanton is the one which just f i t s
into the box. IE the box i s small enough that perturbation theory ia

Justified, then in fact the instanton effects remain very small.

The effect of H' on a 6-gtate i s simple co calculate. Recall

where Y _,(A) and ^-AICA) differ from ? (A) by an additional lump of

gauge configuration somewhere. Thus a 6-vacuum i s an eigenfunction of

H' /V with eigenvalue

2
8,
2
„, , „ . -2V k Cos e / • &
9 e e «, , ? > ( )

2 6 5
U ) •/ X
where in general |(T> might not be equal to |0> but could hj. deformed

into It by a continuous gauge transformation. However physical states

are to be gauge-invariant; hence for states ]s> in that subspace of the

great big Hilbert apace which are physical, we can replace le> by jfi>.

Hence for practical purposes H* is in fact diagonal. We see in particu­

lar that the vacuum energy density ia Q-dependent, with 6-0 being the

state of lowest energy.

 - 131 -

REFEREMCES

If you don't know, you are invited to consult someone else's

lectures.

S. Brodsky, these proceedings.

An introduction to this method can be found in J. C. Taylor, "Gauge

Theories of the Weak Interactions," Cambridge University Press, 1976.

Actually an exception to this might be the plon in the limit of

massless u and d quarks. In thut circumstance pions are believed

to be massless Coldstone bosons. However, n^ne of that appears to

have ouch connection with the phenomenon of QCD confinement.

A recent review is given by S. Weinberg, I. I. Rabi Festschrift,

to be published by the New York Academy of Sciences.

See, for example, D. Robson, Kucl. Phys. B130, 328 (197"/), also

K. Roller and T. Walsh, Nucl. Phys, B140, 449 (1978). A recent

detailed study is given by J. Coyne, P. Fishbane and S- Heshkov,

NBS/Vlrginia preprint (1980).

S. Weinberg, Phys, Rev. Oil, 3583 (1975,,

See Robson, Ref. 6, for a discussion.

Actually an indefinite number of gluons might "bleach" the color

field and produce a color singlet state (of fractional charge).

That this is a theoretical possibility—although not a verv desir­

able one—was emphasized to me by Marvin Uclnstein.

For a review, see J. Weis, Acta. Phys. PoJon. B9. 105] (l'J78).

W. Heisenberg and H. Euler, Zeitschrlft fur Fhysin 9JJ, 714 (1936);

J. Schwinger, Phys. Rev. jJ2_, 664 (1951).

A recent study of this mechanism has been uade by A, Casher,

H. Heuberger, and S. Kussinov, Phys. Rev. 1)20, 179 (1979); also

 - 132 -

Tel-Aviv preprint TAUP-747/79; see also H, Neuberger, preprint

TAUP-721/79.

13. However, the OZI rule does give some motivation for a small coupling.

14. I thank Stanley Mandelstam for pointing this out to me.

15. J. BjorKen, Lecture Notes IJI Physics 56, "Current Induced Reactions,"

ed. by J. Koeraer, G. Kramer, and D. Schildkuecht, Springer-Verlag

(H.v.), 1975.
1
16. Were A* not taken to be traceless, the coefficient of the unit

matrix would be & tlv'l) degree of freedom not coupled to color at all.

There is no evidence for its existence as a physical degree of

freedom.
J
17. A IL'-L: the quantity Tr J J is a color invariant.
2 2

18. The Lagrangian density is 'J* « Tr <E -B ) .

19. L. Faddeev ana V. Popov, Phye. Lett. 25B, 29 (1967).

20. Our notation follows that of Reference 1.

21. S. Frenkel, Phys. Rev. D13, 2325 (J^/6), W. KonCtachny and W. Kuaraer,

Nucl. Phys. B100, 106 (1975); see also J. Willemsen, Phys. Rev. D17,

574 (1978).

22. This can also be explicitly checked without much difficulty.

23. V. N. Gribov, SLAC preprint SLAC-TRANS-176; S. Mandelstam, invited

talk at Washington Meeting of the American Physical Society, 1977.

24. Generalizations of the problems have been made by M. Atiyah and

I. Singer. Quantization in axial gauge (A,=0) provides some

improvement; yet there remains soue residual gauge invarlance, as

well as possible problems with boundary terms. R. Jackiw and

J. Goldstone [Phys, Lett. T^B, 81 (1978)] have presented an SU(2)

 - m-
example of quantization without subsidiary conditions; however the

resultant: formalism contain* some possibly troublesome singular

expressions, as wall as appearing to be rather unmanageable. Sec

also V. Baluni and D. Grossman, Phys. Lett. 78B. 226 (1978).

25. We are thinking here of the LSZ formalism as used in Book II.

26. T. D. Lee and H, Hauenberg, Phys. Rev. 1_32, BL549 (L964);

6 5
T. Kinoshita, J. Math. Phys. 2» ° (1962).

27. S. Drell and T. M. Yan, Phys. Rev. Lett. .25, 316 (1970).

28. See the lectures of B. Stroynowskt, these proceedings, £«r an

up-to-date review.

29. K. Wilson, Phys. Rev. 179, 1499 (1969).

30. G. Sterman and S. Weinberg, Phys. Rev. Letters 39, 1436 (1977).

31. This problem is under study in collaboration with M. Fischler,

32. R. Jackiw and C. Rebbi, Phys. Rev. Letters 37, 172 (1?76);

C. Callan, R. Das*-.an, m d D. Gross, Phys. Letters £3B, 334 (1976).

33. Provided, of course, cne does not travel via field variables A(x)

with x on the boundary of the box.

34. A. Belavin, A. Polyako'?, A. Schwartz, and Y. Tyupkin, Phys. Lett.

S9B, 85 (1975).

35. C. Callan, R. Dashan, and D. Cross, Phys. Rev. ai£, 2717 (1978).

36. S. Coleman, "The Uses of Instantons," Harvard University preprint

to be published in froceedings of the 1977 International School of

Subnuclear Physics "Ettore Majorana."

37. This has been verified in field-theoretical contexts by

J. Willemsen, UC-Santa Cruz preprint (1979).

 - 134 -

38. L. Baulieu, J. Ellis, H. K. Gaillard, and W. Zakrzeuski, Phys.

Lett. 77B, 290 (1978); also N. Andrei and D. Gross, Phys. Rev.

D18, 468 (1978); R. Carlitz and C. Lee, Phys. Rev. D17. 3238 (1978),

39. G. Veneziat:o, Nucl. Phys. B117. 519 (1976).

40. D. Robson £p_. cit.; Reference 6.

41. H. Fri;^sc!i and P, Minkowski, Nuovo Cimento 30A. 393 (1975).

42. R. Jaffe and /.. Johnson, Phys. Lett. bOB, 201 (1976).

43. K. Johnson (private copmunication) argues that the pressure on the

bag from the gluon field implies E > B - The modes considered in

Ref. 42 do not satisfy that constraint.

44. See References 6 and 39; also, P. Freund and Y. Nambu, Phys. Rev.

Letters ^ 4 , 1646 (1975).

45. J. Bolzan, K. Geer, W. Palmer, and S. Pinsky, Phys. Rev. Letters

,35, 419 (1975); also J. Bolzan, W. Palmer, a d S. Pinsky, Phys. Rev.

D14, 3202 (1976).

46. See for example E. Witten, Nucl, Phys. B1S6, 269 (1979) for a ni^e

discussion.

47. I am grateful to R. Brover for emphasizing this to me and for very

useful discussions.

48. M. Chanowitz, Vhya. Rev. D12. 918 (1975); S. Brodsky. D Coyne,

I. DeGrand, and R. Horgan, Phys. Lett, 73B, 203 (1978).

49. R. Barbieri, E. d'Enilio, G. Curchi, and E. Remiddi, Nucl. Phys.

B154, 535 (1979).

50. H. Krammer, Phys. Letters 74B, 361 (1978).

51. A. Billoire,.R. Lacaze, A. Morel, and H. Navelet, Phys. Lett. 80B.

381 (1979).
 - 135 -

52. R. Brandelik et al., phys. Lett. 74B, 292 (1978).

53. J. Schwinger, Phys. Rev. 125, 1043 (1962); 127. 324 (1962). See

also Ref. 59.

54. The approach discussed here has been independencly worked out by

S. Drell and also D. Stump (private communications). They have

studied* the nature of this term in greater depth.

55. This is a consequence of the renormalizability of the theory, a

fact which is not self-evident in this formalism. A direct check

that the next term in the expansion behaves properly is at present

being carried out by S. Drell and R. Hughes (private communication).

56. See for example Chapter 19 of Book II.

57. The history of renormalizability is traced quite completely by

M. Veltman, Proceedings of the 6th International Symposium on

Electron and Photo Interactions at High Energies. ed. H. Rollnik

and V, Pfell, North-Holland (Amsterdam), 1974; p. 429.

58. D.B.T. Jones, Nucl. Phys, B75. 531 (197/ H. Caswell, Phys. Rev.

Lett. 3_3, 244 (1974).

59. T. Appelquist, M. Dine, and I, Muzinich, Phys, Lett. W B , 231

(1977), also Phys. Rev. D8, 207't (1978).

60. H. Thacker, C. quigg, and J. Rosner, Phys. Rev. J318, 287 (1978).

61. J. Richardson, Phys. Lett. JJ2B, 272 (1979), See also B, Hargolis,

R. Roskies, and N. ie Takacsy, McCill University preprint (1978);

R. Levine and Y. Toraozawa, Phys. Rev. D19, 1572 (1979).

62. M. Krammcr and H. Krasemann, preprint DESY 79/20.

 - 136 -

63. For a review, sec J. Jackson, Proceedinfts o£ the 1977_ European

Conference on Particle Physics, Budapest! ed. by L. Jcnik and

1. Montvay (CRIP, Budapest, 1978), p. 601; also K, Gottfried,

Proceedings of the 1977 International. Symposium on Leptons and

Photons at High Energies, Hamburg, ed. F. Gutbrod (DESY, Hamburg,

1978), p. 667.

64. N. Isgur and G. Karl, Phys. Rev. D19. 2653 (1979).

65. A recent review is given by C. Qulgg, Fermiljb preprint FERMILAB-

CONF 79/74 THY, to be published in the Proceedings of the 9th

International Conference on Leptons and Photons at High Energy,

Fenailab, Batavia, 1L (1979).

66. K. Gottfried, Phys. Rev. Lett. 40, 598 (1978); M. Peskin, Nucl.

Phys. B156, 365 (1979); G. Bhanot and M. Peskin, Nucl. Phys. B156.

391 (1979).

67. A. Ore and J. Powell, Phys. Rev. ,75, 1696 (1949).

68. E. Wicten, Harvard preprint HUTP-79/A0Q7 (1979).

69. M. Shifman, A. Vainstein, and V. Zakharov, Nucl, Phys, B147, 385,

448 (1978).

70. V. Novikov, M. Shifman, A. Vainatein, and V, Zakharov, ITEP

preprint (1979).

71. J. Bell and R. Jackiw, Nuovo Cimento 60A, 47 (1969); S. Adler,

Phys. Rev. 177., 2426 (1969).

72. C. 't Hoott, Phys. Rev. Lett. 2]_, 8 (1976); Phys. Rev. D14, 432

(1976).

73. This idea seems to have originated with Callan, Dashen, and Gross,

Ref. 35.
 - 137 -

74. The adlabatic hypothesis is in fact unnecessary. A comprehensive

Study has been made by N. Christ (Columbia University Preprint

CU-Tp-160 (1979)), who relates all this to the work of M. Atiyah

and I. Singer on spectral flow and the index theorem [M. Atiyah,

H. Hitch'en, and 1. Singer, Proc. Nat. Acad. Sci. USA .74, 2662 (1977);

H. Atiyah, V. Patodi, and I. Singer, Math. Proc. Camb. Phil. Soc.

_77, 43 (1975); .79, 71 (1976)].

75. J. KiaUs [Phys, Rev. D18, 3690 (1978)] has explicitly Identified

the levels as they crass zero energy. See also L. Brown, R. Caxlitz,

and C. K. Lee, Pliys. Rev. D16, 417 (1977), A. Schwartz, Phys. Lett.

67B, 172 (1977), R. Jackiw and C. Rebbi, Phys. Rev* D16, 1052 (1977);

H. K. Nielsen and B. Schroer, Kucl. Phys. B127, ',93 (1977).

76. M. Baker, K. Johnson, and R, Willey, Phys. Rev. .163, 1699 (1967);

J. Cornwall and R. Norton, Phys. Rev. DS, 3338 (1973); R. Jaekiw

and K. Johnson, Phys. Rev. D8, 2386 (1973).

77. C. Callan, R. Dashen, and D. Gross, Phye. Rev. D12, 2717 (1978);
1 2 1
see also D. Caldi, Phys. *;av. Lett. J39, (1977); R. Carlitz and

D. Creame-:, Annals of Physics 118, 429 (1979),

78. H. Quinn and M. Ueinstein (private communication).

79. There is a solid current-algebra calculntion connection <j with the

neutron electric dipole moment; R. Crewther, P. diVecchia,

G. Veneziano, and E. Witten, CERN preprint CERN-TH-2735 (1979);

see in this connection V. Baluni, Phys. Rev. D19, 2227 (1979).

80. J. Ellis and H. K. Gaillard, Kucl. Pliys, B150, 141 (1979); B. Morel,

Harvard preprint HUTP-79/A009 (1979).

 - 138 -

81. F. Htlczek, Phys. Rev. Lett. £0, 279 (1978).

82. R. Peccei and H, Qulnn, Phys. Rev. Lett. 3_8_, 1440 (1977).

83. S. Weinberg, Phys. Rev. Lett. 40, 223 (1978).

84. F. Wilczek, Phys. Rev. Lett. 40, 279 (1978).

85. T. Donnelly, S. Freedman, R. Lytel, R. Peccei, and H. Schwartz,

Phy. Rev. Dl*. 1607 (1978).

86. The reader is strongly urged to consult the lectures of Sidney

Coleman, Ref. 36, for a complete and lucid explanation of the

methodology.

87. K. Wilson, Phys. Rev. D10, 2445 (1975).

88. J. Kogut, R. arson, and J, Shigemitsu, Phys. Rev. Lett. 43_, 484

(1979).

89. M. Crcucz, Brookhaven preprint, September 1979.

90. P. Haaenfratz and J. Kuti, Phys. Reports £, 75 (1978). For a

newer review, see R. Jaffe, HIT preprint HIT-CTP-814 (1979).

91. A. Chodoa, R. Jafxc, K. Johnson, C. Thorne, and V. Weisskopf,

Phys. Rev. D9, 3471 (1974); D10, 2599 (1974).

92. C. Callan, R. Dashen, and D. Gross, op. cit., Ref. 35; also Lectures

at La Jolla Institute Workshop on Particle Theory, August 1978

(Princeton preprint), and Phys, Rev. D19, 1826 (1979).

93. H. Niel3Cn and P. Olesen, Hucl. Phys. B61. 45 (1972).

94. For a review, see S. Mandelstam, UC-Bcrkeley preprint UCB-PTH-79/9

(1979), to be published In the Proceedings of the 9th International

Symposium on Lepton and Photon Interactions at High Energyi

Fermilab, Batavia, IL (1979).

 - 139 -

95. K. Anishetty, H. Baker, S. Kim, J. Ball, and F. Zachariasen, Phys.

Lett. 86B, 52 (1979); M. Baker, J. Lucht', P. Lucht, and

F. Zachariasen, Caltech preprint CALT-6S-741 (1979); U. Bar-Gadda,

SLAC preprint SLAC-PUB-2347; S. Mandelstam, Ref. 94 and UC-Berkeley

preprint UCB-PTH 79/8.

96. Y. Natnbu, Proceedings of the 19th International Conference on High

Energy Physics, Tokyo, Japan (1978), p. 971, and references therein.

97. A recent review is given by J. Pati, University of Maryland preprint

r
KDDP-' R-79-066 (1978), to he published in Seoul SynpoBlua on

Elementary Particle Physics, Seoul, Korea (1978).

98. L. B. Okun, H. Voloahin, and V. Zskharov, Moscow preprint ITEP-79

(1979).

99. Our approach roughly follows that of Jackiw and Rebbi (Ref. 32)

and of K. Bitar and S. J. Chang [Phys. Rev. pi7, 486 (i978); D18

435 (1978)].

100, Again, we recommend the Erice Lectures of S. Coleman (Ref. 36)

as an excellent Introduction.

101. J. H. Van Vleck, Proc. Nat. Acad. Sci. _14, 178 (1928).